Differential involvement of Ras-GRF1 and Ras-GRF2 in L-DOPA-induced dyskinesia.

OBJECTIVE: Recent findings have shown that pharmacogenetic manipulations of the Ras-ERK pathway provide a therapeutic means to tackle l-3,4-dihydroxyphenylalanine (l-DOPA)-induced dyskinesia (LID). First, we investigated whether a prolonged l-DOPA treatment differentially affected ERK signaling in medium spiny neurons of the direct pathway (dMSNs) and in cholinergic aspiny interneurons (ChIs) and assessed the role of Ras-GRF1 in both subpopulations. Second, using viral-assisted technology, we probed Ras-GRF1 and Ras-GRF2 as potential targets in this pathway. We investigated how selective blockade of striatal Ras-GRF1 or Ras-GRF2 expression impacted on LID (induction, maintenance, and reversion) and its neurochemical correlates.
METHODS: We used both Ras-GRF1 knockout mice and lentiviral vectors (LVs) delivering short-hairpin RNA sequences (shRNAs) to obtain striatum-specific gene knockdown of Ras-GRF1 and Ras-GRF2. The consequences of these genetic manipulations were evaluated in the 6-hydroxydopamine mouse model of Parkinson's disease. Escalating doses of l-DOPA were administered and then behavioral analysis with immunohistochemical assays and in vivo microdialysis were performed.
RESULTS: Ras-GRF1 was found essential in controlling ERK signaling in dMSNs, but its ablation did not prevent ERK activation in ChIs. Moreover, striatal injection of LV-shRNA/Ras-GRF1 attenuated dyskinesia development and ERK-dependent signaling, whereas LV-shRNA/Ras-GRF2 was without effect, ruling out the involvement of Ras-GRF2 in LID expression. Accordingly, Ras-GRF1 but not Ras-GRF2 striatal gene-knockdown reduced l-DOPA-induced GABA and glutamate release in the substantia nigra pars reticulata, a neurochemical correlate of dyskinesia. Finally, inactivation of Ras-GRF1 provided a prolonged anti-dyskinetic effect for up to 7 weeks and significantly attenuated symptoms in animals with established LID.
INTERPRETATION: Our results suggest that Ras-GRF1 is a promising target for LID therapy based on Ras-ERK signaling inhibition in the striatum.

Introduction

l-DOPA-induced dyskinesia (LID) is one of the unwanted and debilitating motor side effects ensuing from prolonged treatment with l-DOPA, the gold standard for the symptomatic treatment of Parkinson’s disease (PD).1 Indeed, the majority of PD patients develop abnormal involuntary movements (AIMs) within 5–10 years of l-DOPA therapy.2 LID is still a significant clinical problem since no truly effective treatment has been developed so far. Besides the classical pharmacological approach targeting neurotransmitter receptors, accumulating evidence from animal models supports a causative role for dysregulated D1 receptor intracellular signaling in striatal medium-sized spiny neurons of the direct pathway (dMSNs). These observations have opened new perspectives for innovative therapeutic approaches against LID, based on the inhibition of either the canonical PKA/DARPP-32 cascade or the non-canonical Ras-ERK and mTOR pathways.3–11

The Ras-ERK cascade is an evolutionarily conserved neuronal pathway involved in several survival processes and an important regulator of behavioral plasticity.12–19 Its sustained activation leads to synaptic rearrangements requiring de novo gene expression and protein synthesis. In striatal cells, glutamate (GLU) and dopamine receptors interact and provide a route to ERK activation.20–24 Importantly, in animal models of PD, including the unilaterally 6-hydroxydopamine (6-OHDA) lesioned rodent and the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)-treated non-human primate (NHP), the supersensitivity of dopamine D1 receptors leads to aberrant ERK activation in response to l-DOPA, which correlates with LID severity.3–5,25,26 In particular, our recent study indicated that Ras-GRF1, a Ras activator (Ras guanine-nucleotide exchange factor, Ras-GEF) expressed only in mature neurons of the central nervous system, is necessary for the integration of GLU and dopamine signaling that leads to ERK activation.23 Importantly, Ras-GRF1 specifically controls downstream ERK signaling in a neurotrophin-independent manner, suggesting that its inhibition would only affect plasticity-related ERK signaling without altering cell survival mechanisms. Consistently, Ras-GRF1 ablation by conventional gene targeting27 does not affect the ability of 6-OHDA to deplete dopamine-producing cells but strongly attenuates ERK activation and AIMs appearance in the rodent lesion model of LID.28 Notably, whilst ERK activity is required in all striatal cells to induce long-term potentiation (LTP), Ras-GRF1 is necessary only in striatal dMSNs, that is, those cells more directly implicated in LID.11 Moreover, attenuation of Ras-GRF1 and ERK signaling in the NHP model of PD results in a strong reduction in dyskinetic symptoms without compromising the antiparkinsonian effect of l-DOPA, providing a more clinically valuable approach via targeting Ras-ERK, which may ameliorate this pathological condition.28

A recent study showed that in Pitx3-deficient mice, a genetic model of PD, the abnormal activation of ERK surprisingly diminishes in MSNs but increases in the large aspiny cholinergic interneurons (ChIs), upon continuous administration of l-DOPA.29 In addition, in a subsequent study Won and colleagues demonstrated that selective depletion of striatal ChIs via Cre-dependent viral expression of the diphtheria toxin A significantly attenuated LID without affecting the therapeutic efficacy of l-DOPA.30 This evidence prompted us to investigate in the first part of our work whether intermittent but prolonged administration of high doses of l-DOPA could lead to different ERK activation profiles in dMSNs and ChIs and whether this activation in cholinergic interneurons is somehow regulated by Ras-GRF1. Our previous observation that Ras-GRF1 inhibition in the brain only leads to ∼50% reduction in AIMs could indeed suggest that other factors may regulate striatal ERK activity in response to l-DOPA, including a potentially Ras-GRF1-independent ERK activation in ChIs. Moreover, Ras-GRF2, a close homolog of Ras-GRF1 in the striatum, may be an additional factor controlling ERK activity in dyskinesia. To address this point in the second part of the work, we took advantage of short hairpin RNA sequences (shRNA) technology to specifically target the dorsolateral striatum, a key area involved in the development of dyskinetic movements to obtain striatal-specific gene knockdown of Ras-GRF1 and Ras-GRF2. We then analyzed the behavioral consequences of these genetic manipulations in the development of dyskinesia and their molecular and neurochemical correlates. Finally, we analyzed the possibility of reversing an already established dyskinetic state.

