Astrocytes modulate thalamic sensory processing via mGlu2 receptor activation

Astrocytes possess many of the same signalling molecules as neurons. However, the role of astrocytes in information processing, if any, is unknown. Using electrophysiological and imaging methods, we report the first evidence that astrocytes modulate neuronal sensory inhibition in the rodent thalamus. We found that mGlu2 receptor activity reduces inhibitory transmission from the thalamic reticular nucleus to the somatosensory ventrobasal thalamus (VB): mIPSC frequencies in VB slices were reduced by the Group II mGlu receptor agonist LY354740, an effect potentiated by mGlu2 positive allosteric modulator (PAM) LY487379 co-application (30 nM LY354740: 10.0 ± 1.6% reduction; 30 nM LY354740 & 30 μM LY487379: 34.6 ± 5.2% reduction). We then showed activation of mGlu2 receptors on astrocytes: astrocytic intracellular calcium levels were elevated by the Group II agonist, which were further potentiated upon mGlu2 PAM co-application (300 nM LY354740: ratio amplitude 0.016 ± 0.002; 300 nM LY354740 & 30 μM LY487379: ratio amplitude 0.035 ± 0.003). We then demonstrated mGlu2-dependent astrocytic disinhibition of VB neurons in vivo: VB neuronal responses to vibrissae stimulation trains were disinhibited by the Group II agonist and the mGlu2 PAM (LY354740: 156 ± 12% of control; LY487379: 144 ± 10% of control). Presence of the glial inhibitor fluorocitrate abolished the mGlu2 PAM effect (91 ± 5% of control), suggesting the mGlu2 component to the Group II effect can be attributed to activation of mGlu2 receptors localised on astrocytic processes within the VB. Gating of thalamocortical function via astrocyte activation represents a novel sensory processing mechanism. As this thalamocortical circuitry is important in discriminative processes, this demonstrates the importance of astrocytes in synaptic processes underlying attention and cognition.


58
The thalamic reticular nucleus (TRN) is responsible for ensuring synchronous activity across specific 59 thalamo-cortical circuits required for sensory perception or the preparation and execution of distinct 60 motor and/or cognitive tasks. It is therefore imperative to ascertain how inhibition from the TRN to 61 thalamic nuclei is controlled to understand how neurophysiological disease states associated with TRN 62 malfunction precipitate (Huguenard, 1999;Rub et al., 2003;Barbas & Zikopoulos, 2007;Pinault, 2011). 63

64
The TRN surrounds the entire anteroposterior extent of the dorsal thalamus, meaning all thalamo-65 cortical and cortico-thalamic projections must pass through and make connections with its mesh of 66 inhibitory interneurons (Houser et al., 1980;Jones, 1985) (FIG. 1). This strategic localisation between 67 thalamus and cortex enables the TRN to mediate coherent activity patterns within the thalamo-68 cortico-thalamic excitatory loop by providing both feedback and feedforward inhibition to thalamic 69 nuclei upon thalamo-cortical and cortico-thalamic input, respectively (Shosaku et al., 1989) (FIG. 1). 70 The Group II metabotropic glutamate (mGlu) receptors (mGlu2/3) modulate physiologically-evoked 71 responses in the somatosensory ventrobasal thalamic nucleus (VB) by reducing inhibition from the 72 TRN (Salt & Turner, 1998;Copeland et al., 2012), with the mGlu2 component to this Group II effect 73 likely activated by glutamate spillover upon physiological sensory stimulation (Copeland et al., 2012). 74 75 VB astrocytes in vitro can respond to sensory afferent stimulation with an elevation in intracellular 76 calcium (Parri et al., 2010), in accordance with astrocytic activation in other brain regions (Porter & 77 McCarthy, 1996; Grosche et al., 1999;D'Ascenzo et al., 2007). These elevations can initiate release of 78 gliotransmitters including glutamate (Fellin et al., 2004), D-serine (Panatier et al., 2006), adenosine 79 triphosphate (Guthrie et al., 1999) and adenosine (Winder et al., 1996), with subsequent modulation 80 of neuronal excitability and synaptic transmission (Fellin et al., 2004;Serrano et al., 2006). Astrocytic 81 processes co-localise with sensory and TRN afferent terminals around the soma and proximal 82 dendrites of VB neurons (Ralston, 1983 with a holding potential of -70mV was used to record miniature synaptic events until the access 143 resistance and holding current were stable in recording solution only. The slice was then superfused 144 with recording solution containing 10µM 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX; Tocris), 50µM 145 DL-2-Amino-5-phosphonopentanoic acid (DL-APV; Tocris) and 0.5µM tetrodotoxin (TTX; Abcam) to 7 block the AMPA-and NMDA-evoked miniature and large amplitude events due to direct action 147 potential firing of inhibitory neurons, respectively, leaving only the GABA mediated miniature synaptic 148 events (confirmed in preliminary experiments by complete blockade of remaining synaptic events with 149 10µM bicuculline). LY354740, LY341495 and LY487379 (all made in-house) stocks were made in 100% 150 DMSO at 1000X the desired working concentration. Compounds were diluted into the recording 151 solution containing CNQX, APV and TTX immediately before application to the brain slice. All solutions 152 applied to the brain slices contained 0.1% to 0.2% DMSO. DMSO  Slices were prepared as described previously . Briefly, following removal from 176 the skull, the brain was glued with cyanoacrylate adhesive to a metal block and submerged in the bath All experiments were conducted using adult male Wistar rats (340-540g, n=18). Animals (Harlan, UK) 216 were housed on a 12h light/dark cycle with food and water ad libitum. 217 218 2.42 SURGERY 219 Animals were anaesthetised with urethane (1.2g/kg intraperitoneal [i.p.] injection) and were 220 prepared for recording as previously described (Salt, 1987(Salt, , 1989 (Salt, 1987(Salt, , 1989). On each occasion, one of the outer barrels was filled with 1M NaCl were calculated from the start of the gating pulse. Using such an approach it is possible to use air-jets 255 to evoke an excitatory response from stimulation of a single vibrissa, as described previously (Salt, 256 1989). Prior to the beginning of each of the experimental protocols described below, the 'principal' 257 vibrissa (i.e. the vibrissa at the centre of the receptive field) for each neuron was identified. All 258 neurons recorded from were quiescent. subtypes (Kingston et al., 1998;Schoepp et al., 1999). However, the parameters used for LY341495 in 307 this study have been demonstrated previously to produce selective antagonism for the Group II mGlu 308 receptors (Kingston et al., 1998). 309 310

