Common and unique dysconnectivity profiles of dorsal and median raphe in Parkinson's disease

2.2. Clinical and behavioral assessments

All participants underwent clinical assessment of motor symptoms and NMS in the “OFF medication” state. Motor symptoms were evaluated using the MDS Unified Parkinson's Disease Rating Scale Part III (MDS‐UPDRS‐III). The dosage of antiparkinsonian drugs was calculated using the levodopa equivalent daily dose (LEDD). Olfactory function was assessed using the Brief Smell Identification Test (BSIT). Cognitive ability was evaluated using the Montreal Cognitive Assessment (MOCA). The presence of depression, anxiety, apathy, and pain was measured using the Hamilton Depression Scale (HAMD), State‐Trait Anxiety Inventory (STAI), and Apathy Evaluation Scale self‐rated version (AES‐S), and MDS‐UPDRS I‐1.9, respectively. Sleep disturbances were measured using the Rapid Eye Movement Sleep Behavior Disorder Questionnaire (RBDQ‐HK) and Epworth Sleepiness Scale (ESS) (Table 1).

2.3. MRI acquisition and preprocessing

During MRI scanning, participants were instructed to keep eyes closed, stay awake, and not think of anything. All the patients were scanned in the “OFF medication” state. The MRI data were acquired using a Siemens Magnetom Skyra 3T scanner (Erlangen, Germany). Rs‐fMRI images were obtained using a single‐shot spin‐echo echo‐planar imaging (SE‐EPI) sequence with following parameters: repetition time (TR), 2000 ms; echo time (TE), 30 ms; field of view (FOV), 22 × 22 cm²; flip angle, 90°; voxel size, 3.4 × 3.4 × 3 mm²; 35 slices, no gap; 176 repetitions; scanning time, 5 min 52 s. T1‐weighted anatomic images were scanned using a magnetization‐prepared 3D rapid gradient echo (MPRAGE) sequence. The image parameters were as follows: TR/TE/flip angle, 2530 ms/2.98 ms/7°; FOV, 25.6 × 25.6 cm²; voxel size 1 × 1 × 1 mm²; 192 slices, no gap. The acquisition time was 5 min 13 s.

Imaging data were preprocessed and analyzed using DPABI (version 6.0; http://www.rfmri.org/dpabi), including removal of the first 10 volumes, slice timing correction, head motion correction, tissue segmentation, spatial normalization into Montreal Neurological Institute (MNI) space and resampling into 3‐mm isotropic voxels. Considering that, the raphe nuclei are in the brainstem and susceptible to interference from physiological noises, independent component analysis (ICA) was used to correct the fMRI signal for respiratory and cardiac noise (Griffanti et al., 2017). FSL's tool MELODIC (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/) was used to perform ICA to denoise the whole‐brain fMRI data for all subjects. Briefly, single‐subject fMRI data are first decomposed into a collection of independent components (ICs) across the time domain (time series) and spatial domain (maps). Then, the signal and noise are manually identified, and the ICs of the noise are labeled and removed (Griffanti et al., 2017). The characteristics of the signal and noise were summarized in Table S1. Examples of signal and noise components are shown in Figures S1–S3. Subsequently, spatial smoothing with a full width at half maximum of 6 mm Gaussian kernel, and temporal bandpass filtering at 0.01–0.08 Hz were performed using DPABI. Notably, participants were excluded if their mean framewise displacement (FD) was >0.3 mm, or if the translational/rotational movements were >3 mm or 3°.

2.4. Functional connectivity analysis of raphe nuclei

The whole‐brain FC patterns were measured with DR and MR as region of interest (ROI). The ROI for DR or MR is defined as a sphere with a radius of 4 mm. The center of position for DR is x = 2, y = −26, z = −18, and for MR is x = 0, y = −34, z = −20 (MNI space; Bar et al., 2020; Beliveau et al., 2015). For each ROI, a whole‐brain FC map was obtained for each participant by calculating Pearson's correlation coefficients between the mean time series of all voxels in the ROI and the time series of every voxel within the gray matter (GM). The correlation coefficients were converted to z‐scores using Fisher's r–z transformation to improve the normality. Two z‐scored FC maps were generated for each participant.

