1. Introduction
The classic motor symptoms of Parkinson disease (PD) are tremor, rigidity and bradykinesia. They result from the degeneration of dopaminergic neurons in the
substantia nigra (SN) and the resulting dopamine deficiency in the striatum. Accordingly, the cardinal PD motor symptoms are alleviated by the dopamine precursor levodopa and by dopamine receptor agonists [
1,
2]. Motor symptoms remain responsive to these dopaminergic medications throughout the course of PD. In advanced PD, however, symptom control is hampered by the fact that a single dose of medication triggers dyskinesia more easily and lasts for a shorter period of time. Presynaptic mechanisms can explain some features of these motor fluctuations but not their induction by dopamine receptor agonists [
3,
4].
Postsynaptic mechanisms include the fact that neurons generally do not remain passive in the face of major changes in synaptic inputs. Homeostatic plasticity allows neuronal networks to maintain a stable rate of activity when synapses change [
5]. This has been studied mainly during development and learning [
5,
6], but also in the course of neurological and psychiatric diseases [
7,
8,
9,
10]. In humans and animal models, chronic dopaminergic depletion and substitution can induce plastic changes that differ from acute effects and that are potentially mediated by homeostatic plasticity mechanisms. These effects include the “long duration response” (LDR) to dopaminergic substitution, levodopa-induced dyskinesias (LID) and tardive dyskinesias [
11,
12].
The consequences of chronic dopaminergic denervation have been primarily assessed in striatal medium spiny neurons (MSN, also called spiny projection neurons) because they are the main postsynaptic neurons for dopaminergic axon terminals. Effects differ between MSN expressing excitatory D1 dopamine receptors (D1-MSN) and MSN expressing inhibitory D2 dopamine receptors (D2-MSN). For instance, D2-MSN are disinhibited by dopamine depletion. The homeostatic response in the face of chronic dopamine depletion includes a decrease in electrical excitability and a reduced number of glutamatergic synapses [
8]. A lower density of dendritic spines and a pruned dendritic tree was also observed in further animal models [
9,
13,
14,
15,
16,
17,
18] and in striatal MSN of PD patients [
19,
20].
In order to guide dopaminergic substitution in PD patients, it is important to know whether these changes in MSN morphology can be reversible, and what governs reversibility. In previous work, reversibility was tested by administration of levodopa in animals with degeneration of dopaminergic neurons induced by injection of 6-hydroxydopamine (6-OHDA) or in aphakia (Pitx3-/-) mice. The reduction in spine density was found to be reversible in D2-MSN but not in in D1-MSN [
21,
22,
23]. Dendritic pruning, in contrast, was not reversible [
8]. This indicates that reversibility requirements may differ between D1 and D2 MSN and between spine density and TDL. It furthermore raises the question of whether a different regimen of dopaminergic substitution would allow a reversibility of dendritic pruning.
In order to investigate reversibility of MSN morphology under optimal conditions, we used the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) model of PD. In this model, dopaminergic neurons in the substantia nigra and their axon terminals in the striatum degenerate after MPTP administration. Nigrostriatal axon terminals in the striatum recover 3 months after MPTP due to axonal sprouting [
24], which does not happen in the 6-OHDA model. Because dopamine is secreted from the recovering nigrostriatal axon terminals, there is no need to optimize the dosing and timing of levodopa administration to revert morphological changes in the postsynaptic MSN.
In this study we thus confirmed that MPTP-induced dopamine depletion in the striatum alters MSN morphology by using Golgi staining at the peak of dopaminergic denervation—21 days after MPTP administration. We then determined which of these changes are reversible at 90 days after MPTP when dopaminergic axon terminals have regenerated.
2. Methods
2.1. MPTP Mice
C57BL/6J mice (RRID: IMSR_JAX:000664) were housed and handled in a pathogen-free animal facility at 20–24 °C with a 12 h light/dark cycle, food and water ad libitum, in accordance with guidelines of the Federation for European Laboratory Animal Science Associations (FELASA). Procedures were approved by the local authorities (Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen, 84-02.04.2014.A171). Mice were 10–12 weeks of age at the start of the experiment. They received either MPTP hydrochloride in 0.9% saline or saline alone. MPTP was administered at a dose of 30 mg free base per kg body weight i.p. at 24 h intervals over 5 consecutive days. MPTP handling and safety measures were in accordance with published guidelines [
25]. One group of animals was sacrificed 21 days after the last MPTP injection by cervical dislocation under deep isoflurane anesthesia. A second group of animals was sacrificed 90 days after the last MPTP injection. Mice were monitored daily for physical condition and weight loss. One animal died on day 5, all others that started the experiment survived with <20% weight loss and were included into the analyses.
