Transduction of group I mGluR-mediated synaptic plasticity by β-arrestin2 signalling

Conventional signalling by the group I metabotropic glutamate receptors, mGluR1 and mGluR5, occurs through G-protein coupling, but evidence suggests they might also utilize other, non-canonical effector pathways. Here we test whether group I mGluRs require β-arrestin signalling during specific forms of plasticity at hippocampal excitatory synapses. We find that genetic ablation of β-arrestin2, but not β-arrestin1, results in deficits in plasticity mediated by mGlu1 receptors in CA3 pyramidal neurons and by mGlu5 receptors in CA1 pyramidal neurons. Pharmacological studies additionally support roles for Src kinases and MAPK/ERK downstream of β-arrestin2 in CA3 neurons. mGluR1 modulation of intrinsic conductances is otherwise preserved in β-arrestin2−/− mice with the exception of a rebound depolarization, and non-mGluR-mediated long-term potentiation is unaltered. These results reveal a signalling pathway engaged by group I mGluRs to effect changes in synaptic and cell intrinsic physiology dependent upon β-arrestin rather than G proteins. Pharmacological manipulation of mGluRs with effector-biased ligands could lead to novel therapies to treat neurological disease.

G roup I metabotropic glutamate receptors (mGluR1 and mGluR5) function as modulators of neuronal physiology and synaptic transmission and have been the target of therapeutic drug development for pathologies including Fragile X syndrome, epilepsy and pain 1 . mGlu1/5 receptors are seven transmembrane receptors (7TMRs) that differ from other mGluRs in their preferential coupling to Ga q/11 second messenger pathways, yet mGlu1 receptors in particular exhibit an intriguing ability to drive cellular responses even when G-protein activity is inhibited pharmacologically [2][3][4][5][6] . G-proteinindependent signalling by other 7TM receptors (for example, angiotensin type 1A receptors and b 2 -adrenergic receptors) is transduced by kinase cascades whose activity is initiated by 7TM receptor coupling to the cytoplasmic scaffolding proteins b-arrestin1 and 2 7,8 , but how the arrestins participate in mGluR signalling and specifically in modification of neuronal synaptic and intrinsic activity remains unclear. mGlu1 and mGlu5 receptors are widely distributed among excitatory synapses in the central nervous system and respond to glutamate most notably by G protein-mediated stimulation of phosphoinositide hydrolysis, mobilization of calcium from intracellular stores, and activation of protein kinase C. Protein kinase C has a well-established role in some forms of synaptic plasticity that include long-term depression (LTD) of cerebellar parallel fibre-Purkinje cell synaptic transmission 9 and mGluR5-mediated selective long-term potentiation (LTP) of N-methyl-D-aspartate (NMDA) receptor-mediated currents at hippocampal mossy fibre-CA3 synapses 10 . G protein-dependent signalling by mGluR1/5 also alters neuronal excitability, in some instances through slow currents carried by transient receptor potential (TRP)-like channels 11 . Not all mGluR1/5 signalling appears to be mediated by G proteins, however. Some aspects of mGluR1 signalling are intact in conditions where G proteins were pharmacologically inhibited using non-hydrolyzable forms of GDP or N-ethylmaleimide 4-6 . GDPbS-resistant signalling by mGlu1 receptors in CA3 pyramidal neurons, for example, underlies transient potentiation of NMDA currents and intermediate-term potentiation of transmission at mossy fibre synapses 2,3 .
In this study, we asked whether b-arrestin scaffolding proteins serve as molecular substrates of mGluR synaptic and intrinsic plasticity in a G protein-independent manner. We found that at least two distinct forms of group I mGluR-dependent hippocampal synaptic plasticity were abrogated in b-arrestin2 À / À but not b-arrestin1 À / À mice. In pharmacological antagonism studies, we identify mitogen-activated protein kinase (MAPK)/ extracellular signal-regulated kinase (ERK) and Src as downstream kinases in an mGluR1-b-arrestin2 signalling network. Non-synaptic functions of mGlu1 receptors were differentially impacted in b-arrestin2 À / À mice. Our data demonstrate that group I mGluRs employ a b-arrestin2-mediated signalling mode to alter synaptic strength and suggest that selective targeting of mGluR signalling modalities could be a viable therapeutic strategy in the future.
Results mGluR1 plasticity at CA3 synapses is b-arrestin2 dependent. We initially tested the contribution of b-arrestins to a selectively mGluR1-dependent form of intermediate-term potentiation of excitatory post-synaptic currents at mossy fibre-CA3 synapses. mGluR1 dependence was previously characterized using mGluR1 À / À mice and selective antagonists of mGlu1 and mGlu5 receptors 3 . In the current study, we compared electrophysiological recordings from acute hippocampal slices prepared from wild-type and b-arrestin (barr) knockout mice. Baseline mossy fibre-excitatory post-synaptic current(mfEPSC) amplitudes were obtained by delivering pairs of electrical stimuli to mossy fibre axons in the stratum lucidum at a basal frequency (0.05 Hz) for 10 min. mGluR1-dependent potentiation of excitatory postsynaptic currents (EPSCs) was induced by delivering paired pulses at 1 Hz for 2 min (paired pulse low-frequency stimulation, PP-LFS; Fig. 1a). mfEPSCs were then evoked at basal frequency for 30 min after the train. At the conclusion of each mossy fibre-CA3 plasticity experiment, we verified contribution from mossy fibre inputs by selectively inhibiting mossy fibre transmission with group II mGluR agonist (2S,2 0 R,3 0 R)-2-(2 0 ,3 0 -dicarboxycyclopropyl) glycine (DCG-IV, 1 mM). All data collection, posthoc analysis, and imposition of inclusion criteria were performed with the experimenter blinded to genotype.
