L-DOPA responsiveness is associated with distinctive connectivity patterns in advanced Parkinson’s disease

Abstract

Introduction

Neuronal loss and dopamine depletion alter motor signal processing between cortical motor areas, basal ganglia and thalamus resulting in the motor manifestations of Parkinson’s disease. Dopamine replacement therapy can reverse these manifestations with varying degrees of improvement.

Methods

To evaluate functional connectivity in patients with advanced Parkinson’s disease and changes in functional connectivity in relation to the degree of response to L-DOPA, nineteen patients with advanced Parkinson’s disease underwent resting state functional magnetic resonance imaging in the ON medication state. Scans were obtained on a 3-Tesla scanner in 3×3×2.5 mm³ voxels. Seed based bivariate regression analyses were carried out with atlas-defined basal ganglia regions as seeds, to explore relationships between functional connectivity and improvement in the motor section of the Unified Parkinson’s Disease Rating Scale (UPDRS-III) following L-DOPA challenge. False discovery rate (FDR)-corrected p-value was set at <0.05 for two-tailed t-test.

Results

A greater improvement in UPDRS-III scores following L-DOPA administration was characterized by higher resting state functional connectivity between the prefrontal cortex and the striatum (p=0.001) and lower resting state functional connectivity between the pallidum (p=0.001), subthalamic nucleus (p=0.003) and the paracentral lobule (supplementary motor area, mesial primary motor and primary sensory areas).

Conclusion

Our findings show characteristic basal ganglia resting state functional connectivity patterns associated with different degrees of L-DOPA responsiveness in patients with advanced Parkinson’s disease. L-DOPA exerts a graduated influence on remapping connectivity in distinct motor control networks potentially explaining some of the variance in treatment response.

Introduction

Parkinson’s disease is a neurodegenerative disorder that affects multiple systems in the brain in a progressive and, to some extent, a predictable fashion [1]. Considerable degeneration of dopaminergic neurons in the substantia nigra pars compacta leads to dopamine depletion in the striatum. This depletion causes dysfunction within the basal ganglia-thalamo-cortical ‘motor’ circuit resulting in the hallmark motor triad of resting tremor, rigidity and bradykinesia. [2,3].

Numerous animal and human studies have demonstrated structurally and functionally segregated parallel circuits or loops that connect the cortex to the basal ganglia and thalamus [4–6]. This segregation is essential for motor processing and learning. Dopamine deficiency results in loss of this segregation and emergence of synchronized oscillations between basal ganglia structures and cortical areas. [7]. A correlation between motor deficits in Parkinson’ and abnormal synchronized oscillatory activity, especially in the ‘beta’ band width (15-35hz) [8–10], has been shown at multiple levels in the motor circuit which is also suppressed by dopaminergic therapies [11–13]. Alterations of basal ganglia physiology in Parkinson’s disease, therefore, broadly consist of two elements, an increase in the neuronal discharge rate in the STN and globus pallidus internus (Gpi) and the appearance of synchronized neuronal oscillations [14]. A new, dynamic ‘center-surround model’ has been proposed with excitatory input from the motor cortex directly influencing the STN through the ‘hyper-direct’ pathway and affecting the globus pallidus without passing through the striatum [15–18].

Functional connectivity of the brain can be explored using resting state fMRI. This offers a relatively simple and fast approach to map brain functional connectivity and pathological neural network changes. The technique does not rely on an experimental task design, making data analysis streamlined and less vulnerable to experimental bias [19].

The time course of low frequency (<0.1 Hz) fluctuations in blood oxygen level-dependent (BOLD) signal has a high degree of temporal correlation in functionally connected brain areas or ‘nodes’ [20]. Multiple statistical modeling techniques such as seed-based correlation mapping and independent component analysis (ICA) can then be used to examine this functional connectivity [19,21]. Resting state functional connectivity (fcMRI) has been used in various clinical applications [22]. For example, selective changes were found in individuals at risk of Alzheimer’s disease [23] and also documented in patients with major depression [24]. Statistically significant positive correlations have been found between fcMRI and structural connectivity [25–28].