3.2. Functional connectivity of raphe nuclei in PD

3.2.1. Within‐group functional connectivity profiles

Within‐group functional connectivity analysis showed that the DR and MR shared a large amount of common positive connectivity in HC group, including the bilateral prefrontal cortex (PFC), medial and lateral temporal cortices, occipital cortex, limbic system including cingulated cortex, insula, amygdala, and hippocampus, as well as subcortical structures including basal ganglia, thalamus, brainstem, and cerebellum (Figure 1). These nuclei also exhibited differential connections, with DR having more extensive connections with the cerebral cortex and basal ganglia and MR having more intense connections with the brainstem and cerebellum.

Open in a separate window

FIGURE 1

Within‐group functional connectivity (FC) of raphe nuclei. Results are thresholded at FWE‐corrected voxel‐level p < .05 (cluster size ≥10 voxels).

3.2.2. Between‐group functional connectivity profiles

Statistical between‐group comparisons showed that compared to HC, patients with PD demonstrated widespread dysconnectivity of DR and MR (Figure 2a). Commonly decreased FC of these nuclei in PD were found with the sensorimotor cortex (superior and inferior frontal gyri [SFG/IFG], postcentral gyrus [PoCG]), temporal cortex (inferior and middle temporal gyri [ITG/MTG], temporal pole [TP]), occipital cortex, limbic system (hippocampus, parahippocampus, amygdala, insula), left thalamus, putamen, and cerebellum. Distinct PD‐related dysconnectivity patterns were also observed for each ROI. DR had lower functional connectivity with the bilateral ventral medial PFC (vmPFC), orbital frontal gyrus (OFC), anterior and middle cingulate cortices (ACC/MCC), and right thalamus, while MR had lower connectivity with the pons in patients with PD compared to HCs (FDR‐corrected voxel‐level p < .01, cluster size ≥10 voxels; Figure 2b and Table S2).

Open in a separate window

FIGURE 2

Differences in functional connectivity of the raphe nuclei between the PD and HC groups. (a) Significant group differences in FC for dorsal raphe nucleus (DRN) and median raphe nucleus (MRN). Group difference FC maps are thresholded at FDR‐corrected voxel‐level p < .01 (MNI152 space, cluster size ≥10 voxels). (b) Spatial similarities and differences of FC maps between DR and MR. The brain areas with blue are unique to DR, those with green are specific to MR, and those with purple are common to both.

4.1. Altered spatial patterns of FC in the raphe nuclei in PD

Based on a relatively large sample, our results revealed that in the HC group, DR, and MR shared common positive connectivity with widespread cerebral cortex as well as subcortical regions such as basal ganglia, thalamus, brainstem, and cerebellum (Figure 1). Differential connectivity was also observed, with DR having more extensive connectivity with the cerebral cortex and basal ganglia and MR having more intense connectivity with the brainstem and cerebellum, which is consistent with the FC results reported by Beliveau et al. (2015). Previous fiber tracing studies also found overlapping and differential projections from DR and MR (Azmitia & Segal, 1978), with extensive 5‐HT projections from both nuclei to the PFC, cingulate cortex, basal ganglia, amygdala, hippocampus, and thalamus. The projections from DR prefer more laterally located regions (cerebral cortex, striatum, etc.), whereas MR prefer more medially located regions (hippocampus, septum, and brainstem; Charnay & Leger, 2010; Commons, 2016). Our findings of connectivity patterns in the HCs are consistent with those 5‐HT projections, thus suggesting that fMRI may provide an effective in vivo measurement of 5‐HT‐related brain function.

Differences between groups reveal that DR and MR share common hypo‐connectivity in PD in the following aspects: (1) both two raphe nuclei are hypo‐connected with the sensorimotor network (SMN), such as the sensorimotor cortex, left thalamus, and putamen. Functional impairment of the SMN has been reportedly associated with impaired sensory integration and motor deficit in PD (Caspers et al., 2021; Hou et al., 2021). (2) Both two nuclei also showed hypo‐connectivity with the salience network (SN), including the insula, amygdala, and thalamus. Less engagement with the SN has been implicated in reward, emotion disturbances, and cognitive impairment (Chang et al., 2017; Christopher et al., 2015; Navalpotro‐Gomez et al., 2020). (3) The cerebellum is thought to be associated with motor deficits and NMS such as hyposmia and cognitive impairment in PD (Riou et al., 2021; Wu & Hallett, 2013). Dysconnectivity between both raphe nuclei and cerebellum may lead to a range of symptoms mentioned above. (4) Both two nuclei exhibit decreased connectivity with the occipital cortex, which is a hub of memory development (Rosen et al., 2018; Yin et al., 2020). Sawyer and Kuo (2018) also observed that hypometabolism in the occipital regions was associated with dementia.

