bioRxiv preprint doi: https://doi.org/10.1101/411017; this version posted September 6, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1 Serotonin regulates dynamics of cerebellar activity

2 by modulating tonic inhibition

3

4 Elizabeth Fleming and Court Hull

5

6 Department of Neurobiology, Duke University Medical School, Durham, United States

7

8

9 Correspondence:

10 Dr. Court Hull

11 Duke University Medical School

12 Dept. of Neurobiology

13 311 Research Drive

14 Durham, NC 27710

15

16 Email: [email protected]

17 Phone: 919-613-0927

18

19 Running Head: 5-HT regulates synaptic integration

20 Keywords: 5-HT, granule cell, synaptic integration, synaptic inhibition,

21

22 Acknowledgements:

23 This work was supported by grants from the NIH NINDS (5R01NS096289-02), the Sloan 24 Foundation, and the Whitehall Foundation. We thank Dr. Lindsey Glickfeld and members of the 25 Hull and Glickfeld labs for input and technical assistance throughout the project. bioRxiv preprint doi: https://doi.org/10.1101/411017; this version posted September 6, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

26 Abstract

27 Understanding how afferent information is integrated by cortical structures requires identifying the 28 factors shaping excitation and inhibition within their input layers. The input layer of the cerebellar 29 cortex integrates diverse sensorimotor information to enable learned associations that refine the 30 dynamics of movement. Specifically, afferents relay sensorimotor input into the 31 to excite granule cells, whose activity is regulated by inhibitory Golgi cells. To test 32 how this integration can be modulated, we have used an acute brain slice preparation from young 33 adult rats and found that encoding of mossy fiber input in the cerebellar granule cell layer can be 34 regulated by serotonin (5-HT) via a specific action on Golgi cells. We find that 5-HT depolarizes 35 Golgi cells, likely by activating 5-HT2A receptors, but does not directly act on either granule cells 36 or mossy fibers. As a result of Golgi cell depolarization, 5-HT significantly increases tonic 37 inhibition onto both granule cells and Golgi cells. 5-HT-mediated Golgi cell depolarization is not 38 sufficient, however, to alter the probability or timing of mossy fiber-evoked feed-forward inhibition 39 onto granule cells. Together, increased granule cell tonic inhibition paired with normal feed- 40 forward inhibition acts to reduce granule cell spike probability without altering spike timing. These 41 data hence provide a circuit mechanism by which 5-HT can reduce granule cell activity without 42 altering temporal representations of mossy fiber input. Such changes in network integration could 43 enable flexible, state-specific suppression of cerebellar sensorimotor input that should not be 44 learned, or enable reversal learning for unwanted associations.

45

46 New and Noteworthy

47 5-HT regulates synaptic integration at the input stage of cerebellar processing by increasing tonic 48 inhibition of granule cells. This circuit mechanism reduces the probability of granule cell spiking 49 without altering spike timing, thus suppressing cerebellar input without altering its temporal 50 representation in the granule cell layer.

51

52 Introduction

53 The cerebellum plays a key role in motor learning, particularly by harnessing a wide array of 54 sensorimotor inputs to establish learned associations that modify motor output. The first major 55 site of cerebellar sensorimotor integration is the granule cell layer, where excitatory input carried 56 by mossy fibers diverges onto a vast number of granule cells. This large-scale divergence has bioRxiv preprint doi: https://doi.org/10.1101/411017; this version posted September 6, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

57 long been thought to provide a substrate for pattern separation that distinguishes unique 58 sensorimotor inputs necessary for associative learning. According to classical and recent models, 59 granule cell layer pattern separation is tightly regulated by local inhibition provided by 60 called Golgi cells (Albus 1971; Billings et al. 2014; Eccles et al. 1967; Marr 1969). As the sole 61 source of synaptic inhibition onto granule cells, Golgi cells can set the excitability of granule cells 62 and hence the transformation of incoming mossy fiber input into granule cell output (Brickley et 63 al. 1996; Chadderton et al. 2004; Duguid et al. 2012; Hamann et al. 2002; Mitchell and Silver 64 2003). Due to this central role in controlling granule cell excitability, modulation of Golgi cell 65 activity could represent a key mechanism for generating the types of flexible, context-dependent 66 responses that have been observed in the granule cell layer during behavior (Albergaria et al. 67 2018; Courtemanche et al. 2009; Ozden et al. 2012).

68 Neuromodulatory inputs to the granule cell layer are well-suited to assert such context-dependent 69 regulation of Golgi cells and cerebellar processing. Specifically, the cerebellum receives 70 significant serotonergic (5-HT) projections from the reticular and raphe nuclei (Bishop 1985; 71 Dieudonne 2001) that are particularly dense in the granule cell layer (Takeuchi 1983). Moreover, 72 5-HT receptors have been reported to be expressed on both Golgi cells and granule cells, 73 although the latter appears to be restricted to early stages of cerebellar development (Geurts 74 2002; Oostland et al. 2014). Because 5-HT activation in other brain regions has been associated 75 with flexible reconfiguration and reversal of learned associations (Clarke 2004; Matias et al. 2017), 76 these anatomical observations suggest that cerebellar 5-HT inputs could provide a key 77 mechanism for modulating learning, as well as cerebellar sensorimotor integration.

