Asymmetric dopaminergic degeneration and levodopa alter functional corticostriatal connectivity bilaterally in experimental parkinsonism.

Asymmetric dopamine loss is commonly found in early Parkinson's disease (PD), but its effects on functional networks have been difficult to delineate in PD patients because of variations in age, disease duration and therapy. Here we used unilateral 6-hydroxydopamine-lesioned (6-OHDA) rats and controls and treated them with a single intraperitoneal injection of levodopa (L-DOPA) before performing diffusion weighted MRI and resting state functional MRI (rs-fMRI). In accordance with a neurodegeneration of the nigrostriatal dopaminergic pathway, diffusion tensor imaging showed increased radial diffusivity and decreased fractional anisotropy in the lesioned substantia nigra. Likewise a deterministic connectometry approach showed increase of isotropic diffusion values in the medial forebrain bundle. rs-fMRI showed reduced interhemispheric functional connectivity (FC) between the intact and the 6-OHDA lesioned caudate-putamen. Unexpectedly, there was an increased FC between the 6-OHDA lesioned caudate-putamen and sensorimotor cortices of both hemispheres. L-DOPA reversed the FC changes between the dopamine denervated caudate-putamen and the sensorimotor cortices, but not the reduced interhemispheric FC between caudate-putamina. Similarly, L-DOPA induced c-fos expression in both sensorimotor cortices, but only in the dopamine-depleted caudate-putamen. Taken together, these data suggest that asymmetric degeneration of the nigrostriatal dopamine pathway results in functional asynchrony between the intact and 6-OHDA-lesioned caudate-putamen and increased interhemispheric synchrony between sensorimotor cortices. The results also indicate that the initial effect of L-DOPA is to restore functional corticostriatal connectivity rather than synchronize caudate-putamina.

Introduction

Parkinson's disease (PD) is characterized by a progressive and asymmetric loss of nigrostriatal dopaminergic neurons in substantia nigra pars compacta resulting in bradykinesia, rigidity and rest tremor (Kalia and Lang, 2015). PD is symptomatically treated with dopamine replacement, among which the dopamine precursor levodopa (L-3,4-dihydroxyphenylalanine) (L-DOPA), together with a peripheral dopa decarboxylase inhibitor, is the golden standard (Ibid.).

PD diagnosis is based solely on clinical assessment (Postuma et al., 2015), which often leads to misdiagnosis (Adler et al., 2014). Reproducible and robust objective biomarkers to support diagnosis and quantify disease progression would represent a scientific breakthrough (Mehta and Adler, 2016). A promising method for aiding diagnosis is diffusion weighted magnetic resonance imaging (dw-MRI), particularly diffusion tensor imaging (DTI). Indeed, a consistent finding in PD patients is a reduction of fractional anisotropy (FA) and an increase of mean diffusivity in the substantia nigra (e.g. Vaillancourt et al., 2009, Rolheiser et al., 2011, Skorpil et al., 2012, Cochrane and Ebmeier, 2013, Schwarz et al., 2013). Additional changes have also been observed in other brain regions, particularly in different white matter structures, but without consensus. Resting-state functional magnetic resonance imaging (rs-fMRI) is favorable to apply to PD as it requires nearly no effort from the patient. It has therefore been utilized in PD patients, but with conflicting results (e.g. Palmer et al., 2010, Luo et al., 2014, Yu et al., 2013, Agosta et al., 2014). Accordingly, the variety of disease severity, treatment conditions and analysis methods has prevented a conclusive meta-analysis of the results from rs-fMRI in PD (Tahmasian et al., 2015). In contrast to clinical studies, animal models of parkinsonism offer advantages, including well-controlled age, therapy, environmental settings and genetic background.

To experimentally study effects of asymmetric dopamine loss in a highly controlled manner, unilateral injection of the neurotoxin 6-hydroxydopamine (6-OHDA) into the medial forebrain bundle (MFB) of rodents is commonly used (Herrera-Marschitz et al., 2010). 6-OHDA is retrogradely taken up via the dopamine transporter into dopaminergic neurons where it leads to cell death by oxidative stress within a few days. Using structural MRI it has been reported that 6-OHDA injections in the nigrostriatal pathway results in an edema followed by a hypointensity and an increase in T2 relaxation time interpreted as iron accumulation (Kondoh et al., 2005, Soria et al., 2011; Virel et al., 2014). In contrast to the aforementioned data with DTI in PD patients, it was reported that intrastriatal injection of 6-OHDA increases nigral FA in rats (Van Camp et al., 2009).

To elucidate structural and functional connectivity in experimental PD and in response to L-DOPA under controlled conditions, we performed DTI, connectometry and rs-fMRI in the 6-OHDA experimental model of parkinsonism at baseline and following a single injection of L-DOPA.