After decapitation, the brains were rapidly removed and dissected on ice. The rostral part, including the striatum, was processed for Golgi staining (left hemisphere) and HPLC analysis of catecholamines (right hemisphere). The caudal part including the substantia nigra was processed for immunohistochemistry.
2.2. Golgi Staining and Analysis
For quantification of morphological changes, the left hemispheres were stained with a Golgi staining kit (FD Rapid Golgi Stain Kit PK-401, FD Neuro Technologies, Columbia, MD, USA) following the manufacturer’s instructions. In brief, tissue was impregnated for 2 weeks before being cut into 170 μm slices using a vibratome (5100mz-135, Campden Instruments, Loughborough, UK). Slices were washed twice with water for 4 min, incubated in the staining solution for 10 min and washed again. Slices were then dehydrated and coverslipped with Entellan (Merck Millipore).
Four slices of each brain located around Bregma level 0.02 mm [
26] were selected using the anterior commissure as a landmark and imaged using an inverted microscope (IX81S1F, Olympus, Hamburg, Germany). Type 1 MSN were identified by their spine-free, round or oval soma (12–18 μm) and at least 3–4 branches and their densely spined distal dendrites [
27]. Three ventral and three dorsal MSNs were selected per section. For each neuron, images with 20× and 60× objectives (oil immersion, NA 1.35) were acquired as z-stacks with 0.5 μm distance using the software xCellence v2.0 (Olympus, Hamburg, Germany).
Analyses were carried out using ImageJ (FiJi v2.0.0, National Institutes of Health (NIH), Bethesda, Maryland) after minimum intensity projection of the z-stacks. The total dendritic length was measured using the plugin NeuronJ [
28]. The number of branch points and the number of spines in a 10 µm stretch of dendrite were determined manually using the multipoint tool of ImageJ. From the total dendritic length (TDL) and the number of branch points (BP) we calculated the segment length (SL) as SL = TDL/BP.
2.3. HPLC Analysis of Catecholamines
In order to quantify the loss of dopaminergic axon-terminals in MPTP-treated mice, striatal catecholamine concentrations were measured by HPLC with electrochemical detection at 800 mV. Striatal tissue was homogenized with 50 μL of 0.1 M perchloric acid per mg of tissue using ceramic beads and a 30 s pulse of a speed mill (P12, Analytik Jena AG, Jena, German). Cell debris was removed by centrifugation (17,000× g, 20 min at 4 °C) and 20 μL of sample or standard was injected onto the reverse phase column (Prontosil 120-3-C18, Thermo Fisher, Waltham, MA, USA) and kept at 25 °C. The mobile phase consisted of 85 mM sodium acetate, 35 mM citric acid, 0.5 mM octane sulfonic acid, 0.15 mM EDTA, and 10% methanol (pH 4.3). Flow rate was 0.8 mL/min. Retention times were ≈5.5 min for DA, ≈4.7 min for DOPAC and ≈10.9 min for HVA. Peaks were identified manually using Chromeleon software (v6.80, Thermo Fisher, Waltham, MA, USA) and compared to external standards run every 3 samples.
2.4. Immunohistochemistry and Quantification
The posterior parts of the brain, including the SN, were fixed for 24 h in 4% paraformaldehyde (pH 7.4), cryoprotected in 30% sucrose for 2 days at 4 °C, frozen by immersion in isopentane (−45 °C) and then stored at −80 °C until serially cut into 30 µm coronal sections. Every third section spanning the SN was stained for tyrosine hydroxylase (TH). The free-floating brain sections were washed three times in Tris-buffered saline (TBS) with 0.1% TritonX (TBS-T). Endogenous peroxidase was blocked by incubation with 0.3% H2O2 in TBS-T for 30 min followed by three washing steps with TBS-T. The primary anti-TH antibody (rabbit polyclonal, Merck Millipore) was incubated overnight at 4 °C in a dilution of 1:1000 in TBS-T containing 3% normal goat serum (Vector Laboratories). The sections were washed again with TBS-T and incubated with the secondary antibody (biotinylated goat anti-rabbit IgG, Vector Laboratories) in a dilution of 1:200 in TBS-T for 30 min. Subsequently, the sections were washed in PBS, incubated with Avidin-Biotin Complex (Thermo Fisher) for 30 min, followed by another washing step. Visualization was performed via diaminobenzidine (DAB, Vectastain®ABC-Kit Standard PK-4000, Vector Laboratories, Burlingame, CA, USA) in a dilution of 1:20 in PBS for 10 min. Sections were mounted on microscope slides after dehydration sections and coverslipped with Entellan (Merck Millipore, Burlington, MA, USA).
