The local research ethics committee approved this study, and written informed consent was obtained from all of the participants. Patients who were recruited consecutively in our cohort were first diagnosed at the Movement Disorders Outpatient Clinic of West China Hospital of Sichuan University. These patients did not take anti-Parkinson medicines (i.e., drug-naïve) since they were not diagnosed correctly before visiting us. The recruited patients fulfilled the PD Society Brain Bank diagnostic criteria (Hughes et al., 1992). From this cohort, patients were excluded if they had (1) head motion being visible to the naked eyes; (2) cerebrovascular disorders, including previous stroke, history of head injury, history of seizure, hydrocephalus, intracranial mass, previous neurological surgery and other neurologic diseases; (3) anti-Parkinson, antidepressant or neuroleptic treatment before enrollment; and (4) history of dementia. After enrollment, the patients were followed-up for at least 1 year to confirm the diagnosis of PD because patients with PD at enrollment were in the early stage. Additionally, 22 normal subjects [healthy controls (HC) group] with no history of neurologic or psychiatric diseases were recruited and matched for age, gender and education with the PD subjects.

Movement disorder specialists collected data pertaining to handedness, age, sex, disease duration, and education years, and evaluated clinical symptoms using rating scales prior to MRI examination. The severity of PD was assessed using the Hoehn & Yahr staging scale (H&Y) and the Unified PD Rating Scale (UPDRS). The Hamilton Depression Rating Scale (HDRS) and the Hamilton Anxiety Rating Scale (HARS) were used to quantify depression and anxiety, respectively. The Montreal Cognitive Assessment (MoCA) was used to assess global cognitive function. Moreover, all participants underwent a complete neuropsychological battery. Attention/working memory [adaptive digit ordering test (DOT-A) and Golden Stroop test], executive [verbal fluency test (VFT) and clock drawing test (CDT)], memory [Hopkins verbal learning test-revised (HVLT-R) and brief visuospatial memory test revised (BVMT-R)], language [Wechsler intelligence scale for adults-Chinese revised (WAIS-RC) and Boston naming test (BNT)] and VS [Benton line orientation (BLO) and clock copying test (CCT)] functions were tested in all subjects. The cognitive performance scores were converted to z-scores and adjusted for age, sex, and education. Patients were classified as PD-MCI according to the new criteria of the Movement Disorder Society Task Force for mild cognitive impairment in PD. The criterion was met when patients performed approximately 1–2 standard deviation below the normative mean score on at least two neuropsychological tests, represented by either two impaired tests in one cognitive domain or one impaired test in two different cognitive domains (Litvan et al., 2012). In the current study, the PD-CU patients were defined as having a cognitive performance with normal scores or with abnormal scores with less than 1 standard deviation below the normative mean score. A total of 14 patients diagnosed as having mild cognitive impairment (PD-MCI group) with multiple domains and 18 patients diagnosed as cognition unimpaired (PD-CU group) were enrolled. PD-MCI patients were matched with PD-CU patients for PD disease severity. We arranged the enrolled patients to undergo the clinical evaluation and MRI scanning as soon as possible (usually on the second day after making a diagnosis of PD). The clinical assessment and MRI scanning were conducted on the same day prior to the initiation of any treatment. Then the patients started the anti-Parkinson treatment (usually on the same day after the clinical evaluation and MRI scanning).

An MRI was performed on a 3.0 Tesla MRI System (Siemens 3.0 T Trio Tim, Germany) by using an eight-channel phased-array head coil. High-resolution T1-weighted images were acquired via a volumetric three-dimensional spoiled gradient recall sequence. The acquisition parameters were as follows: repetition time [TR] = 1900 ms, echo time [TE] = 2.26 ms, flip angle [FA] = 90°, field of view [FOV] = 256 mm × 256 mm, slice thickness = 1 mm, no slice gap, voxel size = 1.0 mm × 1.0 mm × 1.0 mm, number of slices = 176. The MR images sensitized to changes in the BOLD signal levels (TR = 2000 ms, echo time = 30 ms, flip angle = 90°) were obtained via a gradient-echo echo-planar imaging sequence (EPI). The slice thickness was 5 mm (no slice gap) with a matrix size of 64 × 64 and a field of view of 240 mm × 240 mm, resulting in a voxel size of 3.75 mm × 3.75 mm × 5 mm. Each brain volume comprised 30 axial slices, and each functional run contained 240 image volumes. We provided ear plugs to each subject to reduce the noise interference during scanning and used foam to fix the participants’ head position and avoid head motion. The fMRI scanning was performed in darkness, and the participants were explicitly instructed to relax, close their eyes and not fall asleep (confirmed by subjects immediately after the experiment) during the resting-state MR acquisition.

