Intermittent theta burst stimulation at personalized targets reduces the functional connectivity of the default mode network in healthy subjects

Understanding the mechanisms by which transcranial magnetic stimulation protocols exert changes in the default mode network (DMN) is paramount to develop therapeutically more effective approaches in the future. A full session (3000 pulses) of 10 Hz repetitive transcranial magnetic stimulation (HF-rTMS) reduces the functional connectivity (FC) of the DMN and the subgenual anterior cingulate cortex but current understanding of the effects of a single session of intermittent theta burst stimulation (iTBS) on the DMN in healthy subjects is limited. To reduce the effects of inter-individual variability in functional architectures, we used a novel personalized target selection approach based on each subject’s resting state fMRI for an unprecedented investigation into the effects of a single session (1800 pulses) of iTBS over the DMN in healthy controls. 26 healthy subjects participated in a double-blind, crossover, sham-controlled study. After iTBS to the personalized left dorsolateral prefrontal cortex (DLPFC) targets, we investigated the time lapse of effects in the DMN and its relationship to the harm avoidance (HA) personality trait measure (Temperament and Character Inventory/TCI). Approx. 25-30 minutes after stimulation, we observed reduced FC between the DMN and the rostral anterior cingulate cortex (rACC). About 45 minutes after stimulation the FC of rACC strongly decreased further, as did the FC of right anterior insula (rAI) with the DMN. We also report a positive correlation between the FC decrease in the rACC and the HA domain of TCI. Our results show how iTBS at personalized left-DLPFC targets reduces the FC between DMN and the rACC and rAI, regions typically described as nodes of the salience network. We find that HA scores can potentially predict iTBS response, as has been observed for HF-rTMS.


44
Both the large variability of responses in the treatment of depression by the FDA-approved 10 Hz 45 repetitive transcranial magnetic stimulation (rTMS) protocol has led to a world-wide demand for 46 better techniques or improved protocols. The non-inferior antidepressant efficacy of the 3 47 min/session theta burst protocol 1 compared to 37.5 min/sessions of conventional 10 Hz rTMS 48 protocol has played a role in increasing the use of the theta burst protocol for antidepressant 49 treatment 2-5 . The connectivity changes underlying the effects of intermittent theta burst stimulation 50 (iTBS) delivered at the left dorsolateral prefrontal cortex (DLPFC) remain unexplored. Many factors 51 contribute to inter-individual variability, including natural variation in anatomy and functional 52 connectivity. Here we use a previously validated target selection method to improve precision of coil 53 localization and investigated the effects of iTBS on the relevant brain networks that cover the left 54 DLPFC and the anterior cingulate cortex (ACC). 55 The TBS protocol was developed to mimic rodent 6,7 and human hippocampal activity 8 , where a 56 combination of gamma-frequency spike patterns superimposed on theta rhythms 9 was found. It 57 involves application of a burst of three TMS pulses every 20 milliseconds (50 Hz), which is repeated 58 five times per second (5 Hz) 10,11 . When delivered continuously (continuous TBS -cTBS) for 40 59 seconds, it results in reduced corticospinal excitability, while when administered in an intermittent 60 fashion (iTBS) it results in increased corticospinal excitability 9 . Studies of TBS stimulation on motor 61 cortex have shown plasticity changes beyond the duration of stimulation typically lasting in the 62 range of 30 minutes 11,12 . 63 Beyond local effects under the stimulation coil, plasticity changes in brain's altered functional 64 connectivity away from stimulation point e.g. the DLPFC 13 are likely relevant to the treatment of 65 psychiatric disorders, which has been suggested to result from aberrant brain functional 66 connectivity 14 . The DMN, consisting of the medial prefrontal cortex, posterior cingulate cortex and 67 areas of posterior parietal cortex 15 , is usually hyperconnected to subgenual ACC (sgACC) in depression 15,16 . A reduction of this hyperconnectivity has been related to a reduction of symptoms 69 17 . Using 10 Hz rTMS as antidepressant treatment, a study has recently replicated the prediction of 70 symptomatic alleviation in depression when aberrant sgACC connectivity with the DMN is 71 decreased, which happened in responders but not in non-responders 18 . Furthermore, such effects 72 over networks in healthy subjects have been shown in our previous work 19 . Already after a single 73 session of 10 Hz rTMS (3000 pulses), delivered at personalized left DLPFC sites, a reduction in the 74 connectivity between the sgACC and the DMN was evidenced, most strongly in subjects with lower 75 harm avoidance (HA) scores from the Temperament and Character Inventory (TCI). 76 Given the central involvement of the DMN in the pathophysiology of depression and the importance 77 of a shorter protocol such as iTBS in reducing symptoms, 15,20-31 here we aimed to uncover if a single 78 session of a prolonged iTBS protocol (1800 pulses) in healthy subjects would result in reduced DMN 79 connectivity to the ACC. We applied a single session of iTBS at personalized left DLPFC sites and 80 analyzed the DMN during three time windows after stimulation in a double-blind, crossover, and 81 sham-controlled study. Given that the nature of iTBS differs from 10 Hz rTMS, we were interested in 82 whether the modulation across the sgACC is similar. Moreover, based on a negative correlation seen 83 between HA scores and coupling changes of the sgACC and the DMN after one session of 10 Hz 84 rTMS 19 , we hypothesized a correlation between iTBS induced changes in DMN and HA scores. 85

