MRgFUS Pallidothalamic Tractotomy for Chronic Therapy-Resistant Parkinson's Disease in 51 Consecutive Patients: Single Center Experience

Background: There is a long history, beginning in the 1940s, of ablative neurosurgery on the pallidal efferent fibers to treat patients suffering from Parkinson's disease (PD). Since the early 1990s, we undertook a re-actualization of the approach to the subthalamic region, and proposed, on a histological basis, to target specifically the pallidothalamic tract at the level of Forel's field H1. This intervention, the pallidothalamic tractotomy (PTT), has been performed since 2011 using the MR-guided focused ultrasound (MRgFUS) technique. A reappraisal of the histology of the pallidothalamic tract was combined recently with an optimization of our lesioning strategy using thermal dose control. Objective: This study was aimed at demonstrating the efficacy and risk profile of MRgFUS PTT against chronic therapy-resistant PD. Methods: This consecutive case series reflects our current treatment routine and was collected between 2017 and 2018. Fifty-two interventions in 47 patients were included. Fifteen patients received bilateral PTT. The median follow-up was 12 months. Results: The Unified Parkinson's Disease Rating Scale (UPDRS) off-medication postoperative score was compared to the baseline on-medication score and revealed percentage reductions of the mean of 84% for tremor, 70% for rigidity, and 73% for distal hypobradykinesia, all values given for the treated side. Axial items (for voice, trunk and gait) were not significantly improved. PTT achieved 100% suppression of on-medication dyskinesias as well as reduction in pain (p < 0.001), dystonia (p < 0.001) and REM sleep disorders (p < 0.01). Reduction of the mean L-Dopa intake was 55%. Patients reported an 88% mean tremor relief and 82% mean global symptom relief on the operated side and 69% mean global symptom improvement for the whole body. There was no significant change of cognitive functions. The small group of bilateral PTTs at 1 year follow-up shows similar results as compared to unilateral PTTs but does not allow to draw firm conclusions at this point. Conclusion: MRgFUS PTT was shown to be a safe and effective intervention for PD patients, addressing all symptoms, with varying effectiveness. We discuss the need to integrate the preoperative state of the thalamocortical network as well as the psycho-emotional dimension.


INTRODUCTION
Stereotactic neurosurgery for Parkinson's disease (PD) has a long history, beginning in the 1940s, with targets placed on different sections of the pallidothalamic fiber system. Meyers proved in the early 1940s that basal ganglia surgery could be done without impairing consciousness (1,2). After ablations of anterior parts of the striatum, he moved to ansotomies (3), using first mechanical lesioning and later an early high intensity focused ultrasound system (4)(5)(6)(7). Targeting pallidal efferent fibers was further developed and refined by others (8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18). Bilateral lesional approaches have been intensively explored in the sixties mostly (10,(19)(20)(21)(22)(23)(24) and have shown highly variable side-effects, the description of which often lacked precision in term of intensity. In addition, older techniques were less precise and much larger lesions were performed, with sometimes extensive thalamic damage and risk of capsular encroachement. In the early 1990s, deep brain stimulation came to dominate the field, setting aside in most centers further refinements and developments in ablative stereotaxy. Nowadays, in the context of improved technologies, bilateral interventions are being considered to deserve further studies (25).
In the early 1990s, we undertook an update of the subthalamic approach (26,27), later supported by a three-dimensional histological description of the pallidothalamic fiber system in the subthalamus (28). We proposed to target specifically the pallidothalamic tract just below the thalamus at the level of Forel's field H1, where the ansa lenticularis and fasciculus lenticularis converge. We named this intervention pallidothalamic tractotomy (PTT), a more precise denomination than "campotomy, " as Forel described three fiber fields in the subthalamus named "Haubenfascikeln" H, H1 and H2. The PTT has been performed during 20 years with radiofrequency (26) and since 2011 with the MR-guided focused ultrasound (MRgFUS) technique (29). PTT corresponds to an optimized pallidotomy, because it allows, with very limited tissue ablation, an extensive liberation of the thalamocortical dynamics from pallidal overinhibition (see below), while leaving the thalamus intact. A prospective open-label study and a case report describing similar subthalamic approaches have been recently published (30,31). The limited current interest for fiber tract ablations in the subthalamus (but not for chronic stimulation procedures) may be explained (1) by uncertainties concerning the anatomical course of the pallido-and cerebellothalamic tracts (28,32) and (2) by the necessity to hit a small fiber bundle (and not a larger nuclear, pallidal or thalamic, area) closely surrounded by relevant structures, i.e., the mammillothalamic tract (MTT), the subthalamic nucleus, the internal capsule (IC) and the somatosensory thalamus (12,33). Such topographic conditions require a targeting precision inside the millimeter, which the MRgFUS technology can provide (29,34).
There are pathophysiological arguments to choose pallidal efferent fibers among other targets in the treatment of medically refractory chronic PD (26,27,29,35). In Parkinson, a chain reaction is triggered by the loss of dopamine producing cells in the substantia nigra (36)(37)(38)(39)(40). It results in increase and decrease of cell activities in the striatum followed by an overactivity of the internal pallidum, which chronically overinhibits the pallidal-recipient thalamic relay cells in the thalamus (27,41). This overinhibition has been proposed to be at the source of the initiation and maintenance of a thalamocortical dysrhythmic process (35,(42)(43)(44)(45), measurable thanks to quantitative EEG (46,47) and responsible for the production of the parkinsonian symptoms.
Our first experience with MRgFUS PTT was collected by repeating sonications on the same spot (29). This led to the appearance of partial therapeutic effects in the follow-up of some patients. A next step has been to use longer application times. This approach was not sufficient and required further developments. A recent reappraisal of the histological anatomy of the pallidothalamic tract in H1 was combined with an optimization of our lesioning strategy using thermal dose control, with the goal to refine target coverage and thus provide a further step in the prevention of recurrences or partial therapeutic effects (53,54). We hereby present the clinical results in PD of the first consecutive 56 uni-or bilateral MRgFUS PTTs applying this new approach.