The functional image preprocessing and statistical analysis were conducted by using SPM¹. The first 10 volumes of functional images were discarded for the signal equilibrium and participant adaptation to the scanning noise. The remaining EPI images were preprocessed via the following steps: slice timing, motion correction, spatial normalization to the standard Montreal Neurological Institute (MNI) EPI template in SPM8, and resample to 3 mm × 3 mm × 3 mm, followed by spatial smoothing with a 6 mm full-width at half-maximum (FWHM) Gaussian kernel. In our study, the scrubbing was not performed. However, the headers are modified for each of the input images, which can also reflect the relative orientations of the data. Then the details of the transformation are displayed in the results window as plots of translation and rotation, which can be modeled as confounds within the general linear model. In addition, according to the record of head motions within each fMRI run, all participants had less than 0.5 mm maximum displacement in the x, y, or z plane, and less than 0.5° of angular rotation about each axis.

Functional connectivity was examined by a seed-based correlation approach. Thus, we selected the following areas as seeds in the DMN: anterior, dorsal and ventral medial prefrontal cortex (mPFC), superior frontal gyrus (SFG), inferior frontal gyrus (IFG), posterior inferior parietal lobule (pIPL), precuneus, PCC, temporal parietal junction, anterior temporal lobe (ATL), superior temporal sulcus and HF. We defined the 18 nodes using the MNI coordinates (Spreng et al., 2013) (see Supplementary Table S1) and 10 mm radius circular masks. Using REST², after bandpass filtering (0.01–0.08 Hz) and linear trend removal, the reference time series for each seed region was extracted by averaging the fMRI time series of all voxels within each region of interest. Correlation functional analyses were performed by computing the temporal correlation between each seed reference and the rest of the brain in a voxel-wise manner. Eight nuisance covariates were regressed to remove the possible variances from the time course of each voxel, including the white-matter signal, the cerebrospinal fluid (CSF) signal and six head motion parameters. The global signal regression was not included, because it is a controversial issue (Murphy et al., 2009). However, the results with global signal regression were included in the Supplementary Table S2 and Supplementary Figure S1, and they were largely overlapped with the results without global signal regression. The correlation coefficients in each voxel were then transformed to z-value images by using the Fisher r-to-z transformation to improve normality. Hence, an entire brain z-value map was created for each subject.

In addition, to characterize the abnormal FC within the DMN, we chose these 18 nodes as being representative of the DMN and performed an additional region-wise manner. The mean-time series were extracted by the same way as above and were correlated between these 18 nodes of each subject. A Fisher’s r-to-z transformation was applied to normalize the correlation coefficient. Finally, the z-value was extracted for each participant, and we then compared z-values between the groups.

Differences between groups in terms of demographic and clinical variables were performed by the Pearson χ² test, one-way analysis of variance (ANOVA), or Student’s t-test, as appropriate. A voxel-based comparison of z-value maps among the PD patients with MCI, PD patients with CU, and HC was performed by using a design model of one-way ANOVA with age, sex and education as covariates, followed by post hoc two-sample t-tests. The significance threshold was set at p < 0.001. Family-wise error (FWE) correction for multiple comparisons was also conducted at the cluster level. In addition, a comparison of the correlation values (z-values) among groups was performed by one-way ANOVA, followed by post hoc two-sample t-tests (p < 0.05, two-tailed). To explore the possible relationship between changes in the patterns of network FC and performance in cognitive functions, we conducted Pearson correlations between FC values of significant inter-group differences and cognitive scores, and the significance was set at p < 0.05 (two-tailed). FC values of significant inter-group differences were required from two ways of FC analyses. One is the voxel-wise analysis, and the statistical analysis model was one-way ANOVA among the three groups, followed by post hoc two-sample t-tests. The significance threshold was set at p < 0.001, and family-wise error (FWE) correction for multiple comparisons was performed at the cluster level. The other one is the region-wise analysis. We only chose the FC values showing significant inter-group differences within the DMN in the Pearson correlation analysis. Then, the Pearson correlation analyses were performed between these FC values of significant inter-group differences and cognitive scores in all PD patients.