Materials and methods

Animals and lesioning protocol

Male Sprague-Dawley rats (Charles River Laboratories, Sulzfeld, Germany), weighing 150 g, were used in this study. Experiments were performed in agreement with the European Communities Council (86/609/EEC) and approved by the Stockholm North Ethical Committee (Ethical permit # N245/11). Animals were arranged in two experimental cohorts. Both cohorts included rats lesioned with 6-OHDA application in the MFB. These 6-OHDA-lesioned rats were compared with naive control rats (experimental cohort 1) or sham-lesioned rats (experimental cohort 2 and ex-vivo experiment). To protect noradrenergic neurons, rats were systemically treated (i.p.) with a mixture of 25 mg/kg desipramine (Sigma–Aldrich, Saint-Louis, Missouri, USA) and 5 mg/kg pargyline (Sigma–Aldrich). They were then anesthetized with 80 mg/kg ketamine (i.p.; Pfizer, New-York, USA) and 5 mg/kg xylazine (i.p.; Bayer, Leverkusen, Germany) and mounted in a stereotactic frame. A hole in the skull was drilled and 12.5 μg of 6-OHDA in 2.5 μL of 0.01% ascorbate (Sigma–Aldrich) (6-OHDA groups) or 2.5 μL of 0.01% ascorbate (sham groups) was injected into the MFB of the right hemisphere (AP, − 2.8 mm; ML, − 2.0 mm; and DV, − 9.0 mm relative to bregma) (Paxinos and Watson, 1998). The skin incision was sutured and post-operative pain relief was given using 0.01 mg/kg injections of Temgesic (s.c.; Indivior UK Limited, Slough, UK) directly after the surgery and twice more within 48 h. Two weeks after unilateral 6-OHDA lesioning, rats were administered 1 mg/kg of the dopamine D1/D2 receptor agonist apomorphine (i.p.; Sigma–Aldrich) and observed for 30 min. Since 6-OHDA-lesioning results in supersensitization of dopamine receptors, a contralateral rotational response can be quantitated in successfully dopamine denervated animals (Ungerstedt, 1971, Jorge et al., 1975). Only rats rotating > 100 turns contralateral to the lesioned hemisphere during these 30 min were included in further experiments. MRI scanning was performed four weeks after the 6-OHDA lesioning.

Ex vivo dw-MRI experiment

Animals and tissue preparation

Eight rats were lesioned (four 6-OHDA-lesioned and four sham-lesioned) using the aforementioned protocol. Four weeks after the surgery, the rats were anesthetized with ketamine/xylazine and perfused intracardially with 4% paraformaldehyde to fixate their brains. They were then decapitated and their heads placed in a 4% paraformaldehyde solution for 24 h at 4 °C before being washed and subsequently stored in a phosphate buffer solution at 4 °C. Several hours before the MRI scan, the skin and muscular tissues surrounding the skull were removed so that the heads were tightly fitted in 30 mL syringes (Terumo Europe N.V., Leuven, Belgium) filled with Fomblin (Solvay Solexis, Neder-Over-Hembeek, Belgium). Particular attention was paid in extracting air from the syringe and from tissues to reduce risks of magnetic susceptibility artifacts in the MRI scans. Samples were at room temperature during scanning.

Data acquisition

Data were acquired in a 9.4 Tesla horizontal scanner (Agilent, Yarnton, UK) equipped with a 31 cm wide bore size. A volume coil of the millipede design with a 30 mm inner diameter (Varian, Yarnton, UK) and a gradient system capable of 1000 mT/m (Varian) thermostated to 20 °C was used. Heads were placed horizontally in the scanner and shimming of the static magnetic field was performed manually. A diffusion-weighted spin-echo sequence was used to acquire diffusion images for a total scan time of 40 h (Jones 30 kernel (30 diffusion weighted direction and one non-diffusion weighted scan), b-value = 1250 s/mm2, 100 slices in coronal orientation, 150 μm thick slices, no gap, field of view (FOV) equal to 38,4 × 28,8 mm for a matrix size of 256 × 192 voxels resulting in a 150 μm isotropic resolution and 6 averages). The parameters chosen for the dw-MRI scan were based on a compromise between a sufficient b-value, a high enough spatial resolution, the number of directions and the scanning time (Dyrby et al., 2011).

DTI data processing

b0-maps obtained using dw-MRI were skull-striped in two steps. They were first stripped semi-automatically using the Brain Surface Extractor tool of BrainSuite version 14b1 (brainsuite.org). The masks obtained were then manually corrected using ITK-SNAP 3.2 (www.itksnap.org) (Yushkevich et al., 2006). DTI reconstruction was then performed on all diffusion datasets using DSI Studio (February 8th 2015 build, http://dsi-studio.labsolver.org) before being normalized in three steps. The datasets were first normalized using SPM8's normalize function (Statistical Parametric Mapping, Wellcome Department of Imaging Neuroscience, University of London, UK, http://www.fil.ion.ucl.ac.uk/spm/), which operates as a toolbox for MATLAB (Mathworks, Natick, Massachusetts, USA), using FA maps as source images. The normalized b0-maps were then used as source images for a second step of normalization. The skull-stripped masks were used as weighing images in both cases. Datasets then followed a third step of normalization using SPM8’s DARTEL and performed using normalized FA and mean diffusivity maps obtained from the second normalization step. This normalization procedure was performed to enable the two first steps to give the best possible fit of both white matter and gray matter structures for DARTEL to converge to a solution. Datasets were carefully examined to identify any misalignment between datasets over the caudate putamen (CPu), substantia nigra and sensorimotor cortex as well as the internal and external capsules.