Resting state fMRI connectivity changes may relate to the abnormal synchronized oscillations in Parkinson’s disease. Evidence for this relation has been demonstrated in numerous human and non-human primate studies. Correlation between the slow frequency (<0.1 Hz) of the BOLD signal in resting state fMRI and neural electrophysiological activity at higher frequencies [29], in the alpha band range [30], the gamma band range [31] and in the beta band range [32], provides inference between networks identified using fcMRI and their underlying neurophysiological correlates. This relation can be very useful in studying and understanding Parkinson’s disease neurophysiology.

To date, several fcMRI studies have examined the connectivity profile in patients with Parkinson’s disease compared to healthy controls and the effect L-DOPA therapy has on the connectivity pattern [33–41]. The majority of these studies, with the exception of one [40], were on patients in the early stages of Parkinson’s disease. Limited information is available about the functional connectivity changes that occur later on in the course of the disease when complications of chronic L-DOPA therapy such as levodopa induced dyskinesias (LID) and motor fluctuations emerge.

Furthermore, patients can have varying degrees of response to L-DOPA in advanced Parkinson’s disease. The emergence and progression of L-Dopa refractory signs represents one of the most challenging aspects of treating the disease. Furthermore, the degree of L-Dopa responsiveness is an important factor in patient selection for STN DBS. Greater than 25%–50% improvement in motor UPDRS (UPDRS III) following ‘L-DOPA challenge’ is generally considered a good outcome predictor for STN DBS in Parkinson’s disease [42].

In this study, a seed-based approach to resting state functional MRI is used to examine differences in functional connectivity of the basal ganglia with cortical and subcortical areas in relation to the degree of clinical response to L-DOPA challenge in 19 patients with advanced Parkinson’s disease and motor fluctuations. Our aim is to identify the fcMRI correlates associated with differences in treatment response to better understand the underlying neural network activity in advanced Parkinson’s disease.

Materials and methods

This study received ethical approval by West London NHS Research Ethics Committee (10/H0706/68). All participants provided written informed consent.

Patients

Nineteen patients who met UK brain bank criteria for idiopathic Parkinson’s disease were included (Table 1). Patients were recruited from the surgical waiting list for bilateral STN DBS after selection by a multidisciplinary team of specialized movement disorders neurologists and functional neurosurgeons. All patients underwent a neuropsychological assessment and a structural brain MRI to rule out dementia and significant brain atrophy, respectively. A formal L-DOPA challenge test was conducted, whereby the motor section of the unified Parkinson’s disease rating scale (UPDRS III), was assessed in the OFF state at least 12 hours after omitting Parkinson’s disease medications and repeated 30-60 minutes (or when clinically ON) after administration of the patient’s regular medications supplemented with an additional dose of 50mg/12.5mg dispersible Madopar. Inclusion in the present study was limited to L-DOPA responsive patients with advanced Parkinson’s disease with at least 25% improvement of UPDRS III between off and on medication states and who could tolerate lying flat for the duration of the scan.

Table 1.

Patient information and L-DOPA challenge

		Range	Mean ± Std. Deviation
Age (yr.)	-	41-70	55.5 ± 9.9
Gender	15M, 4F	-	-
Hand dominance	19 RHD
Disease duration (yr.)	-	4-22	11.2 ± 4.4
Duration of motor fluctuations (yr.)	-	1-9	3.4 ± 1.8
UPDRS III OFF (p)	-	20-73	43.8 ± 12.9
UPDRS III ON (p)	-	4-42	17.5 ± 9.9
Left hemibody UPDRS III OFF (p)	-	4-23	14.2 ± 6
Right hemibody UPDRS III OFF (p)	-	4-20	11.8 ± 4.7

RHD=right hand dominant, M=male, F=female, yr.=year, p=points.

Magnetic resonance imaging data acquisition

Imaging was acquired during admission before DBS surgery. Medications were optimised by a specialist neurologist to ensure patients were in the ON- state, with no concurrent head tremor or LID at the time of MRI. 3T Siemens Magnetom Trio TIM Syngo MR B17 was used with a well-padded 32 channel receive head coil to reduce discomfort and head motion during the scan.