78 Here, we demonstrate that 5-HT depolarizes Golgi cells, likely through activation of the 5-HT2A 79 receptor, but does not act directly on either granule cells or mossy fibers in young adult rats. As 80 a result, 5-HT significantly increases tonic inhibition onto both granule cells and Golgi cells. 81 Additionally, although 5-HT produces a net depolarization of Golgi cells that elevates their firing, 82 it does not significantly alter the probability or timing of evoked Golgi cell inhibition onto granule 83 cells. Together, increased tonic inhibition with normal feed-forward inhibition acts to reduce 84 mossy fiber driven spike probability in granule cells without degrading spike timing. These data 85 provide a circuit mechanism by which 5-HT can regulate the reliability of granule cell firing to 86 modify sensorimotor representations. Such changes in network integration could underlie the 87 types of context-dependent gating of sensorimotor input that have been observed in the 88 cerebellum and enable flexible suppression of inputs that should either not be learned or should 89 be unlearned. bioRxiv preprint doi: https://doi.org/10.1101/411017; this version posted September 6, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

90

91 Materials and Methods

92 SLICES AND RECORDINGS

93 All procedures were approved by the Duke University Institutional Animal Care and Use 94 Committee (IACUC), and are in accord with guidelines set by the US National Institutes of Health. 95 Acute sagittal slices (250 µm) were cut from the of postnatal day 20-27 male 96 Sprague Dawley rats. Slices were prepared in an ice-cold solution of 130 mM K-gluconate, 15 97 mM KCl, 0.05 mM EGTA, 20 mM HEPES, and 25 mM glucose (pH 7.4), with 2.5 µM R-CPP. This 98 solution has previously been found to enhance the survival and health of cerebellar Golgi cells 99 (Hull and Regehr 2012; Kanichay and Silver 2008). Slices were then stored in artificial cerebral

100 spinal fluid (aCSF) containing 125 mM NaCl, 26 mM NaHCO3, 1.25 mM NaH2PO4, 2.5 mM KCl,

101 1 mM MgCl2, 2 mM CaCl2, and 25 mM glucose, and equilibrated with 95% O2 and 5% CO2. This 102 solution contains divalent cation concentrations (£ 3 mM) that permit Golgi cell spontaneous 103 packemaking in acute rat cerebellar slices. Slices were incubated at 34°C for 20 minutes after 104 preparation, then kept at room temperature for up to 6 hours. 105 106 Slices were viewed using Dodt Gradient Contrast optics (Scientifica) on an upright microscope 107 (Olympus BX51WI), with a 40x water-immersion objective and a CMOS camera (QImaging, 108 Rolera Bolt). Whole-cell and cell-attached recordings were obtained with patch pipettes (Golgi 109 cells: 3-5 MΩ, granule cells, whole-cell: 6-9 MΩ, granule cells, cell-attached: 10-14 (MΩ) pulled 110 from borosilicate capillary glass (World Precision Instruments) with a Sutter P-1000 micropipette 111 puller. Electrophysiological recordings were performed at 31-33°C. 112

113 Spontaneous IPSCs (sIPSCs) were recorded at 0 mV. Evoked excitatory postsynaptic currents 114 (eEPSCs) were recorded at -70 mV. The reversal potential for evoked inhibitory postsynaptic 115 currents (eIPSCs) in granule cells was determined empirically in each experiment by adjusting 116 the membrane potential until no eEPSC was observed. eIPSCs were recorded at a mean holding 117 potential of 3.8 ± 3.3 mV. Voltage-clamp recordings were performed using an internal pipette 118 solution containing: 140 mM Cs-gluconate, 15 mM HEPES, 0.5 mM EGTA, 2 mM TEA-Cl, 2 mM

119 MgATP, 0.3 mM NaGTP, 10 mM phosphocreatine-tris2, and 2 mM QX 314-Cl. pH was adjusted 120 to 7.2 with CsOH. Membrane potentials were not corrected for liquid-junction potential. bioRxiv preprint doi: https://doi.org/10.1101/411017; this version posted September 6, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

121 Current-clamp recordings were performed with an internal solution containing: 150 mM K- 122 gluconate, 3 mM KCl, 10 mM HEPES, 0.5 mM EGTA, 3 mM MgATP, 0.5 mM GTP, 5 mM

123 phosphocreatine-tris2, and 5 mM phosphocreatine-Na2. pH was adjusted to 7.2 with KOH. As 124 described previously (Hull and Regehr, 2013), membrane hyperpolarization was avoided before 125 and immediately after break-in to whole-cell mode to prevent immediate firing rate potentiation 126 during Golgi cell recordings. Golgi cells were identified according their size, shape, location and 127 their spontaneous pacemaking, which was verified in cell-attached mode before break-in to 128 whole-cell mode (Hull and Regehr 2012). Golgi cell recordings in which action potential height 129 did not change across the recording were used for afterhyperpolarization analysis. 130 131 Current-clamp experiments were performed in synaptic blockers of AMPA receptors (2,3-Dioxo- 132 6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide disodium salt, NBQX, 5 μm), NMDA 133 receptors (3-((R)-2-Carboxypiperazin-4-yl)-propyl-1-phosphonic acid, R-CPP, 5 μm), and GABA 134 receptors (6-Imino-3-(4-methoxyphenyl)-1(6H)-pyridazinebutanoic acid hydrobromide, SR95531, 5 135 μm), unless otherwise noted. For experiments blocking voltage-activated sodium channels, 136 tetrodotoxin citrate (Octahydro-12-(hydroxymethyl)-2-imino-5,9:7,10a-dimethano-10aH- 137 [1,3]dioxocino[6,5-d]pyrimidine-4,7,10,11,12-pentol citrate,TTX,1 µM) was applied to the bath >2 138 minutes before recording. For all experiments using 5-HT (3-(2-Aminoethyl)-1H-indol-5-ol 139 hydrochloride,10 µM), slices were discarded and replaced following each bath application. The 5- 140 HT2A receptor antagonist MDL 100907 ((R)-(+)-α-(2,3-Dimethoxyphenyl)-1-[2-(4- 141 fluorophenyl)ethyl]-4-piperinemethanol, 500 nM) was added to the bath >2 minutes prior to or during 142 recordings. No holding current was injected for current clamp experiments. Notably, we find that 143 even when holding current remains stable and negligible in voltage-clamp, granule cells become 144 more excitable in current-clamp in response to injected current within minutes of break-in. Thus, 145 to measure evoked granule cell spiking, we only considered responses to injected current within 146 2 minutes after break-in. To accommodate this approach, evoked spiking was measured in 147 separate cells for control and 5-HT experiments. For 5-HT evoked granule cell spiking, all 148 measurements were made within 7 minutes of 5-HT application to the bath. All drugs were 149 purchased from Tocris or Abcam. 150 151 Electrophysiological data were acquired using a Multiclamp 700B amplifier (Molecular Devices), 152 digitized at 20 kHz with a Digidata 1440A digitizer (Molecular Devices) and low-pass filtered at 10 153 kHz. Acquisition was controlled using Clampex 10.3 software. Series resistance was bioRxiv preprint doi: https://doi.org/10.1101/411017; this version posted September 6, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

