mGlu1 Receptors Monopolize the Synaptic Control of Cerebellar Purkinje Cells by Epigenetically Down-Regulating mGlu5 Receptors

In cerebellar Purkinje cells (PCs) type-1 metabotropic glutamate (mGlu1) receptors play a key role in motor learning and drive the refinement of synaptic innervation during postnatal development. The cognate mGlu5 receptor is absent in mature PCs and shows low expression levels in the adult cerebellar cortex. Here we found that mGlu5 receptors were heavily expressed by PCs in the early postnatal life, when mGlu1α receptors were barely detectable. The developmental decline of mGlu5 receptors coincided with the appearance of mGlu1α receptors in PCs, and both processes were associated with specular changes in CpG methylation in the corresponding gene promoters. It was the mGlu1 receptor that drove the elimination of mGlu5 receptors from PCs, as shown by data obtained with conditional mGlu1α receptor knockout mice and with targeted pharmacological treatments during critical developmental time windows. The suppressing activity of mGlu1 receptors on mGlu5 receptor was maintained in mature PCs, suggesting that expression of mGlu1α and mGlu5 receptors is mutually exclusive in PCs. These findings add complexity to the the finely tuned mechanisms that regulate PC biology during development and in the adult life and lay the groundwork for an in-depth analysis of the role played by mGlu5 receptors in PC maturation.

impairment of long-term depression (LTD) at parallel fiber-PC synapses associated with a profound defect in the conditioned eyeblink reflex and motor coordination. All these defects are rescued by selective re-introduction of mGlu1α receptors in PCs 11,12 . Interestingly, abnormalities in the expression/activity of mGlu1 receptors or downstream signaling molecules in PCs have been found in genetic mouse models of type-1, -2 -3, -5, and -14 spinocerebellar ataxias [13][14][15][16] , in ataxic moonwalker mutant mice 17 , in mice subjected to experimental autoimmune encephalomyelitis, and in autoptic samples from individuals affected by multiple sclerosis 18 . Neutralizing autoantibodies directed against mGlu1 receptors have been detected in patients with autoimmune ataxia [19][20][21] . All these findings suggest that mGlu1 receptors might be targeted by therapeutic intervention in cerebellar disorders 7 .
As opposed to mGlu1 receptors, mGlu5 receptors are virtually absent in mature cerebellar PCs, and show low expression levels in the adult cerebellar cortex, being mainly localized in Lugaro and Golgi cells 22,23 . This dampened the interest for the study of mGlu5 receptors in the cerebellum, although changes in cerebellar mGlu5 receptor expression were found in autoptic tissues from individuals affected by psychiatric disorders, autism, and Fragile X-associated tremor/ataxia syndrome [24][25][26][27] . We were intrigued by the finding that mGlu5 receptor protein levels in the rat cerebellum are higher in the early postnatal life than in the adult life, as opposed to mGlu1α receptor protein levels, which are more abundant in the adulthood than at PND9 28 . This opposite developmental pattern of expression of mGlu1α and mGlu5 receptors was characteristic of the cerebellum and was not observed in other brain regions including the hippocampus, corpus striatum, cerebral cortex, hypothalamus, and olfactory bulb 28 .
Where precisely the mGlu5 receptor is expressed in the developing cerebellum is unknown, at present. No evidence exists to our knowledge that developing PCs express mGlu5 receptors, although mGlu5 receptor knockdown in the early postnatal life causes a severe impairment in PC maturation 29 .
We now report that mGlu5 receptors are highly expressed by cerebellar PCs in the first 12 days of postnatal life, and that the developmental decline in mGlu5 receptor expression coincides with the appearance and up-regulation of mGlu1α receptors. In addition, we demonstrate that it is the mGlu1 receptor that down-regulates the expression of mGlu5 receptors during the development of PCs and maintains its suppressing activity in the adult life.