Resting state functional MRI

Multi-echo Echo Planar Imaging (2D-MEEPI) sequences were obtained in two successive acquisitions with a total duration of 15 minutes. Spatial resolution=3×3×2.5 mm³, repetition time=60 ms, echo time=30 ms, flip angle=90 degrees, field of view=192 mm × 192 mm, 45 axial slices (2.5 mm thickness), matrix size=64×64 and a total of 512 scans. Patients were given instructions to keep their eyes open and gaze fixed on a cross hair.

Structural MRI

Multiple Parameter Mapping (MPM) sequences [43] were acquired for structural imaging just before the functional MRI scan. In brief, the whole protocol consisted of three 3D FLASH acquisitions performed with T1, Proton Density (PD), and Magnetization Transfer (MT) weighting; these were paired with B1 (transmit field mapping data to correct for the effect of inhomogeneous flip angles on the T1 maps) and B0 field map acquisitions. Spatial resolution=1×1×1 mm³, repetition time=24.5 ms, multiple echo times, field of view=256 mm, flip angle=6 (PD), 21 (T1), 6 (MT) degrees, matrix=256×256, partitions=176; total acquisition time of 26 minutes.

Preprocessing

The structural scans were processed using the voxel based quantification (VBQ) toolbox in SPM12 to generate quantitative maps of multiple tissue properties (MT, PD*, R1 (1/T1) and R2* (1/T2*)) [43]. The MT maps, which provide high contrast to noise ratio, were segmented to generate white matter, grey matter and cerebrospinal fluid (CSF) maps using the ‘Segment’ toolbox in SPM12 [44,45].

The first three scans of each resting state session were discarded. The two sessions were concatenated to produce a single 506 scan time-series. Functional volumes underwent realignment, unwarping and slice-timing correction. The functional scans of each subject were rigidly registered to the anatomical (R1) scans and the corresponding CSF and white matter maps. Anatomical scans were then spatially normalized to MNI space (spatial resolution=2×2×2 mm³); the resultant transformation was further used to normalize the functional data. Functional outlier detection was carried out using Artifact Detection Tools (ART)-based scrubbing (www.nitrc.org/projects/artifact_detect/). The functional volumes were then inspected and smoothing was applied to reduce potential spatial and temporal artifacts using an 8 mm full width at half maximum Gaussian kernel.

In order to reduce spurious sources of variance in the functional data; denoising was carried out using component-based noise correction method (CompCor) [46]. The temporal time series with estimated subject motion (three rotation and three translation parameters, plus another six parameters representing their first-order temporal derivatives), as well as the BOLD time series within the subject-specific white matter mask (three principal component analysis parameters) and CSF mask (three principal component analysis parameters), were used as temporal covariates and removed from the BOLD data using linear regression. Ultra-low frequency fluctuations in the resulting residual BOLD time series were removed using a high-pass filter (1/128s, ≈0.0078 Hz). All processing steps were undertaken in MATLAB-based CONN toolbox [47].

With tremor and dyskinesias as potential sources of motion, the framewise displacement (FD) was calculated in order to precisely quantify the degree of motion correction performed for each volume of each subject [48]. In order to assess the possibility of movement confounding fcMRI analysis, a Pearson correlation was performed between the mean of each patient’s FD values and the fcMRI of interest.

Analysis

Seed based analysis was performed. Regions of interest (ROI) were defined on the FSL Harvard-Oxford cortical structural atlas, the ATAG subthalamic nucleus atlas [49], and the ATAG MNI04 basal ganglia atlas. The right and left caudate, putamen, pallidum, STN, and thalamus were defined as ROIs for seed-to-voxel and ROI-to-ROI analysis. Those nodes largely incorporate the subcortical elements of the cortical-basal ganglia-thalamic loop implicated in motor control [50,51]. The paracentral lobules were used in the ROI-to-ROI analysis to further explore the strong fcMRI shown in the seed-to-voxel analysis.

Bivariate regression coefficients of the ROIs BOLD timeseries were measured before being entered into second-order between subject analysis. The connectivity of each of these network nodes to the rest of the brain was individually explored using seed-to-voxel analysis. FcMRI within cortico-basal ganglia-thalamic circuits was also explored using ROI-to-ROI analysis. Only the BOLD signal within each subject-specific gray matter mask was included in these calculations; normalizing each subject to the aforementioned atlases permitted second order analysis.

