Among non-motor disturbances, cognitive impairment is common in individuals with Parkinson’s disease (PD). A growing body of evidence indicates that PD is hallmarked by cognitive decline in a range of cognitive domains. Decline of attentional and executive functioning is thought to result from disruption of striato-thalamo-frontal pathways, and is often a stable clinical trait already detectable at the earliest disease stages. Deficits of episodic memory and visuospatial skills, on the other hand, are related to dysfunction in temporo-parietal areas and tend to be more common when clinical decline leads to dementia (Williams-Gray et al., 2007; Papagno and Trojano, 2018).

Resting-state functional magnetic resonance imaging (rs-fMRI) holds considerable potential for the investigation of disruption of cognitive circuitries (Prvulovic et al., 2011; Ferreira and Busatto, 2013), by measuring temporal synchronisations in the blood-oxygen-level-dependent (BOLD) signal across brain regions at rest (Fox and Raichle, 2007). This has also been useful in the study of cognitive decline in PD (Wolters et al., 2019). This can be addressed by adopting complementary methodologies such as the extraction of large-scale functional networks (data-driven approach) or the calculation of functional connectivity (FC) maps based on an a priori seed region selection (theory-driven approach).

Several rs-fMRI studies have explored large-scale networks, reporting disruptions in the default-mode network (DMN) and in the fronto-parietal networks (FPNs) in PD with cognitive impairment (Lebedev et al., 2014; Madhyastha et al., 2015; Ruppert et al., 2021). Importantly, the brain regions that are part of these networks are areas involved in sensorimotor integration and in higher cognitive functioning. In healthy controls (HC), reduced DMN connectivity appears associated with decreased memory performance, slower processing speed and worse executive functioning (Andrews-Hanna et al., 2007; Damoiseaux et al., 2008; Vidal-Piñeiro et al., 2014). There has also been significant evidence of DMN disruption in other neurodegenerative disorders such as Alzheimer’s disease, Huntington’s disease and frontotemporal dementia (Zhou et al., 2010; Wolf et al., 2012; Balthazar et al., 2014). Similarly, changes in DMN connectivity have been previously reported in PD (Tessitore et al., 2012; Disbrow et al., 2014; Yao et al., 2014). Decreased FC within the DMN differentiates PD patients with and without cognitive impairment (Wolters et al., 2019). Tessitore et al. (2012) reported decreased DMN connectivity in the bilateral inferior parietal cortex in a cohort of cognitively unimpaired PD patients. They also showed significant positive correlations between DMN connectivity and cognitive performance in tests of memory and visuospatial functioning, suggesting that functional DMN alteration can precede objective cognitive impairment in PD (Tessitore et al., 2012).

The FPNs (also known as executive control networks) follow the dorsal-attentional streams (Fox et al., 2006) and support attentional control (Japee et al., 2015). These networks have intricate functional connections with the basal ganglia, in particular with the caudate nucleus (Seeley et al., 2007). Disruption in FPNs seems to have a critical role in determining cognitive decline in PD. Alterations in FPNs were reported in PD patients with cognitive impairment (Lewis et al., 2003; Caminiti et al., 2015). Furthermore, recent findings have demonstrated that the topological robustness of the FPNs is associated with the absence of cognitive decline in PD individuals, suggesting that the integrity of these networks may help support cognitive performance in PD (Cascone et al., 2021).

In the last decades, enormous progress has been made in understanding which factors may contribute to “resilience” against neurodegeneration. In this respect, reserve-related processes such as cognitive reserve (CR), brain reserve, and brain maintenance, which refers to the mitigation of age-related brain changes by life experiences, are known to play a major role in modulating neurofunctional resources (Stern, 2002; Stern et al., 2018). As initially observed in Alzheimer’s disease (Stern et al., 2018), these factors can account for the apparent lack of direct correspondence between the severity of pathological changes and the clinical manifestations; they might help the understanding of any differential susceptibility to the effects of pathology in PD, mainly in cognitive functions and functionality in daily-living activities.

CR refers to the processing resources accrued over time as a result of being engaged in mentally-stimulating activities, i.e., education, professional attainment, and leisure activities (Stern et al., 2018). To quantify CR, it is possible to rely on “convenience proxies” such as socio-behavioral indices, e.g., education, intelligence quotient, occupational complexity, leisure and physical activity (Stern et al., 2018).

Empirical evidence from studies of PD indicates that CR can modulate cognitive performance and contrast cognitive decline. Higher levels of education were found to be associated with better cognitive performance and slower cognitive decline (Hindle et al., 2014). A study of 35 non-demented PD patients, using the Cognitive Reserve Index questionnaire (Nucci et al., 2012) and the Brief Intelligence Test (Colombo et al., 2000), showed a meaningful and significant effect of CR on patients’ performance in tasks of executive function, the cognitive domain most affected in PD (Ciccarelli et al., 2018). Thus, a higher educational attainment, coupled with a high mentally-stimulating lifestyle, appears to support cognitive performance in PD and, therefore, limit cognitive deterioration.