TH-positive (dopaminergic) cells in the lateral SNc of the right hemisphere were stereologically counted using the optical fractionator method (StereoInvestigator v11, MicroBrightField, Williston, VM, USA) as described previously [
29]. In brief, neurons were manually identified in 50 × 50 µm counting frames presented by the software using an Axioskop 2 microscope (Carl Zeiss Vision) and an oil immersion 63× objective (NA 1.4). Grid size was 50 × 50 µm, and every third section was analyzed. Counts were performed blinded for treatment and timepoint.
2.5. Statistical Analysis and Data Visualization
Statistical analyses were performed using GraphPad Prism 5.0 (GraphPad Software, San Diego, CA, USA). The animal-based values—i.e., the number of TH-positive neurons in the substantia nigra, striatal dopamine concentration and striatal metabolite ratio, were normally distributed (Pearson Omnibus normality test). In the graphs they are represented as mean ± SEM and in addition as markers for each animal. Groups were compared by unpaired t-test or two-way ANOVA as indicated in the text.
The values of TDL, branch points, segment length and spine density for each MSN were not normally distributed in all groups. In the graphs they are therefore represented by box and whisker diagrams and groups were compared by the nonparametric Mann–Whitney test. The number of animals and MSN included in the analyses are listed in
Table 1.
The entire dataset was analyzed by fitting linear models using the function “lm” and comparing their performance using the function “anova” in RStudio 1.3 following a tutorial by Achim Zeileis (Wirtschaftsuniversität Wien). In order to determine whether a factor (e.g., treatment) had a significant effect (e.g., on segment length SL), we performed the trivial fit f1 < −lm(SL ~ 1, data = d) and the fit for treatment fT < −lm(SL ~ Treatment, data = d) and then compared the results using anova(f1, fT). In order to determine whether a second factor, e.g., the striatal dopamine concentration DA, contains significant additional information, we performed the fit for both factors fDT < −lm(SL ~ DA + Treatment, data = d) and then compared the results using anova(f1, fT, fDT). The different models and the
p values resulting from the anova analyses are listed in
Table 2.
In addition, we performed principal component analysis (PCA) using Rstudio 1.3 and the function PCA of the FactoMineR package. Factors were displayed using the function corrplot of the corrplot package.
4. Discussion
In this study we observed morphological alterations in striatal MSN after MPTP-induced degeneration of dopaminergic neurons that are consistent with findings in PD patients and other animal models. Following the recovery of dopaminergic axon terminals, these morphological changes were reversible. Reversibility in spine density has been observed in previous studies with exogenous administration of levodopa. Yet, this is, to our knowledge, the first study showing reversibility of dendritic arborization as reported by TDL and branch points.
To analyze MSN morphology, we used the classical Golgi staining that was also used in studies on PD patients [
19,
20,
30] and in several of the earlier studies on animal models [
9,
13]. The Golgi staining allows reconstruction of dendritic arbors of individual neurons and quantification of dendritic spines without the need to fill individual neurons by a fluorescent dye. This advantage allowed us to analyze a much larger number of MSN than in studies with individually filled MSN [
14,
21,
22,
23]. The major disadvantage of the Golgi method lies in the fact that it cannot be easily combined with immunohistochemistry. Consequently, we were not able to discriminate between D1-MSN and D2-MSN. In previous work by others, notable differences were observed between D1-MSN, D2-MSN, and D1/D2 double positive MSN. In D2-MSN, a reduction in spine density was observed already 2–3 weeks after dopamine depletion whereas in D1-MSN spine density was only reduced with more chronic dopamine depletion [
7,
14,
21,
22,
31,
32].
The density of striatal spines we observed (
Figure 3B) was similar as in previous studies in mice [
13,
14,
15,
21,
23]. The reduction of spine density in response to dopamine depletion we observed after MPTP (
Figure 3B) was also observed in mice with 6-OHDA-induced dopamine depletion [
14,
15,
21], in Pitx3-/-mice [
22], rats [
16,
17,
18], monkeys [
9] and PD patients [
20,
30]. The magnitude of the reduction was smaller in our data than in previous studies. This could be related to the smaller extent of dopamine depletion in the MPTP model as compared to the 6-OHDA models and PD patients.
