|Home | About | Journals | Submit | Contact Us | Français|
DYT1 dystonia is a severe form of inherited dystonia, characterized by involuntary twisting movements and abnormal postures. It is linked to a deletion in the dyt1 gene, resulting in a mutated form of the protein torsinA. The penetrance for dystonia is incomplete, but both clinically affected and non-manifesting carriers of the DYT1 mutation exhibit impaired motor learning and evidence of altered motor plasticity. Here, we characterized striatal glutamatergic synaptic plasticity in transgenic mice expressing either the normal human torsinA or its mutant form, in comparison to non-transgenic (NT) control mice. Medium spiny neurons recorded from both NT and normal human torsinA mice exhibited normal long-term depression (LTD), whereas in mutant human torsinA littermates LTD could not be elicited. In addition, although long-term potentiation (LTP) could be induced in all the mice, it was greater in magnitude in mutant human torsinA mice. Low-frequency stimulation (LFS) can revert potentiated synapses to resting levels, a phenomenon termed synaptic depotentiation. LFS induced synaptic depotentiation (SD) both in NT and normal human torsinA mice, but not in mutant human torsinA mice. Since anti-cholinergic drugs are an effective medical therapeutic option for the treatment of human dystonia, we reasoned that an excess in endogenous acetylcholine could underlie the synaptic plasticity impairment. Indeed, both LTD and SD were rescued in mutant human torsinA mice either by lowering endogenous acetylcholine levels or by antagonizing muscarinic M1 receptors. The presence of an enhanced acetylcholine tone was confirmed by the observation that acetylcholinesterase activity was significantly increased in the striatum of mutant human torsinA mice, as compared with both normal human torsinA and NT littermates. Moreover, we found similar alterations of synaptic plasticity in muscarinic M2/M4 receptor knockout mice, in which an increased striatal acetylcholine level has been documented. The loss of LTD and SD on one hand, and the increase in LTP on the other, demonstrate that a ‘loss of inhibition’ characterizes the impairment of synaptic plasticity in this model of DYT1 dystonia. More importantly, our results indicate that an unbalanced cholinergic transmission plays a pivotal role in these alterations, providing a clue to understand the ability of anticholinergic agents to restore motor deficits in dystonia.
Dystonia is a neurological syndrome of involuntary muscle contractions, causing twisting, repetitive movements and progressive abnormal postures (Fahn, 1988). DYT1 dystonia is the most common form of inherited early-onset generalized dystonia. It is caused by a 3 bp deletion that deletes a glutamic acid residue in the C-terminal coding region of the protein torsinA (Ozelius et al., 1997). The physiological function of torsinA is unclear, but it has sequence homology to the family of ‘ATPases associated with a variety of cellular activities’ (AAA+ proteins), which perform chaperone-like functions and assist in protein trafficking, membrane fusion and in secretory processing (Goodchild et al., 2005; Hewett et al., 2007; Breakefield et al., 2008; Granata et al., 2008). Pathologically, there is no clear evidence for neural degeneration in the few human post-mortem cases of DYT1 dystonia that have been examined, suggesting that functional and/or ultrastructural alterations, rather than neurodegenerative processes underlie the dystonic symptoms (Rostasy et al., 2003). A growing body of evidence suggests that the DYT1 mutation produces dystonia through aberrant neuronal signalling within the basal ganglia, and more specifically in the striatum (Todd and Perlmutter 1998; Rostasy et al., 2003, Breakefield et al., 2008). Over-activation of the basal ganglia circuits is observed in many forms of dystonia (Ceballos-Baumann and Brooks, 1997; Playford et al., 1998). In DYT1 dystonia, abnormal motor plasticity has been reported (Edwards et al., 2006), while clinically unaffected DYT1 gene carriers have subtle abnormalities in motor behaviour with impaired sequence learning (Ghilardi et al., 2003). Therefore, altered motor plasticity appears to play a relevant role in the pathophysiology of dystonia. Although mouse models of DYT1 dystonia do not exhibit overt dystonia, they do display abnormalities of motor learning, reminiscent of the human non-manifesting carrier state (Sharma et al., 2005; Breakefield et al., 2008).
Long-term changes in synaptic efficacy at corticostriatal synapses are thought to be the cellular basis of motor learning and associative memory processes. High-frequency stimulation (HFS) of glutamatergic striatal afferents can induce either long-term depression (LTD) or long-term potentiation (LTP) at this synapse (Calabresi et al., 1992a, b; Lovinger et al., 1993), depending on a variety of complex signalling mechanisms (Kreitzer and Malenka, 2008; Shen et al., 2008). Nevertheless, compelling evidence suggests that cholinergic transmission is a major modulator of the induction process (Wang et al., 2006; Bonsi et al., 2008). Despite the intrinsic nature of acetylcholine (ACh) innervation, cholinergic interneurons profoundly influence synaptic activity and plasticity of medium spiny neurons (MSNs) (Pisani et al., 2007). Recently, we have identified a profound alteration of cholinergic function in transgenic mice that over-express mutant torsinA (hMT mice). In cholinergic interneurons, the normal response to activation of dopamine D2 receptors is inhibition of firing activity. In hMT mice, we observed instead a paradoxical excitatory response, which likely results in an enhanced acetylcholine release (Pisani et al., 2006). Hence, in the present work, we investigated the possibility that an unbalanced cholinergic tone could disrupt synaptic plasticity in this mouse model of DYT1 dystonia. Using a combination of electrophysiological, morphological, biochemical and molecular approaches, we examined glutamatergic synaptic plasticity in slices from either hMT mice or mice expressing human wild-type torsinA (hWT). Our results demonstrate a substantial alteration of corticostriatal synaptic plasticity in hMT mice. More importantly, they show that normalizing striatal cholinergic signalling fully restores physiological synaptic plasticity.