Ethics
All patients treated with this protocol signed an informed consent form after having been fully informed about the treatment, its results and risks. No ethical approval was sought because MRgFUS PTT has been approved by the Federal Office of Public Health (FOPH) of Switzerland and is covered by swiss social insurances.

Study Context
Patients were referred to receive specifically MRgFUS interventions. There was no patient randomization and no blinding. This single center consecutive and prospective case series reflects our current treatment routine of chronic therapyresistant PD with the MRgFUS PTT. It was collected between January 2017 and September 2018. Monitoring for primary outcomes as well as cognitive and adverse event evaluations were performed by senior independent neurologists 3 months and 1 year after the procedure, in the context of a swiss federal registry on functional neurosurgical MRgFUS interventions. All available data of this consecutive series were analyzed cross-sectionally in December 2018, thus explaining variable "n" values given in Results, as not all patients reached the 1 year follow-up at this moment.

Selection Criteria
Selection criteria for MRgFUS PTT included (1) idiopathic PD (diagnosed by neurologists), (2) chronic disease with at least 1 year of therapy resistance, characterized by (a) insufficient efficacy of L-Dopa dosed up to at least 600 mg L-Dopa equivalents per day, with symptom control during maximum 50% of the day, (b) gastro-intestinal or other side-effects, (c) fluctuations (on-off phenomenon), and (d) on-medication dyskinesias (choreoathetosis), (3) intensity of symptoms (intensity of tremor at rest and/or hypobradykinesia of 3/4 or more), (4) strongly diminished quality of life, and (5) Montreal Cognitive Assessment (MoCA) test in the norm or reduced but not below 20/30. Three subtypes of idiopathic PD were considered (55): the tremor dominant (TD), the akinetorigid (AR) and the mixed form (MX). Asymmetry of symptoms was not a selection criterion. No prefixed age limit was set. All patients were examined by a senior internist for any medical contraindications. Antiaggregant therapy was stopped for 10 days before the intervention, and normal coagulation and blood pressure were checked for all patients prior to surgery.

