Early structural and functional plasticity alterations in a susceptibility period of DYT1 dystonia mouse striatum

The onset of abnormal movements in DYT1 dystonia is between childhood and adolescence, although it is unclear why clinical manifestations appear during this developmental period. Plasticity at corticostriatal synapses is critically involved in motor memory. In the Tor1a+/Δgag DYT1 dystonia mouse model, long-term potentiation (LTP) appeared prematurely in a critical developmental window in striatal spiny neurons (SPNs), while long-term depression (LTD) was never recorded. Analysis of dendritic spines showed an increase of both spine width and mature mushroom spines in Tor1a+/Δgag neurons, paralleled by an enhanced AMPA receptor (AMPAR) accumulation. BDNF regulates AMPAR expression during development. Accordingly, both proBDNF and BDNF levels were significantly higher in Tor1a+/Δgag mice. Consistently, antagonism of BDNF rescued synaptic plasticity deficits and AMPA currents. Our findings demonstrate that early loss of functional and structural synaptic homeostasis represents a unique endophenotypic trait during striatal maturation, promoting the appearance of clinical manifestations in mutation carriers.


Introduction
Early-onset generalized torsion dystonia (DYT1) is an autosomal dominant movement disorder, commonly caused by a GAG base-pair deletion in the TOR1A gene coding for torsinA protein, without gross brain structural defects or other detectable neuropathology (Ozelius et al., 1997;Ledoux et al., 2013). Intriguingly, only 30-40% of DYT1 mutation carriers develop dystonia, typically in childhood-early adolescence (Bressman et al., 2000). However, what triggers the clinical onset of symptoms is currently unknown, although the presence of a critical developmental period of susceptibility is highly probable, since mutation carriers that do not develop symptoms in that time-window remain unaffected for their entire life (Pappas et al., 2014).
Plasticity changes include functional and structural synaptic specialization, leading to experiencedependent acquisition of motor skills. However, genetic or acquired alterations may lead to maladaptive plasticity changes. Accordingly, human studies indicate neural processing and synaptic plasticity alterations as major determinants in dystonia pathophysiology (Quartarone and Hallett, 2013). A significantly enhanced responsiveness to plasticity protocols has been reported in dystonic patients (Edwards et al., 2006;Weise et al., 2006;Quartarone et al., 2009). Moreover, patterns of impaired motor learning have been described even in clinically unaffected DYT1 mutation carriers (Ghilardi et al., 2003), further supporting the notion that aberrant plasticity represents a unique endophenotype in dystonia.
Of note, an impairment of striatal plasticity has been demonstrated in a number of different DYT1 models, including transgenic mice and rats overexpressing mutant torsinA (Martella et al., 2009;Grundmann et al., 2012), knock-in mice heterozygous for Dgag-torsinA (Dang et al., 2012;Martella et al., 2014;Rittiner et al., 2016), revealing an impressive similarity with studies of synaptic plasticity in human dystonia. Collectively, these observations support the hypothesis that DYT1 dystonia is a complex neurodevelopmental disorder of abnormal neurochemistry, wiring, and physiology (Goodchild et al., 2013;Pappas et al., 2014).
However, these alterations were observed in adult rodents, and to date, a relationship between age and corticostriatal plasticity in dystonia is still lacking. Furthermore, the question as to whether functional and structural plasticity abnormalities occur early in life or later as adaptive changes remains unknown. We report structural and functional abnormalities occurring in a defined postnatal time-window in Tor1a +/Dgag mice, indicative of a 'premature' and abnormal functional and structural plasticity, which is paralleled by a time-dependent increase in both BDNF levels and AMPAR-mediated currents.
Our findings reveal molecular, functional and structural changes in DYT1 striatal spiny projection neurons (SPNs), emphasizing the link between abnormal plasticity and dystonia. Understanding the key stages at which synaptic circuits are affected could suggest new routes to prevent or treat the disorder.