Region of interest (ROI)-wise and SPM analyses of DTI results

Substantia nigra of both the intact and 6-OHDA-lesioned hemispheres were delineated on normalized datasets obtained using ITK-SNAP 3.2. Mean and standard deviation over the ROI were extracted for each individual subject for FA values. Two-sample t-tests were performed on the obtained mean values. A two-sample t-test was also performed on normalized FA, axial, and radial diffusivity maps using the second-level analysis tool of SPM8 to extract significant changes in DTI metrics in a voxel-wise fashion.

Creation of a template for q-space diffeomorphic reconstruction (QSDR)

A Generalized Q-ball Imaging (GQI) reconstruction was performed on diffusion datasets using DSI Studio (Yeh et al., 2010). The resulting maps were normalized as described above. An average Quantitative Anisotropy map was calculated from the 8 datasets generated in this study as well as 13 additional datasets from another unpublished study, in which the animals underwent the same scanning and analysis procedures. The image was scaled to indicate a voxel size of 1 mm, rotated and cropped to match the matrix size and orientation of the human template (x: left-right, y: anterior-posterior, z: inferior-superior) provided with DSI Studio.

QSDR and connectrometry analysis

The datasets from each animal were scaled to indicate a resolution of 1 mm instead of 150 μm. The diffusion data were then reconstructed with QSDR (Yeh and Tseng, 2011) using DSI Studio (May 5th 2016 build) with a diffusion sampling length ratio of 1.25. The output resolution for the reconstructed datasets was 1 mm.

dw-MRI connectometry (Yeh et al., 2016) was conducted to compare the isotropic value of the diffusion orientation density function (iso), a measure of isotropic diffusion, between the two groups. The group difference was quantified using a percentage measurement (i.e. 2 × (d1 − d2) / (d1 + d2) × %), where d1 and d2 are the group averages of iso. A threshold of 40% difference was used to select fiber directions with substantial difference in iso. A deterministic fiber tracking algorithm (Yeh et al., 2013) was conducted to connect these fiber directions in whole brain regions. Tracks with greater length than 35 mm were collected (corresponding to tracks longer than 5.25 mm before dataset rescaling). The seeding density was 20 seeds per mm3 (meaning 20 seeds per voxel) over the whole brain. To estimate the false discovery rate, a total of 1000 randomized permutations were applied to the group label to obtain the null distribution of the track length.

Functional MRI experiments

Resting-state fMRI acquisition

MRI measurements were performed in the aforementioned MRI scanner. Animals were imaged using a bird cage 72 mm coil (Rapid biomedical GmbH, Würzburg-Rimpar, Germany) as transmitter and a “rat head” phased array surface coil (Rapid biomedical GmbH) as receiver in a gradient system with 120 mm inner diameter and a maximum strength of 600 mT/m. Rats were induced to sleep with 3% isoflurane in an oxygen:air mixture of 3:7 and placed prone in a MR-compatible stereotactic holder with tooth and ear bars to fixate the head. A two meter long catheter was placed subcutaneously on the back of the animal and connected to an infusion pump (PHD 2000 Syringe, Harvard Apparatus, Inc., Massachusetts) for subsequent infusion of medetomidine. The core body temperature was kept at 37 °C using a warm air flow and monitored via a rectal temperature sensor. The temperature, respiration rate and pulse were followed using a monitoring device (SA Instruments, Stony Brook, NY, USA). Animals were then transferred into the scanner and isoflurane was reduced to 1.5% of the gas mixture.

Upon a stabilized physiological state, a bolus of 0.05 mg/kg medetomidine (s.c.; Domitor, Pfizer), an α2 adrenergic agonist with sedative and analgesic properties, was injected using the pump and the catheter. In the experimental cohort 1, isoflurane was interrupted when pulse and respiration rate dropped, due to medetomidine (about 5 min after bolus). A continuous infusion of medetomidine (0.1 mg/kg/h) was started 15 min after the bolus injection. In experimental cohort 2, isoflurane was reduced to 0.25% of the gas mixture instead of being interrupted, allowing the physiological state of the animal to be more stable (Lu et al., 2012). Scout images were acquired using a 3D gradient echo sequence for slice planning. A single-shot gradient echo-planar imaging sequence was used for functional imaging, acquired coronally with a readout direction along the rostral-caudal axis. All parameters were the same for the two experiments but the echo time was 9.32 ms in experimental cohort 1 and 15 ms in experimental cohort 2. The other parameters were as follows: 18 slices, FOV of 32.5 × 32.5 mm2, matrix size 50 × 50 voxels, slice thickness 0.65 mm, repetition time 1000 ms and 300 repetitions, triple reference and 8 dummy scans for a total acquisition time of 5 min and 12 s. Thus, datasets obtained had an isotropic resolution of 0.65 mm. The functional images were acquired about one and a half hour after the medetomidine bolus as recommended by Lu et al. (2012). In experimental cohort 2, following the first functional scan, animals were taken out of the scanner, intraperitoneally injected with L-DOPA (10 mg/kg)/benserazide (7.5 mg/kg) and reinserted in the scanner. A second functional scan was then acquired 30 min after the L-DOPA injection. Subsequent to the imaging procedure, rats received an injection of 0.1 mg/kg atipamizole (s.c.; Antisedan, Pfizer), an α2 adrenergic antagonist, to reverse the effects of medetomidine.