c-Raf kinase facilitates activation of MAPK/ERK via association with barr2 in cell lines 15 and, therefore, might have a role in synaptic plasticity mediated by mGlu1 receptors. We tested this hypothesis by bath applying GW5074 (1 mM), a potent inhibitor of c-Raf and b-Raf kinases, 10 min before and during PP-LFS 12 . This compound did not affect potentiation (mfEPSC amplitudes 137±20% relative to drug baseline, n ¼ 9, compared with 135 ± 12% of baseline for vehicle, n ¼ 18, P ¼ 0.54) (Fig. 2j-l). This was a surprising result given that the preceding u0126 data clearly support a role for MEK. While canonical models of MAPK signalling cascade proceed through sequential activation of Ras, Raf, MEK and ERK, MEK activation through Raf-independent mechanisms, such as mixed-lineage kinases, phosphoinositidedependent protein kinase-1 and protein kinase A, has described in non-neuronal cells [16][17][18] . Our data are also consistent with unconventional activation of MEK as a critical signalling component for induction of PP-LFS-induced plasticity of mossy fibre-CA3 synapses.
Non-synaptic mGlu1 activity and b-arrestin2 signalling. We next tested the contribution of b-arrestin2 to mGluR1-dependent modulation of spiking characteristics in CA3 pyramidal neurons. Pharmacological activation of mGlu1 receptors persistently suppresses fast after-depolarizing potentials (ADP) through an unknown mechanism that is insensitive to Kv7 channel inhibition and buffering of intracellular calcium 20 . Postspiking conductances are important determinants of rate and pattern of neuronal burst firing, such that mGluR1-mediated suppression of the ADP effectively reduces the initial frequency of action potential firing during a prolonged suprathreshold current injection. Activation of mGluRs in CA3 neurons can elicit a biphasic, outward current followed by an inward current 21 , and the inward, depolarizing current is supported by a G protein-dependent mechanism 11 . Through these studies we sought to examine a potential role for b-arrestin2 in non-synaptic processes mediated by mGlu1 receptors, and to concurrently determine whether mGlu1 receptors are present and functional on the plasma membrane of CA3 pyramidal neurons in barr2 À / À mice.
n.s. High-frequency stimulation (HFS) of mossy fibre inputs elicits robust potentiation of mfEPSCs in slice recordings from both barr2 À / À and barr2 þ / þ mice (P ¼ 0.85, n ¼ 6 cells from three wild-type animals, n ¼ 8 cells from five knockout animals). A time course plot and cumulative probability histogram illustrate robust facilitation following HFS (amplitudes during HFS were omitted). Amplitudes deprecate over time but remain elevated relative to baseline 25-30 min post-tetanus. (c,d) mfLTP is intact in barr1 À / À mice. mfEPSC amplitudes measured from barr1 þ / þ and barr1 À / À mice are strongly potentiated immediately following tetanus and slowly deprecate to similar levels above baseline 30 min later (P ¼ 0.56, n ¼ 7 cells from six wild-type animals, n ¼ 9 cells from eight knockout animals). Individual post-train measurements are provided in a cumulative probability histogram. (e) mfLTP potentiation levels in b-arrestin knockout mice and wild-type littermates are summarized by a box plot and 10-90th percentile whiskers. Representative trace calibration: x axes, 10 ms; y axes, 250 pA. Groups were compared by Mann-Whitney tests; n.s., non-significant.
Sch-CA1 PP-LFS LTD requires mGluR5. We next conducted similar experiments with the addition of mGluR1 and mGluR5 antagonists in the b-arrestin2 gene-targeted mice to delineate the induction mechanism of PP-LFS-induced LTD at Sch-CA1 synapses. MTEP, a negative allosteric modulator for mGlu5 receptors 30 , was bath applied to acute slice preparations 10 min before and throughout the 15 min PP-LFS train; mGluR5 antagonism by MTEP prevented induction of LTD and resulted in equal recovery of EPSC amplitudes measured 25-30 min post-train in barr2 þ / þ and barr2 À / À mice (86 ± 4% in wild type, n ¼ 14; 96±5% in knockout, n ¼ 19; P ¼ 0.27, Fig. 6a,b,g). To test for contributions by mGluR1, the competitive mGluR1 Instantaneous frequency measured during vehicle (MTEP) and DHPG application (MTEP þ DHPG) are summarized in a box plot. Representative barr2 þ / þ traces before and after DHPG application are shown with double-headed arrows denoting the intervals used to calculate instantaneous frequency. (b) In barr2 À / À recordings, DHPG activation of mGlu1 receptors reduces the instantaneous frequency of action potential firing during a 500 ms current step (P ¼ 0.0023, n ¼ 14 cells from seven animals). Vehicle (MTEP) and DHPG (MTEP þ DHPG) instantaneous frequencies are summarized in a box plot. Representative traces from a barr2 À / À CA3 pyramidal neuron show action potentials evoked by suprathreshold current injection under vehicle and DHPG conditions. (c) Hyperpolarization among spiking and non-spiking cells elicited during DHPG application is normal in barr2 À / À mice (P ¼ 0.62), as shown in a box plot. (d) barr2 þ / þ and barr2 À / À CA3 neurons exposed to DHPG (black bar) exhibited a brief hyperpolarizing potential. Representative traces of spiking cells are provided, and black arrows denote the peak hyperpolarization. (e) A subset of barr2 þ / þ CA3 neurons exhibit a rebound depolarization with prolonged application of DHPG that did not lead to spiking, but in barr2 À / À CA3 non-spiking neurons, the rebound depolarization is reduced or absent (P ¼ 0.047). Rebound depolarization measured in only the non-spiking population of cells is shown in a box plot. (f) Representative traces from non-spiking barr2 þ / þ and barr2 À / À CA3 neurons are provided. Black arrows denote peak hyperpolarization and subsequent rebound depolarization. antagonist LY367385 (50 mM) was bath applied before and during PP-LFS 31 . EPSCs from neurons in barr2 À / À mice again did not exhibit LTD after PP-LFS (91 ± 3%, n ¼ 14), whereas those from barr2 þ / þ animals were depressed (70±4%, n ¼ 10; P ¼ 0.003, Fig. 6c,d,g). We, therefore, conclude that PP-LFS induces LTD through activation of mGlu5 receptors, but not mGluR1, and that barr2 is required to induce this form of plasticity at Sch-CA1 synapses.