Results
The critical period for symptom onset in DYT1 dystonia matches a time-window of postnatal life when motor memories are shaped by activity-dependent changes in the striatum. Thus, in order to characterize plasticity changes in the early adolescence, Tor1a +/Dgag mice were recorded from postnatal day P15 to P35, in good agreement with the approximate life phase equivalencies between humans and mice, predicting that~4 weeks of mouse age correspond to~14 years in humans (Flurkey et al., 2007).
The pattern of torsinA expression is common to all striatal DARPP-32-labeled neurons (Martella et al., 2009). To unmask potential differences between direct-and indirect-pathway SPNs, recording electrodes were filled with biocytin. Enkephalin staining revealed that neither ENK-positive nor ENK-negative SPNs exhibited LTD, ruling out a possible segregation to a specific population of SPNs ( Figure 2E).
Collectively, these data demonstrate that LTD appeared at P28 in wild-type mice, whereas it could not be elicited during the entire postnatal period of observation in Tor1a +/Dgag mice ( Figure 2E). Moreover, while in Tor1a +/+ mice LTP could not be evoked before P24, in SPNs from Tor1a +/Dgag LTP appeared prematurely at P15 ( Figure 2F).

Increased AMPA receptor function and abundance at corticostriatal synapses during development
Changes in synaptic strength during learning and memory processes implicate an accurate regulation of AMPARs and NMDARs expression at postsynaptic membranes (Bassani et al., 2013;Czö ndö r and Thoumine, 2013). Thus, we performed an electrophysiological and biochemical characterization of AMPARs and NMDARs of SPNs in both Tor1a +/+ and Tor1a +/Dgag mice.
The normalized IV relationship of NMDAR-EPSCs showed the characteristic 'J-shape' (Mayer et al., 1984) in SPNs recorded at P26 from both genotypes ( Figure 3E). No significant difference was found in the voltage-dependence of NMDARs (p>0.05). By analyzing the kinetics of the response at HP =+ 40 mV, we detected a significantly decreased decay time in Tor1a +/Dgag mice compared to controls ( Figure 3F,G; p<0.05), despite a comparable rise time, suggesting a modification of NMDAR subunit composition (Paoletti et al., 2013). In particular, it is well-established that the decay time of NMDAR currents is correlated to the amount of GluN2-type subunits. GluN2A and GluN2B represent the most abundant NMDAR regulatory subunits expressed in SPNs (Chen and Reiner, 1996;Dunah and Standaert, 2003) and are characterized by a fast and slow decay time, respectively (Sanz-Clemente et al., 2013).
Taking into account all the above-described electrophysiological results, we evaluated the levels of AMPAR and NMDAR subunits into TIF fractions purified from striata of both juvenile (P26) and adult (P60) mice by means of WB analysis. We found a significant increase in the levels of both GluA1 and GluA2 AMPAR subunits in the postsynaptic compartment of P26 Tor1a +/Dgag mice compared to controls ( Figure 4A,B; p<0.05), consistent with the observed reduction of the NMDA/ AMPA ratio and the absence of any alteration of the RI (see Figure 3). Interestingly, we also found an increase of phosphorylation at GluA1-Ser845 ( Figure 4A,B;, p<0.05), which is known to be correlated with LTP expression and to prevent endocytosis of GluA1-containing AMPARs (Oh et al., 2006;Bassani et al., 2013). Moreover, in agreement with the reduction of the NMDAR decay time, we observed an increase of GluN2A but not GluN2B subunit at postsynaptic sites of P26 Tor1a +/Dgag mice compared to Tor1a +/+ ( Figure 4C,D; p<0.05). Finally, no modifications of PSD-95, the most abundant scaffolding protein at the excitatory synapse, was observed ( Figure 4C,D). Notably, these alterations of AMPAR and NMDAR subunits were not present in SPNs from P60 Tor1a +/Dgag mice ( Figure 4; p>0.05).
Next, we performed a detailed evaluation of dendritic spine density and morphology in Tor1a +/D gag SPNs, compared to age-matched Tor1a +/+ mice. P26 Tor1a +/Dgag SPNs ( Figure 5A-D) exhibited a higher number of mushroom-type spines ( Figure 5C; p<0.05) and, consequently, a concomitant overall increase of dendritic spine width compared to Tor1a +/+ mice ( Figure 5B; p<0.05), thus suggesting an advanced stage of spine maturation, in agreement with the observed molecular GluN2A/ GluN2B switch (see Figure 4). This event was associated, as expected, to an overall decrease of dendritic spine density ( Figure 5A; p<0.05).
Likewise, in vivo treatment with ANA-12 totally normalized the IV curve of AMPAR-EPSC in P26 Tor1a +/Dgag mice ( Figure 7C; p>0.05). Accordingly, also the RI displayed no significant difference between genotypes ( Figure 7C; p>0.05). These findings suggest that increased BDNF levels are involved in the abnormal developmental expression of AMPARs on SPN postsynaptic membranes, leading to synaptic plasticity alterations in juvenile mice.
Finally, to demonstrate that BDNF alterations occur in a defined time-window, we tested the effect of ANA-12 on corticostriatal LTD expression in adult Tor1a +/Dgag mice. In vivo treatment with ANA-12 (0.5 mg/kg, intraperitoneal, two administrations at 12 hr and 4 hr before the experiment) failed to restore corticostriatal LTD in P60 Tor1a +/Dgag mice ( Figure 7D; p>0.05) confirming that BDNF-dependent alterations are limited to a sensitive period.