Our findings showed that a sustained activation of ERK is present in both dMSNs and ChIs after 4 weeks of intermittent high doses of l-DOPA but in ChIs this event is Ras-GRF1 independent. In addition, only Ras-GRF1, but not Ras-GRF2 was critically involved in the development of LID and in the correlated molecular changes. Importantly, Ras-GRF1 silencing had a long-lasting therapeutic effect, up to 7 weeks and was also effective in alleviating established dyskinesia, demonstrating the potential as a therapeutic intervention.

Material and Methods

For detailed description of material and methods see Data S1.

Animals

In this study, we used the following: (1) C57BL/6 male mice (Charles River Laboratories, Calco, Italy); (2) Ras-GRF1 KO males and their littermates.27 All experimental animal procedures were conducted according to the EU Directive 2010/63/EU and to experimental animal licenses of the Fondazione San Raffaele del Monte Tabor and University of Ferrara approved by the Italian Ministry of Health and the local authorities.

shRNA constructs and LV production

Three shRNA constructs were probed against Ras-GRF1 and Ras-GRF2 genes in primary cultures. The sequence targeting the Ras-GEF region of Ras-GRF1 (sequence 1) and PH domain of Ras-GRF2 (sequence 3) were used in vivo. The Vesicular stomatitis virus (VSV) pseudotyped third-generation lentiviral vectors (LVs) were produced as previously described.31,32

Stereotaxic Surgery and l-DOPA treatments

6-OHDA lesion

Unilateral 6-OHDA MFB lesions were performed as described in Fasano et al.28

LV-shRNA injections

Two weeks post-6-OHDA lesions, mice with confirmed ipsilateral rotations received unilateral striatal injections (2 × 1 μL) of LVs as described in Fasano et al.28

l-DOPA administration

l-DOPA (Sigma Aldrich, Milan, Italy) was injected accordingly to the following protocols: (1) daily injection of 3 (day 1–3), 6 (day 4–6) or 12 (day 7–9) mg/kg in combination with benserazide (Sigma-Aldrich) (12 mg/kg) in the 9 days protocol; (2) twice a week 12 mg/kg with 12 mg/kg benserazide in the maintenance protocol of 4 and 7 weeks and in the reversal protocol of 4 weeks (experimental design 1, 4, and 5, respectively). Mice not treated with l-DOPA received an equivalent volume of Saline. Both drugs were injected i.p. in a total volume of 10 mL/kg body weight.

Behavioral analysis, in vivo microdialysis and postmortem examination were performed as described in refs 28,33–36.

Results

Prolonged and intermittent l-DOPA treatment in dopamine-depleted mice induces persistent and long-lasting ERK activation in both dMSNs and ChIs

In order to define ERK activation profile in dMSNs and ChIs in LID, we first applied different protocols of l-DOPA administration in 6-OHDA lesioned mice: (1) 1 day acute protocol; (2) 9 days dose-escalating protocol and (3) 9 days protocol followed by LID maintenance protocol of either 2 or 4 weeks (Fig.1A). In the acute protocol, mice were injected with 3 mg/kg of l-DOPA and 20 min after they were perfused to detect molecular changes without scoring AIMs. In the 9 days protocol, mice developed a clear dyskinetic state that remained stable over 4 weeks (Fig.1B). A correlation was found between ERK1/2 immunoreactivity (pERK levels) in dMSNs and axial, limb and orolingual (ALO) AIMs at 9 days but not 4 weeks (Fig.1C). In fact, ERK activity in the dMSNs became maximal already after acute l-DOPA challenge and remained high also after 9 days of treatment in concomitance to an exacerbated dyskinetic behavior. However, we detected only a small but significant decrease after 9 days of treatment in comparison to the acute challenge that was not observed after the intermittent administration of l-DOPA in the maintenance protocol (Fig.1D–G). In contrast, in ChIs, ERK activation was evident only after 9 days of l-DOPA administration (Fig.1H and I). The results thus far indicate that ERK activity remains high in dMSNs in response to a prolonged but intermittent l-DOPA administration and appears also in ChIs only after a repeated challenge with l-DOPA.