Treatment and handling of animals was carried out in accord with both the EC and Italian guidelines (86/609/EEC; D.Lvo 116/1992, respectively) and approved by the University of Rome ‘Tor Vergata’ (n. 153/2001A). Transgenic mice (8- to 10-weeks old) were generated as previously described (Sharma et al., 2005). Breeding colonies of both hWT and hMT mice were established at our animal house. For each strain, non-transgenic (NT) littermates were utilized as controls, and are defined as NT mice (C57BL-6). Hence, the genetic background was identical between the control and experimental animals, with the only difference being the presence of the transgene. Age-matched muscarinic M2/M4 knockout (−/−) mice were generated as described (Duttaroy et al., 2002); wild-type littermates of the same mixed genetic background were utilized as controls.
For genotyping, tail DNA from mice was isolated using the Extract-N-Amp Tissue PCR Kit (Catalogue Numbers XNAT2, Sigma-Aldrich, Italy). To amplify a 560 bp segment, two specific primers for human torsinA were utilized (5′-CACATTGCACTTTCCACATGCT-3′ and 5′-GTTTTGCAGCCTTTATCTGA-3′). The samples were prepared according to the ‘Tissue Preparation’ and ‘Reagent Preparation’ protocols (Sigma-Aldrich; 35 cycles; annealing temperature 60°C). The human torsinA coding sequence was identified via 1.5% agarose gel electrophoresis. The mouse genotype was confirmed by restriction digestion with BseRI (New England BioLabs, USA). The human torsinA and human mutant torsinA PCR products were digested with BseRI into different fragments (279, 238, 24, 22 and 279 bp; 259 and 22 bp, respectively). Then, fragment profiles were identified by gel electrophoresis, using 2% SYBR Safe agarose (Invitrogen, Italy) (Fig. 1A).
Mice were sacrificed by cervical dislocation under ether anaesthesia and the brain was immediately removed from the skull. Coronal corticostriatal slices (250–300 μm thick; Fig. 1C) were cut with a vibratome in Krebs’ solution (126 NaCl, 2.5 KCl, 1.3 MgCl2, 1.2 NaH2PO4, 2.4 CaCl2, 10 glucose, 18 NaHCO3, all expressed in milliMolars), bubbled with 95% O2 and 5% CO2. In another set of experiments, parasagittal slices (300 µm) were prepared as described in detail previously (Fig. 1E; Smeal et al., 2007; Ding et al., 2008). In these slices (cut at 9–10°C), a knife-cut was made between the striatum and the thalamus, to prevent contamination from thalamostriatal inputs (Ding et al., 2008). After 30–60 min recovery, individual slices were transferred into a recording chamber (~0.5–1 ml volume), continuously superfused with oxygenated Krebs’ medium, at 2.5–3 ml/min and maintained constantly at 32–33°C.
Current-clamp recordings were performed blindly using sharp microelectrodes filled with 2 M KCl (40–60 MΩ). This electrophysiological approach was used to retain the fidelity of post-receptor signalling pathways. Signal acquisition and off-line analysis were performed with an Axoclamp 2B amplifier and pClamp9 software (Molecular Devices, USA). Striatal neurons were characterized according to their peculiar electrophysiological properties (Table 1). Glutamatergic excitatory post-synaptic potentials (EPSPs) were evoked with a bipolar electrode placed either in the corpus callosum (coronal slices) or in the cortex (V–VI layer, parasagittal slices, Fig. 1C and E), to activate glutamatergic afferents, and were recorded in the presence of bicuculline (10 µM) to prevent GABAA-receptor contamination. Stimuli of increasing intensity (50–500 µA, 30 µs duration, Fig. 2) were delivered at 0.1 Hz and six traces were averaged at each stimulus intensity. Then, peak excitatory post-synaptic potential amplitudes were measured and input–output relationships were obtained (pClampfit 9.2 software). The averaged excitatory post-synaptic potential whose peak amplitude was 50% of the maximum was further analysed for comparison among genotypes. Paired-pulse facilitation was assessed by presenting two stimuli, which evoked synaptic responses at ~50% of maximal amplitude, with an interstimulus interval of 50 ms and measuring the ratio of the peak amplitude of the second excitatory post-synaptic potential (EPSP2), divided by the first (EPSP1). For HFS (three trains: 3 s duration, 100 Hz frequency, 20 s intervals), stimulus intensity was raised to spike threshold level. The amplitude of EPSP was averaged and plotted as percentage of the control amplitude for ~10 min pre-high-frequency stimulation. Magnesium (Mg2+) was omitted from the perfusing solution to relieve the Mg2+-dependent block of NMDA receptors, a procedure required during sharp-microelectrode recordings to optimize LTP induction (LTP protocol) (Calabresi et al., 1992b). Synaptic depotentiation (SD) was induced by a low-frequency stimulation (LFS) protocol (2 Hz, 10 min), applied 20–30 min after LTP induction. One to six neurons per animal were recorded, each electrophysiological measure in the three groups of mice was obtained by pooling data from at least four distinct animals. One cell per slice was used for synaptic plasticity experiments.