Discussion
The critical period for the onset of symptoms in DYT1 dystonia patients matches an early time-window of activity-dependent plastic changes in the striatum, which shape motor memory and learning processes during childhood and early adolescence.
Our systematic analysis of functional and structural synaptic plasticity in DYT1 dystonia demonstrates: (i) The existence of a critical period when SPNs exhibit premature LTP; (ii) A significant increase of AMPAR levels in the postsynaptic compartment which correlates with the reduction of NMDA/AMPA ratio, the increased amplitudes of postsynaptic currents and the rightward shift in the AMPA I-V curve observed in juvenile Tor1a +/Dgag mice; (iii) A BDNF time-dependent increase in expression profile, which parallels the alterations described; (iv) abnormal plasticity is associated with profound changes of dendritic spine density and morphology in juvenile Tor1a +/Dgag ; (v) A rescue of the synaptic plasticity deficits is obtained by in vivo administration of a TrkB inhibitor.
Our electrophysiological assessment of synaptic plasticity identifies a rather narrow time-window, between P15 and P26 when striatal SPNs exhibit a premature LTP, whereas LTD cannot be evoked. Although it has to be cautiously reminded that these alterations match those described in adult DYT1 striata , the time-course performed in the present study indicates their abnormal early appearance, in a developmental phase when normal striatal SPNs do not yet exhibit long-lasting synaptic changes. Of interest, the loss of LTD was observed in a similar time frame in a novel model with a rare missense variant in the Tor1a gene (Bhagat et al., 2016).
Moreover, we describe, along with electrophysiological deficits, molecular and structural changes at striatal synapses that appear to be limited to a specific time-window. Striatal LTP either in mature tissue preparation or in the developing striatum is dependent on the activation of NMDAR (Calabresi et al., 1992b;Partridge et al., 2000), whereas LTD depends on AMPAR (Calabresi et al., 1992a). Our electrophysiological and biochemical characterization demonstrates an increase in currents mediated by AMPAR, consistent with the increased amplitude of mEPSCs, and additionally, the NMDAR/AMPAR ratio was significantly reduced in SPNs from DYT1 mice. A major mechanism regulating synaptic strength involves the balance between synaptic insertion and removal of glutamate receptors into the postsynaptic membrane (Gong and De Camilli, 2008). Specifically, loss of homeostatic regulation of excitatory synapses in distinct neuronal subtypes involve postsynaptic changes in accumulation of AMPAR (Lissin et al., 1998;O'Brien et al., 1998). Accordingly, we observed a significant increase of both GluA1 and GluA2 subunits of AMPARs in the postsynaptic compartment of P26 Tor1a +/Dgag mice compared to controls, suggestive of an increased surface expression of AMPARs. Of interest, the significant increase of the phosphorylation of GluA1-Ser845, a well-established correlate of LTP (Oh et al., 2006;Bassani et al., 2013) is consistent with the abnormal LTP expression measured in DYT1 mice. Moreover, GluA1-Ser845 (Roche et al., 1996) plays a key role in the synaptic delivery of GluA1-containing AMPARs by LTP (Esteban et al., 2003;Bassani et al., 2013) and is involved in surface reinsertion/stabilization of AMPARs (Ehlers, 2000), thus providing a molecular mechanism for the observed increase of AMPARs at postsynaptic membranes in P26 Tor1a +/Dgag mice compared to controls. Thus, we hypothesize that the loss of LTD may be related to the aberrant composition of striatal AMPARs observed in mutant mice.