86
Participants 87 Healthy subjects ages 18-65 were enrolled in the study. We evaluated the subjects with structured 88 clinical interviews and ruled out current or prior psychiatric disorders. We performed the 89 experiments in agreement with relevant guidelines and regulations 32,33 . The Ethics Committee of the 90 University of Medical Center Göttingen approved the study protocol and subjects provided their 91 informed consent before investigation. 92

Study design 93
The study reported here with healthy subjects is a sham-controlled, double-blind (subject and 94 interviewer), crossover study. We conducted the experiments over three sessions (each session on a 95 different day, Figure 1) with each session separated by at least one week. 96 interviewer obtained written informed consent from the subject after completing an evaluation of 105 the inclusion and exclusion criteria. Next, we acquired structural T1-weighted MRI and resting state 106 functional MRI (rsfMRI) scans for our novel method of personalized target selection (see Figure 1 for 107 further details). The process of personalized left DLPFC target selection has been described 108 previously 19 . 109

Session 2 and Session 3 110
To allow wash out of any potential iTBS effects session 2 and session 3 were separated by at least a 111 week. After determining the resting motor threshold (RMT), we applied iTBS, at 80% RMT. We 112 navigated to the personalized left DLPFC target using an online neuronavigation system (Visor 1 113 software, ANT Neuro, Enschede, Netherlands). We obtained a pre-iTBS (baseline) rsfMRI scan (R0) 114 followed by three post-iTBS rsfMRI scans. Subjects completed the Positive and Negative Affect 115 Schedule (PANAS 35 ) at the before and after the experiment on session 2 and session 3. This allowed 116 us to follow any short-term changes in the subjects' mood because of iTBS. Figure 1 pictorially 117 details the study design. sham conditions using a, and report results surviving a statistical threshold of p<0.05 FWE whole 166 brain corrected for multiple testing. We ran Pearson's correlation tests between rACC functional 167 connectivity strengths and the harm avoidance domain of the TCI using MATLAB. We used R to run 168 two-way t-tests to compare the scores from YMRS, HAM-D, MADRS, PANAS, VAS and BDI II for real 169 and sham stimulation sessions. 170

Data availability statement 171
Owing to restrictions in the data sharing consent obtained from the participants of the study, the 172 datasets generated and analysed cannot be made publicly available. 173