Procedure
The procedures were performed in a 3T MR imaging system (GE Discovery 750, GE Healthcare, Milwaukee, WI, USA) using the ExAblate Neuro device (InSightec, Haifa, Israel). Targeting was performed using the stereotactic multiarchitectonic Morel Atlas of the Human Thalamus and Basal Ganglia (56) and its developments (28,53). Three-dimensional stereotactic coordinates were measured on the MR images, based on the intercommisural line and the thalamo-ventricular border. The center of the PTT target was located 6.5 mm from the medial thalamic border (L 6.5), 1 mm posterior to the MCL (MCL-1) and on the intercommissural plane. Sonications had the shortest possible time and the corresponding power in order to provide a thermal dose of 240 CEM at each focal point which represents a conservative value corresponding to a 100% probability of lesion in a volume of 1.5 × 1.5 × 3.0 mm. The detailed technical description of the targeting and realization of the MRgFUS PTT using thermal dose control has been published in a separate work (53).
Two patients had a contralateral centrum medianum thalamotomy (CMT) in addition to PTT during the same session. Mean maximal sonication power was 1,000 ± 265 [W] (range 500-1,400 W) using the shortest possible sonication durations. MR imaging was performed for co-registration preoperatively and 2 days after the treatment. Figure 1 shows a PTT target as seen at the end of the procedure. Target reconstruction was performed for every patient according to Moser et al. (29,34,57,58). All interventions took place in an ambulatory setting with one night stay at a nearby clinic for comfort. In two patients (3.6%) hospitalization was longer than one night, in one case for comfort and in the second because of tiredness and preexistent mobility issues. No patient of this series had to be acutely hospitalized during the first postoperative month except for pre-planned physical rehabilitation programs.
As to bilateral interventions, the second one took place at least 6 months after the first PTT. It was performed if the symptomatology on the untreated side was strong and therapyresistant enough (with an intake between 300 and 600 mg L-Dopa equivalents per day) to justify surgery. Complementation of first PTTs performed before this series (with sonication repetition or long sonication times) was coupled with the treatment of the second side during the same session except in one case.
Baseline preoperative L-Dopa equivalent intake (range of intake between 0 and 1,400 mg/day, not taking into account dopamine agonists) corresponded to the intake during the last weeks prior to treatment and not to the maximum tried dosages along the whole disease history before the intervention. In patients showing small up to no effect to L-Dopa or having suffered from side-effects, it was indeed stopped usually long before the intervention period. The reduction in L-Dopa intake was supervised by the referring neurologist. It was undertaken according to the wish and readiness of patients to perform it.

Clinical Evaluation and Outcome Measures
The primary endpoints at 3 months and 1 year postoperatively were (1) the Unified Parkinson's Disease Rating Scale (UPDRS) scores in on and off states (59, 60), (2) assessment by the patient her/himself of the global symptom relief for the treated body side (GSRt), of the tremor control on the treated side and of the global symptom relief for the entire body (GSRb), (3) reduction in drug intake, (4) off dystonia, (5) on dyskinesias (choreo-athetosis), (6) sleep disturbances, (7) pain, and (8) adverse events.
As to secondary endpoints, the MoCA test (61) was performed in the few months up to 2 days prior to the treatment, and repeated 2 days and 1 year after it. In addition, all patients were asked to fill the questionnaire for the activities of daily living according to Bain et al. (62), the World Health Organization Quality of Life (WHOQOL bref) (63), and the Hospital Anxiety and Depression Scale (HADS) (64) preoperatively, 3 months and 1 year after the treatment.
Follow-up assessments took place 2 days, 3 months, and 1 year after the procedure. On-and off-medication examinations were performed preoperatively as well as at 1 year. At 2 FIGURE 1 | (A) Intraoperative MR axial T2 scan with body coil after PTT lesion, cut through the intercommissural plane. (B) PTT at higher magnification taken intraoperatively with superimposed atlas maps 0.9 mm ventral to the intercommissural plane, modified from the Morel Atlas of the Human Thalamus and Basal Ganglia. AC, anterior commissure; al, ansa lenticularis; fct, fasciculus cerebello-thalamicus; fx, Fornix; GPe, globus pallidus, external segment; GPi, globus pallidus, internal segment; H1, H1 field of Forel; ic, internal capsule; mcl, midcommissural line; mtt, mammillothalamic tract; PC, posterior commissure; PuT, putamen. days and 3 months patients were usually seen in an onstate. For international patients, telemedicine (video and phone conversation) were offered for both postoperative evaluations when unable to travel long distances.

