Using ICA, the sources of signal variation was blindly separated, and 7 large-scale brain networks were extracted. The DMN, VN and SMN were selectively vulnerable in newly diagnosed drug-naïve PD patients with MCI. Moreover, FC between the SMN and limbic network, and between the VAN and VN were more prone to be damaged in PD patients.
The DMN was characterized by reduced activation in task-based situation compared with the resting state [20]. In PD, DMN is the most studied intrinsic connectivity network, which is consistent with previous research on other neurological and psychiatric disorders [21]. The first study to explore the relationship between the resting state FC in DMN and cognitive performance in PD was focused on cognitively unimpaired PD patients while medicated, and suggested reduced FC in the right medial temporal lobe and bilateral inferior parietal cortex within the DMN using the ICA [22]. Our previous study using the same method found lower FC in the left inferior parietal lobule within the DMN in cognitively unimpaired drug-naïve PD patients with akinetic-rigidity subtype [23]. Even though these patients did not meet the criteria for cognitive impairment, the altered FC within the DMN was significantly associated with cognitive function. Various studies investigated the resting-state networks in PD patients with cognitive impairment, and a meta-analysis including seventeen studies reported cognitive impairment in PD was prominently relevant with reduced FC in the DMN [6].
More recent studies indicated connectivity changes in several RSNs rather than a single intrinsic connectivity network. However, changes of intrinsic connectivity network and even the association between the altered network and specific cognitive performance hold a considerable heterogeneity, which is likely due to the variability in the inclusion of patients, appliance of cognitive tests, and preprocessing strategies.
A dynamic functional analysis in PD suggested two discrete connectivity states, within-network state (state I) with more frequent and sparse connectivity, and between-network state (state II) with less frequent and strong interconnectivity [24]. In the state I, sparse connections were located mainly within DMN, VN, and SMN, which might play a vital role in the pathogenesis of PD manifestations. Our study observed abnormal FC within the DMN correlated with deficits in memory function, within the VN correlated with deficits in visuospatial function, as well as within the SMN correlated with the severity of disease. Cortical visual processing regions have been involved in a number of neuroimaging studies about PD [25], reporting occipital-cortical thinning, metabolic deficits, and hypoperfusion. A longitudinal fMRI study demonstrated a progressive loss of FC mainly in posterior parts of brain strongly correlated with decreasing cognitive performance [26]. The visuospatial deficit in PD was regarded as primary posterior cortical pathology rather than dopamine deficits, which might be a sensitive predictor of progression to dementia in PD [27]. Several lines of evidence suggested disrupted sensorimotor integration in PD [28]. The cortico-striatal loops were commonly impaired in previous studies investigating SMN connectivity in PD patients. It can be noted that dopamine deficiency might be one of the potential mechanisms underlying impaired SMN in PD.
According to Braak staging [29], the pathological process of PD occurs primarily in the brain stem, pursues an ascending process, and arrives to the neocortex in the final stage. One previous study reported reduced FC in mesolimbic-striatal and cortico-striatal circuits in drug-naïve PD patients, which reflects pathologic changes of early non-motor and motor deficits, corresponds to Braak staging, and further provides insight into the network integration in the early-state of PD [30]. The structural and functional connectivity in the SMN and limbic networks have been investigated separately in PD, less is known about the correlation between these two RSNs. A connectomic analysis revealed that the amygdala, as a key structure in the limbic system, had a close interplay with areas within the SMN, including the postcentral gyrus, the precentral gyrus and the paracentral lobule, suggesting a limbic–motor interface involved in the emotional modulation of complex functions [31]. The impaired emotion perception can contribute to the damage of smiling mimicry in PD, which is closely related to fibers connection between the amygdala and SMN. Our finding of decreased FC between the SMN and limbic network was in line with previous literature and fostered this idea of existing alterations of limbic–motor interface in newly diagnosed drug-naïve PD patients.
More than a decade ago, two attention systems (DAN and VAN) with distinct anatomy and function has been introduced, and their roles in the visuospatial attention system were mainly described [32]. The DAN can be active when attention is overtly or covertly oriented in space, and the VAN can be active when faced with unexpected stimuli associated with behavior. In advanced PD, the visual processing impairment (e.g., misperception and hallucinations) was closely related to the DMN, DAN, VAN and VN [33]. When the DAN is unable to recruit activation, the DMN and VAN can frequently functionally interactive and relatively overactive. Our finding of decreased FC between the VAN and VN might be an early change in the visuospatial attention system of PD, and subsequent changes and the significance remain to be further studied.
There are some limitations within the present study. First, the level of education differs between the PD-MCI and PD-CU subgroups, which might be an important confounder. In the analyses, the variable was regarded as covariate to reduce the influence. Another concern is the modest sample size, which limited the generalizability of our findings.