154 monitored in voltage-clamp recordings with a 5 mV hyperpolarizing pulse, and only recordings 155 that remained stable during data collection were used for analysis. An ISO-flex stimulus isolation 156 unit (A.M.P.I.) and glass pipettes (0.25-1 MΩ) filled with aCSF and placed in the tract 157 were used to activate mossy fiber inputs. Stimulating electrodes were placed a minimum distance 158 of approximately 40 µm from recorded cells. Stimulus intensity was adjusted to achieve success 159 rates of approximately 50% for whole-cell recordings of eIPSCs. Disynaptic eIPSCs were 160 identified according to latency (Kanichay and Silver 2008) and sensitivity to NBQX. For cell 161 attached recordings, stimulus intensity was set to achieve success rates of approximately 50% to 162 the second stimulus.

163 ANALYSIS AND STATISTICS

164 Mean action potential (AP) waveforms were generated by averaging 2 s of spikes per condition 165 in each experiment, then averaging these mean waveforms across experiments. Changes in AP 166 waveforms were calculated by subtracting the average waveforms for each individual cell per 167 condition, then by averaging subtracted waveforms across cells. eEPSC and eIPSC latencies 168 were calculated from the time of stimulus onset to the time at 20% of the peak evoked current 169 amplitude. Quantification of average 5-HT-evoked changes in membrane potential, spontaneous 170 inhibition, and evoked currents was performed using the time period of maximal 5-HT effect, as 171 defined by the window where Golgi cell spike rates were maximally elevated (within a 3 minute 172 window, beginning 1 minute after 5-HT application). Synaptic potency, defined as the amplitude 173 of success only trials, was calculated using trials where currents exceeded a threshold of 5 pA 174 below baseline (for EPSCs). Data are presented as mean ± SEM. Data distributions were tested 175 for normality using the Shapiro-Wilk test. For data sets that were non-normally distributed (Figs. 176 4C and 2D), non-parametric statistical tests were used, with the specific test indicated in the 177 results for each dataset. For normally distributed data, sample means were compared using 178 paired or unpaired, two-tailed Student’s t-tests. Statistics were calculated using Prism 6.0 179 (Graphpad). Mean values were considered significantly different at p < 0.05. Significant 180 differences are noted by asterisks, with single asterisks representing p < 0.05, two asterisks 181 representing p < 0.01, and three asterisks representing p < 0.001. 182

183 Results

184 To test whether 5-HT acts presynaptically to modulate excitatory input from mossy fibers entering 185 the granule cell layer or postsynaptically to modulate either of the two principal cell types of the bioRxiv preprint doi: https://doi.org/10.1101/411017; this version posted September 6, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

186 granule cell layer, we performed whole-cell recordings from granule cells and Golgi cells in acute 187 cerebellar slices. First, to determine whether 5-HT can alter granule cell excitability by modulating 188 evoked excitation from mossy fibers, we recorded excitatory post-synaptic currents (EPSCs) in 189 granule cells while stimulating the white matter, before and after applying 5-HT (Fig. 1). These 190 experiments revealed no significant effect of 5-HT on evoked EPSC amplitude (Fig. 1C; control 191 = 55.8 ± 12.3, 5-HT = 50.3 ± 12.7 pA, p = 0.75), potency (control = 60.2 ± 12.4, 5-HT = 55.6 ± 192 13.2 pA, p = 0.79), failure rate (control = 0.15 ± 0.1, 5-HT = 0.20 ± 0.1, p = 0.64) or paired-pulse 193 ratio (control = 0.97 ± 0.05, 5-HT = 0.99 ± 0.08, p = 0.84, n = 10). Next, to test whether 5-HT acts 194 directly on granule cells to modulate their excitability, we performed current-clamp recordings in 195 the presence of synaptic transmission blockers (NBQX 5 µM, R-CPP 5 µM, SR95531 5 µM) (Fig. 196 1D-G). Following 5-HT (10 µM) bath application, we observed no change in the membrane 197 potential of recorded granule cells (Fig. 1E; control = -75.1 ± 3.0, 5-HT = -75.4 ± 2.7 mV, p = 198 0.7435, n = 11) and no difference in evoked spiking in response to current injection (Fig. 1F, G; 199 input/output slope: control = 0.302 ± 0.024; 5-HT = 0.301 ± 0.031, p = 0.86). These data suggest 200 that 5-HT does not act on either excitatory glutamatergic mossy fiber inputs or granule cells at the 201 input stage of cerebellar processing. 202 203 Next, we tested whether 5-HT can directly modulate the excitability of Golgi cells, the interneurons 204 responsible for all inhibition of cerebellar granule cells. These experiments revealed that 5-HT 205 produced a robust increase in Golgi cell firing rates in the presence of synaptic transmission 206 blockers (Fig. 2B-D; 156.2 ± 54.0% mean increase, control = 4.0 ± 0.9, 5-HT = 7.6 ± 1.1 Hz, p = 207 0.046, Kolmogorov Smirnov test, n = 10). This increase was followed by a slower return toward 208 baseline despite the continued presence of 5-HT, consistent with the well-documented 209 desensitization of 5-HT receptors (Sullivan Hanley and Hensler 2002). Because 210 immunohistochemistry has suggested 5-HT2A receptors may be expressed by Golgi cells (Geurts 211 2002), we next tested the possibility that these effects on firing rate were mediated by 5-HT2A 212 receptors using the selective antagonist MDL 100907 (500 nM) (Kehne 1996; Sorenson 1993). 213 Based on our finding that 5-HT-induced firing rate increases can begin returning toward baseline 214 within a few minutes, we conducted these experiments by applying 5-HT in the presence of MDL 215 100907. In support of the hypothesis that 5-HT2A receptor activation likely mediates spike rate 216 increases, we observed no significant change in Golgi cell firing rates in the presence of MDL 217 100907 (Fig. 2E, F; Δ spike rate = 0.8 ± 1.0 Hz, n = 4, p = 0.1061). We did, however, observe 218 that Golgi cells were more depolarized in the presence of MDL as compared to control conditions, bioRxiv preprint doi: https://doi.org/10.1101/411017; this version posted September 6, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