Statistics
Statistical analysis of quantitative scores compared with baseline was carried out by repeated ANOVA measures and multiple comparisons were applied using a post-hoc analysis with Bonferroni-Holm testing (Daniel's XL toolbox; https://www. xltoolbox.net/). Statistical significance was set to α < 0.05. Table 1. Fifty-two of 56 (92%) interventions in 47 patients were included in this study. Four patients were lost to follow up (8%). All of them were international patients who could be reached by phone 3 months after the treatment but did not provide at least a video examination. The "n" values in the tables stand for "interventions" as some patients were treated in two sessions. The 3 months follow-up was reached in 52 interventions (47 patients) and the 1 year follow-up in 33 interventions (31 patients) in December 2018 (see Figure 2). At 1 year, one patient was deceased and one was lost for FU. UPDRS total scores were available for all interventions at baseline, the UPDRS III on-medication was missing in two patients and the UPDRS III off-medication in three patients. At 1 year, 26 total UPDRS scores were collected and five patients were interviewed with telemedicine. The variability in "n" values in Table 2 is due to the inclusion of clinical data obtained through telemedicine. UPDRS III on-medication scores were missing in five patients and off-medication scores in two patients.

Patient characteristics are summarized in
Mean age at the time of the treatment was 67 ± 10 years. Mean symptom duration was 10.0 ± 5.3 years. The mean and median follow-up times were 9.8 ± 5.8 and 12 months, respectively. Mean Hoehn and Yahr stage was 2.6 ± 0.7. Of the three subtypes of idiopathic PD, 72% were MX, 18% TD and 10% AR. Forty-one over 56 interventions were unilateral. Fifteen patients received bilateral PTT either in one session (n = 2) or staged. PTT target complementation in patients operated before this series was performed in 10 cases.
Mean skull density ratio (SDR) was 0.57 ± 0.11 (median: 0.57, range: 0.31-0.77). SDR ratio below 0.3 was an exclusion criterion. One patient only was denied treatment during the duration of this study because of a low SDR value. This SDR distribution was in accordance with our previous experience with a population dominantly Caucasian.

Secondary Outcome Measures
Secondary outcome measures are presented in Table 5   year (p < 0.005) and for WHOQOL item 1 (how would you rate your quality of life) at 3 months (p = 0.002).

Postoperative Dynamics
Six patients (11%) showed fluctuations of consciousness (increased sleepiness up to confusional states) which all resolved completely within 24 h. One patient displayed a short-lived (<12 h) fluctuating corticospinal syndrome (paresis with Babinski sign) without evidence of capsular involvement according to the applied thermal doses as well as to the intraoperative post-lesion MR-examination.

Adverse Events
There was no bleeding, no infection and no ballism. Sonications were painful for a few seconds in seven patients (13%). No patient reported lasting significant headache >6 h after the procedure. One patient developed a scalp hypoesthesia around one pin fixation which had fully recovered after 3 months. One patient suffered from a short-lived intense anxio-depressive episode from which he rapidly and completely recovered, with no relapse till more than 1 year postoperatively. At 3 months follow-up, seven patients reported increased (two patients) or new (five patients) speech difficulties (UPDRS II, item 5), at 1 year they were 2 and 4, respectively. In these six cases, hypophonia was increased by one point in half of them and two points in the other half. An objective increase in hypophonia (UPDRS III, item 18) at 1 year in off-medication state (compared with preoperative offmedication state) was present in three patients (by one point only) over 24. Out of the four patients among them treated bilaterally and controlled at 1 year, one was improved, one unchanged and the last two worsened by one point. There was no clear-cut dysarthria in the whole series. One patient was seen in an outpatient clinic for an L-Dopa over-dosage. One patient suffered from a hiccup as well as difficulties for breathing and speech lasting over months but regressive at 10 months. One patient had a fall 1 month after surgery and broke his hip. He underwent hip surgery and was walking without aid at 3 months follow-up. One patient died 5 months after PTT from a gastrointestinal occlusion.