General linear models were generated to assess the relationship between the degree of improvement after L-DOPA (% improvement UPDRS-III) and connectivity. Thresholds for voxel-level height and cluster-level extent were set to a false discovery rate (FDR)-corrected P-value of <0.05. Seed-level correction was used to apply FDR separately for each seed ROI by implementing both a voxel-level height threshold and a cluster level extent threshold. Given the changes that occur in fcMRI with age (Ferreira, 2015), the primary effect of change in UPDRS III score with medication on functional connectivity was investigated while controlling for patient age in the GLM. In order to account for any effects related to differences in disease severity at the time of testing, the primary effect of L-dopa related change in UPDRS III score on cortico-basal ganglia-thalamic networks was investigated while controlling for UPDRS III OFF-medication score.

Results

Patients

Scanning proceeded with no adverse effects. The mean improvement in UPDRS III following L-DOPA administration was 26.3 points (60%, SD=20%) [95% confidence interval (CI): 21.5–31.1, P<0.0001]. There was no significant difference between the left and right UPDRS III OFF hemi-body scores [mean difference 2.4 points, P=0.12]. One patient had L-DOPA responsive tremor-dominant Parkinson’s disease (UPDRS III OFF=33, ON=10). Mean Mini Mental State Examination Score (MMSE)=29.5/30 (SD=0.6). Modified Hoehn and Yahr Staging was 3-4 for all patients. Mean levodopa equivalent dose (LED) was 1330mg (SD=484.5). No significant correlation was found between daily LED and UPDRS-III OFF scores (r=0.16, P=0.5) or UPDRS-III improvement during L-DOPA challenge (r=0.32, P=0.18 and r=0.25, P=0.31 for raw and percentage change respectively). Detailed patient information can be found in Table 1.

Quantification of Motion Correction

Mean framewise displacement (FD) values ranged from 0.3mm to 2.8mm (1.6 ± 0.7mm). No volumes were discarded in preprocessing. There was no statistically significant correlation between FD measures and fcMRI calculations. [Table 2]

Table 2.

Analysis of potential effects of motion correction (framewise displacement, FD) on resting state functional connectivity (fcMRI). For each ROI-to-ROI analysis, a Pearson correlation was performed between mean FD values for each patient and the calculated fcMRI values.

	GP-PL	Thal-PL	STN-PL	Put-PL	GP-M1	Thal-M1	STN-M1	Put-M1	Caud-MFPC	Caud-LPFC	Put-LPFC
R	-0.05	0.44	0.37	-0.01	-0.16	0.27	-0.01	0.04	0.13	0.34	0.39
p-value	0.88	0.18	0.27	0.98	0.64	0.43	0.99	0.90	0.71	0.31	0.24

GP = globus pallidus; Thal = thalamus; STN = subthalamic nucleus; Put = putamen; Caud = caudate; PL = paracentral lobule; M1 = motor cortex; MFPC = medial prefrontal cortex; LPFC = lateral prefrontal cortex

Basal ganglia and thalamic functional connectivity

Greater improvement in UPDRS III scores following L-DOPA administration was associated with higher fcMRI between the caudate and dorsolateral and medial prefrontal cortices and between the putamen and the inferior part of the left dorsolateral prefrontal cortex (DLPFC), centered on the inferior frontal gyrus. Greater improvement in UPDRS III scores was also associated with lower fcMRI between the basal ganglia and thalamus and the sensorimotor cortex, specifically, in the mesial areas with the portions of M1, SMA, and S1 residing in the paracentral lobule and superior portion of the cortical convexity. This was particularly notable for the pallidum and thalamus, and less prominent for the caudate and putamen. Table 3

Table 3.

Functional connectivity of basal ganglia and thalamus with paracentral lobule in correlation with UPDRS III reduction following L-DOPA administration

	Paracentral lobule				t-score	FDR corrected P-value
	MNI coordinates			Cluster Size
	x	y	z
STN	+32	-16	+56	9445	-3.46	0.0032
Pallidum	+24	-48	+58	10407	-4.03	0.0010
Caudate	+14	-52	+62	2672	-3.72	0.0018
Putamen	-02	-12	+74	2123	-3.36	0.0040
Thalamus	+32	-30	+74	8093	-3.77	0.0017

MNI coordinates represent the centre of gravity of the cortical cluster

FDR=false discovery rate

The cluster for STN negative correlations with M1 and S1 was more diffuse. Greater improvement in UPDRS III scores was also associated with lower fcMRI between the thalamus and the posterior parietal cortex (PPC), between pallidum and PPC, between pallidum and lingual and fusiform gyri (visual association area), between STN and right PPC, between STN and left superior temporal gyrus, between STN and lingual gyrus, cuneus and precuneus (left) and between STN and posterior and caudal anterior cingulate cortices. Figure 1

Figure 1.