Results
Complementary expression of mGlu1α and mGlu5 receptors in developing cerebellar PCs. We examined mGlu1α receptor expression in cerebellar tissue from mice at 3, 7, 9, 12, 16, and 18 PNDs by combining immunoblot analysis and immunofluorescent staining. mGlu1α receptor protein levels in the cerebellum were very low in the first twelve PND and increased substantially from PND12 to PND16, remaining high at later stages of development (Fig. 1A). The transcript of mGlu1 receptors increased by >2 fold from PND9 to PND18 (Fig. 1B), and this was associated with a trend to a reduction in Grm1 promoter methylation (Fig. 1C). Confocal analysis showed that expression of mGlu1α receptors was faint in calbindin-positive PCs at PND9 and increased substantially at PND16 (Fig. 1D). Interestingly, the mGlu5 receptor showed a complementary pattern of expression in the developing cerebellum, with mGlu5 receptor protein levels being high between PND9 and PND12 and decreasing afterwards ( Fig. 2A). The transcript of mGlu5 receptors was dramatically reduced from PND9 to PND18 (Fig. 2B), and this was associated with a significant increase in Grm5 gene promoter methylation (Fig. 2C). Confocal microscopy analysis of mGlu5 receptors was performed in two separate experiments at different time points after birth. Developing PCs labelled for carbonic anhydrase-8 (Fig. 2D, upper panel) or calbindin (Fig. 2D, lower panel) expressed mGlu5 receptors at PND7, 9, and 12, but were no longer decorated with mGlu5 receptor antibodies after PND14 (Fig. 2D). After PND14 weak labelling for mGlu5 receptors remained in carbonic anhydrase-8 negative dendritic elements of putative interneurons.
Developmental switch from mGlu1 and mGlu5 receptor signaling in the postnatal cerebellum.
We used cerebellar slices challenged with the mGlu1/5 receptor agonist, DHPG, in the presence of mGlu1 or mGlu5 receptor blockers for the analysis of receptor-stimulated PI hydrolysis. As expected, maximally effective concentrations of DHPG (100 μM) largely increased [ 3 H]inositolmonophosphate (InsP) formation in PND3-12 cerebellar slices. Stimulation of PI hydrolysis decreased at PND16 and became very low at PND21 (Fig. 3A). In the first 12 days of postnatal life, DHPG-stimulated PI hydrolysis was highly sensitive to the mGlu5 receptor negative allosteric modulator (NAM), MPEP, whereas only the mGlu1 receptor NAM, JNJ16259685, was able to antagonize the action of DHPG at PND16 (Fig. 3A). To localize mGlu1 or mGlu5 receptor signaling at cellular level, we assessed intracellular Ca 2+ release in PCs from cerebellar slices prepared from PND9, PND17, and PND20. At PND9, DHPG-stimulated Ca 2+ release was abrogated by MPEP (Fig. 3B). In contrast, the action of DHPG was antagonized by JNJ16259685, but not by MPEP, at PND17 and PND20 (Fig. 3B). mGlu1 receptors drive the developmental decline of mGlu5 receptors in PCs. To examine whether the decline of mGlu5 receptors in PCs was causally related to the appearance of mGlu1α receptors during postnatal development, we used pharmacological and genetic approaches. In a first set of experiments, we treated mice daily with the mGlu1 receptor NAM, JNJ16259685 (2.5 mg/kg, i.p.) during a developmental period that corresponds to the appearance of mGlu1 receptors, i.e., from PND9 to PND16. Pharmacological blockade of mGlu1 receptors during this time window prevented the decline of mGlu5 receptor protein, which was still expressed in PCs at PND16 (Fig. 4A,B). Grm5 gene promoter methylation was reduced in the cerebellum after treatment with JNJ16259685 (Fig. 4C). In addition, MPEP retained the activity of antagonizing DHPG-stimulated PI hydrolysis in cerebellar slices after in vivo treatment with JNJ16259685 (Fig. 4D).
Alternatively, we treated mice with the selective mGlu1 receptor positive allosteric modulator (PAM), Ro0711401 (10 mg/kg, s.c., daily), from PND7 (for Western blot analysis) or PND10 (for functional analysis) to PND12, when expression of mGlu1 receptors in PCs is still low. This treatment accelerated the drop in mGlu5 receptor protein in the cerebellum, which was already reduced at PND12 (Fig. 5A). When DHPG-stimulated PI hydrolysis was monitored in cerebellar slices, MPEP lost its antagonistic activity at PND12 after systemic treatment with Ro0711401 (Fig. 5B). These data suggested that endogenous activation of mGlu1 receptors was responsible for the developmental drop of mGlu5 receptors in PCs.
To further demonstrate the developmental link between mGlu1 and mGlu5 receptors in PCs, we used conditional mGlu1 receptor knockout (cKO) mice. In these mice, exposure to doxycyclin (added to the drinking water of lactating mothers) from PND0 to PND14 suppressed the expression of mGlu1 receptors in PCs of the offspring (Fig. 6A). Doxycylin-induced knockdown of mGlu1 receptors was associated with a large increase in mGlu5 receptor expression, which was still prominent in the PCs at PND16 (Fig. 6A,B).