When measured at 90 days after MPTP, spine density recovered almost completely (
Figure 5B). This is noteworthy given the moderate recovery of dopamine in this cohort (
Figure 1E). Yet, we assume that the recovery of nigrostriatal axon terminals was more pronounced than the recovery of dopamine. In previous experiments [
24,
29], nigrostriatal axon terminals returned almost to baseline levels. This is consistent with the normalized ratio of dopamine metabolites (
Figure 1F) and can explain the pronounced recovery of spine density (
Figure 5B). In the 6-OHDA model and in Pitx3-/-mice, reversibility of spine density is observed in D2-MSN but not in D1-MSN after 1–2 weeks of levodopa administration [
21,
22,
23]. It will therefore be interesting to determine in future studies whether reversibility of spine density is different between D1-MSN and D2-MSN at 90 days after MPTP. Of note, reversibility of spine density was previously observed in the MPTP model after a treadmill exercise [
13]. This finding is reminiscent of a finding from the ELLDOPA study on PD patients where the long-term beneficial effects of levodopa medication were more pronounced in the dominant hand, i.e., in the hand that receives more exercise [
33]. It indicates that exercise can to some extent mimic the effects of levodopa administration.
Our values for TDL and branch points were measured in reconstructions of minimum intensity projections obtained from coronal slices. They were in a similar range as that obtained from maximum intensity projections of MSN filled with lucifer yellow in coronal slices [
14]. Both values are 2–3 times shorter than TDL obtained with 3D analysis of two-photon images from parasagittal slices of a similar thickness [
21]. This difference indicates that we substantially underestimate the extent of the dendritic tree with 2D projections. The number of dendritic branch points we observed was similar to the 3D analysis [
21], confirming the validity of our analysis. Because we are primarily interested in relative changes, the systematic underestimation of TDL is not problematic. It nonetheless illustrates how the dendritic arbor is altered when neurons grow on coverslips or plates instead of the 3D environment of the brain. A much larger dendritic arbor could explain the quite substantial differences recently observed between matrigel-based 3D neuronal cultures as compared to traditional 2D cultures [
34,
35].
With MPTP-induced dopamine depletion we observed a reduction in TDL (
Figure 2B), consistent with previous findings in mice using MPTP [
13] or 6-OHDA [
14,
21], and with findings in monkeys [
9] and PD patients [
19,
20,
30]. In contrast to the recovery of TDL we observed in our MPTP model, TDL was not reversible after levodopa administration in the 6-OHDA model [
21]. There are several possible explanations that will be interesting to explore in future studies. First, dopamine depletion was more pronounced in the cited 6-OHDA models than in our MPTP model. This might affect the possibility for TDL recovery. Furthermore, levodopa dosing in the cited studies was chosen to induce dyskinesias and has not been optimized for TDL recovery. Finally, secretion of endogenous dopamine in our study might have different effects to exogenous administration of levodopa. Exogenous levodopa is at least partially converted in serotonergic terminals [
4] and glutamate is co-released from nigrostriatal terminals [
36,
37]; such co-transmitters could thus be important for the maintenance and recovery of MSN morphology. In addition, adaptation in cortico-striatal glutamatergic signaling, which is known to affect MSN spine density [
38], could underlie the changes in MSN spine density after dopamine depletion and recovery.
Taken together, our findings indicate that all consequences of dopamine depletion on the morphology of striatal MSN can be reversible. Of course, further research is required to mechanistically determine whether MSN changes after degeneration and recovery of dopaminergic axon terminals really result from the changes in dopaminergic neurotransmission or from other factors. Similarly, further work will be required to determine the molecular pathways that govern the adaptation of MSN arborizations and spine density. Additional interventions such as acute or chronic administration of dopamine agonists and antagonists in animals with different extents of dopaminergic denervation will be required. Still, this encouraging finding suggests that it should be possible to design regimens of dopamine substitution for PD patients that avoid structural changes in striatal MSN. Such regiments could potentially also avoid the development of dyskinesias. It is already established that pronounced depletion of striatal dopamine is necessary for levodopa-induced dyskinesias in rodents. If the smaller extent of dopamine depletion in the MPTP model is confirmed as the critical factor for reversibility, these findings indicate that dopamine substitution needs to be started early in the disease course to avoid long-term motor complications.