Chemical LTD does not require b-arrestin2. We tested if barr2 is required for the induction of a chemical form of LTD induced by the group I mGluR agonist DHPG; this form of plasticity is thought to share some signalling elements with that induced by electrical stimulation 22,26 . DHPG (100 mM, 5 min) application to acute hippocampal slices from juvenile (P16-P25) mice reduced Sch-CA1 EPSC amplitudes to 67 ± 7% of control in wild-type mice (n ¼ 13) and 72±8% of control amplitudes in barr2 À / À littermates (n ¼ 8), which was not statistically different (P ¼ 0.86, Fig. 6e-g). Thus, barr2 is not essential to DHPG-induced LTD, in contrast with its role in PP-LFS LTD.
mGlu5 surface expression is not altered in barr2 À / À mice. Because barr2 has a well-characterized role in internalization of many GPCRs, we evaluated whether mGluR5 surface localization was affected in barr2 gene-targeted mice. We incubated hippocampal wild-type and barr2 À / À brain tissue from acute slice preparations in biotin, and enriched for biotinylated surface membrane proteins using streptavidin Sepharose beads. Western blotting of the biotinylated fraction from wild-type and barr2 À / À tissue showed equivalent levels of hippocampal, surface-expressed mGluR5 when normalized to kainate receptor subunit GluK2 (108 kDa; representative immunoblotting of three wild-type and 3 barr2 À / À mice is shown in Fig. 6h, uncropped blots are provided in Supplementary Fig. 3). GluK2 is a surface and membrane-bound protein not known to associate with barr2 and was, therefore, utilized as a loading control. mGluR5/GluK2 optical density was 0.90±0.19 a.u. for wild type, and 0.90±0.18 a.u. for barr2 À / À (n ¼ 3 pairs, P ¼ 1, Fig. 6h). Notably, mGluR5 and GluK2 immunoreactivity was low in the non-biotinylated fraction compared with the biotinylated fraction, but immunoreactivity for b-tubulin (50 kDa), an intracellular cytoskeletal protein, was greater in the non-biotinylated fraction than the biotinylated  In contrast, Sch-CA1 EPSCs from barr2 À / À animals recover more than wild types 25-30 min after PP-LFS (P ¼ 0.03, n ¼ 17 cells from 10 animals). For illustrative purposes, every third event was plotted during basal stimulation, and every ninth event was plotted during 1 Hz PP-LFS. A cumulative probability histogram illustrates separation of post-PP-LFS amplitudes between barr2 þ / þ and barr2 À / À recordings. (d,e) PP-LFS induces equivalent depression of Sch-CA1 EPSCs in barr1 À / À and barr1 þ / þ (P ¼ 0.39, n ¼ 12 cells from seven wild-type animals, n ¼ 17 cells from 10 barr1 À / À mice), shown in time course and cumulative probability histograms. (f) Summary box plot of PP-LFS LTD post-train amplitudes from barr2 and barr1 groups. (g) Baseline PPRs measured using barr2and barr1-null mice were not different from wild-type littermates (P ¼ 0.46 for barr2 þ / þ and barr2 À / À , P ¼ 0.56 for barr1 þ / þ and barr1 À / À ). Median and interquartile ranges are provided. Calibration of representative traces: x axes, 10 ms; y axes, 100 pA. Groups were compared by Mann-Whitney tests. Asterisks denote significance (*Po0.05); n.s., non-significant.
fraction. These results indicate that streptavidin immunoprecipitation was strongly enriched for biotinylated membrane proteins and that mGluR5 expression is unaltered in barr2 À / À mice. Expression of mGluR5 in total and membrane-enriched fractions also is not grossly affected in barr2 knockout animals ( Supplementary Fig. 4), and thus it is unlikely that the deficit in PP-LFS LTD in barr2 À / À mice can be attributed to loss of surface-expressed mGluR5.  ), an mGluR1 antagonist, PP-LFS LTD is observed in barr2 þ / þ , but not barr2 À / À mice. A time course plot illustrates depression of EPSCs in barr2 þ / þ following PP-LFS in the presence of LY367385 and recovery to baseline levels in barr2 À / À animals (P ¼ 0.003, n ¼ 10 cells from six barr2 þ / þ mice, n ¼ 14 cells from eight barr2 À / À mice); data is summarized in a cumulative histogram.