mGlu1 receptors retain the suppressive activity on mGlu5 receptor expression in adult PCs.
mGlu1α receptor cKO mice were also used to examine whether mGlu1 receptors could maintain the mGlu5 receptor suppressed in PCs also in the life. Chronic treatment with doxycyclin (200 μg/ml in the drinking water) from PND24 to PND76 caused the expected suppression of mGlu1α receptor protein associated with a significant increase in mGlu5 receptor protein levels in the cerebellum (Fig. 7A). Remarkably, cerebellar PCs were highly decorated with mGlu5 receptors after chronic mGlu1α receptor knockdown (Fig. 7B).

Discussion
Our findings demonstrate for the first time that mGlu5 receptors are physiologically expressed by cerebellar PCs in the early stages of postnatal development, and largely mediate group-I mGlu receptor-stimulated PI hydrolysis in the first two weeks of postnatal life. Interestingly, the developmental pattern of expression of mGlu1α and mGlu5 receptors in PCs was specular, with mGlu1α receptors entirely replacing mGlu5 receptors during the first two weeks after birth. Gene methylation analysis demonstrated that the developmental switch between mGlu5 and mGlu1 receptors was driven by an epigenetic mechanism. Of particular interest was the demonstration that the appearance of mGlu1α receptors caused the developmental decline of mGlu5 receptors in PCs. Pharmacological experiments with mGlu1 receptor ligands administered to mice in restricted developmental  These findings raise a number of fundamental questions: (i) why is the mGlu5 receptor highly expressed by PCs in the first twelve days of postnatal life?; (ii) why a robust stimulation of PI hydrolysis is needed during that period?; (iii) why the mGlu1 receptor monopolizes the control of PC physiology by eliminating the mGlu5 receptor after the first two weeks of postnatal life?; and, (iv) are the two receptors different considering that they belong to the same subgroup and are both coupled to PI hydrolysis as a primary signal transduction mechanism?
Hydrolysis of phosphatidylinositol-4,5-bisphosphate generates inositol-1,4,5-trisphosphate (InsP 3 ) and dyacylglycerol (DAG), which stimulates intracellular Ca 2+ release and activates protein kinase C (PKC), respectively 30 . Experiments carried out in heterologous expression systems have shown that mGlu1 and mGlu5 receptors differ in the kinetics of intracellular Ca 2+ release in response to receptor activation. Glutamate elicits single-peaked intracellular Ca 2+ mobilization in cells expressing mGlu1 receptors and induces Ca 2+ oscillations in cells expressing mGlu5 receptors 31 . This difference relies on a threonine residue of mGlu5 receptor (T845) that is localized in the G-protein interaction domain and is phosphorylated by PKC 31 . Although PCs are generated before birth, they undergo dramatic morphological changes in a postnatal time window that corresponds to mGlu5 receptor expression, particularly after PND7 when PCs develop their typical dendritic trees 32 . Dendritic development in rodent PCs proceeds through an initial phase in which the primitive dendritic tree is retracted and multiple filopodia-like processes are formed around the cell body (see Fig. 2D at PND7). After PND7, these processes disappear and the final dendritic tree starts to grow. This second phase of dendritic growth is driven by both intrinsic and extrinsic factors, including the glutamate release from afferent fibers 33 . Interestingly, a phase of developmental PC death precedes the onset of the second phase of dendritic growth 33 . Activation of mGlu5 receptors with ensuing stimulation of PI hydrolysis and intracellular Ca 2+ oscillations might have a key role in any of these early postnatal events and might provide a survival signal for those Purkinje neurons that undergo early innervation by climbing and parallel fiber. Activation of mGlu5 receptors might also be involved in the early elimination phase of supranumerary climbing fibers (from PND7 to PND15), the underlying mechanism of which is still unknown 7,34 .
Activation of mGlu1 receptors after PND12 might provide a more localized Ca 2+ signal, which, together with PKC-γ activation might be instrumental for the late phase of elimination of supranumerary climbing fibers via the induction of semaphorin-7A or other mechanisms 9 , as well as for parallel fiber elimination and territory segregation to the distal dendrites of PCs occurring after PND15 10 . mGlu1 receptors might optimize the signal-to-noise ratio in intracellular Ca 2+ release during an active period of synaptic refinement by eliminating a competitor receptor (the mGlu5 receptor) that generates Ca 2+ oscillations and waves that may spread across the cytoplasm. In addition, elimination of mGlu5 receptors might serve to deprive PCs of a morphogenetic signal that becomes inappropriate during a period of active synaptic refinement.