Functional connectivity changes between cortical areas and subthalamic nucleus (A), globus pallidus (B), caudate (C), putamen (D) and thalamus (E) in relation to percentage improvement in UPDRS III with L-DOPA administration using a seed-to-voxel analysis.

ROI-to-ROI analysis showed that UPDRS III improvement with L-DOPA was positively associated with higher fcMRI between the caudate and the thalamus [t=3.27, corrected-p=0.03] and the caudate and the pallidum (t=3.38, corrected-p=0.01), on the left. A similar trend was also seen on the right side between the caudate and the thalamus but did not reach statistical significance after correcting for multiple comparisons (t=2.60, corrected-p=0.07 (see figure 2).

Figure 2.

Connectivity graph showing ROI-to-ROI analysis (t-score) against improvement in UPDRS III scores following L-DOPA administration. Line colors and width reflect the strength and polarity of connectivity as designated in the heat bar (correlations represented in orange and red; and anti-correlations represented in blue). The nodes on the perimeter of the connectivity circle are labeled and their colors correspond to those illustrated on the adjacent brain images.

Discussion

In this work, resting state functional MRI was used to examine basal ganglia and thalamic functional connectivity in patients with advanced Parkinson’s disease. Employing seed-to-voxel and ROI-to-ROI approaches, associations with the degree of response to L-DOPA were demonstrated. After adjusting for age and disease severity (baseline UPDRS III), significant associations were evident between L-dopa responsiveness and fcMRI between basal ganglia and cortex, and within basal ganglia and thalamus. The degree of response to L-DOPA, as measured by improvement of UPDRS III scores, was associated with increased fcMRI between the striatum and thalamus (caudate-pallidum and caudate-thalamus) and between the striatum and the PFC. The degree of response to L-DOPA was inversely associated with increased fcMRI between the thalamus, pallidum and STN and the paracentral lobule and parieto-occipito-temporal association areas. As there was no significant correlation between movement correction measures and fcMRI, it is more likely that these findings are the result of inherent differences in the basal ganglia-thalamic-cortical networks than the result of differences in patient movement.

Daily LED did not appear to correlate significantly with the degree of UPDRS-II response during the L-DOPA challenge or with the baseline UPDRS-III scores. This is likely due to the unpredictable bioavailability of dopamine in the brain following oral administration of levodopa mediated by erratic intestinal absorption, metabolism and transport across the blood brain barrier [52–54]. Levodopa dose was therefore discounted from the analysis.

A consistent pattern of change in metabolic activity in Parkinson’s disease has been described previously in PET studies. Disease progression is often associated with increasing metabolism in the STN, GPi and thalamus as well as in the dorsal pons and M1, which progresses to declining metabolism in the PFC, SMA and inferior parietal regions. This pattern is shown to be reproducible and is referred to as a Parkinson’s disease related covariance pattern (PDRP) [55–57]. Studies that used ‘experimental design’ fMRI to examine the activation patterns of the motor network in Parkinson’s disease patients showed cortical activation abnormalities centered on a fronto-parietal cortical network comprising the pre-SMA, M1, and posterior parietal cortex, partially reversed with L-DOPA. Inconsistent results, potentially a consequence of disparities in selection and timing of the applied tasks, were found in those studies. [58].

Other studies have looked at group changes in activation in early or advanced PD patients in the presence and absence of L-dopa therapy but none have sought to directly analyze the relationship between functional connectivity and degree of L-dopa response.