Parasagittal hippocampal slices (300–400 μm) were prepared as previously described (Errico et al., 2008). Extracellular field excitatory post-synaptic potentials (fEPSPs) were recorded in stratum radiatum of the area CA1. The recording electrode [glass microelectrodes filled with artificial cerebral spinal fluid (ACSF)] was placed in the pyramidal cell layer of CA1 and a bipolar electrode was placed in stratum radiatum to activate the Schaeffer collaterals. Data were filtered at 3–10 kHz using an Axoclamp 2B Amplifier (Molecular Devices, USA), and collected for online analysis at a sampling rate of 20 kHz using LTP software (Anderson and Collingridge, 2007). The slope of the initial rising phase of the fEPSP was used as a measure of synaptic efficacy and the stimulus intensity was set to give a basal response with slope measurement ~40% of that at which a population spike was first observed. All data were normalized to the baseline preceding the HFS protocol (1 train, 100 Hz, 1 s). Data are expressed as mean ± SEM. Statistical significance was assessed using the Student's t-test. Drugs were from Tocris Cookson (UK), except for muscarine, trihexyphenidyl, hemicholinium-3 and bicuculline (Sigma Aldrich, Italy), and were applied by switching the control perfusion to drug-containing solution.
Histological samples were obtained both from perfused animals and biocytin-loaded slices from the three groups of mice: NT, hWT and hMT. Mice were group-housed in standard cages and kept under a 12 h light/dark cycle in a conditioned facility. Food and water were provided ad libitum. The animals were transcardially perfused under deep anaesthesia (ketamine and xylazine, 0.2 ml/10 g body weight, i.p.) with 150 ml of 0.9% saline at room temperature (RT) followed by 200 ml of cold 4% paraformaldehyde in 0.1 M pH 7.4 phosphate buffer (PB). Brains were dissected, post-fixed for 2 h at RT and cryoprotected in 30% PB/sucrose at 4°C. They were then frozen with dry ice and cut into 40 µm-thick transverse sections with a freezing microtome. Randomly, recording electrodes were also loaded with 2% biocytin to examine the morphology of the recorded neurons. Immediately after recording, slices containing biocytin-loaded cells were fixed by immersing in 4% paraformaldehyde in 0.1 M PB overnight at +4°C. The slices were subsequently immersed in 30% PB/sucrose at 4°C. They were then frozen with dry ice and cut into 35 µm-thick transverse sections. Sections derived from biocytin-loaded slices were immediately incubated with Cy2- or Cy3-conjugated Streptavidin (1:200; Jackson Immunoresearch, Laboratories, USA) in PB 0.3% Triton X-100 for 1 h at RT. Sections derived from perfused animals were incubated with a solution containing rabbit anti-dopamine and cAMP regulated phosphoprotein (DARPP)-32 (1:200; Millipore, cat. number AB1656, Billerica, USA), mouse anti-torsinA (1:300, Cell Signalling, USA), guinea pig anti-Substance P (SP, 1:200, Abcam, UK) primary antibodies in PB 0.3% Triton X-100 overnight at +4°C. After three washes in PB, sections were incubated with a solution containing a Cy2-conjugated donkey anti-mouse, Cy3-conjugated donkey anti-rabbit, Cy5-conjugated donkey anti-guinea pig IgGs (1:200; Jackson Immunoresearch, Laboratories) secondary antibodies. After three washes in PB, sections derived from all histological material were mounted on slides, air-dried and coverslipped with GEL MOUNT (Biomeda, USA). Fluorescence was examined under a confocal scanning laser microscope (CSLM Leica SP5, Leica Microsystems, Germany). Final images were projections of z-stack series taken through a total thickness of 20 µm. Figures were generated by adjusting only brightness and contrast, and composed by using Adobe Illustrator 10. After a visual inspection of the labelling throughout the sections, all images were acquired from the dorso-lateral striatal region where recordings were performed.
The quantitative analysis of double immunolabelled neurons was performed on digital images acquired through CLSM using a 40X objective at a 0.7 zoom factor. Three separate fields (dorso-lateral, central, medial) in three digital squared frames (200 µm per side) of five rostro-caudally spaced sections were examined (n = 5 animals).
Cellular co-localization of double-labelled neurons was analysed by counting and characterizing cell labelling off-line through the CLSM proprietary image analysis programme. The features of immunolabelled neurons were analysed by zooming on the cells and by serially excluding each channel. Double- and single-immunolabelled neurons were then digitally marked, recorded and the material was stored in a data archive. All data were expressed as mean ± SEM.
NT, hWT and hMT mice (n = 8 per genotype) were killed and the striata dissected out, sonicated in 200 μl of 1% sodium dodecyl sulfate (SDS) and boiled for 10 min. Equal amounts of total protein (30 μg) from each sample were loaded onto 10% polyacrylamide gels. Proteins were separated, transferred and membranes immunoblotted as described earlier (Errico et al., 2008). Selective antibodies against NR1 (1:1000; Sigma, St Louis, MO, USA), NR2A (1:1000; Sigma), NR2B (1:1000; Upstate, Lake Placid, NY, USA), GluR1 (1:5000; Chemicon, Temecula, CA), GluR2/3 (1:1000; Upstate) and DARPP32 (1:1000; Cell Signaling Technology, Beverly, MA, USA) were used. Blots were then incubated in horseradish peroxidase-conjugated secondary antibodies and target proteins visualized by electrogenerated chemiluminescence (ECL) detection (Pierce, Rockford, IL, USA), followed by quantitation by Quantity One software (Biorad). Optical density values were normalized to DARPP-32 for variations in loading and transfer. Normalized values were then averaged and used for statistical comparisons (Student's t-test).