Figure 1

LID profile and ERK activation in dMSNs and ChIs during chronic l-DOPA administration. (A) Experimental design 1: 2 weeks after 6-OHDA lesion, different groups of C57BL/6 mice were treated with l-DOPA: one group (n = 8) received an acute dose (3 mg/kg, 1 day) of l-DOPA (hereinafter 1 day) and was perfused 20 min thereafter without performing AIMs; one group (n = 11) was treated with a 9 days protocol (hereinafter 9 days) (days 1–3, 3 mg; days 4–6, 6 mg; days 6–9, 12 mg/kg) and lastly, two groups after the 9 days protocol underwent a dyskinesia maintenance protocol consisting of twice a week injections (12 mg/kg l-DOPA plus 12 mg/kg of benserazide) for either 2 weeks (n = 11) or 4 weeks (n = 11). For each time point, a small group of mice (n = 3) was treated with saline as internal control for subsequent postmortem analysis. (B) Temporal profiles of all groups throughout the 9 days, 2 weeks, and 4 weeks protocol expressed as a sum of Axial, Limb, and Orolingual AIMs (ALO AIMs) are shown. No differences in dyskinesia expression levels were found amongst groups in either the initial 9 days period (P = 0.259) or throughout the maintenance period (2 weeks vs. first 2 weeks of 4 weeks P = 0.577; vs. last 2 weeks of 4 weeks P = 0.238). (C) Pearson’s correlation index revealed a positive, linear correlation between pERK levels in dMSNs and AIMs in the 9 days, but not in the 4 weeks group (9 days: r = 0.802, P < 0.05; 4 weeks: r = 0.084, P = 0.878). (D) Representative micrographs of ERK1/2 immunoreactivity (pERK) in the dorsolateral part of the striatum after completion of different l-DOPA treatments (1 day, 9 days and maintenance protocol). Scale bar 30 μm. (E) Quantification of pERK levels observed in all l-DOPA-treated groups with statistical difference between 1 day and 9 days protocol (one way ANOVA, treatment effect F(4, 38) = 20.884, P < 0.0001, Tukey’s post hoc 1 day vs. 9 days P < 0.05). (F) Immunofluorescent micrographs of ERK activation in the dMSNs: pERK (green), DR1 (red) and NeuN (blue) cells in dorsal striatal. Scale bar 20 μm. Small insets of split channels of pERK and DR1-positive cells are also reported. (G) Quantification of pERK-positive cells in D1R-positive and -negative cells, with statistical difference after 1 day and 9 days protocol (one way ANOVA, treatment effect F(3, 34) = 8.2302, P < 0.001 Tukey’s post hoc 1 day vs. 9 days P < 0.001). (H) Immunofluorescent micrographs of ERK activation in cholinergic interneurons: pERK (green), ChAT (red) and NeuN (blue) cells. Scale bar 30 μm. Small insets of split channels of ChAT and pERK-positive cells are also reported. (I) A raise of pERK levels was observed in ChIs after 9 days (Kruskal–Wallis Test P < 0.0001; Mann–Whitney Test 1 day vs. 9 days, 2 weeks and 4 weeks P < 0.01) and persisted through the prolonged administration of l-DOPA (Mann–Whitney Test 9 days vs. 2 weeks P = 0.124, 9 days vs. 4 weeks P = 0.787, 2 weeks vs. 4 weeks P = 0.916). *P < 0.05, **P < 0.0.01, ***P < 0.001. LID, l-DOPA-induced dyskinesia; l-DOPA, l-3,4-dihydroxyphenylalanine; 6-OHDA, 6-hydroxydopamine; AIMs, abnormal involuntary movements.

Ras-GRF1 controls ERK signaling in dMSNs but not in ChIs

As described earlier, Ras-GRF1 KO animals are significantly less dyskinetic than controls and exhibit lower levels of ERK activation and FosB/ΔFosB expression. This evidence was not due to a different expression of Ras-GRF1 in striatonigral MSNs but to the specific engagement of Ras-GRF1 by D1 receptors.11,28

Thus, we next examined whether Ras-GRF1 may also regulate ERK signaling in ChIs using the Ras-GRF1 KO model. We found that pERK was significantly reduced in dMSNs (Fig.2A and B) but not in ChIs (Fig.2E and F) of Ras-GRF1 KO animals upon a 4 weeks dyskinesia maintenance protocol. Similarly, downstream phosphorylation of acetyl-histone H3 (AcH3), also known to be associated with LID but never measured in ChIs4,25 was only attenuated in dMSNs (Fig.2C and D) but not in ChIs (Fig.2G and H). These data strongly indicate that Ras-GRF1 does not control ERK activity in ChIs of the striatum, suggesting that other factors may regulate ERK signaling in these cells.

Figure 2

Reduced ERK signaling in dMSNs but not ChIs population in Ras-GRF1 KO mice. (A) Immunofluorescent photomicrographs of ERK activation in the dMSNs: pERK (green), DR1 (red) and NeuN (blue) cells in dorsal striatal area of WT and Ras-GRF1 KO mice after the 4 weeks regimen of l-DOPA. Scale bar 20 μm. (B) pERK-positive cells colocalizing with D1R-positive cells were counted, with statistical difference between WT and Ras-GRF1 KO mice (Independent-samples t-test, P < 0.001). (C) Immunofluorescent micrographs of acetyl-histone H3 activation (pAcH3) in the dMSNs: pAcH3 (green), DR1 (red) and NeuN (blue) cells in dorsal striatal area of WT and Ras-GRF1 KO mice after the 4 weeks protocol of l-DOPA. Scale bar 20 μm. (D) pAcH3-positive cells colocalizing with D1R-positive cells were counted, with statistical difference between WT and Ras-GRF1 KO mice (Independent-samples t-test, P < 0.05). (E) Immunofluorescent micrographs of ERK activation in ChIs: pERK (green), ChAT (red) and NeuN (blue) cells in dorsal striatal area of WT and Ras-GRF1 KO mice after 4 weeks regimen of l-DOPA. Scale bar 20 μm. (F) pERK-positive cells colocalizing with ChAT-positive cells were observed, without statistical difference between WT and Ras-GRF1 KO mice (Independent-samples t-test, P = 0.415). (G) Immunofluorescent micrographs of AcH3 activation in the ChIs: pAcH3 (green), ChAT (red) and NeuN (blue) cells in dorsal striatal area of WT and Ras-GRF1 KO mice after the 4 weeks regimen of l-DOPA. Scale bar 20 μm. (H) pAcH3-positive cells colocalizing with ChAT-positive cells were observed, without statistical difference between WT and Ras-GRF1 KO mice (Independent-samples t-test, P = 0.295). Data are mean ± SEM of 8–10 animals for each group. *P < 0.05, ***P < 0.001.