311
Group II mGlu receptor activation has been previously demonstrated to depress VB neuron inhibitory 312 postsynaptic potentials (IPSPs) evoked upon stimulation of the TRN (Turner & Salt, 2003), and an 313 mGlu2 component to this Group II effect was recently described in an in vivo study (Copeland et al., 314 2012). Therefore, we first determined whether mGlu2 receptor activation is able to modulate 315 inhibitory synaptic transmission at the TRN-VB synapse. One component that would contribute to IPSP 316 depression is direct inhibition of GABAergic vesicle fusion with the presynaptic TRN membrane. By 317 recording miniature inhibitory postsynaptic currents (mIPSCs) it is possible to examine the frequency 318 of spontaneous presynaptic quantal release events and so detect changes in transmitter release in the 319 14 absence of evoked synaptic activity. In the absence of endogenous mGlu2 receptor activation, a 320 sub-maximal concentration (30nM) of the Group II agonist LY354740 was able to reduce mIPSC 321 frequency compared to baseline when applied alone (10.0±1.6% reduction compared to control, n=6 322 from 6 slices , FIG. 2). Application of the mGlu2 PAM LY487379 alone had no effect on mIPSC frequency 323 (data not shown). By nature of design, PAMs potentiate the action of orthosteric agonists, without 324 themselves possessing any intrinsic agonist activity (Johnson et al., 2003). This lack of effect of the 325 PAM in this preparation is therefore unsurprising as there is likely no baseline activation of mGlu2 326 receptors under these conditions. However, when the mGlu2 PAM was co-applied with the sub-327 maximal concentration of Group II agonist, a significant additional reduction in mIPSC frequency was 328 observed (30nM LY354740 & 30μM LY487379: 34.6±5.2% reduction, n=6 from 6 slices, p<0.001, FIG.  329 2), comparable to that seen upon maximal agonist effect (100nM LY354740: 39.1±4.7% reduction 330 compared to control, n=6 from 6 slices, p<0.001; FIG. 2). The Group II mGlu receptor effect on mIPSC 331 frequency was confirmed by its reversal upon Group II orthosteric antagonist LY341495 co-application 332 (100nM LY354740 & 100nM LY341495: 6.6±7.5% reduction in mIPSC frequency compared to control, 333 n=6 from 6 slices, p<0.01, FIG. 2). Taken together these data indicate that there is indeed an mGlu2 334 component to the Group II mGlu receptor effect on GABAergic transmission at the TRN-VB synapse. 335 Ultrastructural studies indicate that TRN terminals exclusively express the mGlu3 receptor subtype 336 (Tamaru et al., 2001), while VB astrocytes express both mGlu2 and mGlu3 (Ralston, 1983;Ohara & 337 Lieberman, 1993;Liu et al., 1998;Mineff & Valtschanoff, 1999). We therefore sought to confirm 338 functional expression of astrocytic mGlu2 receptors. 339 340