Pre-processing of fMRI data

All fMRI datasets were preprocessed using SPM8. Time series of images were realigned, normalized and smoothed as detailed below. The realignment step adjusted the images so that the brain is in the same position in every image, thus correcting virtually any motion. A mean image was calculated in this step, which was skull-stripped using ITK SNAP v.2.4 (http://www.itksnap.org). To account for inter-subject differences in size, shape and orientation, individual brains were spatially transformed into a standardized template which consisted of a mean image from an arbitrarily chosen control subject from experimental cohort 1. In this step, realigned images were normalized using mean images obtained during the realignment step as the source image and a mask generated from the skull-stripping as the source weighing images. Data were smoothed using a Gaussian filter with a full-width-at-half-maximum of 1 mm. BOLD signal has a low frequency of 0.01 to 0.1 Hz, therefore the data were band-pass filtered to eliminate lower and higher frequency hence suppressing potential physiological interferences and system drift using the time filter tool from the REST toolbox (http://restfmri.net) (Song et al., 2011).

Data processing

Voxel-wise correlation analysis was performed on time-filtered datasets using the REST toolbox (Song et al., 2011). Several ROI were defined as seeds in the CPu, the motor and the somatosensory cortices of both hemispheres. The correlation between the signal from each of these seeds and the signal from all voxels within the brain was then calculated resulting in maps of correlation coefficients. To fulfill the normality requirement of the t-tests, Fisher transformation was performed on the correlation value maps obtained, allowing statistical analysis to be performed using SPM.

Statistical analysis

One sample two-tailed t-tests were used to characterize the functional networks relative to each of the seeds which were consistent within the groups, while two-sample t-tests were used to identify functional networks differences between the two groups of animals, 6-OHDA versus control or 6-OHDA versus sham for experiments 1 and 2, respectively (cf. Table 1). A paired t-test was used to identify functional changes in the data after L-DOPA injections compared to baseline. All these tests were applied on Fisher transformed correlation coefficient maps, as these datasets are normally distributed, using SPM8's second level analysis tool.

Table 1

Statistical tests performed on the datasets obtained with resting-state functional MRI.

Experiment 1
One sample t-tests	Group tested		p-Value threshold
	6-OHDA		0,01
	Control		0,01
Two sample t-tests	Group 1	Group 2	p-Value threshold
	6-OHDA	Control	0,05
Experiment 2
One sample t-tests	Group tested		p-Value threshold
	6-OHDA drug naive		0,01
	6-OHDA acute treated		0,01
	Sham drug naive		0,01
Two sample t-tests	Group 1	Group 2	p-Value threshold
	6-OHDA drug naive	Sham drug naive	0,05
Paired t-tests	Group 1	Group 2	p-Value threshold
	6-OHDA acute treated	6-OHDA drug naive	0,05

Autoradiography of the dopamine transporter (DAT) to assess dopamine denervation

After all the functional MRI scans were performed, animals were killed using CO2 and then decapitated, their brain dissected out and fresh frozen in isopentane chilled in dry ice. The brains were then sectioned (14 μm) with a cryostat. Slide mounted sections were preincubated in 50 mM Tris-HCl/120 mM NaCl (pH 7.5) for 20 min. Incubation in binding buffer (50 mM Tris–HCl/120 mM NaCl, pH 7.5/1 μM fluoxetine) was conducted with 50 pM [125I] RTI-55 (Perkin-Elmer Life Sciences, Boston, USA) for 60 min. For nonspecific binding, 100 μM nomifensine (Sigma Aldrich) was added to the assay. The slides were washed 2 × 10 s in ice-cold binding buffer, rapidly dipped in deionized water, dried, and exposed to autoradiographic films (BioMax MR, Merck Eurolab, Sweden) at − 20 °C for 7 days. The films were then developed (using Kodak D19 and Kodak Unifix; Kodak, Rochester, New York, USA). Autoradiograms were digitized using a Dia-Scanner (Epson Perfection 4870 PHOTO; Seiko Epson Corporation, Suwa, Japan), and optical density values were measured using Scion Image for Windows (version alfa 4.0.3.2; Scion Corporation).

Detection of c-fos mRNA by in situ hybridization

In a separate experiment, a cohort of animals were 6-OHDA- (n = 5) or sham-lesioned (n = 5). Four additional 6-OHDA-lesioned animals were also treated with L-DOPA (10 mg/kg, i.p.)/benserazide (7.5 mg/kg, i.p.) 30 min before being killed. Their brains were rapidly dissected out, snap-frozen with isopentane chilled in dry ice and cut into 14 μm sections with a cryostat. A 35S-labelled anti-sense cRNA probe against c-fos was prepared by in vitro transcription from a cDNA clone corresponding to fragments of c-fos (Curran et al., 1987). The transcription was performed from 100 ng of linearized plasmid using 35S-UTP (1000 Ci/mmol) and SP6 RNA polymerase. In situ hybridization was performed as previously described (Zhang et al., 2007). Briefly, cryostat sections were postfixed in 4% PFA for 5 min at room temperature, rinsed twice in 4 × sodium chloride–sodium citrate buffer (SSC) and placed into 0.25% acetic anhydride in 0.1 M triethanolamine/4 × SSC (pH 8) for 10 min at room temperature. After dehydration in graded alcohols, the sections were hybridized overnight at 55 °C with 35S-labelled c-fos probe in 50 μl of hybridization solution (20 mM Tris–HCl/1 mM EDTA/300 mM NaCl/50% formamide/10% dextran sulphate/1 × Denhardt's/250 μg/ml yeast tRNA/100 μg/ml salmon sperm DNA/0.1% SDS/0.1% sodium thiosulphate). The slides were washed in 4 × SSC (5 min, four times), RNAse A (20 μg/ml) (20 min, at 37 °C), 2 × SSC (5 min, twice), 1 × SSC(5 min), 0.5 × SSC (5 min) at room temperature, and rinsed in 0.1 × SSC at 65 °C (30 min, twice) (all washes contained 1 mM DTT), before being dehydrated in graded alcohols. The slides were then exposed on X-ray films for 4 to 28 days and analyzed using the NIH Image 1.63 software (https://rsb.info.nih.gov/nih-image/).