219 suggesting a resting serotonergic tone in the slice (Fig. 2E; Vm control = -54.8 ± 0.8 mV; Vm MDL: 220 -59.1 ± 1.8 mV; p = 0.0328). 221 222 5-HT2A receptor activation has been reported to modulate neuronal excitability by several 223 mechanisms, including directly depolarizing the membrane either by opening a depolarizing 224 conductance (Hannon and Hoyer 2008) or attenuating resting potassium conductance (Villalobos 225 et al. 2005). Thus, we first tested whether 5-HT modulates the resting membrane potential of 226 Golgi cells by recording in current-clamp in the presence of the voltage-activated sodium channel 227 blocker tetrodotoxin (TTX, 1 µM). In these recordings, 5-HT produced a significant depolarization 228 of the Golgi cell membrane (Fig. 2E, F; 7.6 ± 2.4 mV depolarization, control = -55.1 ± 0.7, 5-HT = 229 -47.6 ± 2.9 mV, p = 0.024, n = 7). This depolarization was absent when 5-HT was applied in the 230 presence of MDL 100907 (Fig. 2E, F; 1.1 ± 1.9 mV hyperpolarization, MDL = -58.0 ± 3.8, +5-HT 231 = -59.1 ± 1.9 mV, p = 0.6063, n = 4). Consistent with the hypothesis that this membrane 232 depolarization is mediated by a reduction in resting potassium conductance rather than activation 233 of a depolarizing conductance, we measured an increase in membrane resistance following 5-HT 234 application that did not occur in the presence of MDL (Fig. 2G; 312.9 ± 225.6 MΩ increase, control 235 = 360.6 ± 60.3, 5-HT = 679.7 ± 274.6 MΩ, p = 0.0411, n = 6; MDL: control = 588.8 ± 106.9, 5-HT 236 = 590.5 ± 86.4 MΩ, p = 0.9480, n = 5).

237 Previous work has demonstrated that Golgi cell spike rate can also be modulated by a plasticity 238 mechanism termed firing rate potentiation (FRP) (Hull et al. 2013). This FRP plasticity mechanism 239 involves a kinase-dependent, bi-directional regulation of BK-type potassium channels 240 participating in the action potential afterhyperpolarization (AHP) (Hull et al. 2013). Specifically, 241 CaMKII activation is thought to increase BK channel open probability, enhancing AHPs and 242 decreasing spike rates, while PKC is thought to decrease BK channel open probability, leading to 243 smaller AHPs and elevated spiking (van Welie and du Lac 2011). Because 5-HT2A receptors act 244 via a Gq-coupled pathway to increase PLC, it is possible that they can also engage this same 245 FRP plasticity mechanism by activating PKC to modulate BK channels and enhance Golgi cell 246 spike rates. To test whether 5-HT can engage FRP in addition to depolarizing Golgi cells, we 247 measured the effect of 5-HT on Golgi cell AHPs in control conditions and after inducing FRP (Fig. 248 3). The rationale for this approach is that, if 5-HT acts in part through an FRP mechanism to 249 modulate BK channels, it should have a smaller effect on AHPs after FRP has already been 250 maximally engaged. Hence, we began by measuring the effects of 5-HT on Golgi cell AHPs in 251 control conditions from a subset of our experiments where there was no change in the amplitude 252 of action potentials across the duration of the experiment. Consistent with previous data showing bioRxiv preprint doi: https://doi.org/10.1101/411017; this version posted September 6, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

253 a small change in the AHP in response to depolarization (Hull et al. 2013), we measured a 1.4 ± 254 0.2 mV change in the AHP following 5-HT application (Fig. 3A). Next, we compared the change 255 in the Golgi cell AHP induced by a depolarization protocol previously established to produce 256 maximal FRP with the change in the AHP when 5-HT was applied after maximal FRP. Consistent 257 with previous results, FRP induction produced a significant change in the AHP (Fig. 3C; D; Δ = 258 2.4 ± 0.8 mV, control = 13.6 ± 1.9, FRP = 11.0 ± 1.5 mV, p = 0.0199, n = 4). Following maximal 259 FRP induction, 5-HT produced an additional reduction in the AHP that was not different from the 260 AHP change induced by 5-HT in control conditions (Fig. 3C; 5-HT only = 1.6 ± 0.06, 5-HT with 261 FRP = 1.2 ± 0.2 mV, p = 0.1701, n = 4). Because 5-HT produces the same change in the AHP 262 in control and after FRP induction, we conclude that 5-HT does not act through the previously 263 described FRP mechanism to modulate Golgi cell spike rates.