DISCUSSION
This prospective case series of 56 consecutive interventions for PD performed in 2017 and 2018 is placed in the context of an 8-year long experience with the MRgFUS PTT against PD. Symptom recurrences or partial symptom control led us to develop a target protocol to optimize target coverage, including (1) a histological reappraisal of the antero-posterior extension of the pallidothalamic tract taking into account interindividual variability, and (2) the use of preplanned focal point displacement, shortest sonication duration and thermal dose control. Indeed, neither repetition of sonications nor prolonging sonication duration on the same focal point brought sufficient consistency in patient symptom control (53). This evolution integrates itself in a learning curve development: for example, the tremor relief on the operated side was estimated by the patients to be 52% at the beginning of our experience, moving up to 60% when we repeated sonications on the same spot [see Magara et al. (29)], 70% when we increased the sonication times, and finally 88% currently, as described in this study. Patients reported an 88% mean (95% median) tremor, 82% mean (90% median) global symptom relief on the operated side and 69% mean (80% median) global symptom relief for the whole body at 1 year. These results compare well with our scale improvement ratings: at 1 year the off-medication state was compared with the baseline on-medication state and revealed a percentage reduction of the mean of 84% for tremor, 70% for rigidity, and 73% for peripheral hypobradykinesia. Patients'  Table 2).
reports on their own surgical outcomes cover their whole daily living time. They provide thus a useful complement to the clinical snapshots collected during the short time frame of control examinations (49,65).
PTT achieved suppression of on-medication dyskinesias on the operated side in all patients as well as strong reduction in pain, dystonia and sleep problems. Reduction of L-Dopa intake was performed slowly and respecting patient readiness. Reduction of the mean L-Dopa intake at 1 year was 55% with a median intake of 75 mg L-Dopa equivalent/day. Dopamine agonists were reduced more readily and were only taken by three patients after 1 year.
The most resistant symptoms in this series were speech and axial symptoms, as they did not improve significantly at 1 year follow-up, although the trend was toward improvement for both. No capsular or thalamic dysarthria was recorded. Increase of hypophonia along subjective and objective estimations concerned a limited number of patients and was slight in intensity. Although, the number of patients who received bilateral PTT and have been controlled at 1 year is small (n = 4), we observed similar results for this group compared to the whole patient group (n = 24) for axial items (see Figure 6) and speech. The bilateral PTT group is too small to allow firm conclusions at this point.
The whole and the bilateral PTT patient groups showed a stable mean MoCA score 2 days after surgery. MoCA reductions at 2 days can be attributed first to postoperative fatigue and second to reductions of thalamocortical reserves (34) in the absence of capsular or MTT encroachments. A valuable way to explore the presence or not of hemispheric significant distributed cell losses is to couple MR imaging (analysis of hemispheric atrophy) with the analysis of cognitive functions, as paralimbic/multimodal areas are very extensive and widespread throughout both brain hemispheres. We have some evidence (not quantified) that postoperative cognitive reductions correlate with the magnitude of preoperative thalamocortical atrophy. The presence of postoperative cognitive reductions in a given study will therefore depend on the limit set by the neurosurgical team as to what are acceptable preoperative cognitive deficits.
In the single patient of this series having experienced symptom recurred, re-analysis of the intervention showed partial coverage of the pallidothalamic tract due to technical difficulties. He reported a 50% symptom relief on the operated side at 3 months, but most of his symptoms had reappeared at 1 year. Figures 5, 6 shows that for this patient (green dot) some clinical benefit was nevertheless obtained. This underlines the central importance of optimal thermal dose coverage of the target (53,54).
Two patients had a CMT on the contralateral side to PTT. This adjunction did not influence most results, as they were collected on the side treated by PTT. However, a discrete effect only on the total UPDRS cannot be excluded. There is current interest for the CMT (66,67) in the treatment of movement disorders. It is being explored in our center as a secondary target to PTT for bilateral interventions.
In a context of brain surgery and particularly of lesioning procedures, an intervention should in our opinion provide better results than drug therapy, thus justifying surgical risks. Hence our central selection criterion of therapy resistance. As shown in Figure 4 and Table 2, postsurgical status is indeed better than preoperative medication status. We are aware that this is at variance from criteria of DBS studies (68,69). To follow our line of thought, we have compared in this study postoperative offmedication state with preoperative on-medication state, allowing us to assess directly the superior relief obtained by surgery as compared to medical treatment.