Results

Efficacy of the unilateral 6-OHDA lesioning of the MFB

The efficacy of the unilateral 6-OHDA lesioning was evaluated in vivo by apomorphine-induced rotations. Rats exhibiting > 100 contralateral rotations were included in further studies. In addition, ex-vivo analysis of the efficacy of the unilateral 6-OHDA lesioning was confirmed by [125I] RTI-55 binding to DAT (Fig. 1). The 6-OHDA lesioning (n = 8) caused a near-complete and highly significant (96.8 ± 1.8%, Two-Sample t-test, p < 0.0001) abolishment of DAT at the level of the CPu ipsilateral to the lesion in experimental cohort 1. Likewise, in experimental cohort 2, 6-OHDA lesioning (n = 8) caused a highly significant reduction (90.8 ± 2.4%, p < 0.0001, Bonferroni post-hoc test) of DAT in the lesioned hemisphere, while the sham injection (n = 5) caused a much smaller, but still significant, reduction (11.6 ± 3.6%, p < 0.05, Bonferroni post-hoc test) of DAT.

Fig. 1

6-OHDA injection in the right medial forebrain bundle caused a near-complete dopaminergic denervation of the corresponding caudate putamen. Autoradiograms of DAT as detected by [125I]RTI-55 at the level of the caudate putamen (a and b) and quantified relative to the intact hemisphere for experiments 1 (a) and relative to the intact hemisphere of sham-lesioned animals for experimental cohort 2 (b). DAT, Dopamine Transporter. Number of animals per group: a) 6-OHDA lesioned rats (n = 8); b) 6-OHDA lesioned rats (n = 8) and sham-lesioned rats (n = 5). *p < 0.05; ***p < 0.0001. Intact, intact hemisphere; Lesion, lesioned hemisphere.

Ex vivo dw-MRI

An ex vivo dw-MRI experiment was conducted to identify changes in diffusion properties due to the loss of dopaminergic neurons.

ROI-wise analysis and SPM analysis

As a first approach, a ROI-wise analysis was conducted on the metrics obtained using DTI. A non-significant decrease in FA of 8.1 ± 5.3% was observed in the substantia nigra in response to 6-OHDA lesioning. We therefore decided to directly compare FA, axial and radial diffusivity maps of the different individuals after spatial normalization using voxel-based statistical parametric mapping (Fig. 2). Using this approach, a significant increase (p < 0.05) in radial diffusivity in the substantia nigra pars compacta was found in the 6-OHDA-lesioned hemisphere (Fig. 2a). However, a significant decrease (p < 0.05) in FA was also found in the lateral part of the substantia nigra, particularly pars reticulata, in the 6-OHDA-lesioned hemisphere (Fig. 2b).

Fig. 2

Injection of 6-OHDA in the right medial forebrain bundle increases radial diffusivity (RD), but decreases fractional anisotropy (FA), in the substantia nigra. Clusters with a significant increase of RD (a) respectively decrease of FA (b) in 6-OHDA lesioned rats (n = 4) compared to sham-lesioned rats (n = 4) as revealed by statistical parametric mapping. p < 0.05 by one-tailed two-sample t-test and clusters larger than 20 voxels are considered significant. FA, fractional anisotropy; RD, radial diffusivity; SNr, substantia nigra pars reticulata; SNc, substantia nigra pars compacta; 6-OHDA, 6-hydroxydopamine.

Connectometry analysis

As a complementary approach, changes in iso over white matter tracts were evaluated using a connectometry analysis based on a deterministic fiber tracking algorithm (Fig. 3). It revealed that iso increased by > 40% along the lesioned MFB in 6-OHDA-lesioned animals compared to controls. A permutation test with 1000 permutations estimated that the false discovery rate for these results was 0.41%, i.e. only 0.41% of the tracts with a significant increase in iso may be false positives.

Fig. 3

Connectometry and QSDR analyses reveal tracts with increased isotropic value of the diffusion orientation density function (iso) following a 6-OHDA injection in the MFB of the right hemisphere. Area where the tracts with > 40% increased Iso are passing, overlaid on a rat quantitative anisotropy template (a). Tracts with > 40% increased iso as measured by QSDR corresponds to tracts along the right MFB between CPu and substantia nigra (b and c). Tracts with 6-OHDA (n = 4) iso superior to 140% of Sham (n = 4) iso and longer than 5,25 mm. False discovery rate (FDR) evaluated using a permutation test with 1000 permutations and equal to 0.41%. MFB, medial forebrain bundle; 6-OHDA, 6-hydroxydopamine; iso, isotropic diffusivity; QSDR, Q-space diffeomorphic reconstruction, CPu, Caudate Putamen; SN, substantia Nigra.