174
Twenty-nine healthy subjects (11 females, mean age of 28 years +/-8 years) signed up for the study. 175 Two subjects (both females) were dropped from the study due to failure to locate their personalized 176 left DLPFC target and one subject (male) dropped out of study due to discomfort from stimulation. 177 Thus 26 subjects were included in final analysis, none of whom reported any adverse effects during 178 or after stimulation. 179 Our previous work has identified a negative relationship between harm avoidance scores on TCI and 221 the changes induced by 10 Hz rTMS in the right sgACC during R2 rsfMRI compared to R1 rsfMRI 19 . 222 We hence explored if such a relationship existed also between the harm avoidance scores of 223 subjects in the current study and the observed decrease in the functional connectivity of rACC during 224 R2 rsfMRI compared to R1 rsfMRI. We identify a positive correlation between the harm avoidance 225 between the task positive and task negative network is required to quickly reallocate resources 254 towards internal or external stimuli according to immediate demands. It has been shown that this 255 "circuit breaker" role is played by the salience network (SN) 43 , with the dACC and right AI as the main 256 network nodes. We identified these regions as being decoupled from the DMN in healthy subjects 257 after a single session of a prolonged iTBS protocol (1800 pulses) (Figure 3). Although the changes 258 evidenced here may not directly translate to the context of psychopathology, it was very promising 259 to see these nodes as part of our results. The AI is considered to be the essential hub of the SN 260 because it mediates the information flow across the brain to different networks and switches 261 between central-executive and default-mode networks 42,43,46 . 262 In depression, the SN shows aberrant functional connectivity to the DMN and CEN 23,25,47 . One of the 263 network-based hypotheses of depression conjectures that the increased interaction between the SN 264 and the DMN results in pathologically increased allocation of resources to negative information 265 about the self, e.g. ruminative thoughts 30,48 . Considering the proven efficacy of iTBS for treatment of 266 depression and speculating that the effects seen in healthy participants would extend to patients, 267 the mechanism by which iTBS may initially influence the symptomatology of depression could be by 268 Their results stem from seed-based analysis of the sgACC after iTBS. One possible reason for the 278 discrepancy between their and our results may be the method of analysis, as seed-based analysis 279 focuses on the functional connectivity of a predefined ROI while ICA allows exploration of functional 280 connectivity changes of the whole brain without having to pre-define a ROI. In contrast to our 281 previous work that identified the sgACC as the main region decoupled from the DMN after a single 282 session of personalized 10 Hz rTMS 19 , the strongest changes in connectivity after iTBS are not with 283 the sgACC, but rather the rACC and dACC as mentioned above. However, due to the relevance of the 284 sgACC, we further explored the beta weights from this region and evaluated its relation to the left 285 DLPFC at baseline and up to 50-min after stimulation. We evidenced a shift of correlation between 286 these regions from negative to positive within the observation time (Figure 4, green dotted curve). 287 This suggests the participation of the sgACC in the effects driven by iTBS, even though it is not 288 directly engaged by it. The striking similarity between the red curves seen in the rACC after iTBS 289 ( Figure 4) and in the sgACC after 10 Hz TMS ( Figure 5 in Singh et al. 19 ) suggests that sgACC is rather 290 the first target after 10 Hz rTMS (under standard dose of 3000 pulses). Another important aspect 291 that might have contributed to differences between our and the results of Baeken et al. 50 is that we 292 stimulated functionally relevant sites within the left DLPFC, as opposed to their structural selection. 293 Of course, the most profound difference is that our study closely evaluated connectivity changes 294 after one session of iTBS in healthy subjects, whereas Baeken et al. 50 evaluated patients with 295 depression after 20 stimulation sessions. It must be considered that the complexities associated with 296 the underlying pathophysiology of depression could have contributed to differences in how iTBS 297 interacts with brain regions and networks. Our results shed light on other relevant regions that 298 respond to a single session of iTBS in the healthy brain. Future work examining brain networks in 299 patients before and after 20 iTBS treatment sessions should close these knowledge gaps. 300 We also evidenced a positive correlation between changes in the HA score on the TCI and the 301 functional connectivity to the rACC in the DMN ( Figure 5A). This indicates that the higher the 302 subjects scored on the HA domain, the stronger the reduction in functional connectivity. This 303 correlation indicates that it might be possible to utilize the HA to predict the extent of DMN-rACC 304 coupling changes induced by iTBS. Interestingly, the correlation between connectivity changes and 305 HA scores replicates the time window in which this was seen in an independent sample using 10 Hz 306 rTMS 19 , although in opposite direction and in a different brain region, the sgACC. Considering the 307 rACC and the sgACC are the regions whose connectivity to the DMN is affected by iTBS and 10 Hz 308 TMS respectively, we speculate that HA scores may facilitate identification of participants who will 309 present stronger DMN changes in response to TMS protocols. If this holds true for clinical samples, 310 we hypothesize that such a measure could be used to determine beforehand who would benefit 311 most from one stimulation protocol or the other. We hope future research in precision medicine will 312 investigate this using different TMS protocols, considering the direct clinical application and 313 potential relevance to improving treatment response. 314 There are limitations to this study. It should be noted that our choice of a single session of 1800 315 pulses iTBS for uncovering its effect on the DMN in healthy brains was based on ethical reasons, as 316 applying 20 sessions of iTBS as done in patients 51 would not be prudent. While the results presented 317 have been controlled for using a sham stimulation, it must be noted that the method used for sham 318 stimulation allowed some lingering current on the sham side. The strength of this current was low 319 and not enough to elicit motor response. However, the fact remains that the sham condition used 320 was not completely passive and hence could be interpreted as an active sham. Lastly, the diseased 321 state of the brain, e.g. in depressive state, is likely to influence interactions between brain networks 322 in response to multiple sessions of iTBS. Therefore, assumptions based on healthy samples must be 323 made cautiously. 324 In conclusion, by means of a double-blind, sham-controlled crossover study involving healthy 325 subjects, we show that a single session of iTBS results in decoupling of the rostral/dorsal ACC, 326 followed by the mPFC and the right AI, with the DMN. The interaction between the personalized 327 sites of stimulation at the left DLPFC and the rACC shows a progressive shift from negative to 328 positive correlation. Lastly, connectivity changes in the rACC induced by a single real session of iTBS 329 in the healthy brain positively correlated with the HA score on the TCI scale. 330