To the best of our knowledge, to date no study has investigated the modulatory role of CR on FC in PD. Initial evidence from studies of HC suggests that education and CR might have a positive effect on FC networks. Arenaza-Urquijo et al. (2013) examined a cognitively healthy older cohort (60–80 years) and described better brain metabolism, larger gray matter (GM) volumes as well as enhanced FC in regions such as the anterior cingulate cortex, right hippocampus, right posterior cingulate cortex, left inferior frontal and left angular gyri in individuals with higher education (Arenaza-Urquijo et al., 2013). Similarly, Marques et al. (2015) examined the relationship between education and FC and found that individuals with higher education had wider connectivity networks in all lobes of both hemispheres. These authors suggested that increased connectivity might moderate the effects of age (Marques et al., 2015). Moreover, Marques et al. (2016), in a study of a cohort of 120 elderly HC, demonstrated that demographic characteristics (especially years of education) were associated with higher FC, in particular with higher clustering, local efficiency and strength in parietal and occipital regions. These findings, collectively, indicate that individuals with higher education rely on different neural processing (Marques et al., 2016). Amongst the main large-scale networks, it has been demonstrated that higher CR is associated with increased brain activity in the DMN in elderly HC (Bosch et al., 2010).

Further evidence of a modulatory role of CR on FC has been obtained from individuals with other neurodegenerative conditions (Bozzali et al., 2015; Franzmeier et al., 2017a,b; Fuchs et al., 2019). Bozzali et al. (2015) investigated whether CR modulates FC in healthy, amnestic mild cognitive impairment, and Alzheimer’s disease individuals. The authors found that individuals with Alzheimer’s disease and higher education levels had greater FC in the DMN compared with individuals with Alzheimer’s disease and lower education levels (Bozzali et al., 2015). Some of the amnestic mild cognitive impairment patients had similar connectivity strength, suggesting that education and, more in general, CR, fosters mechanisms of compensation and limits progression of atrophy. A pioneering study in patients with multiple sclerosis used premorbid verbal intelligence as a proxy for CR and network-based measures to demonstrate that patients with higher CR had more preserved FC despite having GM atrophy (Fuchs et al., 2019). These authors hypothesized that preservation of network FC attenuates the impact of structural network disruption on cognition (in particular on cognitive processing speed and visual/spatial memory) in patients with multiple sclerosis.

This study tested the hypothesis that, in PD patients, FC alterations can be detected in large-scale and seed-based resting-state brain networks. It also tested the hypothesis that the patterns of alteration would be modulated by CR. First, we explored if CR is associated with the activity of the main large-scale functional cognitive networks, namely the anterior and posterior DMN (aDMN, pDMN, respectively), the left and right FPN (lFPN, rFPN, respectively) and the salience network (SalN) in a group of PD patients and one of HC. The DMN and SalN were chosen based on their well-established association with cognitive performance (Esposito et al., 2009; Menon and Uddin, 2010). FPNs were selected because of their online role in executive control (Zanto and Gazzaley, 2013). We also tested the impact of CR on two additional large-scale networks: the sensorimotor network (SMN) and the occipital visual network (OVN). The SMN was selected because of its documented disruption in PD and because it is typically associated with cardinal motor symptoms (Tessitore et al., 2014). The OVN was instead chosen as a non-cognitive control network. Second, we explored if CR modulates FC of key seed regions (dorsolateral and ventrolateral prefrontal cortex and substantia nigra) in PD and HC groups. We hypothesized that FC would be reduced in PD patients more than in HC in fronto-parietal regions, and that alterations of FC would be greater in patients with low CR.

This study tested the hypothesis that in PD patients FC alterations can be detected in large-scale and seed-based resting-state brain networks, and that CR might contribute to modulating these patterns of alterations.

One of the most intriguing results of the present study is that CR may exert a modulatory effect on FC involving basal ganglia and executive-attentional fronto-parietal networks.

We found evidence that CR modulates FC in PD patients by using both an ICA and a seed-based approach. The ICA approach revealed that low-CR PD patients showed lower FC within the basal ganglia (putamen and caudate nucleus) and anterior cingulate. Moreover, the seed-based approach showed lower FC in low-CR PD patients between bilateral frontal and parietal regions, but at the same time stronger FC between the left middle frontal gyrus and medial and superior frontal gyri. The modulation of FC offered by CR supports the hypothesis that lifelong cognitive enrichment may exert a neuroprotective role (Stern, 2002) and may mitigate functional down regulation induced by neurodegeneration (Brayne et al., 2010). An alternative (and largely complementary) explanation would suggest that low-CR individuals may have less resources to cope with the functional changes that occur in the presence of neurodegeneration. It is also worth noting that patients with higher CR showed slightly less motor impairment as evaluated with the MDS-UPDRS-III, suggesting a protective role of CR not only on cognitive but also on motor function, in agreement with previous studies that have reported less severe motor symptoms in PD individuals with higher CR (Guzzetti et al., 2019).