In a study that examined fcMRI changes in Parkinson’s disease, a reduction in connectivity between the SMA, left dorsolateral PFC and left putamen was demonstrated. This connectivity pattern correlated with UPDRS III and was relatively normalized following L-DOPA administration [33]. Decreased connectivity between the PFC and the left putamen, right insula, right PMA, and left inferior parietal lobule was also demonstrated in another study of 18 Parkinson’s disease patients and matched controls [36]. We showed in the present study that an increase in resting functional connectivity between the PFC, mostly in the inferior frontal gyrus (IFG), and the striatum was associated with a better response to L-DOPA. The IFG role is well described in the behavioral (proactive) inhibition network [59–65]. It has also been implicated in the pathogenesis of L-DOPA-induced dyskinesias (LID), possibly a reflection of motor inhibition network failure. Transcranial magnetic stimulation (TMS) over the right IFG reduces the extent of dyskinesia induced by a supra-maximal single dose of L-DOPA [40].

Changes in fcMRI between the STN, M1, SMA and PMA have been previously described in Parkinson’s disease compared to controls. In a study of fcMRI of the STN in 31 Parkinson’s disease patients with early stage disease during the medication-off state compared to 44 healthy controls, connectivity was increased to M1, SMA and PMA. A tremor-dominant subgroup showed localized increase in connectivity between the STN and hand area of M1 and the primary sensory cortex. In a non-tremor subgroup, increased connectivity was found between the STN and mesial cortical motor areas including the SMA [35].

The pattern of fcMRI changes shown in our study is supported by findings from a recent study that looked at differences in ‘networks’ identified on independent components analysis (ICA) in 27 Parkinson’s disease patients and 16 controls. Significant functional connectivity increase in patients, between the sensorimotor network and the spatial attention network in the parietal cortex was found and was correlated with UPDRS III [41]. Furthermore, L-DOPA has been shown to restore fcMRI in the sensorimotor network, especially in the SMA, when administered to drug-naïve patients [38].

Our analysis did not show a statistically significant correlation between improvement in UPDRS III and change in fcMRI between the putamen and the inferior parietal cortex. However, changes in fcMRI between these areas have been shown in Parkinson’s disease patients when compared to controls [34].

We have shown that reduced resting functional connectivity between the striatum, pallidum and the thalamus is associated with a worse response to L-DOPA. It is conceivable that this finding is a result of the breakdown of segregation within the basal ganglia and the emergence of oscillatory activity within the sensorimotor network with dopamine depletion [7]. This pattern of fcMRI abnormality was also demonstrated in a study that focused on the functional connectivity between the striatum, thalamus and the brainstem. In that study, 13 patients with advanced Parkinson’s disease compared with 19 controls showed lower functional connectivity patterns with the thalamus, midbrain, pons and cerebellum. A progressive ‘gradient’ of altered connectivity (posterior putamen > anterior putamen > caudate) was observed in keeping with the pattern of dopaminergic loss in the pathogenesis of PD [37].

The shared findings of studies that investigated the way fcMRI is influenced by L-DOPA in Parkinson’s disease [33–39,41], point towards an abnormal fcMRI pattern which is relatively restored following L-DOPA administration. We propose that the contrasting patterns of functional connectivity that we have observed in association with L-DOPA responsiveness are a reflection of the relative degrees of normalization of functional connectivity with dopamine therapy. On the one hand, improvement of motor symptoms with L-Dopa was associated with increased resting functional connectivity between the striatum and the DLPFC and IFC; and on the other hand with decreased resting functional connectivity of the STN and pallidum with the paracentral lobule and parieto-occipito-temporal association areas. It has been previously shown that PD patients can suppress involuntary movements albeit for short periods of time [66]. Given that we examined functional connectivity in the resting state, when patients were instructed to remain at rest and refrain from making any movements, the observed pattern of increased caudate-DLPFC/IFC functional connectivity with greater L-Dopa responsiveness may potentially reflect engagement of a goal-directed proactive inhibition network to withhold voluntary movement [65,67]. By contrast, the negative association between L-Dopa responsiveness and functional connectivity of the STN and pallidum with the M1 and posterior association areas may reflect engagement of a more ‘reactive’ inhibition pathway via the hyperdirect pathway [65,67] for withholding movements during the resting MRI in those with a worse response to levodopa medication. However, further work is needed to support this proposition.