Striata from NT and both hWT or hMT mice were weighed and homogenized in conical glass/glass homogenizer in a 1:20 w/v ratio. Homogenization was performed in 0.1 M sodium PB pH 7.0 containing 1 M NaCl, 10 mM EDTA, 0.5% Triton X-100, 40 mg/ml leupeptin and 20 mg/ml pepstatin. Cholinesterase extraction was carried out with three cycles of 20 strokes each, with intervening freezing and thawing of the samples. After centrifugation at 12 000 g for 30 min at 4°C, the enzymatic activity was assayed on the supernatant, according to Ellman and coworkers (1961), using 1 mM final concentration of acetylthiocholine iodide as substrate. In order to evaluate the contribution of acetylcholinesterase to the total cholinesterase activity, appropriate inhibitors were also used. Pseudocholinesterase or acetylcholinesterase activity was inhibited by pre-incubating tissue extracts for 1 h with 14 μM ethopropazine or 10−5 M BW284c51, respectively. Enzyme activity is expressed as specific activity in units per milligram of protein, 1 U corresponding to 1 µmol of substrate hydrolyzed per min at 30°C, pH 8.0.
Electrophysiological results are presented as mean ± SEM. Student's t-test and non-parametric Mann–Whitney test were used to compare means pre- and post-HFS/drug. Analysis of variance (ANOVA) test with a post hoc Tukey test were performed among groups (P < 0.05; α = 0.01). For all analyses, a P < 0.05 was considered statistically significant.
Passive and active intrinsic membrane properties such as resting membrane potential (RMP), input resistance, action potential amplitude and duration, delay to spike threshold and firing frequency allowed us to identify MSNs. Intrinsic properties of NT MSNs (n = 59) were not significantly different from those measured from both hWT (n = 62) and hMT neurons both from coronal and parasagittal slices (n = 76, 27, respectively) (Table 1), and did not differ from those previously described for mouse MSN in tissue slice preparations (Goldberg et al., 2005; Bonsi et al., 2008). MSNs were silent at rest and depolarizing current pulses caused tonic action potential discharge and strong inward rectification that did not differ among genotypes (Fig. 1B). RMP was kept constant throughout the recording session, by injecting negative current through the recording electrode (0.3–100 pA), whenever required. Membrane potential drift was monitored and subtracted at the end of each recording session. Biocytin labelling confirmed that all the recorded neurons were indeed MSNs with similar appearance among genotypes. Neurons from both coronal and parasagittal slices had medium-sized soma (10–20 µm), an extensive dendritic tree densely studded with spines (Fig. 1D and F).
The immunohistochemical characterization demonstrated that torsinA immunoreactivity was widely distributed throughout the striatal areas examined without apparent variation in the labelling intensity. The cellular staining was confined to the cytoplasm and sometimes to proximal dendrites of MSNs, as previously shown (Sharma et al., 2005) (Supplementary Fig. S1). We never observed labelling in cell nuclei or fibres. In NT, hWT and hMT mice, torsinA immunostaining was punctate and in the cytoplasm, consistent with localization in the endoplasmic reticulum. On morphological grounds torsinA labelling appeared to be confined to neurons without any evidence of glial labelling. In hMT striata, most of the DARPP-32-positive neurons were also immunoreactive for torsinA (96.4 ± 1.9%), ruling out the possibility that torsinA expression could be functionally segregated in a specific population of D1- or D2 receptor-containing MSNs (Supplementary Fig. S1). However, not all MSNs expressing DARPP-32 were Substance P (SP)-positive, consistent with the distinct neurochemical profiles of striatal MSNs (62.8 ± 0.36%). Neurons labelled for torsinA were also DARPP-32 immunoreactive (89.8 ± 1.8%). This pattern of torsinA expression was common to NT, hWT and hMT mice (data not shown, P > 0.05) and is in accordance with the description of a widespread, uniform torsinA labelling of the striatum (Konakova and Pulst, 2001; Walker et al., 2002; Sharma et al., 2005).
Cortically evoked EPSPs were fully blocked by a combination of N-methyl-d-aspartic acid (NMDA) and α-amino-3-hydroxyl-5-methyl-4-isoxazole-propionate (AMPA) glutamate receptor antagonists, Dizocilpine (MK-801; 30 µM) and 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX; 10 µM) both in coronal and parasagittal slices. MSNs from NT coronal slices exhibited similar mean amplitude and duration of the EPSPs as compared with MSNs from hWT (n = 16) and hMT (n = 20) mice (Fig. 2A; P > 0.05). The input–output relationship did not show statistically significant differences among groups (Fig. 2A; ANOVA P > 0.05, followed by post hoc Tukey test). Paired-pulse facilitation is considered to reliably parallel modifications in transmitter release probability (Schulz et al., 1994). Therefore, we measured paired-pulse ratio (PPR) as an indicator of pre-synaptic glutamatergic activity. No significant differences in the PPR were found between neurons from NT (n = 11, 1.36 ± 0.3), hWT (n = 11, 1.4 ± 0.6) and hMT mice (n = 17, 1.36 ± 0.06) (Fig. 2B; ANOVA P > 0.05).
The absence of possible changes in glutamatergic transmission was further confirmed by our analysis of both NMDA and AMPA receptor subunit expression levels. Indeed, western blotting analysis revealed unchanged expression levels for NMDA and AMPA receptor subunits in both hWT and hMT mice, compared with NT mice (Fig. 3; P > 0.1, per each protein; Student's t-test).