264 The primary role of Golgi cells is to provide synaptic inhibition to granule cells at the input layer 265 of cerebellar processing. Specifically, GABA release from Golgi cells mediates both tonic and

266 phasic synaptic inhibition of granule cells via distinct types of GABAA receptors (Kaneda et al. 267 1995; Puia et al. 1994). To test how 5-HT-mediated changes in Golgi cell firing alters each type 268 of granule cell inhibition, we next performed voltage-clamp recordings from granule cells in the 269 absence of synaptic transmission blockers. Following 5-HT bath application, the frequency of 270 spontaneous inhibitory post-synaptic currents (sIPSCs) was significantly increased in recorded 271 granule cells (Fig. 4C; 374.1 ± 148.4% mean increase, control = 2.8 ± 0.9, 5-HT = 9.6 ± 4.0 Hz, 272 p = 0.016, Wilcoxon matched-pairs signed rank test, n = 13), with no associated change in sIPSC 273 amplitude (control = 5.9 ± 0.6 pA, serotonin = 5.7 ± 0.6 pA, p = 0.67, n = 13). In addition, we 274 measured a significant increase in the tonic holding current that reflects GABAergic inhibition via

275 non-desensitizing GABAA receptors (Brickley et al. 1996; Kaneda et al. 1995; Rossi and Hamann 276 1998; Wall and Usowicz 1997) (Fig. 4D; 83.0 ± 29.1%, control = 8.3 ± 1.3, 5-HT = 13.3 ± 1.8 pA, 277 p = 0.02, n = 9). These large increases in both IPSC frequency and holding current did not occur 278 when 5-HT was applied in the presence of SR5531 (Fig. 4C, D; -2.6 ± 0.1% holding current 279 decrease, SR95531 holding current = 12.5 ± 2.4 pA; SR95531 + 5-HT holding current = 12.0 ± 280 2.6 pA; p = 0.6308, n=8; no sIPSCs detected in SR95531), or when 5-HT was applied in the 281 presence of MDL 100907 (17.9 ± 26.2% mean sIPSC rate increase, control = 23.7 ± 7.7 Hz, 5- 282 HT = 32.9 ± 16.6 Hz, p = 0.39, n = 5; 21.2 ± 18.8% mean holding current increase, control = 19.6 283 ± 4.6 pA, 5-HT = 22.3 ± 4.3 pA, p = 0.67, n = 7, respectively). These data suggest that the 5-HT- 284 mediated increases in Golgi cell firing rate can significantly elevate granule cell tonic inhibition. 285 bioRxiv preprint doi: https://doi.org/10.1101/411017; this version posted September 6, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

286 Golgi cells also make inhibitory synaptic connections onto neighboring Golgi cells (Hull and 287 Regehr 2012), thus 5-HT should also increase inhibition onto Golgi cells. Such a circuit 288 configuration could serve as a form of negative feedback to restrict the magnitude of granule cell 289 inhibition. To test this hypothesis, we next measured how 5-HT modulates synaptic inhibition 290 onto Golgi cells by performing voltage-clamp recordings without blocking synaptic transmission. 291 In these experiments, 5-HT significantly elevated Golgi cell sIPSC frequency (Fig. 5C; 433.9 ± 292 75.9% mean increase, control = 3.4 ± 0.5, 5-HT = 17.1 ± 1.5 Hz, p = 0.0008, n = 5) consistent 293 with a previous report that had attributed such effects to inputs from Lugaro cells (Dieudonne and 294 Dumoulin 2000). To test the hypothesis that this increase in Golgi cell inhibition serves to regulate 295 the magnitude of 5-HT-induced changes in Golgi cell spontaneous spiking (and hence granule 296 cell spontaneous inhibition), we next performed current-clamp recordings in the absence of 297 synaptic blockers. These experiments revealed that Golgi cell spike rates increased in the 298 presence of 5-HT (Fig. 5D; 79.2 ± 30.5% mean increase, control = 6.2 ± 0.9, 5-HT = 10.8 ± 1.6 299 Hz, p = 0.081, n = 3), but to a much smaller degree than when synaptic transmission was blocked 300 (156.2 ± 54.0% mean increase, control = 4.0 ± 0.9, 5-HT = 7.6 ± 1.1 Hz, p = 0.046, n = 10). We 301 conclude that the increase in granule cell layer inhibition mediated by 5-HT is restricted by a 302 negative feedback mechanism regulated at least in part by Golgi cell to Golgi cell , and 303 possibly also by additional 5-HT-induced input from Lugaro cells.

304 In addition to regulating tonic inhibition of granule cells, Golgi cells provide evoked feed-forward 305 and feedback inhibition to granule cells in a manner that is thought be important for regulating 306 spike timing (Kanichay and Silver 2008). Thus, we next tested whether 5-HT acts to regulate 307 evoked (phasic) granule cell inhibition by performing voltage-clamp recordings from granule cells 308 while electrically stimulating mossy fibers (Fig. 6). These experiments revealed no statistical 309 differences in the amplitude (Fig. 6C, D; stim 1: control = 19.2 ± 1.7, 5-HT = 17.7 ± 1.9 pA, p = 310 0.93; stim 2: control = 22.1 ± 2.1, 5-HT = 17.7 ± 1.6 pA, p = 0.48) or latency (stim 1: control = 3.3 311 ± 0.3, 5-HT = 3.3 ± 0.2 ms, p = 0.66; stim 2: control = 4.0 ± 0.4, 5-HT = 5.0 ± 0.8 ms, p = 0.10, n 312 = 18) of eIPSCs before and after applying 5-HT. Additionally, the failure rate (Fig. 6E, F; stim 1: 313 control = 0.7 ± 0.0, 5-HT = 0.8 ± 0.0, p = 0.30; stim 2: control = 0.8 ± 0.0, 5-HT = 0.8 ± 0.0, p = 314 0.64) and paired-pulse ratio (control = 1.2 ± 0.1, 5-HT = 1.1 ± 0.1, p = 0.37, n = 18) of eIPSCs 315 were not statistically different before and after applying 5-HT. These data suggest that, while the 316 depolarization evoked by 5-HT is sufficient to increase spontaneous Golgi cell spiking, it is not 317 sufficient to alter the probability or timing of Golgi cell spiking in response to incoming mossy fiber 318 input. Hence, these data reveal that 5-HT acts to regulate spontaneous, tonic granule cell 319 inhibition without significantly altering evoked inhibition. bioRxiv preprint doi: https://doi.org/10.1101/411017; this version posted September 6, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