Validation of LV-assisted gene knockdown of Ras-GRF1 and Ras-GRF2

Ras-GRF2 is a close homolog of Ras-GRF1, and is expressed, although to a lesser extent, in the striatum.23 To address the possibility that Ras-GRF2 could play a role in LID, we used a sophisticated viral-mediated approach to knockdown both Ras-GRF1 and Ras-GRF2 gene in striatal cells, in vitro and in vivo. Expression cassettes containing 3 distinct Ras-GRF1 and Ras-GRF2-specific short hairpin RNAs (shRNA) were transferred into green fluorescent protein (GFP)-tagged LV in order to test their inhibitory potential in neurons on p140Ras-GRF1 and p135Ras-GRF2, respectively. We prepared in vitro neuronal cultures from newborn (P1) wild type mice, which were subsequently infected with equal amounts of LV. Then, cells were processed by Western blot analysis using specific antibodies against Ras-GRF1 and Ras-GRF2 proteins (Fig.3A and C). All selected sequences were able to reduce the expression of Ras-GRF proteins in comparison with control sequence (hereinafter sh-Ctr). We selected the most effective sequences, that is, sequence 1 of LV-shRNA/Ras-GRF1 (hereinafter sh-Ras-GRF1) and sequence 3 of LV-shRNA/Ras-GRF2 (hereinafter sh-Ras-GRF2), for further in vivo experiments (Fig.3B and D respectively). To confirm in vivo efficacy of the selected shRNA sequences, we performed unilateral stereotaxic injections of sh-Ras-GRF1, sh-Ras-GRF2 or sh-Ctr in the motor striatum of wild type mice, and 3 weeks later, protein content in dorsal striata was determined by Western blot (Fig.3E). Expression level of p140Ras-GRF1 was reduced by sh-Ras-GRF1 with no alterations in Ras-GRF2 levels. Similarly, sh-Ras-GRF2 significantly decreased the expression level of p135Ras-GRF2 without affecting Ras-GRF1 levels (Fig.3F). The observed in vivo reduction in p140 and p135 was likely to be an underestimation due to the significant presence of non-infected cells in the tissue samples (see also Fig. 7). Importantly, no changes in either ERK1/2 protein levels or basal phosphorylation were observed (Fig.3G and H).

Figure 3

Ras-GRFs-specific gene silencing in vitro and in vivo. Ras-GRF1 and Ras-GRF2 residual protein expression levels in neurons were determined by Western blotting 48 h after LVs infection with three different shRNA cassettes (Seq 3, 2, 1 for (A) Ras-GRF1 and Seq 1,2,3 for (C) Ras-GRF2) or the relative control sequence (Seq Ctr). Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as a loading control. (B and D) Quantification of three independent experiments showed that all sequences were able to reduce the expression of Ras-GRF proteins. (E) For in vivo delivery, Seq 1 of sh-Ras-GRF1 and Seq 3 of sh-Ras-GRF2 were injected in striatum of wild type mice. Representative immunoblots of p140 Ras-GRF1 and p135 Ras-GRF2 are shown. (F) Densitometry analysis demonstrated that p140 Ras-GRF1 was reduced by sh-Ras-GRF1 (one way ANOVA, main effect F(2, 9) = 10.517, P < 0.01, Bonferroni’s post hoc sh-Ctr vs. sh-Ras-GRF1 P < 0.01) without alterations in Ras-GRF2 protein levels (Bonferroni’s post hoc sh-Ctr vs. sh-Ras-GRF2 P = 1.000). Similarly, p135 Ras-GRF2 was reduced by sh-Ras-GRF2 expression (main effect F(2, 9) = 5.45, P < 0.05, Bonferroni’s post hoc sh-Ctr vs. sh-Ras-GRF2 P < 0.05) with no variation in Ras-GRF1 protein levels (Bonferroni’s post hoc sh-Ctr vs. sh-Ras-GRF1 P = 0.282). (G) Immunoreactivity to both anti-phospho-ERK and anti-ERK antibodies in striata of sh-Ras-GRF1, sh-Ras-GRF2 and sh-Ctr injected mice. (H—I) Quantification of normalized p44 and p42 bands intensities revealed no changes in either protein levels or basal phosphorylation state of ERK1/2. *Non-specific band. *P < 0.05, **P < 0.0.01.

Ras-GRF1 and Ras-GRF2 gene knockdown does not interfere with basal motor behavior in 6-OHDA lesioned mice

Next, we asked whether Ras-GRF1 or Ras-GRF2 knockdown could affect the course of 6-OHDA lesioning. Mice were hemilesioned with 6-OHDA then sorted by comparable levels of rotational behavior 2 weeks later, and finally unilaterally injected with LVs in the dorsal striatum. Three weeks later, lesion-induced motor deficits were examined using spontaneous rotations, rotarod, and drag test28,33 (Fig.4A). All four shRNA groups showed a similar number of spontaneous ipsilateral rotations during a 10 min session, suggesting an equivalent extent of 6-OHDA damage, regardless of the injected short hairpins (Fig.4B). Motor abilities were also evaluated on the rotarod during a 5-days training protocol. Latency to fall was similar in sh-Ras-GRF1, sh-Ras-GRF2 and sh-Ctr animals at day 1, and significantly increased to the same extent in all groups at day 5, indicating a significant effect of training (Fig.4C). A marked reduction in stepping activity at the contralateral forepaw in the drag test was observed in sh-Ras-GRF1, sh-Ras-GRF2 and sh-Ctr mice at both day 1 and day 5 of training, demonstrating a uniform effect of 6-OHDA across groups (Fig.4D).