Despite the common dysconnectivity, there is also differential dysconnectivity between DR and MR in PD. Specifically, DR had unique hypo‐connectivity with the bilateral vmPFC, OFC, ACC, and MCC. These hypo‐connected areas are collectively referred to as the default mode network (DMN; Raichle, 2015). Decreases in DMN activity have been reportedly associated with olfactory impairment (Karunanayaka et al., 2017; Lu et al., 2019), depression (Xiao et al., 2022), cognitive impairment (Ruppert et al., 2021), sleep problem (Hong, Fallon, & Friston, 2021), and pain (Alshelh et al., 2018). Additionally, MR had decreased connectivity with the pons, which may lead to dysfunction of the cortico‐ponto‐cerebellar loops and impaired sensory‐motor control (Glickstein, 1998; Kamali et al., 2010; Palesi et al., 2017).

Notably, the ROIs of the DR and MR used in this study are very small and located in the brainstem, which makes them challenging to accurately normalize from raw space to the MNI152 template space. In this study, the following steps were done to improve the accuracy of normalization. First, the origin of individual raw structural and functional image was manually adjusted to the origin of the MNI152 template (anterior commissure) using SPM12. Second, the functional images normalized to MNI152 space were superimposed on the MNI152 T1 template to check the accuracy of normalization (Figure S4).

4.2. Common and unique roles of raphe nuclei dysconnectivity in PD

Not surprisingly, common and unique roles of DR and MR on motor symptoms and NMS are observed. Particularly, hypo‐connectivity of DR with the left amygdala and MR with the left thalamus are jointly associated with depression. Previous studies also found structural and functional alterations in the amygdala and thalamus in depression (Hong, Li, et al., 2021; Miller et al., 2017; Sibille et al., 2009; Williams, 2017), thus dysconnectivity of the raphe nuclei with the amygdala and thalamus may be one of the mechanisms of depression in PD.

In addition, unique roles of DR in PD are as follows (Table 2): (1) hypo‐connectivity with the left postcentral gyrus is negatively correlated with daytime sleepiness. Modulating role of serotonin in sleep–wake behavior has been widely reported (Monti, 2011; Vaseghi et al., 2022), thus, impaired connectivity between DR and sensory cortex may be a reason for daytime sleepiness. (2) Hypo‐connectivity with the bilateral ACC/MCC is associated with anxiety. ACC is an important emotional processing hub and is involved in social anxiety (Duval et al., 2013). In addition, Li et al. (2021) found that serotonergic circuit from DR to ACC regulates emotion and sociability. Our findings suggest that anxiety may be associated with abnormalities of the serotonergic neural circuitry between DR and ACC.

Moreover, unique roles of MR in PD were also observed (Table 2). First, we found that decreased connectivity with the left cerebellum and bilateral pons was negatively correlated with the severity of motor impairment (MDS‐UPDRS III scores and H&Y stage), suggesting that disrupted connectivity of the cortico‐ponto‐cerebellar loop may be a reason of motor deficits. Next, we observed that reduced connectivity with the left cerebellum was positively correlated with MoCA scores. As mentioned earlier, cerebellum is a hub of cognition (Jacobi et al., 2021), working memory and executive tasks are accompanied by activation of bilateral cerebellum (Guell et al., 2018). Cerebellar lesions can lead to cognitive deficits (Argyropoulos et al., 2020). Therefore, reduced connectivity between MR and cerebellum could be related to cognitive impairment in PD. Besides, dysconnectivity between MR and the left insula was related to RBD. The insula has strong connection with wake and sleep‐promoting regions such as hypothalamus and brainstem, and regulates sleep and wake activity (Chen et al., 2016). Therefore, impaired 5HT projection from MR to the insula may promote the development of RBD symptom. Finally, hypo‐connectivity between MR and the left superior frontal gyrus was associated with pain. The superior frontal gyrus is located at the superior part of the PFC. 5‐HT has been reported to play a crucial role in pain modulation, and 5‐HT receptors is widely distributed in the PFC, making the PFC important in pain processing (Liu et al., 2020; Ong et al., 2019).

This work was supported by the National Natural Science Foundation of China (82071423).