Resting-state fMRI

Experimental cohort 1

To characterize functional consequences of the 6-OHDA lesioning, 6-OHDA-lesioned (n = 8) and control rats (n = 8) were scanned using a BOLD signal sensitive MRI sequence. The datasets were realigned spatially to a common template and were analyzed using a correlation analysis with seeds placed in CPu and secondary somatosensory cortex of both hemispheres. One-sample t-tests were used to characterize the functional networks of each group relative to the seeds (Fig. 4a. and Fig. 4b) (p < 0.01) and a two-sample t-test was used to compare the functional networks of the two groups and reveal statistically significant differences (p < 0.05) (Fig. 4c).

Fig. 4

Experimental cohort 1: Resting-state functional connectivity (FC) decreases between the two caudate putamina and increases between the caudate putamen of the lesioned hemisphere and sensorimotor cortices of both hemispheres. FC network of control rats (n = 8) (a) and 6-OHDA lesioned rats (n = 8) (b) as measured using a seed-based correlation analysis with seeds located in four different locations: i, the caudate putamen in the intact hemisphere; ii, the caudate putamen in the lesioned hemisphere; iii, the secondary somatosensory cortex of the intact hemisphere; and iv, the secondary somatosensory cortex of the lesioned hemisphere. Significant differences between the FC of 6-OHDA rats and the FC of control rats (c) for the same seeds' locations as a and b. Significant t-value maps are overlaid on a representative structural MRI image. z corresponds to coordinates of the slices relative to bregma (in mm). Red colors represent significant positive t-values and blue colors represent significant negative t-values. Yellow dots indicate the location of the seeds. a) and b): p < 0.01 by two-tailed One-Sample t-test, c): p < 0.05 by two-tailed two-sample t-test on Fisher correlation values.

Seeds placed in the CPu of either hemisphere showed that the CPu of control animals are functionally connected between hemispheres. Similarly, seed-based correlations of secondary somatosensory cortices showed that both hemispheres are highly interconnected. The spatial distribution of functional connectivity (FC) to either CPu appeared similar in 6-OHDA-lesioned and control rats (Fig. 4a and Fig. 4b). However, the bilateral FC between the intact and lesioned CPu was lower in 6-OHDA lesioned animals (Fig. 4b). This observation was confirmed by the two-sample t-test made for the seeds placed in the lesioned CPu which showed a decrease of FC between the two CPu. Furthermore, when the seed was placed in the lesioned CPu an increase of FC from the seed to the motor cortex and secondary somatosensory cortex of both hemispheres was revealed. In contrast, no similar change in FC was found when the seed was placed in the intact CPu. When seeds were placed in either of both secondary somatosensory cortices, an increase of FC was observed in the lesioned, but not intact, CPu. In contrast, increases of FC were found from either secondary somatosensory cortex to motor cortices of both hemispheres (Fig. 4c).

Experimental cohort 2

We reasoned that having two independent functional MRI experiments showing similar results is a robust way to distinguish true changes from false positives. It is, indeed, highly unlikely to obtain false positives in the same area in two independent experiments. A second experiment was therefore performed for replication purposes and to assess a sham-lesioned group instead of naïve controls. This experiment also investigated the acute effect of systemic L-DOPA on FC.

6-OHDA versus Sham - replication of experimental cohort 1

A cohort of sham-(n = 5) and 6-OHDAlesioned (n = 8) animals presented very similar FC networks as the control and 6-OHDA-lesioned animals of experimental cohort 1 (Fig. 5a compared to Fig. 4). In agreement with experimental cohort 1, 6-OHDA-lesioned rats showed a loss of FC between the two CPu when compared to sham animals. Furthermore, and in accordance with experimental cohort 1, there was an increase of FC between the lesioned, but not intact, CPu and motor cortices and secondary somatosensory cortices of both hemispheres (Fig. 5b).

Fig. 5

Experimental cohort 2: Resting-state functional connectivity (FC) increase between the caudate putamen in the lesioned hemisphere and sensorimotor cortices of both hemispheres is counteracted by a single injection of L-DOPA. FC network of sham-lesioned rats (n = 5) (a) as measured using a seed-based correlation analysis with seeds located in the four same locations as in Fig. 4: i, the caudate putamen in the intact hemisphere; ii, the caudate putamen in the lesioned hemisphere; iii, the secondary somatosensory cortex of the intact hemisphere; and iv, the secondary somatosensory cortex of the lesioned hemisphere. Significant differences between the FC of 6-OHDA rats (n = 8) and the FC of sham-lesioned rats for the same seeds' locations (b). Significant differences between 6-OHDA rats treated with a single injection of L-DOPA/benserazide i.p. 30 min prior to scan and the same rats before the injection as calculated using a two-tailed paired t-test (p < 0.05) (c). Significant t-value maps are overlaid on a representative structural MRI image. z corresponds to coordinates of the slices relative to bregma (in mm). Red colors represent significant positive t-values and blue colors represent significant negative t-values. a): p < 0.01 by two-tailed One-Sample t-test, b): p < 0.05 by two-tailed two-sample t-test on Fisher correlation values.