(e,f) DHPG induces LTD in barr2 þ / þ and barr2 À / À mice. barr2 þ / þ , and barr2 À / À recordings have similar time courses for depression and post-train amplitude distributions (P ¼ 0.86, n ¼ 13 cells from six wild-type mice, n ¼ 8 cells from four knockout mice). (g) A box plot shows recovery of all barr2 À / À conditions from PP-LFS LTD but not DHPG LTD; PP-LFS LTD is abrogated in barr2 þ / þ recordings only by MTEP. (h) Monomeric mGluR5 (yellow arrow) and GluK2 immunoreactivity was detected in both the barr2 þ / þ and barr2 À / À surface biotinylated fractions by western blotting, and less so in non-biotinylated fractions (n ¼ 3 pairs of barr2 þ / þ and barr2 À / À mice). b-tubulin immunoreactivity indicated that intracellular proteins were enriched in the non-biotinylated fraction. Quantitated monomeric mGluR5 optical density normalized against the GluK2 signal is provided in a bar graph illustrating the mean and s.e.m. (i) mGluR5 co-immunoprecipitates with b-arrestin2 in the cortex and hippocampus of wild-type mouse brains (n ¼ 5 blots). Pull-down by an IgG isotype antibody yields little mGlu5 receptor immunoreactivity.b-arrestin2 is strongly detected in the immunoprecipitation and input conditions. Representative traces calibration: x axes, 10 ms; y axes, 100 pA. Groups were compared by Mann-Whitney tests. Asterisks denote significance (*Po0.05, **Po0.01); n.s., non-significant. mGlu5 receptors and b-arrestin2 associate in brain tissue. In the final set of experiments, we evaluated whether b-arrestin2 and mGlu5 receptors biochemically associate in brain tissue. Immunoprecipitation of barr2 protein from wild-type mice yielded positive immunoreactivity for mGluR5 monomers (150 kDa) in western blots from cortex and hippocampus but not cerebellum. Monomeric mGluR5 protein was not detected following immunoprecipitation by an IgG isotype antibody (n ¼ 5, representative blot in Fig. 6i, uncropped blot provided in Supplementary  Fig. 5). Similarly, mGluR5 was not detected in co-immunoprecipitation studies performed using tissue from mGluR5 À / À mice (n ¼ 3, representative blot in Supplementary Fig. 6). mGlu5 receptor immunoreactivity in lysate preparations from wild-type and mGluR5 À / À tissue confirmed the absence of mGluR5, while barr2 was detected in the immunoprecipitation and input conditions for both genotypes. These data indicate that barr2 and mGluR5 associate in the hippocampus and cortex.

Discussion
Here we report that b-arrestin2 scaffolding proteins are required for some forms of mGluR-dependent excitatory synaptic plasticity at two hippocampal synapses. The observation that non-canonical G protein-independent signalling shapes hippocampal synaptic plasticity expands our understanding of the molecular substrates for learning and memory. b-arrestin2 transduces a form of mGluR1-dependent plasticity of mossy fibre-CA3 synaptic currents that requires MAPK/ERK and Src kinases as downstream effectors. In contrast, classical mGluR-independent mossy fibre LTP did not require either b-arrestin isoform. Some cell-intrinsic responses to mGlu1 receptor activation in CA3 pyramidal neurons, specifically reduced initial spike frequency and transient hyperpolarizing potentials, also were not dependent on b-arrestin2, whereas elimination of b-arrestin2 altered rebound depolarization with prolonged DHPG application. In the Sch-CA1 hippocampal pathway, LTD induced by synaptic stimulation is impaired in b-arrestin2 gene targeted mice and prevented by mGluR5 antagonism. These results show that neuronal mGluR stimulation of b-arrestin signalling occurs at multiple sites in the central nervous system and comprise alternate molecular pathways that could be targeted to treat neuropathological conditions involving group I mGluRs. Why has group I mGluR-b-arrestin2 signalling at central synapses largely escaped detection to date? First, signal transduction by arrestins has gradually gained wider recognition through pioneering studies with a number of family A and family B 7TMRs over the last 15 years, but analogous evidence linking mGluRs and other family C 7TMRs with arrestins is limited: calcium sensing receptor and mGlu7 receptor function are sensitive to b-arrestin knockdown or overexpression [32][33][34] , whereas GABA B receptors do not appear to associate with or initiate b-arrestin-mediated trafficking or signal transduction 35 . Second, in previous studies G protein-independent signalling has been associated experimentally with mGlu1 receptors rather than mGluR5 (ref. 2), and the latter has received significantly greater scrutiny in part because of its linkage with Fragile X Syndrome 36 . Third, the convergence of G protein-and b-arrestin-mediated pathways on common downstream effectors, especially MAPK and Src kinases, likely obscured contributions from b-arrestin signalling in some experimental models. Fourth, the group I mGluR agonist DHPG appears to preferentially activate G protein signalling while minimally stimulating b-arrestin pathways downstream of mGlu1 receptors 37,38 , which could also account for the divergence between DHPG-and synaptic stimulation-induced effects observed in our studies. It will be of interest to determine what roles b-arrestins have in other forms of G protein-independent signalling by mGluRs that have been described in hippocampal pyramidal cells 2,3,5 , hippocampal interneurons 4 and cerebellar interneurons 6 . mGluR1-arrestin studies in non-neuronal heterologous expression systems and primary neuronal cultures provide important insights regarding the relationship between mGluR1 and b-arrestin1 and are useful to interpret the present observations in slices. Recombinant b-arrestin1 translocates to and co-localizes with recombinant mGlu1 receptors with agonist stimulation 39 , and co-localization could reflect association between b-arrestin1 and mGlu1 receptors based on co-immunoprecipitation studies 39,40 . b-arrestin1 also contributes to ERK1 and ERK2 phosphorylation stimulated by receptor activation, and sustained pERK, mediated by b-arrestin1, has been linked to promoting cell survival 13,14 . Collectively these studies point towards a role for b-arrestin1 in mGlu1 receptor signal transduction that could exist in parallel with the mGlu1-b-arrestin2 signalling manifested as changes in plasticity and intrinsic excitability that we described in the current study. Moreover, splice variations in the carboxy terminus of the mGluRs could confer an additional level of specificity in their interactions with b-arrestin isoforms to differentially impact distinct aspects of cell signalling. While still speculative for the b-arrestins, an analogous degree of specificity occurs for the interacting proteins Homer and tamalin, which specifically associate with the mGluR1a splice variant due to the presence of a binding site not found in other receptor isoforms 41,42 .