Pathophysiological Considerations
In view of the pathophysiological data discussed elsewhere and summarized in the Introduction, the PTT provides the possibility to liberate the thalamocortical network from pallidal overinhibition while keeping the thalamus unoperated. In our opinion, PTT is an optimized form of pallidotomy, because (1) it provides a maximal amount of suppression of overinhibiting pallidal output to the thalamus in a much smaller volume, (2) it avoids the lenticulostriatal domain, thus reducing the vascular risk of pallidal surgery, and (3) the surgical risk for the optic tract is suppressed and reduced for the internal capsule. This allows symptom relief without reduction of the thalamocortical dynamics, which is essential for our whole sensory, motor and complex mental functions. The absence of predictable and unavoidable deficits in our study after the interruption of the pallidothalamic tract confirms this statement. In our opinion, the pallidothalamic tract may be considered as an ideal target because (1) it is in position to disturb the thalamocortical dynamics, and (2) pallidal overinhibition of the thalamus may have made this pathway useless. Preexisting normal pallidothalamic functions may have been taken over by the powerful corticothalamic network, in the context of extensive redundancy and plastic abilities of the whole thalamocortical system (70). Because the PTT interruption lies upstream of the thalamus, it gives the possibility for the whole thalamocortical system to normalize, thus allowing in principle all symptoms to recede. This process needs variable amounts of time to be implemented, in view of the functional complexity of interacting thalamic and cortical partners. Clinical observation shows indeed progression in symptom controls, which are variable from symptom to symptom. The tremor for example needs an average of 2-3 months to progressively disappear, as tremor waves become shorter in duration and lower in amplitude and frequency (26).
The short but sometimes intense transient phenomena observed postoperatively (see Results/Postoperative Dynamics) correlate inversely with the amount of thalamocortical reserves (34) and are thought to be related to a physiological rebound caused by the readjustment of the thalamocortical network, as it is abruptly liberated from the incoming pallidothalamic overinhibition (26).

The Paralimbic/Multimodal Psycho-Emotional Dimension
After PTT, improvements of axial functions were limited and hypophonia increased in some patients. This voice worsening was present in 13% of patients but never more than one UPDRS point in intensity. According to peroperative data and postoperative target reconstructions, it cannot be explained by an encroachment of the treatment zone on the internal capsule or on the thalamus. The explanation currently offered is that this would be the sign of a further unchallenged disease progression. We tend to look rather in the direction of psycho-emotional factors. Indeed, gait and voice functions are complex and  bilateral, and are at the base of two essential human functions: independence and communication. The voice is surely the most sensitive of the two, having a small effector serving minute movements, and being tightly driven by emotions. The main if not exclusive speech symptom we observed is hypophonia, a symptom also often seen in depressive patients. We consider the emotional load of neurodegeneration set on the shoulders of parkinsonian patients as huge. In view of the presence of rich interconnections between the paralimbic/multimodal and motor systems, particularly but not exclusively at the level of the cingulate motor areas (71,72), a significant effect of ideas and emotions on motricity should be considered. Such a strong psychomotor effect could well be at the source of the difficulties arising typically after the intervention for the second side, in a moment where the patients wonder intensely how things will proceed now that the surgical treatment is finished. The tensions inside the emotional network can lead to psychomotor abnormal outputs mainly in the domain of axial, voice and gait hypobradykinesia, opening the way to misinterpretations of a psychomotor effect for a sign of disease progression.

Limitations of This Study
This consecutive surgical case series presents the obvious limitation of not being sham-controlled. Patients were not all followed for 1 year because the study was a crosssectional one with a stop in December 2018. This provides irregular final follow-ups for the different patients. The small group of bilateral PTTs at 1 year follow-up (four patients) shows similar results as compared to unilateral PTTs but does not allow to infer firm conclusions at this point concerning the feasibility and risk profile of the bilateral PTT.

CONCLUSION
This case series, which will require longer follow-up and more patients receiving bilateral treatment, supports the PTT as a safe and highly effective surgical option in the treatment of chronic therapy-resistant PD. We compared patient's on-medication preoperatively with their off-medication postoperative state, providing justification for surgery. Improvements could be observed for all symptoms. We discussed the importance of the sparing of the thalamus, the necessity to integrate the preoperative thalamocortical state as well as the relevance of the psycho-emotional dimension.

DATA AVAILABILITY STATEMENT
The datasets generated for this study are available on request to the corresponding author.

ETHICS STATEMENT
All patients treated with this protocol signed an informed consent form after having been fully informed about the treatment, its results and risks. No ethical approval was sought because MRgFUS PTT has been approved by the swiss health state department and is covered by swiss social insurances.

AUTHOR CONTRIBUTIONS
MG: conception and design of the study, data acquisition and analysis, interpretation of the data, and co-drafted the manuscript. DJ: conception and design, data acquisition, interpretation of the data, and co-drafted the manuscript. DM and CF: co-drafted the manuscript. FR, AM, MS, RB, MK, PP, and CD: data acquisition. All authors read and approved the final manuscript.

FUNDING
CF was supported by the Swiss National Science Foundation.