HFS of glutamatergic afferents led to the induction of robust LTD in MSNs in both coronal and parasagittal slices from NT mice, similar in magnitude and duration to that reported previously in mouse brain slices (Goldberg et al., 2005; Bonsi et al., 2008) (Fig. 4A and B; 51.4 ± 6.3% of control, from coronal slice, and 51.7 ± 7.1% from parasagittal slices, measured 25 min post-HFS; n = 34 and 10, respectively; t-test P < 0.001 and P < 0.05). Similarly, in MSNs from hWT mice, HFS was able to cause an LTD that was indistinguishable from that measured in NT mice (Fig. 4A and B; 50.5 ± 5.4% and 58.2 ± 9.7%, from coronal and parasagittal slices, respectively, measured 25 min post-HFS; n = 39 and 6; Mann–Whitney P < 0.001). In slices from hMT mice, however, HFS failed to cause any LTD (Fig. 4A and B; 94.9 ± 7%, n = 81 from coronal slices; LTD was not elicited in 70 cells out of 81; t-test P > 0.05; 101.4 ± 10.7%, n = 6 from parasagittal slices, P > 0.05).
The LTP induction protocol elicited LTP in NT as well as in hWT mice in both slice preparations (Fig. 4C and D; NT: 151.2 ± 5% of pre-HFS; 71 out of 84 from coronal slices exhibited LTP; t-test P < 0.0001; 158.3 ± 8.1%, n = 10 from parasagittal slices, P < 0.001; hWT: 157.1 ± 2.7%, n = 56, in six cells no LTP was observed, coronal slices; t-test P < 0.0001; 162.5 ± 13.1%, n = 5 from parasagittal slices, P < 0.0001). Notably, in MSNs from hMT mice, HFS led to a potentiation of synaptic activity that was significantly higher in magnitude as compared with NT or hWT littermates (Fig. 4C and D; 212.6 ± 6.4%, n = 49, two cells not showing LTP; Mann–Whitney P < 0.001; 198.1 ± 15.2%, n = 18 from parasagittal slices, P < 0.0001).
The higher magnitude of LTP prompted us to further analyse this phenomenon. Thus, we performed saturation experiments by using a well-established protocol, consisting of four consecutive HFS trains (20 min apart, using a stimulation intensity producing half-maximal response) (Yin et al., 2009). Pathways were considered saturated if the difference between the two states of LTP inductions was not significantly different (P > 0.05). Interestingly, a lower number of trains was needed to reach saturation in hMT mice, as compared with NT mice (Supplementary Fig. S2A and B). However, the ‘ceiling’ of LTP did not change (NT = 45.03 ± 14.9 mV, n = 6; hWT = 45.2 ± 11.9, n = 4; hMT = 43.9 ± 12.1, n = 6; P > 0.05), suggesting that the range of synaptic changes for LTP induction was not modified (Rioult-Pedotti et al., 2000).
Once striatal LTP is stabilized, a LFS (2 Hz, 10 min) protocol can reverse synaptic strength to resting levels, a phenomenon defined as SD (Fujii et al., 1991; Larson et al., 1993). SD is considered an active cellular mechanism aimed at erasing redundant synaptic information (Stäubli and Chun, 1996; Chen et al., 2001). LFS was able to cause SD in NT and hWT mice (Fig. 4C and D; NT: 100.4 ± 5.2%, n = 26, t-test P < 0.0001, coronal slices; 104.1 ± 11%, n = 6 from parasagittal slices; hWT: 100.4 ± 2.1% and 106.2 ± 13%, n = 46 and n = 6, from coronal and parasagittal slices, respectively; t-test P < 0.001), but failed to revert LTP to resting levels in hMT mice (Fig. 4C and D; 193.1 ± 5.5%, n = 20 from coronal slices, t-test P = 0.42; 193.6 ± 14%, n = 9 from parasagittal slices, Mann–Whitney P > 0.05).
Next, to test for the possibility that requirements for SD induction were changed in hMT mice, we modified the LFS protocol, either by changing frequency and duration (0.5 Hz, 10 or 15 min; 1 Hz, 10 or 15 min) or by reducing the stimulation intensity before LTP induction (Supplementary Fig. S2C). However, these modifications did not result in a rescue of SD in hMT mice (Supplementary Fig. S2C–E; n = 18; P > 0.05).
Finally, to investigate whether the observed synaptic plasticity impairment is specific for corticostriatal glutamatergic synapses, we measured synaptic transmission and plasticity at another glutamatergic synapse, the Schaeffer collateral pathway of the hippocampus. No differences in the PPR were observed between the groups (data not shown; P > 0.5 at all interpulse intervals). We then studied the effect of the DYT1 mutation on LTP at CA1 synapses. The post-tetanic potentiation, measured as the peak response elicited by HFS, was similar between genotypes, as well as the resulting LTP (data not shown; fEPSP slope percentage of baseline measured 60 min after HFS, hWT versus hMT: 123.3 ± 16% versus 135.3 ± 14%, P > 0.05, n = 6).