Taken together, the finding of reduced FC in parietal regions and increased FC in prefrontal cortex detected in the low-CR PD patients supports earlier observations from functional neuroimaging studies that report age-related reductions in the activation pattern of posterior regions as well as increases in anterior regions, a potential mechanism at the basis of the “posterior–anterior shift in aging” neurocompensation model (Davis et al., 2008). This model postulates that the recruitment of anterior regions, i.e., prefrontal cortex, might sustain maintenance of cognitive performance in the presence of a reduction in posterior activity (Davis et al., 2008; Park and Reuter-Lorenz, 2009; Grady, 2012). Nevertheless, increased activity in prefrontal regions may also reflect less specific or less efficient functioning, rather than compensation (West, 1996; Morcom and Henson, 2018). Our data are in line with both hypotheses. Further investigations of prefrontal activity and how this correlates with cognitive performance will help shed light on the mechanisms involved.

A main effect of CR was also observed, with low-CR individuals showing more FC between left middle frontal regions and the right cerebellum. Functional neuroimaging studies found that cerebellar regions co-activate with fronto-parietal cortices during cognitively demanding tasks (Balsters et al., 2014) and that cerebellar regions receive input from prefrontal and parietal regions through cortico-subcortical pathways (Krienen and Buckner, 2009; Buckner et al., 2011). This heightened connectivity might represent a compensatory mechanism that may mitigate less efficient neural processing in low-CR individuals. It has been proposed that the cerebellum is intrinsically capable of self-compensation and restoration, and these abilities are referred to as cerebellar reserve (Bordignon et al., 2021).

A final remark concerns structural data. Although no GM differences emerged between HC and PD groups in the VBM analysis, high-CR individuals showed greater volumes in a bilateral frontal cluster. Our findings are consistent with previous studies that have explored the link between CR and structural integrity of older adults’ brains (Anatürk et al., 2018) reporting a positive association between engagement in social-intellectual activities and GM volumes of frontal and temporal areas (Bartrés-Faz et al., 2009; Arenaza-Urquijo et al., 2017). Greater GM volumes in individuals with higher CR may correspond to better tolerance of age-related damage (Stern, 2002; Mortimer et al., 2003; Bartrés-Faz and Arenaza-Urquijo, 2011; Stern et al., 2018), with GM loss concentrated in the prefrontal cortices and subcortical structures, including the hippocampus (Raz et al., 2004; Allen et al., 2005; Driscoll et al., 2009; Taki et al., 2013; Schippling et al., 2017).

Several limitations should be taken into account when interpreting the results of the present study. First, most PD patients were on dopaminergic medication. Studies of drug-naïve patients would exclude the effects of dopaminergic medications on functional examination. Second, we did not test the relationships between FC and clinical profiles inclusive of neuropsychological performance and disease severity, as a very modest difference in MoCA scores was found between the two groups. This modest difference is most likely reflective of the limited psychometric properties of this screening test, since there were no significant differences in scores on the extended neuropsychological assessment between patient subgroups with high and low CR. Although this additional evidence suggests that cognitive variability between our PD subgroups is of marginal relevance to the mechanisms under examination (as cognitive variability also depends on brain networks), this is an aspect that certainly deserves more attention from researchers. Third, pathophysiological factors other than dopaminergic dysfunction might contribute to FC alterations in individuals with PD, e.g., alterations to cholinergic pathways (Bohnen et al., 2022) or TAU pathology (Pan et al., 2021). The link between distinct pathophysiological mechanisms and specific FC abnormalities, however, still needs to be clarified. On this note, we did not include a fine-grained characterization of the profiles of motor symptoms shown by patients (e.g., their type and lateralization), nor did we focus on the impact of reserve on specific cognitive domains such as, for instance, attentional processes and their respective functional (i.e., non data-driven) networks. Future studies will have the opportunity to explore this aspect in more detail, in order to describe the effects of variables of neurological importance (such as CR) as a function of a pathology-informed pattern of FC alterations. Fourth, we used a proxy of CR that did not include other relevant aspects of reserve such as lifelong enriching activities and experiences of leisure time. This information was available as part of the CRIq scale (and all CRIq subscores were significantly correlated with our CR predictor), but missing data prevented us from applying this instrument to the entire cohort. Fifth, although a range of denoising methodologies was applied, it is still possible that the findings might have been, in part, influenced by non-neural sources of signal variability. This methodological consideration is of central importance when PD is studied, as individuals with this condition may show significantly higher levels of in-scanner motion. The framewise displacement values calculated across each individual run, however, were not different between the two diagnostic groups. This indicates that in-scanner motion was not a major cause for concern. It is fair to acknowledge, however, that other, more sophisticated methods (i.e., such as scrubbing or CompCor) are available to researchers to control for motion and physiological artifacts in a more fine-grained manner. It is also important to point out that, while in-scanner motion has a detrimental impact on signal quality, this effect appears to be more pronounced with magnetic fields of higher strengths (Duyn, 2012). In this respect, although a 1.5 T magnetic field strength provides a lower spatial resolution, it offers the advantage of being less negatively influenced by motion, which is a central issue when PD is studied. Finally, as no body of studies exists on functional brain networks and CR in PD, we decided to rely on a conservative method to define the threshold of statistical significance (i.e., relying on a Family-Wise Error correction). Future studies will have the opportunity to adopt alternative approaches (e.g., threshold-free cluster enhancement methods) to limit false negatives.