The identification of increased AMPAR subunit levels in the postsynaptic compartment offers new opportunities to identify potential regulators of AMPAR turnover. Neurotrophins have been implicated in glutamatergic synapse development and plasticity, suggesting a potential role in postsynaptic proteins distribution (Causing et al., 1997;McAllister et al., 1997;Kong et al., 2001). Previous work elucidated the role of BDNF in the regulation of AMPAR expression and function, including synaptic AMPAR subunit trafficking (Narisawa-Saito et al., 1999;Jourdi and Kabbaj, 2013). Indeed, BDNF treatment acutely controls both AMPAR subunits and their scaffolding proteins trafficking, thereby modifying the strength of synaptic activity (Minichiello et al., 1999;Mauceri et al., 2004). Remarkably, we observed an enhancement of pro-BDNF and BDNF protein level in P26 Tor1a +/Dgag mice, which appears critical for the onset of abnormal neurophysiological phenotype in DYT1 dystonia. Consistently, we obtained a functional rescue of synaptic plasticity and AMPA-mediated currents with the competitive antagonist of BDNF TrkB receptor ANA-12 (Cazorla et al., 2011).
Activity-dependent synaptic plasticity as well as composition and activity of NMDARs and AMPARs strictly govern modifications of dendritic spine morphology, leading to a long-lasting structural plasticity. Yet, BDNF also plays a major role in spine maturation in several brain regions, including the striatum (Baquet et al., 2004;Rauskolb et al., 2010). Thus, the abnormal increase in BDNF expression fits with the abnormalities in spine morphology we observed. In P26 Tor1a +/Dgag mice, we measured an increase in mushroom spines, suggestive of a 'premature' maturation process accompanied by an overall decrease in the density of dendritic spines. It is well-known that expression patterns of the GluN2 subunits of NMDARs at dendritic spines change during the first postnatal weeks. In particular, GluN2A expression increases from the second postnatal week to become widely expressed and abundant throughout the brain (Bellone and Nicoll, 2007;Gray et al., 2011). Yet, in agreement with the reduction of the NMDAR decay time, we found an increase of postsynaptic GluN2A in P26 Tor1a +/Dgag mice suggesting a 'premature' GluN2A/GluN2B switch, thus indicating the existence of a molecular and morphological early maturation of the excitatory synapse in this DYT1 model. Moreover, the existence of a close coordination between spine size and AMPAR levels at synaptic membranes has been previously reported (Kopec et al., 2007;Malinverno et al., 2010) and spine volume has been positively correlated with the strength of AMPAR-mediated synaptic transmission. Accordingly, in Tor1a +/Dgag mice we found a significant increase of spine head width, an increase in mushroom spines and a concomitant increase of both GluA1 and GluA2 subunits of AMPARs.
Most of the molecular and structural alterations described in juvenile DYT1 mice were not confirmed at our analyses performed in adult (P60) mice. Indeed, inhibition of BDNF with ANA-12 did not offset the plasticity deficits in adult mice. Additionally, the anticholinergic agent pirenzepine failed to rescue the plasticity deficits in juvenile animals, contrarily to what reported in adults (Dang et al., 2012;Martella et al., 2014), indicating that distinct mechanisms sustain the abnormal patterns of synaptic activity at different developmental ages. Future work is required to address the precise mechanisms governing this switch.
Collectively, we demonstrate that the rise of BDNF, in a restricted time-window, drives AMPA receptor composition changes and, consequently, structural modifications in spine morphology, resulting in the loss of homeostatic regulation of synaptic plasticity early in postnatal life.
Our hypothesis is also consistent with the clinical observation that the beneficial effects of Deep Brain Stimulation (DBS) in dystonic patients is more effective in young patients, as compared to patients implanted later in life (Isaias et al., 2008). Additionally, compared to the prompt efficacy observed in Parkinson's disease patients, weeks are commonly required to obtain symptomatic relief following DBS, and improvements may continue to be manifest over time (Vercueil et al., 2001;Krauss, 2002;Vidailhet and Pollak, 2005). It is plausible that severity of abnormal plasticity is related to disease duration, thus justifying the longer time required to erase aberrant plasticity patterns.
In a therapeutic perspective, these sensitive periods might be considered as temporal windows of opportunity, during which specific molecular steps could be targeted to prevent aberrant plasticity to develop.