Ambient acetylcholine modulates bidirectional striatal synaptic plasticity (Pisani et al., 2007). Accordingly, loss of striatal muscarinic M2/M4−/− autoreceptors, a condition that causes an increase in endogenous acetylcholine (Zhang et al., 2002), impairs LTD but not LTP (Bonsi et al., 2008). In hMT mice, we have previously described a paradoxic excitation of cholinergic interneurons in response to dopamine D2 receptor activation (Pisani et al., 2006), an effect that is likely to enhance acetylcholine release and lead to increased striatal cholinergic tone. Thus, we evaluated the hypothesis of an involvement of a disrupted acetylcholine signalling in the hMT mice, using three different pharmacological approaches: (i) by decreasing striatal acetylcholine levels with hemicholinium-3, a depletor of endogenous acetylcholine (Parikh and Sarter, 2006); (ii) by enhancing cholinergic tone with AF-DX384, a selective M2/M4 autoreceptor antagonist, that blocks the self-inhibitory mechanism by which acetylcholine regulates its own release and (iii) by selectively blocking M1 muscarinic receptors, key modulators of both LTD and LTP (Calabresi et al., 2000; Wang et al., 2006; Bonsi et al., 2008). The selectivity of M1 antagonists was confirmed by testing the efficacy of either pirenzepine or trihexyphenidyl to block the M1-mediated membrane depolarization induced by muscarine (60 µM) or by the selective M1 agonist McN-A343 (3 µM) (Bonsi et al., 2008). Both pirenzepine (100 nM) and trihexyphenidyl (3 µM), given separately, blocked the depolarizing response to M1 receptor activation (data not shown; hMT: n = 6; hWT: n = 9, P > 0.05).
Hemicholinium-3 (10 µM, 20 min) did not modify the EPSP amplitude, the intrinsic properties of MSNs per sè, nor did it affect LTD in NT or hWT mice, but this treatment completely rescued LTD in hMT mice (Fig. 5A; hMT: 47 ± 2%, n = 10; t-test P < 0.001). Similarly, both pirenzepine (100 nM, 20 min) or trihexyphenidyl (1, 3 µM, 20 min) did not alter basal EPSP amplitude or LTD induction in NT or hWT mice, but fully restored LTD in hMT mice (Fig. 5B and C; pirenzepine: 52.3 ± 1.4%, n = 11; Mann–Whitney P < 0.001; trihexyphenidyl 60.7 ± 3.12%, n = 9; Mann–Whitney P < 0.0001). In addition to these experiments with acute treatments of the slices in vitro, we also examined the effect of chronic treatment with trihexyphenidyl in vivo (3 days, 20 mg/kg i.p., last injection 100 min before sacrifice), prior to the preparation of striatal slices. This experimental condition, which more closely mimics the use of anti-cholinergics in humans was also able to rescue LTD (Fig. 5B; 56.5 ± 3%, n = 8; t-test P < 0.0001).
AF-DX384 (300 nM, 20 min) caused no significant effect on control EPSP, but prevented LTD induction in MSNs from both NT and hWT mice (Fig. 5D; NT: 96.5 ± 8%, n = 6; t-test P > 0.05; hWT: 104.3 ± 1.8%, n = 5; Mann-Whitney P > 0.05), supporting the view that enhanced striatal cholinergic tone disrupts striatal LTD in this slice system.
Endogenous acetylcholine promotes LTP induction, through the activation of muscarinic M1 receptors (Calabresi et al., 2000). This suggests the possibility that the enhanced LTP which we observed in hMT mice could result from an over-activation of M1 receptors. However, since full blockade of muscarinic M1 receptors with pirenzepine prevents LTP induction in NT mice (data not shown), experiments were carried out in the presence of lower doses of this muscarinic M1-preferring antagonist. In slices pre-treated with 50 nM pirenzepine, HFS did not prevent LTP in hMT mice, but restored its normal magnitude (data not shown; 151.7 ± 9.5%, n = 6; P < 0.05).
Finally, we hypothesized that acetylcholine was also involved in the impairment of SD observed in hMT mice. We tested this possibility by genetic and pharmacological manipulations. We first verified that cholinergic excess would lead to blockade of SD under our experimental conditions. We utilized slices from mice lacking muscarinic M2/M4 autoreceptors, and observed that LFS failed to revert the LTP to baseline-synaptic levels (Fig. 6A, 184.1 ± 8%, n = 9, Mann–Whitney P > 0.05). A further confirmation was obtained by utilizing the selective M2/M4 receptor antagonist AF-DX384 in slices from NT and hWT mice. Under these conditions, LFS failed to cause SD in both genotypes (Fig. 6A, NT: 177.7 ± 4%, n = 9, Mann–Whitney P = 0.68; hWT: 176.1 ± 7.4%, n = 11, Mann–Whitney P = 0.6).
In a final set of experiments, slices from hMT mice were pre-treated with hemicholinium-3 or M1 receptor antagonists. As both drugs are able to prevent LTP induction per sè, they were added after LTP induction, but before delivering LFS (15–20 min). In hMT slices, hemicholinium-3 (10 µM, 15 min) restored SD (Fig. 6B; 95.3 ± 3.5%, n = 12, t-test P < 0.001). Likewise, either pirenzepine (100 nM) or trihexyphenydil (1–3 µM) rescued SD in hMT mice (Fig. 6C; pirenzepine: 95.3 ± 3.4%, n = 14, t-test P < 0.0001; trihexyphenydil: 99.7 ± 2V, n = 6, t-test P < 0.0001).
Finally, we measured the activity of striatal cholinesterases, responsible for acetylcholine degradation. The cholinesterase activity was measured in tissue extracts of striata from both NT and transgenic mice expressing hWT or hMT. Acetylcholinesterase activity did not significantly differ in striata from NT or hWT mice. Instead, acetylcholinesterase activity was significantly increased (P < 0.05) in hMT mice (Fig. 7). Ethopropazine, an inhibitor of pseudocholinesterase, did not affect cholinesterase activity in the tissue extracts; conversely BW284c51, a specific acetylcholinesterase inhibitor, decreased cholinesterase activity to background levels in all the samples. These observations indicate that the cholinesterase activity determined under our experimental conditions is almost entirely due to acetylcholinesterase (data not shown), and suggests that hMT mice have an adaptive enhancement of acetylcholinesterase activity in compensation for an increase in endogenous acetylcholine.