Acute effects of L-DOPA on FC

The acute functional effect of L-DOPA in 6-OHDA lesioned rats was mapped by comparing FC from rs-fMRI datasets obtained 30 min after L-DOPA with baseline in the same animals (Fig. 5c). A significant decrease in FC was observed between the lesioned CPu and the motor cortices. When seeds were placed in secondary somatosensory cortices, decreases of FC towards the lesioned, but not intact, CPu was observed. However, unexpectedly, regardless whether the seed was placed in the intact or lesioned CPu, no significant change in FC was observed between the two CPu (Fig. 5c).

c-fos mRNA expression in caudate-putamina and cortices

Four weeks post-lesioning, a cohort of 6-OHDA rats was treated with a single injection of L-DOPA and killed after 30 min. L-DOPA increased c-fos expression in the motor and secondary somatosensory cortices of both hemispheres (Fig. 6). L-DOPA significantly increased the expression of c-fos in the 6-OHDA-lesioned, but not intact, CPu (Fig. 6). Thus, in agreement with the FC data, L-DOPA had bilateral effects on c-fos in somatosensory cortices, but differential effects on this immediate gene in caudate-putamina.

Fig. 6

L-DOPA increases c-fos mRNA in motor and secondary somatosensory cortices of both hemispheres but only in the lesioned caudate putamen (CPu) in 6-OHDA rats. c-fos mRNA in sham-lesioned (n = 5), 6-OHDA lesioned (n = 5) and 6-OHDA lesioned and acutely L-DOPA treated (n = 4) rats in the motor cortex, the caudate putamen and the secondary somatosensory cortex. Boxes indicate the location of measurements. *p < 0.05 between the lesioned side of 6-OHDA + L-DOPA and all other groups by two-sample t-test. #p < 0.05 between 6-OHDA + L-DOPA intact side and 6-OHDA intact side by two-sample t-test. Intact, intact hemisphere; Lesion; lesioned hemisphere; CPu, Caudate putamen; S2, secondary somatosensory cortex. 6-OHDA + L-DOPA, 6-OHDA lesioned rats acutely treated with L-DOPA.

Discussion

Here, we provide novel evidence that MRI studies in rodents pose a viable alternative to clarify the complex roles of dopamine denervation and acute L-DOPA actions on structural and functional network connectivity relevant to PD.

To study structural connectivity, we performed ex vivo dw-MRI. This experiment showed that the loss of dopaminergic neurons in the substantia nigra was accompanied by a decrease of FA and an increase of radial diffusivity in the substantia nigra. This is in agreement with several aforementioned clinical studies (e.g. Vaillancourt et al., 2009, Rolheiser et al., 2011, Skorpil et al., 2012, Cochrane and Ebmeier, 2013, Schwarz et al., 2013) and the results in a MPTP (1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine mouse model (Boska et al., 2007). MPTP is a precursor of MPP +, which interferes with complex I of the electron transport chain of the mitochondrial metabolism. This results in degeneration of dopaminergic neurons of the substantia nigra. Interestingly, more recently a decrease of FA and an increase of radial diffusivity in the substantia nigra have also been found in MitoPark mice, a genetic mouse model of parkinsonism (Cong et al., 2016). However, our data disagree with a study by Van Camp et al. (2009), which reported an increase of FA in the substantia nigra in a rat model where 6-OHDA was injected into the CPu. It is worth noting that the statistical changes occur predominantly in the most lateral part of the substantia nigra, in line with the degree of dopaminergic loss observed in both PD patients and animal models of experimental parkinsonism (Langley et al., 2016, Damier et al., 1999, Fearnley and Lees, 1991). The decrease of FA in the substantia nigra pars reticulata found here could be related to a drastic loss of the dentritic tree of dopaminergic cells innervating this area extending ventrally from the substantia nigra pars compacta. Witkovsky et al. (2009) describe dopaminergic dendrites in the substantia nigra pars reticulata as “prominent” and “running in bundles”. It is, thus, likely that the diffusion of the substantia nigra is influenced by these dendrites. The connectometry analysis also showed an increase in isotropic diffusion along the MFB in line with the loss of dopaminergic axons, which could be interpreted as an elimination of constraints for the diffusion of water molecules along the medial forebrain bundle.