Interestingly, mGlu1 receptors maintained the suppressive action on mGlu5 receptors in the adult life, as shown by the reappearance of mGlu5 receptors in PCs after genetic deletion of mGlu1α receptors. At least four functional splice variants of mGlu1 receptors (named mGlu1α or a, -β1 or b, -β2 or f, and -γ or d) have been described 35 , of which mGlu1α predominates in the cerebellum, and mGlu1α and mGlu1β1 are expressed in PCs 36,37 . The long C-terminal domain exclusive of mGlu1α receptors allows interaction with Homer proteins 38 and better phospholipase-C coupling efficacy 39,40 and is also required for perisynaptic targeting of mGlu1 receptors, inositol-1,4,5-trisphosphate-mediated Ca 2+ mobilization, elimination of supranumerary climbing fibers, LTD, and motor learning 37 . We showed that epigenetic regulation of mGlu5 receptor expression is dependent on mGlu1α variant. The role of mGlu1β1 receptor in reguration of mGlu5 receptor would be elucidated by use of mice in which mGlu1β1 transgene is introduced in PCs of mGlu1 receptor knockout mice.
A re-expression of mGlu5 receptors in PCs has been shown in pathological conditions characterized by a reduced expression or activity of mGlu1 receptors in PCs, i.e., in mice modelling type-1 spinocerebellar ataxia (SCA1) 15 , in mice developing experimental autoimmune encephalomyelitis (EAE) 18 , and in autoptic cerebellar samples from patients affected by multiple sclerosis 18 . This is one of the several examples of developmental proteins that are re-expressed in degenerating neurons. At least in SCA1 and EAE mice systemic treatment with mGlu5 receptor PAMs or NAMs did not affect cerebellar motor symptoms, suggesting that re-expressed mGlu5 receptors do not substitute for mGlu1 receptors in the regulation of motor learning and motor coordination. However, we cannot exclude that re-expressed mGlu5 receptors support the survival of degenerating PCs. If so, continuous treatment with a selective mGlu5 receptor PAM could slow the progression of cerebellar disorders by amplifying the otherwise sterile pro-survival program driven by mGlu5 receptors. This interesting hypothesis warrants further investigation. C57BL/6 male mouse pups were used at postnatal day (PND) 3, 9, 12, 16, 18, 21, the day of birth was designed as PND zero.

Materials and Methods
Generation of mGlu1 −/− mice bearing L7-tTA and TRE-mGlu1α transgenes (mGlu1α receptor cKO mice; C57BL/6 N background) proceeded as described previously 41 . mGlu1α receptor cKO mice were untreated or treated with 200 µg/ml of Dox (doxycycline hyclate; Sigma, St Louis, MO) in drinking water between 0 and 14 days of age or between 24 and 76 days of age. The Dox-water was delivered in dark bottles to protect Dox from light and changed twice a week. At 16 or 76 days of age, brain samples of these mGlu1α receptor cKO mice were collected. For immunohistochemical analysis, these mice were perfused with 4% paraformaldehyde in 0.1 M phosphate buffer.
Western blot analysis. The cerebella were dissected out and homogenized at 4 °C in RIPA buffer containing protease inhibitors cocktail (Merck Millipore, Milano, Italy) for 30 min and an aliquot was used for protein determination.
Equal amounts of proteins (20 μg) from supernatants were separated by 8% SDS polyacrilamide gel at 100 V for 1 hour for the detection of mGlu1α receptors or mGlu5 receptors, using a mini-gel apparatus (Bio-Rad Mini Protean II cell, Milano, Italy). Proteins were than electroblotted on Immuno PVDF membranes (Bio-Rad) for 7 min using Trans Blot Turbo System (Bio-Rad). Filters were washed three times and blocked for 1 hour in Tris-Tween buffered saline (TTBS) containing 5% non-fat dry milk.
One hour later, the incubation was stopped by the addition of 900 μl of methanol/chloroform (2:1). After further addition of 300 μl of chloroform and 600 μl of water, samples were centrifuged at low speed to facilitate phase separation, and the upper aqueous phase was loaded into Dowex 1-X-8 columns resin (100-200 mesh, formate form; Dow Chemical Company, Midland, MI) for the separation and quantification of [ 3 H]InsP.
Columns were washed twice with water, once with a solution of 5 mM sodium tetraborate and 40 mM sodium formate to elute cyclic InsP and glycerophosphoinositols, and then with 6.5 ml of 0.2 M ammonium formate and 0.1 M formic acid for the elution of [ 3 H]InsP. The remaining aqueous phase and the organic phase were dried under a continuous nitrogen stream, and 0.5 N NaOH was added to each sample. Calcium imaging analysis. Parasagittal slices from the cerebellum were prepared from C57BL/6J mice at PND9, 17 and 20.
Slices were recovered in ACSF for 30 min at 34 °C, then loaded for 45 min with the green fluorescent calcium dye OGB1-AM (10 μM) in the presence of 0.02% pluronic, and finally recovered for 45 min at room temperature.