An abnormal synaptic plasticity of the motor system is considered as an important determinant in the pathophysiology of primary dystonia. In the current work, we provide evidence for the loss of corticostriatal LTD and SD in hMT mice expressing the mutant human torsinA protein which causes DYT1 dystonia. In addition, our findings clearly show that, at least in part, these alterations can be ascribed to an aberrant striatal acetylcholine signalling, occurring through muscarinic M1 receptors.
Traditional models of basal ganglia organization are based on anatomical and functional evidence predicting the existence of distinct, parallel loops that integrate cortical with basal ganglia nuclei, in order to facilitate voluntary movements and to inhibit competing, unwanted movements (Albin et al., 1989; DeLong 1990; Mink, 1996). The imbalance between these pathways is believed to account for the hyper- and hypo-kinetic manifestations of movement disorders. Several clinical and experimental studies in animal models have linked dystonia to dysfunction in the basal ganglia (Todd and Perlmutter, 1998; Mink, 2003; Rostasy et al., 2003; Jinnah et al., 2005; Breakefield et al., 2008). The current hypothesis is that dystonia may indeed result from a deficient ‘surround inhibition’ of competing motor patterns, coupled to an expansion of the facilitatory centre, which would lead to the ‘overflow’ phenomenon, consisting of the involuntary co-contraction of the groups of muscles (Berardelli et al., 1998; Mink, 2003; Sohn and Hallett, 2004; Edwards et al., 2006). Accordingly, either trauma or surgical intervention to a given body part, which has been shown to enhance LTP in the corresponding somatotopic cortical area, can precipitate dystonia in genetically susceptible individuals (Jankovic 2001).
Similar results were observed in patients with focal dystonia by using transcranial magnetic stimulation, a non-invasive tool to assess excitability of the motor system (Hallett, 2000). Indeed, transcranial magnetic stimulation methods that probe mechanisms of synaptic plasticity in the motor cortex showed an abnormal plasticity of sensory-motor circuits in patients with dystonia. These studies have examined the effects produced by an LFS protocol in dystonic patients with a temporary relief of motor symptoms, showing that a dual alteration in plasticity processes is observed in dystonia. An excessive propensity to form associations between sensory inputs and motor outputs, reminiscent of an enhanced LTP, was coupled to a failure to remove associations that were previously consolidated (impaired depotentiation) (Quartarone et al., 2005, 2006; Hallett, 2006; Tamura et al., 2009).
In human DYT1 dystonia, only about a third of the gene carriers develop overt dystonia, but even those apparently unaffected (‘non-manifesting carriers’) demonstrate subtle abnormalities, which point to abnormal striatal plasticity. In particular, non-manifesting DYT1 carriers have been shown to have abnormalities in motor task learning, a function strongly associated with the striatum (Ghilardi et al., 2003). In addition, recent studies of network activity in non-manifesting DYT1 carriers have supported the view that the striatum is a central part of the network, which is dysregulated in the human disorder (Carbon and Eidelberg, 2009).
The hMT mouse model that we have studied is one of several animal models of DYT1 dystonia, which have been generated. These include a homozygous knock-in of the ΔGAG mutation or knockout of torsinA (both of which are lethal at birth); heterozygous knock-in of ΔGAG and transgenic over-expression of mutant human torsinA; and selective knockouts of torsinA in cortex and striatum (Breakefield et al., 2008; Zhao et al., 2008). Interestingly, none of these mouse models exhibit overt dystonia, but many display significant abnormalities of motor behaviour or learning, reminiscent of the human non-manifesting carrier state.
Our findings unequivocally demonstrate abnormalities of plasticity at striatal glutamatergic synapses in mice expressing human mutant torsinA protein. Morphologically identified spiny projection neurons of these mice lack LTD and synaptic depotentiation, while conversely, they express an enhanced LTP. The long-term plasticity deficits are specific for the striatum, because synaptic transmission and LTP in the Schaeffer collateral pathway of the hippocampus are normal, thereby arguing against a generalized glutamatergic defect. If LTP subserves memory formation and consolidation, SD is thought to represent the synaptic underpinning of the processes of inhibition of memory (re)consolidation and a means to erase redundant, unnecessary information (Stäubli and Chun, 1996; Chen et al., 2001). In hMT mice, corticostriatal glutamatergic synapses show an increased propensity to facilitation (LTP-like), coupled to a loss of inhibitory processes (LTD- and SD-like).
The inability to revert synaptic strength from the potentiated state to pre-LTP levels may indeed result in the loss of the skill to select from competing motor patterns, and therefore account for the so-called ‘overflow’ phenomenon. At the cellular level, we hypothesize that recruitment of more receptors (muscarinic M1) to existing synapses might lead to larger LTP and hamper the physiological processes underlying SD (Rioult-Pedotti et al., 2000). Interestingly, there is evidence that the lack of striatal SD parallels the development of levodopa-induced dyskinetic involuntary movements in an experimental model of parkinsonism, supporting the notion that abnormal information storage at striatal glutamatergic synapses is indeed linked with the development of aberrant motor patterns (Picconi et al., 2003).
Our recordings did not reveal significant differences in the intrinsic and synaptic properties of MSNs from hMT mice, suggesting that changes in basal neuronal excitability cannot account for the observed alterations in synaptic plasticity. Likewise, we did not find any difference in NMDA and AMPA receptor subunit composition, excluding a major involvement of glutamatergic transmission. Rather, our study identifies a fundamental change in striatal cholinergic signalling, which may well justify such abnormalities.