341
Are mGlu2 receptors themselves able to directly activate astrocytes? To address this question, we 342 monitored intracellular calcium levels in both VB neurons and astrocytes in an acute in vitro thalamic 343 slice preparation. In the presence of TTX to block neuronal activity, a sub-maximal concentration of 344 the Group II orthosteric agonist induced increases in intracellular calcium levels compared to baseline 345 when applied alone (300nM, ratio amplitude 0. 016±0.002, n=56 astrocytes from 5 slices, FIG. 3a-c). 346 15 Application of the mGlu2 PAM alone had no effect on intracellular calcium levels (data not shown). 347 Upon co-application of the mGlu2 PAM with the agonist there was a significant potentiation in 348 astrocytic intracellular calcium levels in the same astrocytes (300nM LY354740 plus 30μM LY487379, 349 ratio amplitude 0.035±0.003, n=56 astrocytes from 5 slices, p<0.001; FIG. 3a-c). This Group II mGlu 350 receptor effect could be reversed upon co-application of 1µM of the Group II antagonist LY341495 351 (1µM LY354740, 2.11±0.45 ΔF% change; 1µM LY354740 plus 1µM LY341495, 0.28±0.17ΔF% change, 352 n=10 astrocytes from 5 slices, p<0.01; FIG. 3d). Co-application of the Group II agonist with 5μM 2-353 Methyl-6-(phenylethynyl)pyridine (MPEP) and 100μM suramin had no effect (1µM LY354740 alone,  FIG. 3e). The initial Ca 2+ peak 373 induced by glutamate is abolished in the IP3R2 knock-out preparation, and can likely be attributed to 374 Group I mGluR activation, whose signal transduction pathway is mediated via Gq. The remaining 375 glutamate effect in the IP3R2 knock-out preparation can be attributed largely to activation of 376 ionotropic glutamate receptors (Höft et al., 2014), as application of ionotropic glutamate receptor 377 antagonists (NBQX and D-AP5) reduced the calcium associated fluorescence by 79% (data not shown); 378 whilst the initial Ca2+ peak, abolished, and this can likely be attributed to Group I mGluR activation. 379 Together, these data show that mGlu2 receptors elicit functional astrocyte responses via IP3R2 380 mediated calcium release; an effect traditionally associated with Gq/11 coupled metabotropic 381 receptors, as opposed to the Gi/o coupled mGlu2 receptor. However, the same metabotropic receptor, 382 when expressed in different cell types/brain areas, is able to couple with alternate G-proteins: GABAB 383 receptors have been reported to couple to both Gi/o and Gq (Gould et al., 2014;Mariotti et al., 2016) 384 and D1 receptors to Gs, Golf and Gq proteins (Lee et al., 2004). Furthermore, GABAB receptor activation, 385 usually assumed to be coupled via Gi/o, induces calcium elevations in VB thalamus (Gould et al., 2014). 386 As well as confirming an astrocytic locus for thalamic mGlu2 action, this represents a novel mechanism 387 of astrocytic activation. 388 389

390
Does this mechanism modulate thalamocortical responses to sensory stimulation in an in vivo system 391 (FIG. 4A)? To test this question, we first assessed whether astrocytes contribute to the generation of 392 VB neuron responses to physiological somatosensory stimulation. A recording electrode with 393 iontophoretic capabilities was advanced into the VB of rats, and vibrissae were deflected as required 394 to generate physiologically relevant activity. Observed waveforms were similar to those previously 395 published (Salt, 1989), and were not perturbed under experimental conditions. Fluorocitrate 396 selectively inhibits glia by interfering with the astrocytic tricarboxylic acid cycle (Fonnum et al., 1997), 397 which is used to generate energy in the form of guanosine triphosphate (GTP). Upon local application 398 of fluorocitrate, a reduction in neuronal responses to repetitive 10Hz stimulation (1s duration) of the 399 principal vibrissae was observed. Specifically, the maintained component of the neuronal response 400 profile was significantly reduced (68±4% of control, n=16 from 9 rats, p<0.001), whereas the initial 401 component remained unaffected (101±3% of control, n=16 from 9 rats, p>0.05) (FIG. 4B,C). The 402 maintained component of neuronal responses to vibrissa stimulation comprises an NMDA-mediated 403 contribution under normal physiological conditions (Salt, 1986). However, neuronal responses to 404 exogenous NMDA application were unaffected in the presence of fluorocitrate (102±4% of control, 405 n=11 from 5 rats, p>0.05) (FIG. 4B RIGHT PANELS; FIG. 4C) .05; FIG. 5B). Fluorocitrate inhibits formation of the energy 424 source GTP (Fonnum et al., 1997), which is required for mGlu2 receptor signal transduction 425 (Niswender & Conn, 2010). From this selective attenuation of the mGlu2 PAM effect upon inhibition 426 of astrocyte function we can infer that mGlu2 receptor modulation of the TRN-VB synapse function is 427 astrocyte-dependent. Furthermore, we can attribute the remaining Group II mGlu receptor 428