Using a co-activation-based approach, these data showed that dynamic connectivity was different between FND patients and HC in terms of entry rates into two brain states : 1) a co-activation of the right TPJ with the DMN and temporo-parietal network (CAP3 DMN-TemporoParietal with patients entering this state less often) and 2) a co-activation with the somatomotor network (CAP4-SomatoMotor with patients entering this state more often).
Our subgroup analyses revealed that patients with symptoms of weakness (FW) stayed for a longer duration in a brain state characterized by co-activation of the right TPJ and dorsal/ventral attentional networks (CAP2-SalienceAttention) and somatomotor regions compared to patients with other motor FND.
We will discuss the relevance of these salience-attention and somatomotor networks followed by a detailed elaboration on the contribution of DMN coupling.
Altered network coupling distinguishes FND patients from healthy controls.
Comparing characteristics in temporal connectivity from the rTPJ to the brain between FND patients and HC, we first identified less entries in patients in a network engaging the DMN and temporo-parietal regions (Fig. 1, CAP3-DMN-TemporoParietal, positive contributions), concomitant with deactivation of the executive network (Fig. 1, CAP3-DMN-TemporoParietal, negative contributions). Temporo-parietal regions and the SMA included in the CAP3-DMN-TemporoParietal are of particular interest, as those areas have been associated with the formation of the SoA41. The findings of reduced entries in this state in patients could thus suggest that a pathophysiological mechanism in FND may involve less ability to turn down the executive control and enter a state where the network underpinning the SoA is active together with self-referential areas of the DMN, such as the precuneus for example, which is important in self-monitoring26.
Second, patients entered more often than controls a brain state characterized by rTPJ co-activation patterns with the somatomotor network (Fig. 1, CAP4-SomatoMotor, positive contributions), involving the SMA, concomitant with an anti-coupling with the DMN and executive control network (Fig. 1, CAP4-SomatoMotor, negative contributions). This tends to suggest that a pathophysiological mechanism in FND may involve abnormally high activity of the motor networks at rest compared to controls.
The role of the salience network and temporo-parietal regions
The first interesting pattern is the altered coupling behaviour of the rTPJ with the salience network and temporo-parietal areas with concomitant DMN coupling and anti-coupling of the ECN. Noteworthy, the connectivity between the salience network with the rTPJ has previously been implicated in stimulus-driven attention and detection of relevant internal and external cues42. It has been shown that the salience network plays a central role when switching between large-scale networks43, such as for example when directing the attention to a salient stimulus by modulating the engagement of the ECN and the simultaneous disengagement of the DMN43. Furthermore, the DMN and ECN – usually anti-correlated networks – represent competing networks with regards to external versus internal processing of stimuli44,45, for which usually an increase in ECN-Sal activity is observed during stimulus-driven processes46. The herein reported alterations in brain states (i.e., CAP3 engaging the DMN, the salience, the temporoparietal areas and disengaging the ECN) might indicate difficulties in flexibility switching between brain states in FND.
The role of the somatomotor network
The second interesting finding is the increased interaction between the rTPJ and sensorimotor regions in FND patients. Only one other study18 assessed resting-state FC derived from the rTPJ as seed region in 35 motor FND patients and compared the data to 35 age- and sex-matched controls. In this study, FMD (mixed motor symptoms, 31% presenting with functional weakness) showed decreased FC between rTPJ and right sensorimotor cortex, cerebellar vermis, bilateral SMA and right insula, which the authors subsumed as impaired feed-forward signals and altered sensory feedback from sensorimotor integration areas that might alter patients’ SoA. It is of note, that this previous work was based on static FC measures, which might explain the apparent inverse directionality of our results (our data showing increased coupling), as our method brings additional information on temporal fluctuation of brain activity. In fact, both studies points towards abnormal connectivity between the rTPJ and the somatomotor network. Findings from Maurer, et al.18 demonstrate that overall, during a period of rest there is decoupling of the rTPJ with the somatomotor network, when findings from our study demonstrate that over a period of rest patients switch more often to a state where the rTPJ is coupled with the somatomotor network and decoupled with the executive and DMN networks. This supports again the hypothesis that there might be impaired executive control or feedforward signals on motor control in FND. With regards to motor symptoms, altered coupling behaviour between the DMN and sensorimotor networks have been associated to gait impairments in neurological (neurodegenerative) patients47, such as in normal pressure hydrocephalus (NPH)37, where altered interactions between the DMN, the ECN and the salience network were significantly associated with gait symptomatology. Moreover, gait disturbances were found to normalize together with these altered brain dynamics after cerebrospinal fluid tap test, indicating functional plasticity mechanisms that directly affect gait performance. Similarly, reduced brain dynamics between the DMN and the somatomotor network was found in elderly healthy controls accompanying motor disturbances often observed in older age48, for which the exact mechanism still remains elusive. These previous findings emphasize how dynamic brain network interactions can add to motor performance in health and disease and support the herein reported results. Therefore, abnormal dynamic network interactions in FND might contribute to the observed motor symptomatology.