We propose that PP-LFS induces formation of post-synaptic mGluR1, b-arrestin2, MEK, and Src complexes as a requisite step to strengthen mossy fibre-CA3 synaptic transmission, as b-arrestin2 scaffolding of MAPK/ERK 8,43 and Src 44 is regulated by other 7TMRs, in addition to other kinases such as JNK 45 and Akt/GSK3 (ref. 46). Unexpectedly, we found no evidence that c-Raf underlies mGluR1-and b-arrestin2dependent plasticity, despite the requirement for ERK activation and scaffolding of c-Raf by b-arrestin2 in in vitro systems 15 . One micromolar GW5074, which we used to inhibit c-Raf and b-Raf kinases, significantly attenuates pERK upregulation in slices stimulated by phorbol ester application (Supplementary Fig. 7). The data suggest that other GW5074-insensitive MEK activators, such as the mixed-lineage kinases, protein kinase A or PKLD [16][17][18] , have a predominant role in regulating the ERK module in this context.
How might MAPK/ERK and Src kinase activation produce their effects on mossy fibre EPSCs? In light of a possible presynaptic, mossy fibre terminal locus of expression for intermediate-term potentiation 3 , Src or ERK might target trans-synaptic signalling pathways to modulate vesicle release. For example, Src or MAPK/ ERK may target the Eph-ephrin pathway, which has been proposed in classical mfLTP as a critical bidirectional transduction mechanism that bridges post-synaptic induction machinery with a presynaptic locus of expression 47 . MAPK/ERK could also stimulate rapid protein synthesis that ultimately modifies preand post-synaptic function. In CA3 neurons, ERK promotes phosphorylation of eukaryotic initiation factor 4E (eIF4E), NMDA receptor 2B (GluN2B) protein levels and transcription of ephrinB2 (ref. 48), thereby achieving indirect modulation of post-synaptic ion channel density and trans-synaptic signalling mechanisms to impact synaptic transmission.
Transduction of mGluR1-dependent synaptic plasticity by b-arrestin2 could also occur through signalling events restricted to the post-synaptic compartment. In CA3 neurons, calcium influx via L-type calcium channels or release from intracellular stores is required for potentiation of mfEPSCs by low-frequency synaptic stimulation 3 , but the precise relationship between post-synaptic calcium mobilization and b-arrestin-mediated signalling mechanisms remains unclear. Calcium influx might contribute to a parallel, G protein-or Homer-mediated pathway to impact plasticity; alternatively, b-arrestin2 association with the calcium effectors calmodulin and CaMKII (ref. 49) could have a role in signal transduction through mechanisms we have not explored in this study. mGluR1-b-arrestin2-Src signalling, perhaps facilitated by an increase in intracellular calcium, could modulate synaptic ionotropic receptor function. For example, transient potentiation of NMDA receptor-mediated currents by pharmacological activation of mGluR1 is GDPbS-resistant and requires Src tyrosine kinases 2 , and PP-LFS at mossy fibre-CA3 synapses produces sustained potentiation of NMDA receptormediated currents 3 while potentiation of the AMPA/kainate mediated component of the mfEPSC requires post-synaptic Src (Fig. 2). mGlu1 receptor regulation of NMDA receptor currents in CA3 pyramidal neurons might, therefore, occur through b-arrestin2 and Src via direct, post-translational modification (for example, phosphorylation) of NMDA receptors to enhance their conductance and thereby influence other forms of mossy fibre-CA3 plasticity 50 . Alternatively, cytoskeletal reorganization of the post-synaptic compartment induced by mGluR1 and b-arrestin2 could occur through regulation of cofilin. mGlu7 receptors, for example, elicit b-arrestin2-dependent ERK phosphorylation and subsequent NMDA receptor current reduction through reduced NMDA receptor surface expression mediated by actin and cofilin 51 . mGluR-b-arrestin2-cofilin regulation of the surface localization of ionotropic receptors might therefore serve as another mechanism of plasticity.