320 Tonic inhibition from Golgi cells regulates granule cell responses to excitatory inputs (Mitchell and 321 Silver 2003), such as those initiated by sensory stimuli (Duguid et al. 2012). Since 5-HT enhances 322 tonic inhibition onto granule cells, we hypothesized that 5-HT may act to selectively reduce 323 efficacy of granule cell responses to mossy fiber input. To test this hypothesis, we performed 324 cell-attached recordings of granule cells while delivering brief bursts of high-frequency stimuli to 325 the mossy fibers to simulate physiological, sensory-evoked cerebellar input (Chadderton et al. 326 2004) (Fig. 7). In most recordings (4 of 6), we observed a marked decrease in the probability of 327 granule cell spiking in the presence of 5-HT (Fig. 7D; stim 1: control = 0.031 ± 0.019, 5-HT = 0.0 328 ± 0.0, p = 0.10; stim 2: control = 0.075 ± 0.040, 5-HT = 0.038 ± 0.030, p = 0.03; stim 3: control = 329 0.144 ± 0.019, 5-HT = 0.113 ± 0.022; p = 0.02, n = 4), that quickly recovered above control after 330 application of MDL 100907. The increase in spiking above control levels in MDL is consistent with 331 the finding of a resting serotonergic tone in the slice (Fig. 2E). There was no change in spiking 332 in the remaining 2 cells (not shown; stim 1: control = 0.038 ± 0.12, 5-HT = 0.038 ± 0.12, p = 0.5; 333 stim 2: control = 0.075 ± 0.050, 5-HT = 0.075 ± 0.075, p = 0.99; stim 3: control = 0.062 ± 0.062, 334 5-HT = 0.037 ± 0.037; p = 0.5, n = 2). This result is consistent with the finding that 5-HT did not 335 increase tonic inhibition onto all granule cells (Fig. 4). In contrast with spike probability, however, 336 granule cell spike timing was not altered by 5-HT application (Fig. 7E; stim 2: control = 4.0 ± 1.2, 337 5-HT = 4.5 ± 1.2 ms, p = 0.88; stim 3: control = 4.3 ± 0.60, 5-HT = 4.6 ± 0.60 ms, p = 0.88; n = 4), 338 in agreement with our observations that 5-HT does not alter evoked feed-forward inhibition onto 339 granule cells (Fig. 6). These data reveal that 5-HT can act to selectively reduce reliability of 340 granule cell spiking in response to mossy fiber input in a manner that does not alter the timing of 341 granule cell spiking.

342

343 Discussion

344 Here we demonstrate a circuit mechanism by which 5-HT regulates granule cell layer excitability 345 by modulating the activity of cerebellar Golgi cells. By providing a modest depolarization of the 346 Golgi cell membrane that is likely mediated by 5-HT2A receptors, 5-HT acts to selectively 347 modulate the spontaneous spiking of these interneurons without altering their evoked responses 348 to incoming mossy fiber input. As a result, 5-HT acts indirectly to decrease granule cell excitability 349 within a range capable of lowering the probability of granule cell spiking in response to mossy 350 fiber input. However, because 5-HT does not alter evoked inhibition onto granule cells, granule 351 cell spike timing is preserved. bioRxiv preprint doi: https://doi.org/10.1101/411017; this version posted September 6, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

352 Mechanistically, we find that 5-HT acts to increase inhibition onto both Golgi cells and granule 353 cells by rapidly depolarizing Golgi cells in a cell-autonomous manner, likely through 5-HT2A 354 receptor activation. While the concentration of MDL 100907 we used does not preclude some

355 contribution from other 5-HT receptors, the IC50 for the next closest 5-HTR has been reported to

356 be 700 nM (5-HT2C), and all other 5-HT receptors have IC50s for MDL that are significantly higher 357 than this (Kehne 1996). Thus, because MDL completely occluded the effects of 5-HT on Golgi 358 cells, it is most likely 5-HT2A receptors are responsible for Golgi cell depolarization. Our results 359 also are consistent with a previous anatomical study that identified 5-HT2A receptors on Golgi 360 cells using histochemical methods (Geurts 2002).

361 5-HT2A receptors have been suggested to produce neuronal depolarization and increased firing 362 rates via multiple Gq protein-dependent mechanisms, including opening of non-specific cation 363 channels (Hannon and Hoyer 2008) and reducing spike AHPs by decreasing calcium-activated 364 potassium conductances (Villalobos et al. 2005). Our results, however, are not consistent with 365 either of these mechanisms, as 5-HT acts to decrease the Golgi cell membrane conductance and 366 does not alter the spike AHP more than would be predicted by the increased spike rate induced 367 by membrane depolarization (Hull and Regehr 2012). While hyperpolarization of Golgi cells 368 drives a long-term increase in firing rate that results from modification of calcium-activated 369 potassium currents underlying the spike AHP (Hull et al. 2013), our results suggest that the 370 increase in firing rate seen in the presence of 5-HT does not interact with this mechanism.

371 Instead, the increase in membrane resistance we observe is consistent with reports that 5-HT2A 372 receptors can reduce resting potassium conductances. In particular, these receptors have been 373 shown to reduce an inward-rectifying potassium conductance in cortical fast spiking interneurons, 374 thus depolarizing the membrane potential at rest (Athilingam et al. 2017). 5-HT2A receptors have 375 also been shown to depolarize feed-forward interneurons in the auditory system by reducing a 376 resting potassium conductance (Tang and Trussell 2017). Golgi cells are similar to cortical fast- 377 spiking interneurons, in that they fire narrow action potentials that can achieve high spike rates, 378 they are gap junctionally-coupled within their population, they make a dense, highly divergent 379 axon plexus to release GABA near the of their postsynaptic targets, and they provide feed- 380 forward inhibition to their targets. Thus, our data may support the general rule that has been 381 demonstrated in the neocortex and elsewhere that FS-like interneurons commonly express 5- 382 HT2A receptors (Jakab 1998; Kruglikov and Rudy 2008).