Figure 4

Striatal-specific gene knockdown of Ras-GRF1 or Ras-GRF2 does not impact on basal motor behavior of 6-OHDA lesioned mice. (A) Experimental design 2: 2 weeks after 6-OHDA lesion mice performed spontaneous rotations test and were split in four groups, injected with sh-Ras-GRF1, sh-Ras-GRF2 and sh-Ctr, respectively. Three weeks post-shRNA injection, mice performed spontaneous rotations, rotarod and drag test. (B) Spontaneous full rotations were found equivalent among the experimental groups (one way ANOVA, F(3, 35) = 0.282, P = 0.838). (C) Motor learning during 5 days on the rotarod was measured. A significant effect of training but not of treatment was observed in sh-Ras-GRF1 and sh-Ctr groups (Repeated measures ANOVA, F(1, 17) = 56.336, P < 0.0001). Similarly, sh-Ras-GRF2 and sh-Ctr mice displayed only a significant effect of training (Repeated measures ANOVA, F(1, 17) = 30.073, P < 0.0001). (D) Akinesia was evaluated using drag test. On day 1, the number of steps of the contralateral forepaw was significantly less than the one performed with the ipsilateral forepaw in all experimental groups (Paired-samples t-test, sh-Ras-GRF1: t(8) = −11.300, P < 0.0001; sh-Ctr: t(9) = −8.899, P < 0.0001; sh-Ras-GRF2: t(9) = −5.349, P < 0.001; sh-Ctr: t(9) = −5.365, P < 0.0001). On the subsequent test on day 5, the number of contralateral steps continued to be significantly lower than the one of ipsilateral steps in all groups (Paired-samples t-test, sh-Ras-GRF1: t(8) = −5.635, P = 0.001; sh-Ctr: t(9) = −12.825, P < 0.0001; sh-GRF2: t(9) = 5.312, P < 0.001; sh-Ctr: t(9) = 4.589, P < 0.001). Data are mean ± SEM of 9–11 animals for each group. ***P < 0.001, ****P < 0.0001.

Ras-GRF1 but not Ras-GRF2 striatal silencing attenuates LID

To evaluate whether specific knockdown of striatal Ras-GRF1 and Ras-GRF2 would affect behavioral responses to repeated l-DOPA treatment, we applied the 9 days l-DOPA protocol.28 A gradual development of dyskinesia in both sh-Ras-GRF1 and sh-Ctr injected animals was observed. However, LID scores were significantly lower in sh-Ras-GRF1 mice than in controls (Fig.5A and B). Further analysis indicated that in sh-Ras-GRF1 mice all AIMs subtypes were weakened, confirming the prominent role of Ras-GRF1 in LID development (Fig.5C). On the contrary, time course and intensity of the l-DOPA response in sh-Ras-GRF2 mice were similar to those observed in control animals, ruling out the involvement of Ras-GRF2 in LID formation (Fig.5D–F).

Figure 5

Striatal downregulation of Ras-GRF1 but not Ras-GRF2 is able to attenuate LID expression. Four weeks after sh-injection, 9 days l-DOPA protocol was administered and AIMs were daily scored. (A) A gradual development of dyskinesia was observed (repeated measure ANOVA, time effect, F(2.182, 47.999) = 29.703, ε = 0.273 P < 0.0001). However, sh-Ras-GRF1 mice developed dyskinesia to a lesser extent in comparison to their controls (repeated measure ANOVA, shRNA effect, F(1, 22) = 6.944, P < 0.05), with the difference becoming evident on day 4 of l-DOPA exposure and persisting till day 9 (P < 0.05). (B) The analysis of the area under curve (AUC) of AIMs score confirmed this reducing effect of sh-Ras-GRF1 (mean ± SEM) (Independent-samples t-test: t(22) = 2.618, P < 0.05). (C) Sh-Ras-GRF1 expression lessened the Limb and Axial components of AIMs in comparison to sh-Ctr mice (ANOVA, shRNA effect, P < 0.05) while the Orolingual subtype was found only almost significantly different from sh-Ctr animals (ANOVA, shRNA effect, P = 0.06). (D) Sh-Ras-GRF2 injected mice showed rising dyskinesia levels over time but the intensity displayed was equal to sh-Ctr mice (repeated measure ANOVA, shRNA effect, F(1, 20) = 0.497, P = 0.489). (E) Analysis of AUC (mean ± SEM) (Independent-samples t-test: t(20) = −0.773, P = 0.448) and (F) individual components of ALO AIMs in Ras-GRF2 silenced mice were similar to those observed in sh-Ctr animals (ANOVA, shRNA effect, P > 0.05). Data are mean ± SEM of 9–11 animals for each group. *P < 0.05.

On day 10, mice were challenged with a final dose of 12 mg/kg of l-DOPA and transcardially perfused 20 min later in order to analyze the involvement of Ras-GRF1 and Ras-GRF2 in the downstream signaling associated with the severity of AIMs. Only sh-Ras-GRF1 mice showed a significant reduction in ERK activation and FosB/ΔFosB immunoreactivity (Fig.6A, B, G and H) whilst knockdown of Ras-GRF2 did not affect these molecular changes (Fig.6D, E, J and K). Furthermore, positive correlations between the severity of AIMs and ERK phosphorylation or FosB/ΔFosB accumulation were observed (Fig.6C, F, I, and L).