The mGluR1-and barr2-dependent plasticity we describe here represents a mechanism for synaptic strengthening of mossy fibres that is distinct from previously described mGluR5-dependent forms of post-synaptic NMDA-LTP in several key aspects 10,52 . The latter is elicited by a higher-frequency pattern of synaptic stimulation (25 or 50 Hz bursts for B1 s), requires activation of NMDA receptors, mGlu5 receptors and A 2A adenosine receptors, is G protein-dependent, and leads to a selective and stable potentiation of NMDA receptor EPSCs at mossy fibre synapses 10,52 . In contrast, mGlu1 receptor-dependent plasticity is evoked by a longer, low-frequency patterned stimulation in which the precise timing of the paired stimuli appears to be an important parameter, is independent of NMDA receptor activation, requires b-arrestin2, strengthens both AMPA and NMDA receptor components of the mossy fibre EPSC, and is only operative for o1 h following train stimulation 3 . Thus, these forms of plasticity can be viewed as complementary with regards to the duration of their effects on synaptic strength, the molecular mechanisms of induction, and the pattern of mossy fibre stimulation required for potentiation. Within the CA3 pyramidal neuron, there are points of convergence in signalling: both require Src kinase as a necessary component and both lead to enhanced NMDA contributions to mossy fibre EPSCs. Potentiation of mossy fibre NMDA-EPSCs leads to metaplasticity of both mossy fibre 53 and associational/ commissural synaptic inputs 54 , and, therefore, the complementary mGlu receptor-dependent signalling pathways are likely to be key mechanisms for dynamic modulation of excitatory synapses in CA3 pyramidal neurons.
b-arrestin2 also appears to be necessary for synaptic stimulation-induced LTD at Sch-CA1 pyramidal cell synapses. Activation of post-synaptic group I mGluRs by PP-LFS causes LTD that is at least partially supported by presynaptic mechanisms 55 , whereas internalization of AMPA and NMDA receptors occurs post-synaptically 56 . Our data suggest that b-arrestin2 is engaged downstream of mGluR5 activation; the biochemical association that we observed between b-arrestin2 and mGlu5 receptors is consistent with a potential post-synaptic scaffolding role for b-arrestin2. mGluR5 and post-synaptic b-arrestin2 could promote receptor internalization through MAPK/ERK and JNK and by regulation of local synthesis of endocytic machinery such as Arc; MAPK/ERK, JNK1, Arc, and protein synthesis have all been implicated in mGluR LTD 22,26,27,57 . Post-synaptic mGlu5 receptors directly associated with b-arrestin2 could also target trans-synaptic Eph-ephrin pathways 58 or stimulate endocannabinoid release in the hippocampus by a yet unidentified mechanism 59 . Alternatively, or even concurrently, mGlu5 receptors could stimulate G proteincoupled mechanisms to produce endocannabinoids that then activate CB1 receptors located on excitatory presynaptic terminals 60 . Presynaptic b-arrestin2 could have a critical role in conjunction with CB1 receptors to depress neurotransmitter release, either by mediating receptor internalization 61 or by formation of an active signalling complex. The multifaceted signal transduction roles that b-arrestin2 could assume in conjunction with mGlu5 receptors, which might involve interplay with endocannabinoid signalling systems and Homer proteins, could have broader implications for neuropsychiatric diseases such as Fragile X syndrome 62,63 and, therefore, merits additional investigation 64 .
Generation of biased ligands that preferentially promote G protein-or arrestin-dependent signalling afford the possibility of unprecedented fine-tuning of signalling downstream of 7TMs. Exploration of ligand bias for mGluR signalling is at a nascent stage, however. In cell culture models of mGluR1 regulation of cell viability, compounds such as DHPG, quisqualate and ACPD (1-amino-1,3-dicarboxycyclopentane) preferentially activate phosphoinositide hydrolysis (indicative of G-protein signalling) and minimally stimulate b-arrestin pathways. In the experiments presented here, DHPG LTD was intact in b-arrestin2 knockout mice, which is consistent with its induction relying on G protein-dependent mechanisms. However, it is also possible that DHPG activation of non-synaptic group I mGluRs led to induction of LTD via a divergence of signalling mechanisms compared with those regulated by PP-LFS. In contrast to DHPG, quisqualate, and ACPD, glutamate and aspartate elicit 'balanced' G protein and arrestin responses, whereas succinate and glutarate were identified as arrestin-'biased' agonists 37,38 . We also attempted to test if the putative arrestin-biased agonist succinate altered mf-CA3 EPSCs, but 1 mM succinate application had no effect on mfEPSC amplitudes evoked at basal frequency or on the potentiation of EPSCs induced by PP-LFS ( Supplementary Fig. 8). Further development appears warranted given our evidence for an important role of b-arrestins in multiple forms of group I mGluR signalling.
Regulation of divergent cellular responses by distinct mGluR signalling modalities suggests that selective targeting of effector pathways with novel drug compounds could be a useful therapeutic strategy. Biased signalling by mGlu1 receptors via G-protein or b-arrestin pathways has been proposed to be regulated by ligands via specific interactions within the ligand binding domain 37 . Whereas selective targeting of mGluR1-b-arrestin1 signalling could be used to regulate cell cycle and potentially treat some forms of cancer 38,65 , targeting of specific mGluR1 signalling modes in the nervous system might have some utility in treating conditions such as cerebellar ataxia 66 and chronic pain 67 . Drugs that interfere selectively with mGluR5-barrestin effectors, should those prove to exist, could have potential in therapeutic targeting of Fragile X syndrome 68 . In addition to orthosteric ligands, allosteric modulators may lead to biased signalling properties as well. For example, an allosteric modulator for mGlu5 receptors with antipsychotic properties exhibited stimulus bias, potentially through enhanced coupling to G q , was recently described 69 . There are no analogous biased allosteric modulators for mGlu1 receptors available, and consequently additional tool development will be required in order to test the viability of biased allosterism as a therapeutic approach.