383 Notably, our results differ from an earlier study that suggested no direct effect of 5-HT on Golgi 384 cells in acute slices from the rat cerebellum (Dieudonne and Dumoulin 2000). Though it is unclear bioRxiv preprint doi: https://doi.org/10.1101/411017; this version posted September 6, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

385 why this study did not reveal any effect of 5-HT on Golgi cell spiking, we note that both studies 386 identified increases in synaptic inhibition onto Golgi cells in response to 5-HT. Our results are, 387 however, consistent with the circuit arrangement predicted from the more recent finding that Golgi 388 cells can inhibit one another (Hull et al. 2013). Moreover, our finding that 5-HT reliably increases 389 inhibition onto granule cells is consistent with the depolarizing effect of 5-HT on Golgi cells, as 390 these are the only interneurons in the cerebellar cortex known to inhibit granule cells. In addition, 391 our results are not in conflict with the previously demonstrated result that 5-HT can depolarize 392 Lugaro cells (Dieudonne and Dumoulin 2000). While a physiological connection from Lugaro 393 cells to Golgi cells remains to be demonstrated, inhibition from Lugaro cells in the presence of 5- 394 HT could also contribute to the increases in Golgi cell spontaneous inhibition we observe.

395 We also note that our results differ from a previous study that showed no effect of 5-HT on granule 396 cell tonic inhibition (Rossi et al. 2003). This study used recording solutions that, in our experience, 397 do not promote Golgi cell spontaneous spiking (methods). Thus, it is possible that conditions in 398 this previous study were not specifically tailored for revealing an effect of 5-HT on granule cell 399 spontaneous inhibition. We therefore speculate that methodological differences may account for 400 the differences between our work and this study.

401 At the circuit level, it is somewhat surprising that our cell-attached recordings revealed no 402 significant change in the timing of evoked granule cell spiking in 5-HT, even though the time to 403 drive granule cells to spike threshold should be prolonged when inhibition is enhanced. However, 404 we suggest that the rapid arrival of feed-forward inhibition, which remains intact in 5-HT, maintains 405 granule cell spike timing (Duguid et al. 2015; Kanichay and Silver 2008). Specifically, while early 406 spikes remain possible in response to large synaptic inputs, feed-forward inhibition prevents late 407 spikes from occurring. As a consequence, we speculate that the lower spike probability in 5-HT 408 results at least in part from mossy fiber input that cannot drive the granule cells to spike threshold 409 before the arrival of feed-forward inhibition. In support of this view, the time window for granule 410 cell spike generation is dramatically enhanced when feed-forward (and spontaneous) inhibition is 411 blocked with SR95531. Moreover, while it might be expected that early spikes should arrive even 412 sooner due a decrease in the membrane time constant when tonic inhibition is elevated, granule 413 cells have unusually high membrane resistance (<1 GW) and small membrane capacitance (~ 3 414 pF) and, thus, a rapid membrane time constant (~5-10 ms). As a result of these passive 415 properties, elevating tonic inhibition at levels measured here (with resistance decreases of 416 approximately 70 mW) should only decrease the membrane time constant by hundreds of 417 microseconds. bioRxiv preprint doi: https://doi.org/10.1101/411017; this version posted September 6, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

418 Granule cell spike timing is thought to be an important feature of sensorimotor encoding at the 419 input stage of cerebellar processing (D'Angelo and De Zeeuw 2009; Kennedy et al. 2014). Recent 420 work has demonstrated that granule cells can receive multimodal input from distinct mossy fibers 421 with different strengths and short-term plasticity (Chabrol et al. 2015). As a result, individual 422 granule cells exhibit diverse spike patterns in response to unique patterns of mossy fiber input 423 that are best distinguished by their timing. Accordingly, it is thought that temporal coding, along 424 with population identity, provides a key mechanism for establishing sensorimotor representations 425 and allowing pattern separation in the granule cell layer (Medina et al. 2000). In this context, our 426 results are consistent with a model in which 5-HT can reduce the strength of a given sensorimotor 427 representation by reducing the reliability of granule cell spiking without altering identity of that 428 representation by preserving temporal coding. Moreover, the precise timing and reliability of 429 granule cell inputs to Purkinje cells determines their potential for undergoing long-term synaptic 430 plasticity (D'Angelo and De Zeeuw 2009). Thus, by selectively reducing the reliability of granule 431 cell spiking, we speculate that 5-HT can decrease associative learning for sensorimotor inputs 432 that behavioral context dictates should not be learned by downstream Purkinje cells. Such a 433 mechanism could also promote reversal learning for an association that has already been made.

434 Indeed, 5-HT signaling in other brain regions has been associated with reversal learning and 435 behavioral changes that are required for adaptation to new environmental 436 conditions. Specifically, both Dorsal Raphe activation and endogenous 5-HT are necessary for 437 some forms of reversal learning, with disruptions in either resulting in perseverative errors (Clarke 438 2004; Matias et al. 2017). Moreover, Dorsal Raphe respond to both positive and 439 negative prediction errors, as well as aversive stimuli and changes in motosensory gain (Cohen 440 et al. 2015; Kawashima et al. 2016; Matias et al. 2017), suggesting that 5-HT may serve as a 441 broad signal for behavioral adaption. 5-HT has also been shown to regulate sensory processing, 442 notably by both suppressing auditory inputs and enhancing multisensory inputs in the cerebellar- 443 like dorsal cochlear nucleus (Tang and Trussell 2017; 2015).

444 The cerebellum receives its serotonergic input from reticular and raphe nuclei (Bishop 1985; 445 Dieudonne 2001), regions that also innervate several other regions of the brain. Thus, 5-HT may 446 be released in the granule cell layer during the same conditions and affect similar learning goals 447 as have been described elsewhere. Given the extensive variety of sensory, motor, and cognitive 448 contextual information converging on the granule cell layer, such adaptive 5-HT signals are poised 449 to influence motor output in response to environmental changes. Hence, by regulating granule 450 cell layer excitability, the circuit mechanisms described here may allow 5-HT to regulate patterns bioRxiv preprint doi: https://doi.org/10.1101/411017; this version posted September 6, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

451 of granule cell activity according to behavioral context in order to regulate motor learning and 452 motor output.