Figure 6

Striatal-specific gene knockdown of Ras-GRF1 but not Ras-GRF2 reduces ERK phosphorylation and FosB/ΔFosB accumulation associated with LID. (A) Photomicrographs of ERK1/2 immunoreactivity in the dorsal striatum of sh-Ctr and sh-Ras-GRF1 groups after 9 days l-DOPA or Saline treatment. (B) Quantification of pERK-positive cells revealed a significant effect of l-DOPA in both shRNA injected groups in comparison to saline (two way ANOVA, treatment effect, F(1, 29) = 41.890, P < 0.0001). However, a significant reduction of pERK was observed in sh-Ras-GRF1 mice (two way ANOVA, interaction genotype × treatment effect, F(1, 29) = 6.319, P < 0.05). (C) Pearson’s correlation index revealed a strong, linear correlation between ERK levels and AIMS in both experimental groups (sh-Ctr: r = 0.935, P < 0.001; sh-Ras-GRF1: r = 0.914, P < 0.0001). (D) Photomicrographs of pERK in sh-Ctr and sh-Ras-GRF2 groups after 9 days l-DOPA protocol. (E) Quantification of pERK-positive cells revealed a significant effect of l-DOPA (two way ANOVA, treatment effect, F(1, 26) = 84.954, P < 0.0001) without difference between sh-Ctr and sh-Ras-GRF2 mice (two way ANOVA, shRNA effect, F(1, 26) = 0.026, P = 0.874). (F) Pearson’s correlation index revealed a positive, linear correlation between ERK levels and AIMS in both experimental groups (sh-Ctr: r = 0.850, P < 0.001; sh-Ras-GRF2: r = 0.895, P < 0.0001). (G) Photomicrographs of FosB/ΔFos Bimmunoreactive cells in the dorsal striatum of sh-Ctr and sh-Ras-GRF1 groups after 9 days l-DOPA or Saline treatment. (H) Ras-GRF1 knockdown significantly reduced FosB/ΔFosB accumulation (two way ANOVA, interaction genotype × treatment effect, F(1, 29) = 4.402, P < 0.05). (I) A positive, linear correlation was found between FosB levels and AIMs in both experimental groups (sh-Ctr: r = 0.931, P < 0.001; sh-Ras-GRF1: r = 0.921, P < 0.0001). (J) Photomicrographs of FosB/ΔFosB-immunoreactive cells in the dorsal striatum of sh-Ctr and sh-Ras-GRF2 groups after 9 days l-DOPA or Saline treatment. (K) Equivalent levels of FosB/ΔFosB accumulation were induced by l-DOPA treatment (two way ANOVA, treatment effect, F(1, 26) = 83.857, P < 0.0001) without difference between sh-Ras-GRF2 mice and their controls (two way ANOVA, shRNA effect, F(1, 26) = 0.353, P = 0.557). (L) A positive, linear correlation was found between FosB levels and AIMs in both experimental groups (sh-Ctr: r = 0.850, P < 0.001; sh-Ras-GRF2: r = 0.895, P < 0.0001). Scale bar 30 μm. Data are mean ± SEM of 9–11 animals for each group. *P < 0.05.

To confirm that reduction in ERK signaling was specific to silenced cells we examined the involvement of Ras-GRF1 and Ras-GRF2 in the phosphorylation of AcH3. Triple-labeling experiments showed that reduction in AcH3 phosphorylation was only restricted to sh-Ras-GRF1 silenced cells (Fig.7A and B). Indeed, no differences were found in non-infected cells of sh-Ctr and sh-Ras-GRF1 mice though the percentage of infection was similar in both groups and accounted for ∼35% of cells (Fig.7C and D). Not surprisingly, sh-Ras-GRF2 silencing did not alter AcH3 activation in l-DOPA-treated mice (Fig.7E–H). Altogether, these data not only confirm a prominent role of Ras-GRF1 in LID formation and ERK regulation in response to chronic l-DOPA but also rule out an involvement of Ras-GRF2 in this process.

Figure 7

Silencing of Ras-GRF1 in the striatum dampens the l-DOPA-induced phospho-acetylated histone H3. (A) Photomicrographs of pAcH3 (red), shRNA constructs (green) and NeuN (blue) in the dorsal striata of sh-Ras-GRF1 mice after the 9 days l-DOPA protocol. (B) A significant reduction in pAcH3 was found in successfully silenced cells of mice treated with sh-Ras-GRF1 in comparison to sh-Ctr injected mice (two way ANOVA, shRNA effect, F(1, 20) = 7.765, **P < 0.001). (C) PAcH3 quantification among non-infected neurons (sh-cells- no GFP) was instead found equivalent in l-DOPA-treated mice (two way ANOVA, shRNA effect, F(1, 20) = 0.01, P = 0.980). (D) Percentage of sh-infected cells counted on total NeuN-positive neurons was found comparable in both experimental groups (two way ANOVA, shRNA infection effect, F(1, 20) = 0.666, P = 0.424). (E) Equivalent photomicrographs of pAcH3 in sh-Ras-GRF2 mice after the 9 days l-DOPA protocol. (F) Identical activation of AcH3 was found in successfully silenced cells of mice treated with sh-Ras-GRF2 in comparison to sh-Ctr injected mice (two way ANOVA, shRNA effect, F(1, 20) = 0.024, P = 0.878). (G) PAcH3 quantification among non infected neurons (sh-cells- no GFP) was found equivalent in l-DOPA-treated mice (two way ANOVA, shRNA effect, F(1, 20) = 0.342, P = 0.565). (H) Percentage of sh-infected cells counted on total NeuN-positive neurons was found comparable in both experimental groups (two way ANOVA, shRNA infection effect, F(1, 20) = 0.010, P = 0.922). Scale bar, 50 μm. Data are mean ± SEM of 10–11 animals for each group ** P < 0.001.

Ras-GRF1 but not Ras-GRF2 striatal knockdown reduces l-DOPA-induced GABA release in the substantia nigra

The alterations caused by Ras-GRF1 gene knockdown on behavioral and cellular correlates of LID prompted us to determine its possible impact on substantia nigra pars reticulata (SNr) and globus pallidus (GP) neurochemical changes after l-DOPA administration. Indeed, previous work showed that an increase in γ-Aminobutyric acid (GABA) and glutamate (GLU) levels in the SNr but not in the GP correlated with the intensity of dyskinesia.34,37 After the 9 days protocol, mice underwent dual probe microdialysis implantation and received a challenge of l-DOPA 24 h later: GABA and GLU release was monitored simultaneously with AIMs scoring (Fig.8A). Sh-Ras-GRF1 mice showed less severe dyskinesia in response to l-DOPA than controls (Fig.8B), this effect being paralleled by a milder increase in GABA levels in SNr (Fig.8D). Conversely, sh-Ras-GRF2 animals were as dyskinetic as sh-Ctr mice (Fig.8C) and, consistently, no difference in the extent of the GABA levels rise was observed between the two groups (Fig.8E). Moreover, GLU overflow in the SNr of sh-Ras-GRF1 mice did not show any increase over time, contrary to what can be observed in sh-Ctr mice (Fig.8F). On the other hand, both sh-Ras-GRF2 mice and sh-Ctr animals showed a significant increase in the GLU overflow over time (Fig.8G). Finally, no changes in either GABA or GLU levels were detected in the GP of any group of mice in response to l-DOPA (Fig.8H–K).