Striatal cholinergic interneurons exhibit autonomous pacemaker activity, providing a constant acetylcholine tone in the striatum (Zhou et al., 2002). Maintenance of acetylcholine levels is regulated by acetylcholine degrading enzymes, and by muscarinic M2/M4 autoreceptors that negatively modulate acetylcholine release (Yan and Surmeier, 1996; Calabresi et al., 1998). Loss of such autoreceptor function has been shown to increase cholinergic tone, and to impair striatal LTD (Wess et al., 2007; Bonsi et al., 2008). Our results convincingly demonstrate that an exaggerated cholinergic tone may indeed be responsible for the selective corticostriatal plasticity deficits which we observed in hMT mice. First, we were able to restore both LTD and SD by lowering striatal acetylcholine with hemicholinium-3, which blocks the high-affinity choline transporter, thereby preventing acetylcholine re-synthesis (Parikh and Sarter, 2006) and reducing cholinergic tone. On the contrary, pharmacological manipulation aimed at increasing acetylcholine levels prevented both LTD and SD in mice expressing normal torsinA, as well as in NT littermates.
Second, we obtained evidence for the specific involvement of muscarinic M1 receptors in abnormal plasticity in hMT mice. Endogenous acetylcholine has been shown to modulate the excitability of MSNs primarily through muscarinic M1 receptors that are strategically positioned at corticostriatal synapses (Hersch et al., 1994). Activation of M1 receptors results in a reduction in both KCNQ and Kir2 potassium currents (Galarraga et al., 1999; Shen et al., 2005, 2007), thereby favouring LTP induction (Calabresi et al., 2000; Bonsi et al., 2008). On the other hand, acetylcholine may modulate LTD by reducing M1 receptor activation through disinhibition of L-type calcium channels (Wang et al., 2006). Interestingly, we found that both tryhexyphenydil, commonly utilized in clinical practice to treat dystonia, or pirenzepine, both of which are preferential M1 receptor antagonists (Dörje et al., 1991) were able to rescue LTD and synaptic depotentiation, and to normalize LTP induction in hMT mice. These data are consistent with clinical experience, as one of the most effective medical therapies for generalized dystonia is treatment with anti-cholinergic drugs (Jankovic, 2006; Pisani et al., 2007).
The present data points to an abnormality of striatal cholinergic tone as the primary defect induced by presence of the mutant torsinA protein, but does not directly identify the cause for this increased cholinergic tone. In a prior study, we found an abnormality within the cholinergic neurons themselves. Normally, activation of dopamine D2 receptors reduces the activity of cholinergic interneurons (DeBoer et al., 1996; Pisani et al., 2000). However, in interneurons from hMT mice, D2 receptor activation dramatically increased, rather than decreased, spike rate (Pisani et al., 2006). This paradoxical activation appears to be related to changes in high voltage activated calcium channels, and would be expected to lead to an increase in striatal acetylcholine tone (Pisani et al., 2006). Although indirect, our observation of an enhanced acetylcholinesterase activity is in line with this hypothesis, and indicates that adaptive changes in enzymatic activity may represent an attempt to counterbalance the excess in acetylcholine levels.
Thus, the simplest mechanistic explanation for the alteration in cholinergic function, as well as for the abnormal plasticity described here, involves the coupling of D2 receptors to their effector mechanisms. The site of this abnormal coupling may be within the cholinergic neurons themselves. Alternatively, it is possible that impaired D2 receptor signalling may be a more general property of striatal neurons in the hMT mouse model, as suggested by our recent studies showing abnormalities in the coupling of D2 signals to striatal GABAergic function (Sciamanna et al., 2009). This is a useful working hypothesis, although it must be interpreted cautiously. The hMT mouse expresses the mutant protein throughout the brain, and thus it is difficult to separate local from network defects. These hypotheses may be further refined through the creation of mouse models with more selective expression of the mutant torsinA protein. In this respect, the development of transgenic mice, in which D1- or D2 receptor-expressing MSNs are labelled with green fluorescent protein neurons, allows us to perform targeted recordings and a more detailed analysis of striatal direct and indirect pathways (Kreitzer and Malenka, 2008; Shen et al., 2008; Shuen et al., 2008). The possibility to generate mice with mutant torsinA in a selected pathway or cellular subtype, that could be identified by a specific fluorophore, remains to be explored. In addition, it will certainly be of interest to replicate these experiments by utilizing different stimulation protocols such as intrastriatal microstimulation and a spike-timing-dependent plasticity protocol (Kreitzer and Malenka, 2008; Shen et al., 2008).
The data presented here provide convincing evidence that the mutant torsinA protein can lead to abnormal striatal plasticity, and support the view that abnormalities in striatal signalling may be an important cause of human dystonia. Our results suggest that approaches directed at modulating cholinergic excess, or intervening more directly in the mechanisms of striatal LTD and LTP, may prove useful in developing novel therapies for the disorder.
Supplementary material is available at Brain online.
Dystonia Medical Research Foundation and Bachmann-Strauss Dystonia & Parkinson's Foundation (to A.P.); Ministero Salute (Prog. Finalizzato and Art. 56 to G.B. and A.P.); Istituto Superiore Sanità (Malattie Rare to A.P.); Agenzia Spaziale Italiana (DCMC grant to G.B.); USPHS grant (P50NS37409 to N.S. and D.G.S.).
We wish to thank Drs D. Centonze, R. Nisticò and J. Wess for critical reading of the manuscript. Muscarinic M2/M4−/− mice were kindly provided by Dr J. Wess.