Likewise, also using dynamic functional measures with the posterior cingulate cortex as a seeding region, altered network dynamics have been reported in patients with functional hyperkinetic movement disorders in which patients transitioned more often between states associated with the DMN49 as compared to HC, which has been interpreted as a potential compensatory mechanism of altered self-referential processes as previously observed in FND50,51.
Different coupling duration in patients with functional weakness compared to abnormal movements.
Dividing the patient cohort into symptom phenotypes allows investigating potential phenotype-specific characteristics. Our data demonstrated a difference between patients with functional weakness (FW) compared to patients with other types of abnormal movements (no-FW) with longer duration of CAP2 in FW patients. This means that patients with FW display increased coupling duration between rTPJ and the salience and attention networks (dorsal and ventral) as well as somatomotor network and executive network (Fig. 2, CAP2-SalienceAttention, positive contributions). A concomitantly reduced coupling duration in FW patients between the rTPJ and the DMN was found compared to no-FW patients (Fig. 2, CAP2-SalienceAttention, negative contributions). This is highly relevant in deciphering the pathophysiology of FND, as it suggests a mechanism of abnormally enhanced coupling between the rTPJ involved in the SoA and somatomotor networks that would be more pronounced in FW. Previous static connectivity analyses from Maurer, et al.18 performed on a cohort of mixed FND with predominant abnormal movements and only a third of FW, revealed a decreased coupling between the rTPJ and the somatomotor areas. This evidence together with our data could suggest that patients with no-FW have decreased coupling whereas patients with FW have increased coupling between rTPJ and other brain motor network nodes.
Another recent study, specifically looked at differences between FW and no-FW symptoms in resting state fMRI26. This study including 48 mFND patients and 65 HC comparable in age and sex, identified increased network centrality in left TPJ and precuneus (related to DMN) in FW relative to no-FW and HC. Comparing FW to HC, a reduced brain network centrality in the insula and the SMA was found. This hyperconnectivity in the TPJ did furthermore correlate with symptom severity in FW, which makes the authors posit the finding as potential biomarker for this subtype of FND. The left TPJ seems involved in attention (non-spatial, motor, monitoring of attention) and is (together with rTPJ) linked to predictions within a Bayesian Inference Model, assumingly updating and adjusting top-down predictions. As we did select the TPJ on the right hemisphere, which has been consistently associated to social cognition and authorship attribution, we can extend the neuropathophysiological differentiation between subtypes of FW and no-FW in regard to self-referential processes including the SoA or attention as both are described as relevant constructs in the multi-network model of FND2.
As we observed higher DMN anti-coupling duration in FW (Fig. 2, CAP2-SalienceAttention, negative contributions), the role of the DMN was similarly highlighted recently in a sample of seven FND patients with stroke-like symptoms reflected in unilateral paresis and hypoesthesia19. Compared to 15 controls, they identified an increased FC strength within the DMN network during RS-fMRI, which they supposed to represent elevated self-referential processing, that eventually stands in interference with motor activity and thus contributes to the phenomenology of weakness. Our analyses did reveal a subtype difference with regards to a reduced coupling duration in CAP2-SalienceAttention (derived from FND) involving predominantly the DMN in FW patients compared to no-FW. As such, looking at the subgroup of FW that display similar phenomenology as the subjects in Monsa, et al.19, we found a longer duration of DMN co-deactivation in FW compared to no-FW, sustaining this previous observation. Likewise, a longer duration of prefrontal (laying within the DMN) and visual regions with enhanced salience network activity has been suggested to be reflected as a decrease in cognitive control consequently with a heightened focus on self-reflective information52. Previously, a higher internally focused attention has been identified in monopolar and bipolar mood disorders, where it has been linked to somatoform bodily complaints, decreased cognitive flexibility as well as emotional dysregulation53. Adding to previous findings, a desynchronization of the DMN – in particular also with visual networks (CAP2-SalienceAttention) – might depict an imbalance between external- versus internal based processes, whereas internally-focused attention in mood disorder patients might reflect the inability to flexibly switch away from self-related processes54.