Figure 8

Ras-GRF1 reduces l-DOPA-induced GABA release in the substantia nigra. (A) Experimental design 3: after the 9 days l-DOPA protocol, sh-Ctr, sh-Ras-GRF1 and sh-Ras-GRF2 dyskinetic mice underwent surgery for microdialysis probe implantation, and were challenged with l-DOPA (12 mg/kg, i.p.) or saline 24 h later. Three baseline samples were collected before the time point 0, indicated with an arrow, corresponding to l-DOPA i.p. injection. ALO AIMs were scored every 20 min over 120 min after l-DOPA administration. (B) A significant reduction in LID was observed in l-DOPA-treated sh-Ras-GRF1 in comparison to sh-Ctr mice (two way ANOVA, shRNA effect F(1, 60) = 41.14, P < 0.05). (C) Comparable levels of dyskinesia were observed in sh-Ras-GRF2 and sh-Ctr mice (two-way ANOVA, shRNA effect F(1, 114) = 3.445, P = 0.066). (D) A significant reduction in GABA levels in SNr of l-DOPA-treated sh-Ras-GRF1 mice compared to sh-Ctr animals was observed (two-way ANOVA, shRNA effect F(1, 81) = 29.82, P < 0.0001). (E) Equivalent levels of GABA in SNr were displayed by l-DOPA-treated sh-Ras-GRF2 and sh-Ctr mice (two-way ANOVA, shRNA effect F(1, 126) = 1.927, P = 0.1675). (F) A reduction in GLU levels in SNr of l-DOPA injected sh-Ras-GRF1 mice compared to sh-Ctr group was found (two-way ANOVA, shRNA effect F(1, 81) = 4.074, P < 0.05). (G) In SNr of sh-Ras-GRF2 groups, GLU levels were equally altered by l-DOPA treatment (two-way ANOVA, shRNA effect F(1, 108) = 0.7117, P = 0.4008). No relevant differences were observed among groups for GABA (H and I) and GLU (J and K) levels in GP. Data are mean ± SEM of 6–12 animals. *P < 0.05.

These data clearly indicate that Ras-GRF1 inhibition and the associated attenuation of LID are linked to significant neurochemical changes in the substantia nigra.

Gene knockdown of Ras-GRF1 provides long-term beneficial effects on LID

Finally, we investigated whether Ras-GRF1 gene knockdown could either attenuate LID in a prolonged l-DOPA protocol or could affect already established dyskinetic state. First, we applied the 9 days protocol followed by twice a week l-DOPA treatment for 7 weeks (Fig.9A): as expected, sh-Ras-GRF1 silenced animals showed a lower progression of LID that was maintained throughout the experiment (Fig.9B and C). The beneficial effect was confirmed on the last day of treatment showing a reduced and shortened ALO profile (Fig.9D). Second, we treated 6-OHDA-lesioned mice with l-DOPA accordingly to our 9 days protocol. Then mice were equally balanced in two groups following an unbiased design and submitted to intrastriatal LVs injection (Fig.9E). After a short recovery period mice were then challenged with 12 mg/kg of l-DOPA twice a week for 4 weeks. The dyskinetic profile before LVs injection was the same in both groups. However, 3 weeks post-LVs injection, AIMs were significantly attenuated in sh-Ras-GRF1 mice and were comparable to the extent observed in the initial phase of the treatment (Fig.9F–H).

Figure 9

Ras-GRF1 inhibition reduces appearance of dyskinesia over prolonged l-DOPA treatment and also alleviates its expression. (A) Experimental design 4: 3 week post-sh-injection, mice underwent first the 9 days l-DOPA protocol followed by 7 weeks maintenance protocol consisting of twice a week injections (12 mg/kg l-DOPA plus benserazide). (B) Daily scoring of AIMs during the 9 days protocol showed a significant reduction in dyskinetic behavior in sh-Ras-GRF1 silenced mice over time (repeated measure ANOVA, shRNA effect, F(1, 22) = 15.163, P < 0.001, time effect, F(8, 176) = 34.263, P < 0.0001; Independent sample T-test, from day 4 to day 9 P < 0.01). This effect persisted over 7 weeks of treatment (repeated measure ANOVA, shRNA effect, F(1, 22) = 41.585, P < 0.001, independent sample T-test, all days of maintenance P < 0.001). (C) The analysis of the area under curve (AUC) of AIMs score during the maintenance protocol confirmed the protective effect of sh-Ras-GRF1 (mean ± SEM) (Independent-samples t-test: t(22) = 5.360, P < 0.001). (D) Time profile of the sum of ALO AIMs on the last day (day 56) of treatment. A significant effect of shRNA, F(1, 22) = 19.412, P < 0.001 and time, F(6, 132) = 141.383, P < 0.001 was found. (E) Experimental design 5: after the 9 days l-DOPA protocol, dyskinetic mice were divided in two equivalent groups and subsequently underwent surgery for LVs-injection in the dorsal striatum. Few days later, mice were subjected to a dyskinesia maintenance protocol of 4 weeks (12 mg/kg of l-DOPA, twice a week). (F) the expression of LID upon 9 days protocol was equivalent in both groups before LVs injection. However, during the maintenance protocol, sh-Ras-GRF1 showed a significantly reduction in dyskinetic behavior (repeated measure ANOVA, shRNA effect, F(1, 22) = 6.7733, P < 0.05, independent sample T-test, from day 27 to day 41 P < 0.05. (G) The analysis of the area under curve (AUC) of AIMs score confirmed the antidyskinetic effect of sh-Ras-GRF1 (mean ± SEM) (independent-samples t-test: t(22) = 2.564, P < 0.05). (H) Time profile of the sum of ALO AIMs on the last day (day 41) of treatment. A significant effect of shRNA, (F(1, 22) = 5.991, P < 0.05) and time, (F(7, 154) = 42.073, P < 0.05) was found. Data are mean ± SEM of 10–12 animals for each group. *P < 0.05, **P < 0.01, ***P < 0.001.

These data demonstrated that Ras-GRF1 inhibition not only provides a long-term beneficial therapeutic effect in attenuating LID but also can significantly ameliorate a severe dyskinetic state.