Animal model
Studies were carried out in juvenile (P15-P35) and adult (P60-P75) knock-in Tor1a +/Dgag mice heterozygous for DE-torsinA, a mutation that removes a single glutamic acid residue (DE) from the torsinA protein, and in their wild-type Tor1a +/+ littermates (Goodchild et al., 2005). Genotyping was performed as described (Ponterio et al., 2018). Animal breeding, on a C57Bl/6J background, and handling were performed in accordance with the guidelines for the use of animals in biomedical research provided by the European Union's directives and Italian laws ( /63EU, D.lgs. 26/201486/609/CEE, D.Lgs 116/1992). The experimental procedures were approved by Fondazione Santa Lucia and University Tor Vergata Animal Care and Use Committees, and the Italian Ministry of Health (authorization #223/2017-PR).

Experimental design
Age-and sex-matched wild-type and mutant littermates were randomly allocated to experimental groups. Investigators performing experiments and data analysis were blind to knowledge of genotype and treatment. Each observation was obtained from an independent biological sample. For electrophysiology, each cell was recorded from a different brain slice. All data were obtained from at least two animals in independent experiments. Biological replicates are represented with 'N' for number of animals and 'n' for number of cells. Sample size for any measurement was based on the ARRIVE recommendations on refinement and reduction of animal use in research, as well as on our previous studies.

Electrophysiology
Brain slice preparation  . The AMPAR and NMDAR IV relationships were measured in the presence of PTX plus MK-801 or CNQX, respectively. The RI was calculated as ratio of the mean EPSC amplitudes measured at +40 mV and À70 mV.

Sharp-electrode recordings
Current-clamp recordings of SPNs were performed with intracellular electrodes filled with 2M KCl (30-60 MW). Corticostriatal EPSPs were recorded in PTX (50 mM). HFS (three trains 100 Hz, 3 s, 20 s apart) was delivered at suprathreshold intensity to induce LTD. Magnesium was omitted to optimize LTP induction (Calabresi et al., 1992b). The EPSP amplitude was averaged and plotted over-time as percentage of control pre-HFS amplitude.

Gene expression analysis
P26 Tor1a +/+ and Tor1a +/Dgag mouse cortex was collected in PCR clean tubes and stored at À80˚C. Total RNA was isolated using TRI-reagent (Sigma Aldrich), quantified and treated with DNAase I (Invitrogen). Integrity was confirmed by 1% agarose gel electrophoresis. RNA was reverse-transcribed using random hexamer primer and anchored-oligo ( (Livak and Schmittgen, 2001).

Immunohistochemistry
To identify direct-and indirect-pathway SPNs electrodes were loaded with biocytin, as described (Martella et al., 2009). Briefly, slices were fixed with 4% PFA in 0.12 M PB and 30 mm thick sections were cut from each slice with a freezing microtome, then dehydrated with serial alcohol dilutions to improve antigen retrieval and reduce background (Buchwalow et al., 2011). We used the following primary antibodies: goat anti-DARPP-32 (1:500 AF6259, R and D system), mouse anti-Enkephalin (1:1000 MAB350, Millipore), and secondary antibodies: anti-goat alexa 647 (Invitrogen), anti-mouse cyanine 3 (Jackson ImmunoResearch) and streptavidin-conjugated alexa 488 (Life Technologies). All sections used for analysis were processed together. Images were acquired with a LSM700 Zeiss confocal laser scanning microscope and analyzed with ImageJ software (NIH; Schneider et al., 2012). Noise was reduced by applying background subtraction in ImageJ.