In our study, diffeomorphic anatomic registration through an exponentiated lie algebra algorithm (DARTEL) (Ashburner, 2007) was used to improve registration of the MR images. Before segmentation, scanner artifacts and gross anatomic abnormalities for each subject were checked, and the image origin was set to the anterior commissure. Images were then segmented into gray matter (GM), white matter and CSF using the unified segmentation model in SPM8. In the next step, a GM template was generated through an iteratively non-linear registration. The GM template was normalized to MNI space, and the resulting deformations were applied to the GM images of each participant. Finally, spatially normalized images were modulated to ensure that the overall amount of each tissue class was not altered by the spatial normalization procedure and smoothed with a 10 mm FWHM Gaussian kernel. Voxel-based comparisons of GM volume were performed between groups using two-sample t-tests with age, sex and education as covariates. The significance thresholds were at voxel-wise p < 0.001 and the cluster level of p < 0.05 corrected by FWE correction.

The demographic and clinical features of all subjects are listed in Table 1. Handedness, age and sex were not significantly different among the HC, PD-MCI and PD-CU groups. Education was not significantly different between total PD patients and controls, whereas, the PD-CU group had significantly more education years than the PD-MCI group. No significant differences in the disease duration, H&Y stage, UPDRS score and HDRS/HARS score between PD-CU and PD-MCI groups were found. By definition, the MoCA scores of PD-CU patients were significantly higher than those of PD-MCI patients (p < 0.000). The scores of all neuropsychological tests of all subjects are provided (Supplementary Table S3).

To characterize the abnormal FC in the DMN, we chose 18 seed regions as representatives of the DMN. Relative to HC, the PD-MCI group showed significantly reduced connectivity of the DMN with a set of regions in the brain cortex. Specifically, the key hubs, mPFC, SFG and PCC, showed a decreased connectivity pattern with the right precentral gyrus and bilateral insula. Also, the mPFC and SFG showed a reduced coupling with the anterior inferior parietal lobule (aIPL); the mPFC and PCC showed a reduced coupling with the bilateral middle frontal gyri (MFG). FC analysis of the bilateral ATLs also showed a decreased connectivity pattern with the bilateral middle temporal gyri (MTG). The PD-MCI group presented significantly reduced connectivity of the DMN with the MFG and MTG compared to the PD-CU group. There was no significant difference between the PD-CU group and HC (see Table 2, Figure 1).

Comparing the mean connectivity values, ordered reductions (HC > PD-CU > PD-MCI) were observed in intra-DMN connectivity, although there was no significant difference between the PD-CU group and HC. Compared with HC, the PD-MCI group showed significant intra-DMN reductions predominantly between HF and IFG, between ATL and IFG, and between PCC and pIPL. The PD-MCI group showed a significantly lower FC between ATL and IFG than the PD-CU group. The means and SDs of z-values are shown in Figure 2 and Supplementary Table S4.

To determine the relationship between the FC values and neuropsychological scores of PD patients, we performed the analysis in all PD patients. The FC of ATL and MTG was positively correlated with Golden Stroop Test score (a: r = 0.423, p = 0.016), and the functional coupling of HF and IFG was significantly correlated with the HVLT-R total score (b: r = 0.429, p = 0.014) (see Figure 3).

There were no significant differences among the three groups in the GM volume, indicating that the altered FC was not caused by anatomic changes.