To study functional connectivity, we performed rs-fMRI in two cohorts of rats. Despite the fact that there are no structural connections between caudate-putamina, we found a high FC between them. This result agrees with previous reports and depends upon interactions with afferent brain regions to CPu (Sierakowiak et al., 2015, Di Martino et al., 2008). CPu receives afferents from many brain regions including cortex (in particular from sensorimotor cortex), thalamus, substantia nigra, amygdala and pallidum (Gerfen and Surmeier, 2011). In unilaterally 6-OHDA-lesioned rats, we found a lower FC between the two CPu, suggesting impairments in circuitries connecting them. At the same time, we also observed an increase of FC between the lesioned CPu and sensorimotor cortices of both hemispheres. In contrast, there is no alteration of FC between the intact CPu and sensorimotor cortex. The two latter findings could underlie the reduced interstriatal FC observed in response to 6-OHDA lesioning. A possibility is that endogenous dopamine efficiently gates the altered cortical input in the intact hemisphere. However, in the lesioned hemisphere, the increase of FC between the CPu and the sensorimotor cortex could, at least partly, be due to dysfunctions of multiple tonic inhibitory gate actions of dopamine D2 receptors (Gerfen and Surmeier, 2011). The lack of inhibition would lead to increased synchrony and FC between corticostriatal neurons and medium size spiny neurons of the CPu. Using microdialysis, several studies (e.g. Maeda et al., 1999, Meissner et al., 2006) have shown that a systemic L-DOPA injection increased striatal dopamine concentration to normal physiological level in the lesioned hemisphere but only marginally increase its levels in the intact hemisphere. Given these data, we expected to observe a normalization of FC between the two CPu by L-DOPA. However, we found that a single L-DOPA injection did not normalize the interstriatal FC. Meanwhile, L-DOPA counteracted the increase in FC between the lesioned CPu and the motor and somatosensory cortices of both hemispheres, suggesting that normalization of corticostriatal transmission is a primary action of L-DOPA. This could be due to activation of presynaptic D2 receptors on corticostriatal neurons and inhibition of corticostriatal neurotransmission (Bamford et al., 2004). The action of L-DOPA may also result from stimulation of D2 receptors of medium spiny neurons of the indirect pathway, inhibiting their activity, but also stimulating the release of endocannabinoids which act retrogradely on CB1 receptors located on glutamatergic corticostriatal terminals to induce long-term depression (LTD) (Mathur and Lovinger, 2012). Such LTD would reduce the cortical input to the striatum. However, the role of dopamine neurotransmission on interstriatal synchrony remains to be fully elucidated.

Several previous electrophysiological studies are consistent with our FC data. A low FC between the intact CPu and secondary somatosensory or motor cortex agrees with the observation that the neuronal activities of the motor cortex and the CPu are not synchronized under healthy conditions (e.g. Bar-Gad et al., 2003). Using electrophysiology and the 6-OHDA lesioned model, Dejean et al. (2012) found an increase in synchrony of neuronal activity between CPu and the motor cortex. Moreover, an increased beta wave activity, as simultaneously recorded by electrocorticogram of the sensorimotor cortex and electrophysiology in the subthalamic nucleus, was reported in the 6-OHDA lesioned rat model (Mallet et al., 2008). Our present findings also agree with previous electrophysiological work showing a decrease in synchrony of neuronal activity in the basal ganglia and the cortex by L-DOPA in PD patients (Priori et al., 2004) and MPTP-lesioned monkeys (Heimer et al., 2002). In this study, we also report that L-DOPA induced c-fos bilaterally in the sensorimotor cortices, but only in the lesioned CPu. These c-fos inductions occur in similar regions as the FC alterations, indicating a possible correlation between these two phenomena. To conclusively explain the molecular mechanism(s) underlying the observed changes in FC and c-fos are beyond the scope of this study. FC is a measure of the synchrony of the neuronal activity between distant regions, it does not measure the activity per se within these regions. In theory, two regions can exhibit low activity, but have a high synchrony or exhibit a high activity and not be synchronized. We can, therefore, observe, at the same time, a decrease of FC due to L-DOPA and an increase of biochemical activity as measured by c-fos. As mentioned above, possible mechanism whereby L-DOPA reduces FC in corticostriatal loops are activation of presynaptic D2 receptors on corticostriatal neurons and inhibition of corticostriatal neurotransmission (Bamford et al., 2004) as well as stimulation of postsynaptic D2 receptors on medium-sized spiny neurons resulting in release of endocannabinoids acting retrogradely on CB1 receptors on corticostriatal and dampening corticostriatal neurotransmission (Mathur and Lovinger, 2012). The stimulation of cortical c-fos by L-DOPA may involve activation of striatal D1 and D2 receptors resulting in stimulation of striatothalamocortical loops (Gerfen and Surmeier, 2011). The fact that both FC and c-fos are altered bilaterally in cortex may relate to prominent cortico-cortico neuronal connections. The unilateral stimulation of c-fos in striatum by L-DOPA most likely relates to activation of supersentitized D1 receptors in the 6-OHDA lesioned hemisphere (Gerfen and Surmeier, 2011).

In conclusion, we show that the near complete loss of dopaminergic neurons in the substantia nigra increased radial and isotropic diffusivity in the nigrostriatal pathway. Our data with near complete dopaminergic asymmetry only partly mirror the clinical situation with a modest asymmetry in dopamine devervation. A cautionary approach needs therefore to be taken when translating our animal data to the clinical setting. Nonetheless, these data provide further evidence that dw-MRI studies properly measure the integrity of nigrostriatal dopamine neurons. We also found increases of FC between the dopamine-depleted CPu and the sensorimotor cortices of both hemispheres and that this increase can be reversed by L-DOPA. Increased FC between the putamen and the sensorimotor cortex of PD patients has been reported in some (Yu et al., 2013, Kwak et al., 2010, Agosta et al., 2014), but not all (Palmer et al., 2010, Luo et al., 2014) studies. Our data suggest that additional well-controlled studies should be done to firmly establish whether alterations of FC between CPu and sensorimotor cortices is occuring in, at least some, PD patients. Our findings provide novel evidence that dw-MRI and rs-fMRI might be a viable approach to study oscillatory signaling and network connectivity in experimental parkinsonism. Future studies could be extended to animal models presenting a milder and more progressive loss of nigral dopaminergic neurons.