These data suggest that unconventional signalling by both group I mGluR subtypes can be found at hippocampal synapses. Further delineation of the synaptic and non-synaptic roles played by b-arrestin1 and b-arrestin2 downstream of group I mGluR activation, enabled by future development of orthosteric ligands or allosteric modulators, could yield new targets for treatment of neurological disease and provide new insights into the molecular and cellular mechanisms of learning and memory.

Methods
Hippocampal slice preparation and electrophysiological recordings. For electrophysiological recordings, acute hippocampal slices were prepared from juvenile (P16-P22 for mossy fibre-CA3 patch-clamp recordings, P16-P25 for Sch-CA1 recordings) C57bl/6, barr1 À / À , or barr2 À / À mice of either gender, in accordance with Institutional Animal Care and Use Committee-approved protocols. barr1 and barr2 gene targeted mouse lines were generously provided by the laboratory of Dr Robert Lefkowitz of Duke University. Genotyping was performed by Transnetyx (Memphis, TN) and heterozygous by heterozygous breeders were used to raise mixed wild-type and knockout litters. In all of the comparative physiological studies, experimenters were blinded to mouse genotype or drug condition until data analysis was completed.
Mice were transcardially perfused with ice-cold sucrose-rich slicing artificial cerebrospinal fluid (aCSF) containing 85 mM NaCl, 2.5 mM KCl, 1.25 mM NaH 2 PO 4 , 25 mM NaHCO 3 , 75 mM sucrose, 25 mM glucose, 10 mM DL-APV, 100 mM kynurenate, 0.5 mM Na L-ascorbate, 0.5 mM CaCl 2 , and 4 mM MgCl 2 , and oxygenated and equilibrated with 95% O 2 /5% CO 2 . Following perfusion, mice were decapitated and brains were sliced in sucrose-aCSF. Horizontal slices (350 mm) preserving the ventral hippocampus were obtained using a Leica VT1200S vibratome (Leica Biosystems, Buffalo Grove, IL). Hippocampal sections were warmed to 30°C in sucrose-aCSF and returned to room temperature; they were then transferred to oxygenated aCSF containing 125 mM NaCl, 2.4 mM KCl, 1.2 mM NaH 2 PO 4 , 25 mM NaHCO 3 , 25 mM glucose, 2 mM CaCl 2 , and 1 mM MgCl 2 and maintained under these incubation conditions for at least 1 h before recording. Following incubation, individual slices were transferred to a recording chamber and continuously perfused with oxygenated aCSF at room temperature (25°C). Hippocampal CA1 and CA3 pyramidal neurons were visually identified using a BX51WI fixed-stage upright microscope (Olympus, Center Valley, PA) and subsequently used for whole-cell patch clamp recordings with a MultiClamp 700A amplifier (Molecular Devices, Sunnyvale, CA).
For voltage clamp recordings, recording electrodes were pulled borosilicate glass with tip resistances of 3-7 MO and filled with an internal recording solution containing 95 mM CsF, 25 mM CsCl, 10 mM Cs-HEPES, 10 mM Cs-EGTA, 2 mM NaCl, 2 mM Mg-ATP, 10 mM QX-314, 5 mM TEA-Cl and 5 mM 4-AP, and adjusted to pH 7.3 with CsOH. All cells were held at À 70 mV for the duration of voltage-clamp experiments; in addition, series resistance was continuously monitored and compensated 65-70%.
Baseline mossy fibre-CA3 and Sch-CA1 EPSC amplitudes were obtained by delivering paired (40 ms interstimulus interval) electrical stimulation at a basal frequency (0.05 Hz) using a glass monopolar electrode, positioned in the stratum lucidum or the stratum radiatum, respectively (Figs 1a and 5a). In all of the recordings excitatory transmission was isolated using GABA A receptor antagonists bicuculline (10 mM) and picrotoxin (50 mM), and NMDA receptor antagonist D-APV (50 mM). Mossy fibre inputs were verified by short and stable latency, robust paired pulse facilitation during baseline recording (41.8), and inhibition (465%) by group II mGluR agonist DCG-IV (1 mm), bath applied at the end of each recording.
Plasticity of Sch-CA1 and mossy fibre-CA3 synapses was induced electrically using previously characterized stimulation protocols. Protocols were executed using pClamp9 software (Molecular Devices) and an A310 Accupulser driving a D53 constant current isolated stimulator (Digitimer, Welwyn Garden City, United Kingdom). Intermediate-term potentiation of mossy fibre EPSCs was induced by lowfrequency stimulation consisting of a single train of paired stimuli (40 ms interstimulus interval) delivered at 1 Hz for 2 min, while mossy fibre long-term potentiation was induced by HFS (three 100 Hz trains, 1 s each delivered in 10 s intervals). mGluRdependent LTD of Sch-CA1 synapses was induced by PP-LFS (1 Hz, 15 min, 50 ms interstimulus interval) or bath application of DHPG (100 mM, 5 min) while in the presence of NMDA receptor antagonist D-APV and GABA A R antagonists.
In current clamp recordings, a potassium gluconate internal solution containing 130 mM K-gluconate, 20 mM KCl, 1 mM K-HEPES, 0.2 mM EGTA, 0.3 mM Na-GTP, 4 mM Mg-ATP pH 7.3 was loaded into recording electrodes. After break-in, CA3 neurons were equilibrated with the internal solution for 10 min and maintained in aCSF containing MTEP and CNQX. Membrane potential was held at À 70 mV between suprathreshold current injections; the amount of current injected was calibrated to generate between 4-6 action potentials during each pre-DHPG step. Ten trains of action potentials were evoked by current steps provided at 0.05 Hz. Cells were then measured in gap-free for 5 min before DHPG application; during this time the initial current required to hold at À 70 mV was not changed. After 5 min of DHPG application, cells were held at À 70 mV and 10 trains of action potentials were elicited by the same amount of current injection used before DHPG application.