Discussion

The Ras-ERK cascade appears to be particularly promising for the treatment of LID since its involvement has been confirmed by various independent studies. However, ERK signaling is not only involved in plasticity processes, like MSNs dysregulation in LID, but also in cell survival mechanisms. Therefore, identification of key components of this signaling pathway selectively involved in LID would be highly advantageous since the repeated use of available MEK inhibitors in PD patients may result in intolerable side effects, including an exacerbation of dopamine cell loss in the SN.38

Here we found that an acute dose of l-DOPA was able to trigger maximal ERK activation in dMSNs and dyskinesia appearance, while a prolonged and intermittent drug administration resulted in a sustained ERK activity in dMSNs and a stable dyskinetic behavior. Differently from Ding and colleagues, we observed an engagement of the ChIs upon chronic l-DOPA treatment not paralleled by a concomitant ERK downregulation in dMSNs.29 These discrepancies are likely due to a different dose and regimen of l-DOPA administration. Moreover, we did not observe a reduction in ERK activation after 4 weeks maintenance protocol as reported by Santini et al., in dyskinetic NHPs after 3 months of l-DOPA administration likely for the same reasons.7 Nevertheless, we considered the contribution of ERK signaling in these two cell populations. Thus, we explored the involvement of Ras-GRF1 in MSNs and ChIs activity using the Ras-GRF1-deficient mice. Surprisingly, we found that Ras-GRF1 controls ERK activity exclusively in dMSNs. Indeed in dopamine-depleted striatum, untreated or acutely exposed to l-DOPA, ERK signaling in ChIs is not altered, but increases only upon repeated l-DOPA administration, in a Ras-GRF1-independent manner. This finding may explain the presence of residual dyskinetic symptoms in Ras-GRF1-deficient mice, possibly due to the intact ERK activity in ChIs. Thus, unlike dMSNs, where ERK activation upon D1 and GLU receptor interplay requires Ras-GRF1,23 in cholinergic cells the route of activation may be distinct and the underlying mechanisms are yet to be elucidated.

Regardless of the exact mechanism controlling ERK activity in ChIs, the reduction in ERK signaling in dMSNs of Ras-GRF1 knockout and knockdown mice not only correlates well with LID attenuation, but also confirms that blockade of Ras-GRF1 mediated signaling in the dorsostriatal dMSNs provides a significant antidyskinetic effect. However, the reduction in ERK activity observed in Ras-GRF1-deficient dMSNs is not complete, strongly suggesting the involvement of other factors. We previously showed that a suboptimal dose of a MEK inhibitor, SL327, could further reduce AIMs (up to 80%) in Ras-GRF1 KO mice indicating that LID symptoms are largely ERK dependent.28 For this reason, we hypothesized the involvement of a close Ras-GRF1 homolog, Ras-GRF2. Unfortunately, our results provide compelling behavioral, histochemical and neurochemical evidence excluding its involvement in this process. This finding is in line with the idea that despite their homology, Ras-GRF1 and Ras-GRF2 display non-overlapping functional roles.39

In addition to the molecular control of ERK at the input level, that is, through Ras-GRF1 or other Ras-GEFs, other downstream mechanisms have been involved in LID. The first is the D1R mediated cAMP-PKA-DARPP32 cascade whose activation leads to ERK enhancement.4,22 The second, attractive pathway relies on the protein tyrosine phosphatase Shp-2 that also acts downstream to D1R to control ERK activity, both in normal and dyskinetic states.9,40

Future work will assess the potential contribution of these two additional pathways in the Ras-GRF1-independent regulation of ERK activity in LID. Finally, it is important to note that additional levels of ERK, histone H3 and/or ΔFosB regulation could be achieved through other mechanisms, either involving mGluR5/PLC/PKC modulation, cAMP/PKA/DREAM-dependent transcriptional repression or NO/cGMP signaling as recently described.10,41,42

One final important point refers to the poorly explored relationship between changes in striatal cell signaling and neurotransmitter release. Previous in vivo microdialysis studies in dyskinetic rats and mice have shown that AIMs appearance following a dyskinetogenic dose of l-DOPA is associated with an increase in GABA and GLU levels in SNr.34,37,43 These changes are consistent with dMSNs hyperactivity since they are prevented by intrastriatal perfusion with a D1 receptor antagonist.43 In line with these findings, silencing Ras-GRF1 in dMSNs caused LID attenuation and no surge in nigral amino acid levels upon l-DOPA challenge. Interestingly, however, we found that also perfusion with a D1 receptor antagonist in the SNr of dyskinetic rats provides attenuation of both LID and the accompanying increase in nigral GABA.43 Consistently, D1 receptor agonist induced [3H]-GABA release was found to be upregulated in ex vivo nigral slices obtained from dyskinetic rats.44 This suggests that also nigral D1 receptors contribute to LID, and that the l-DOPA induced increase of nigral GABA release might be achieved through activation of both striatal postsynaptic D1 receptors located on dMSNs and nigral D1 receptors located on striato-nigral GABA afferents. Silencing Ras-GRF1 in dMSNs would therefore prevent up-regulation of both striatal and nigral D1 receptors, making l-DOPA unable to elevate nigral GABA release.

LID is a chronic disorder and, once established, it will be affecting PD patients throughout their life. Thus, in order to translate our initial findings in clinically relevant observations, we investigated two important aspects associated with the Ras-GRF1 role in this disorder. First, we confirmed that Ras-GRF1 attenuation ameliorates LID symptoms also in the maintenance phase, up to 7 weeks of chronic l-DOPA treatment. Second, we found that already established dyskinesia could be significantly attenuated by Ras-GRF1 knockdown. These data strongly support the therapeutic potential of Ras-GRF1 modulation and are in line with our previous work on the NHP model.28

Altogether, our work provides important insights for a better interpretation of the molecular mechanisms underlying LID and also relevant information for devising successful therapeutic approaches based on intracellular signaling inhibition in the striatum.