Subcellular fractionation and western blotting (WB)
To obtain a preparation that contains selectively proteins of the post-synaptic density (PSD), subcellular fractionation of striatal tissue was performed as reported (Gardoni et al., 2006;Paillé et al., 2010) with minor modifications. Briefly, striata were homogenized with a Teflon-glass potter in icecold 0.32M sucrose containing 1 mM HEPES pH 7.4, 1 mM MgCl 2 ,1mM EDTA, 1 mM NaHCO 3 , 0.1 mM phenylmethanesulfonylfluoride (PMSF) in the presence of a complete set of proteases and phosphatase inhibitors (Complete TM Protease Inhibitor Cocktail Tablets and PhosSTOP TM Phosphatase Inhibitor Cocktail, Roche Diagnostics). The homogenized tissue was centrifuged at 13,000 g for 15 min. The pellet was re-suspended in a buffer containing 75 mM KCl and 1% Triton X-100 and spun at 100,000 g for 1 hr. The final pellet, referred to as Triton-insoluble postsynaptic fraction (TIF), was homogenized in a glass-glass potter in 20 mM HEPES supplemented with Complete TM tablets and stored at À80˚C until use. Protein samples were separated onto an acrylamide/bisacrylamide gel at the appropriate concentration, transferred to a nitrocellulose membrane and immunoblotted with the appropriate primary and HRP-conjugated secondary antibodies. For WB analysis, the following unconjugated primary antibodies were used: polyclonal anti-GluN2A antibody (Sigma-Aldrich); monoclonal anti-GluN2B antibody (NeuroMab); polyclonal anti-GluA1 antibody (Merck Millipore); polyclonal anti-phospho-GluA1 (Ser845; Merck Millipore); monoclonal anti-GluA2 antibody (Neuro-Mab); monoclonal anti-PSD-95 antibody (NeuroMab); monoclonal anti-a-tubulin antibody (Sigma-Aldrich). Membrane development was performed with the reagent Clarity Western ECL Substrate (Bio-Rad) and labeling was visualized by Chemidoc Imaging System and ImageLab software (Bio-Rad). For quantification, each protein was normalized against the corresponding a-tubulin band.
WB analysis of BDNF on mouse striatum was performed as described (Sciamanna et al., 2015;Ponterio et al., 2018). Protein extracts (15-30 mg) were loaded with page LDS sample buffer (Invitrogen, Waltham, Massachusetts, USA) containing DTT and denatured at 95˚C for 5 min. Proteins were separated on 15% SDS-PAGE, and transferred onto 0.45 mm polyvinylidene fluoride (PVDF) membranes. The following primary antibodies were utilized: rabbit anti-BDNF (1:200 sc-546, SantaCruz Biotechnology) and mouse anti-b-actin (1:20.000 A5441, Sigma Aldrich), as loading control, followed by anti-rabbit or anti-mouse horseradish peroxidase (HRP)-conjugated secondary antibodies. Immunodetection was performed by ECL reagent (GEHealthcare) and membranes were exposed to film (Amersham). Quantification of the band intensity on scanned filters was achieved by ImageJ software.

Spine morphology
Carbocyanine dye DiI (Invitrogen) was used to label neurons as previously described (Kim et al., 2007;Stanic et al., 2015). Images were taken using an inverted LSM510 confocal microscope (Zeiss). For morphological analysis, cells were chosen randomly for quantification from four to eight different coverslips; images were acquired using the same settings/exposure times, and at least 10 cells for each condition were analyzed. Morphological analysis was performed with ImageJ software to measure spine density and size. For each dendritic spine the length, the head and neck width were measured, which was used to classify spines into categories (thin, stubby and mushroom) (Harris et al., 1992).

Statistical analysis
Data were analysed with ClampFit 9 (pClamp, Molecular Devices), Origin 8.0 (Microcal) and Prism 5.3 (GraphPad) softwares. All data were obtained from at least two independent experiments and are represented as mean ± SEM. Statistical significance was evaluated, as indicated in figure legends, using paired and unpaired Student's t test, and one-way ANOVA with post-hoc Tukey test and two-way ANOVA with Bonferroni posttest for group comparisons. Statistical tests were twotailed, the confidence interval was 95%, and the alpha-level used to determine significance was set at p<0.05. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.