Previously, enhanced ventromedial prefrontal cortex activity – as part of the DMN – has been found in patients which was attributed to enhanced self-monitoring that negatively interact with motor pathways, i.e., reflecting a dysfunction in motor initiation in FND51,55. In subsequent studies, this dysfunction in motor initiation was extended to an impaired awareness of voluntary motor initiation6,7, associated to increased activity between the prefrontal cortex and the SMA7. In line with this, it could be shown that voluntary, straight movements of functional tremor patients worsened when attention was explicitly shifted towards visual feedback but improved with a shift of the attention to different aspects of the movement, e.g., increasing its velocity56. Thus, altered network connectivity might point towards an impairment already upstream of motor intention and action selection, and is thought to be involved in the inhibition of already ongoing motor programs57, which aligns with our findings on altered entry rate and CAP duration - for example by constantly interfering and updating ongoing motor programs (no-FW patients) or by remaining for too long in a certain state, thus not updating the ongoing program (FW patients). Therefore, altered salient and attentional processes in interaction with the DMN and somatomotor network might interfere with the implicit (“automatic”) execution of a motor program by replacing it with an explicit control of movement, which is then normally less smooth and less well performed58. As a consequence, the attribution of self-authorship gets distorted, with the rTPJ as comparator region being inadequately matching original intention and sensory feedback which is leading to the perception of involuntariness. The neural correlates of the involuntariness during functional symptoms was investigated earlier, in a study during which brain activity of FND patients was examined in the very same moment they experienced their motor symptoms, i.e. tremor13. As compared to the activation during voluntarily mimicking their symptoms, a hypoactivation in the right TPJ was found in patients during the involuntarily experienced symptoms. These findings all help connecting the herein reported results on altered rTPJ coactivation pattern interacting with networks such as the DMN, the salience network, the ECN and SomMot within the framework of an impaired awareness of motor intention and control of movement in patients with FND. In summary, these findings emphasize that functional symptoms might arise as a pathology of the attentional/salient networks, the DMN and ECN rather than – or in combination with a dysfunction of the motor system.
Limitations
Regarding the selection of a seed, we defined the rTPJ as a rigid sphere around literature based MNI-coordinates. However, the rTPJ as neural correlate of SoA has been reported with slightly different localizations59 and our seed might therefore not optimally cover each participants functional equivalent. Furthermore, as caveat of RS based studies, the translation of FC at rest to brain processes during interactions or task performance (e.g., attention, agency attribution, control of movement) remain speculative. On a technical level, seed-based co-activation analyses bear the risk that co-activation with the seed might occur at chance level without direct functional importance34 and might be susceptible to noise. Furthermore, the selection of the optimal cluster size based on consensus might be common practice but could result in the selection of a less optimal cluster. We counter steered this limitation by examining also other cluster sizes, which helps understanding how the different clusters were built. Also, the implementation of a dimensionality reduction step might cause the loss of meaningful signal. In our previous work35, we could show that the PCA step provides a valuable trade-off between high computational load and potential loss of weaker networks. There is no clear consensus on the physiological interpretation of altered temporal characteristics of co-activation patterns in the brain. These results therefore might have no or only little clinical relevance regarding the symptom development in FND. Regarding clinical characteristics, although we focused on patients with motor symptoms only, the symptom types differ across the FND population, which affects the generalizability of the results. Additionally, we did not perform a systematic psychiatric evaluation and thus, cannot exclude that the results were driven by psychiatric comorbidities commonly observed in FND60. Similarly, we did not exclude patients who were currently under psychotropic medication, which might influence functional alterations in the brain. However, we corrected our analyses for a potential effect arising from depression, anxiety, or intake of psychotropic medication.