Increased functional connectivity between the thalamus and the sensorimotor cortex has been previously demonstrated fcMRI studies comparing PD patients to healthy controls [68]. This is consistent with primate models of PD where there is increased activity in the ventrolateral thalamus [51]. In our study, we demonstrated a negative correlation between thalamic-sensorimotor fcMRI and the response to L-DOPA. This once again supports the hypothesis that in poor responders there is reduced fcMRI normalisation to dopamine than in good responders.

The validity of negative correlations or anticorrelations as neurophysiological findings in fMRI data has been a subject of debate. Analytic artifacts may arise from carrying out global signal regression to remove confounds due to noise in the BOLD timeseries leading to anticorrelations even when they are not truly present [69–71], [72]. This had led to a widely accepted consensus not to interpret those findings when using global signal regression [69,73]. In our analysis, however, we applied the CompCor method of noise reductions which does not rely on global signal regression and which has been shown to have higher sensitivity and specificity [74]. It is believed that the use of the ComCor method can generate valid anticorrelations between brain networks in fMRI [47]. We consider that the findings of negative correlations between subcortical and cortical areas in relation to the degree of response to L-DOPA here may have sound neurophysiological basis.

Limitations

All our patients were on optimal L-DOPA medication at the time of the MRI scan. For this reason, we cannot make any inferences on the acute effects of L-DOPA on the underlying pathophysiology as was demonstrated in other reports that compared changes in fcMRI before and after L-DOPA administration. [33–41]. In contrast with the majority of these reports our patients had advanced Parkinson’s disease with severe motor fluctuations and, not infrequently, freezing of gait. We opted to only scan the patients in the ON state with their usual treatment regime firstly to reduce patient discomfort during the MRI scan and secondly to reduce motion artefact, which may degrade the quality of the resulting fMRI scans.

Another limitation associated with fcMRI techniques is that the technique is largely limited to group-level analysis. While this is useful in exploring group-wise changes, inferences on the individual level cannot be readily made, especially on a diagnostic/predictive capacity. This may limit the clinical application of the technique in individual patients.

Our analysis focused on fcMRI changes between cortico-subcortical structures in Parkinson’s disease. We employed prior knowledge to define basal ganglia structures as ROIs for connectivity analysis. We have not explored cortico-cortical fcMRI changes and our findings, therefore, are limited to networks that include those predefined ROIs.

Conclusion

Differences in functional connectivity patterns of the basal ganglia, as mapped using resting state fMRI, are associated with different degrees of response to L-DOPA therapy in patients with advanced Parkinson’s disease, at the group level. This is consistent with the hypothesis that the clinical effects of dopamine are a result of remapping of functional connectivity. Networks linked to cognitive (proactive) motor inhibition show relatively higher connectivity whilst networks linked to reactive motor inhibition show lower connectivity with better dopamine response. Furthermore, connectivity is relatively stronger in between basal ganglia structures with better dopamine response. Future studies may be able to validate these results and explore markers at the ‘individual level’ by employing machine learning algorithms to build predictive models of response to treatment, thus validating, or even corroborating the L-DOPA challenge test. The next step will be to explore the functional connectivity correlates predictive of response to STN DBS in this group once data become available. This could potentially aid with patient selection and help with understanding the mechanism of action of DBS.

Abbreviations

ART

Artifact Detection Tools

BOLD

blood oxygen level- dependent

CompCor

component-based noise correction method

DBS

deep brain stimulation

DCM

dynamic causal modelling

DLPFC

dorso-lateral prefrontal cortex

fcMRI

Resting state functional connectivity

FDR

false discovery rate

fMRI

functional MRI

GPe

globus pallidus externus

GPi

globus pallidus internus

ICA

independent component analysis

LID

levodopa induced dyskinesia

M1

primary motor area

MEEP

multi-echo echo planar imaging

MNI

Montreal Neurological Institute

MT

magnetization transfer

PD

proton density

PDRP

Parkinson’s disease related covariance pattern

PET

positron emission tomography

PFC

prefrontal cortex

PMA

premotor area

PPC

posterior parietal cortex

S1

primary sensory area

SMA

supplementary motor area

SN

substantia nigra

SNr

substantia nigra pars reticulata

STN

subthalamic nucleus

UPDRS III

unified Parkinson's disease rating scale motor score