The findings of this study complement published literature that has described alterations of resting-state brain activity in PD. The ICA approach revealed that PD patients had significantly less FC within the aDMN in the right primary motor, somatosensory and superior parietal cortices, and within the SMN in the right premotor and supplementary motor areas, regions involved in motor preparation and execution (Lee et al., 1999). Reduced FC at rest in the supplementary motor areas has been previously reported in PD using ICA (Canu et al., 2015; Laganà et al., 2020) and network models based on graph theory (Wu et al., 2009). Conversely, PD patients showed hyper-connectivity in the right inferior frontal gyrus (BA47), anterior cingulate (BA32), and caudate nucleus within the rFPN. Thus, our results on large-scale networks demonstrate that alterations in FC were specifically located in regions considered important hubs of the DMN (Raichle et al., 2001) and FPNs (Japee et al., 2015), areas involved in higher cognitive processes (Andrews-Hanna et al., 2007). A recent meta-analysis (Wolters et al., 2019) reported reduced FC in the DMN and FPNs when PD patients with cognitive impairment were compared with HC. DMN disruption was associated with deficits of perception and executive functions in these patients (Tahmasian et al., 2017). However, abnormal FC of the DMN, significantly correlated with cognitive parameters, was also documented in a rs-fMRI study that included cognitively unimpaired PD patients only (Tessitore et al., 2012), suggesting that DMN alteration may have a role in the development of cognitive decline in PD.

The seed-based approach demonstrated lower FC in PD between the bilateral middle frontal gyrus and bilateral parietal regions, covering the inferior parietal lobule and precuneus, major hubs of the attentional network (Coull, 2004). At the same time, PD patients showed higher FC between the left middle frontal gyrus and a cluster including the posterior cingulate. Recent meta-analyses (Pan et al., 2017; Tahmasian et al., 2017) and reviews (Tessitore et al., 2019) have reported that one of the most consistent findings in PD is an abnormal intrinsic functional pattern in the inferior parietal lobule, as confirmed by ICA, graph theory and ‘amplitude of low-frequency fluctuations’ analyses. Altered function of the rostral inferior parietal cortex (BA40) in PD was also observed in task-related fMRI and PET studies (Samuel et al., 1997; Sabatini et al., 2000). Additionally, exploration of anatomical connectivity through probabilistic tractography indicates that rostral inferior parietal areas are strongly connected with inferior frontal, motor, premotor, and somatosensory regions involved in higher motor functions, whereas caudal inferior parietal areas are predominantly connected with posterior parietal, primary visual and temporal areas typically related to spatial attention and language processing (Caspers et al., 2008).

In conclusion, we found abnormal FC across fronto-parietal circuits in PD patients, and we obtained evidence that CR exerts a relatively strong modulatory effect on FC in executive-attentional networks, typically impaired in PD. Future studies are required to evaluate longitudinal FC modifications to establish if these measures may help prediction of cognitive decline in PD, and if CR is linked to slower disease progression or rehabilitation-related changes. Different trajectories of decline may characterize individuals with high or low CR. Finally, forthcoming research may help gain an understanding of how FC may be the outcome of several cellular and molecular mechanisms related to CR building, including genetic polymorphisms (such as variants of the BDNF gene), epigenetic changes, neurogenesis and synaptic plasticity.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by IRCCS Don Carlo Gnocchi Foundation Ethics Committee (Ref No. 3_1/7/2015). The patients/participants provided their written informed consent to participate in this study.

SDT, AV, and MCS conceived the work. FB and SDT collected data. SDT and MDM performed the analyzes. SDT, AV, and MCS interpreted the results. SDT, AV, and MDM wrote the draft of the manuscript. All authors read and reviewed the final version of the manuscript.

This research was supported by the Italian Ministry of Health (Ricerca Corrente).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.