Biochemical associations of endogenous proteins. To assay associations between native mGlu1 and mGlu5 receptors with b-arrestin2 proteins, mice were anesthetized with isoflurane and decapitated. Brains were dissected in ice-cold phosphatebuffered saline. Tissue from cortex, hippocampus, and cerebellum were homogenized in lysis buffer containing 25 mM Tris, 150 mM NaCl, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 1 mM Na 3 VO 4 , 1 mM NaF, 1% Triton X-100, and protease inhibitor cocktail (P2714, Sigma-Aldrich, St Louis, MO) and rotated end over end for 25 min. Solubilized homogenates were then clarified by centrifugation and lysate protein content was measured by Bradford assays using a spectrophotometer (Eppendorf, Hauppauge, NY). One milligram of protein from each sample was then pre-cleared using 40 ml protein A/G beads (20421, Thermo Scientific, Bannockburn, IL) for 1 h, rotating end over end at 4°C. Pre-clearing beads were discarded and 50 ml of protein A/G bead slurry and rabbit polyclonal anti-barr2 antibody (0.02 mg ml À 1 ; ab31294; Abcam, Cambridge, MA) or rabbit IgG antibody (0.02 mg ml À 1 ; DA1E; Cell Signaling Technology, Danvers, MA) were added to the supernatant. Co-immunoprecipitation by barr2 or rabbit IgG was performed overnight at 4°C rotating end over end. Proteins immobilized on the antibodyconjugated beads were eluted with a mixture of 2 Â Laemmli buffer with b-mercaptoethanol and separated by SDS-polyacrylamide gel electrophoresis in an 8% polyacrylamide gel. Following this, proteins were transferred to a polyvinylidene difluoride membrane, blocked in 5% milk in tris-buffered saline, 0.1% Tween 20 (TBS-T), and immunoblotted using mouse anti-mGluR1 (0.05 mg ml À 1 ; 610964; BD Biosciences, San Jose, CA), rabbit anti-mGluR5 (0.2 mg ml À 1 ; AB5675; EMD Millipore, Billerica, MA) or rabbit anti-barr2 (0.125 mg ml À 1 ) antibodies. Pulldown of proteins was visualized using HRP-conjugated goat anti-mouse and goat anti-rabbit secondary antibodies (GE Healthcare, Little Chafont, United Kingdom) with enhanced chemiluminescence (ECL) substrate (Pierce ECL Western Blotting Substrate, Thermo Scientific, Bannockburn, IL).
Surface biotinylation. Acute brain slices were prepared as described for physiology. After 1 h of incubation at room temperature, slices were transferred into 6-well plates containing 0.5 mg ml À 1 sulfo-NHS-LC-Biotin (21335; Pierce) in aCSF, on ice. Dissected hippocampi were biotinylated for 30 min and were subsequently washed twice with NH 4 Cl (50 mM)-supplemented aCSF and twice with aCSF, then homogenized in 1 ml of homogenization buffer (10 mM Tris, 320 mM sucrose, protease inhibitors, pH ¼ 7.4). Samples were clarified by centrifugation at 1,000g for 5 min and supernatants were collected for centrifugation at 100,000g for 1 h. Pellets were then resuspended in 500 ml lysis buffer containing 20 mM HEPES, 150 mM NaCl, 2 mM EDTA, 1% Triton X-100, 0.1% SDS and protease inhibitors with a pH of 7.4 and sonicated. Samples were rotated end over end for 2 h at 4°C; following solubilization, samples were centrifuged for 40 min at 100,000g. The supernatants were subsequently collected and the protein concentration was measured. Threehundred micrograms of protein brought to a 300 ml final volume from each mouse was then incubated with 40 ml of washed streptavidin-Sepharose beads (17-5113-01; GE Healthcare) overnight at 4°C. Streptavidin immunoprecipitation samples were centrifuged at 1,000g for 3 min and the supernatant (non-biotinylated fraction) was collected. The Sepharose beads bound to the biotinylated fraction were washed three times with lysis buffer and 150 ml of 2 Â Laemmli sample buffer (Bio-Rad) with b-mercaptoethanol was added. The protein samples were agitated and boiled at 95°C for 5 min and samples were frozen or used immediately for gel electrophoresis. Antibodies raised against mGluR5 (0.2 mg ml À 1 ), GluK2 (1:10,000; clone NL9, 04-921; EMD Millipore), and b-tubulin (1:10,000) were used to measure immunoreactivity for the respective proteins by western blotting. Film obtained following western blotting was scanned and optical densities were measured using ImageJ.
Statistical analysis. Non-parametric Mann-Whitney tests were used to compare mean post-train amplitudes and mean baseline PPRs between genotypes or drugtreated groups; Wilcoxon tests were used to compare baseline versus peak DHPG application responses. Mean ± s.e.m. is provided throughout the text. EPSC amplitude time courses are represented as mean ± s.e.m., while baseline PPRs are shown with median and inter-quartile ranges. All box and whisker plots are represented as median, interquartile range and 10-90% percentiles. Data availability. The data that support the findings of this study are available from the authors on reasonable request.