453 Author Contributions

454 EF and CH designed the study, conducted experiments, and wrote the manuscript.

455 References

456 Albergaria C, Silva NT, Pritchett DL, and Carey MR. Locomotor activity modulates associative learning in 457 mouse cerebellum. Nat Neurosci 21: 725-735, 2018. 458 Albus JS. A theory of cerebellar function. Mathematical Biosciences 10: 25-61, 1971. 459 Athilingam JC, Ben-Shalom R, Keeshen CM, Sohal VS, and Bender KJ. Serotonin enhances excitability 460 and gamma frequency temporal integration in mouse prefrontal fast-spiking interneurons. Elife 6: 2017. 461 Billings G, Piasini E, Lorincz A, Nusser Z, and Silver RA. Network Structure within the Cerebellar Input 462 Layer Enables Lossless Sparse Encoding. 83: 960-974, 2014. 463 Bishop GA, Ho, R.H. The distribution and origin of serotonin immunoreactivity in the rat cerebellum. 464 Brain Res 331: 195-207, 1985. 465 Brickley SG, Cull-Candy SG, and Farrant M. Development of a tonic form of synaptic inhibition in rat 466 cerebellar granule cells resulting from persistent activation of GABAA receptors. J Physiol 497 ( Pt 3): 467 753-759, 1996. 468 Chabrol FP, Arenz A, Wiechert MT, Margrie TW, and DiGregorio DA. Synaptic diversity enables 469 temporal coding of coincident multisensory inputs in single neurons. Nat Neurosci 18: 718-727, 2015. 470 Chadderton P, Margrie TW, and Hausser M. Integration of quanta in cerebellar granule cells during 471 sensory processing. Nature 428: 856-860, 2004. 472 Clarke HF, Dalley, J.W., Crofts, H.S., Robbins, T.W., Roberts, A.C. Cognitive inflexibility after prefrontal 473 serotonin depletion. Science 304: 878-880, 2004. 474 Cohen JY, Amoroso MW, and Uchida N. Serotonergic neurons signal reward and punishment on 475 multiple timescales. Elife 4: e06346, 2015. 476 Courtemanche R, Chabaud P, and Lamarre Y. Synchronization in primate cerebellar granule cell layer 477 local field potentials: basic anisotropy and dynamic changes during active expectancy. Front Cell 478 Neurosci 3: 6, 2009. 479 D'Angelo E, and De Zeeuw CI. Timing and plasticity in the cerebellum: focus on the granular layer. 480 Trends in 32: 30-40, 2009. 481 Dieudonne S. Serotonergic Neuromodulation in the Cerebellar Cortex: Cellular, Synaptic, and Molecular 482 Basis. Neuroscientist 7: 207-219, 2001. 483 Dieudonne S, and Dumoulin A. Serotonin-driven long-range inhibitory connections in the cerebellar 484 cortex. J Neurosci 20: 1837-1848, 2000. 485 Duguid I, Branco T, Chadderton P, Arlt C, Powell K, and Hausser M. Control of cerebellar granule cell 486 output by sensory-evoked Golgi cell inhibition. Proc Natl Acad Sci U S A 112: 13099-13104, 2015. 487 Duguid I, Branco T, London M, Chadderton P, and Hausser M. Tonic inhibition enhances fidelity of 488 sensory information transmission in the cerebellar cortex. J Neurosci 32: 11132-11143, 2012. 489 Eccles J, Ito M, and Szentágothai J. The Cerebellum as a Neuronal Machine. Berlin: Springer, 1967, p. 490 335-335. 491 Geurts FJ, De Shutter, E., Timmermans, J. Localization of 5-HT2A, 5-HT3, 5-HT5A and 5-HT7 receptor- 492 like immunoreactivity in the rat cerebellum. J Chem Neuroanat 24: 65-74, 2002. bioRxiv preprint doi: https://doi.org/10.1101/411017; this version posted September 6, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

493 Hamann M, Rossi DJ, and Attwell D. Tonic and spillover inhibition of granule cells control information 494 flow through cerebellar cortex. Neuron 33: 625-633, 2002. 495 Hannon J, and Hoyer D. Molecular biology of 5-HT receptors. Behav Brain Res 195: 198-213, 2008. 496 Hull C, and Regehr WG. Identification of an inhibitory circuit that regulates cerebellar Golgi cell activity. 497 Neuron 73: 149-158, 2012. 498 Hull CA, Chu Y, Thanawala M, and Regehr WG. Hyperpolarization induces a long-term increase in the 499 spontaneous firing rate of cerebellar Golgi cells. J Neurosci 33: 5895-5902, 2013. 500 Jakab RL, Goldman-Rakic, P.S. 5-hydroxytryptamine2A sertonin receptors in the primate cerebral 501 cortex: possible site of action of hallucinogenic and antipsychotic drugs in apical 502 . Proc Natl Acad Sci U S A 95: 735-740, 1998. 503 Kaneda M, Farrant M, and Cull-Candy SG. Whole-cell and single-channel currents activated by GABA 504 and glycine in granule cells of the rat cerebellum. J Physiol 485: 419-435, 1995. 505 Kanichay RT, and Silver RA. Synaptic and cellular properties of the feedforward inhibitory circuit within 506 the input layer of the cerebellar cortex. J Neurosci 28: 8955-8967, 2008. 507 Kawashima T, Zwart MF, Yang CT, Mensh BD, and Ahrens MB. The Serotonergic System Tracks the 508 Outcomes of Actions to Mediate Short-Term Motor Learning. Cell 167: 933-946 e920, 2016. 509 Kehne JH, Baron, B.M., Carr, A.A., Shaney, S.F., Elands, J., Feldman, D.J., Frank, R.A., Van Giersbergen, 510 P.L.M., McCloskey, T.C., Johnson, M.P., McCarty, D.R., Poirot, M., Senyah, Y., Siegel, B.W., Widmaier, 511 C. Preclinical characterization of the potential of the putative atypical antipsychotic MDL 100,907 as the 512 potent 5-HT2A antagonist with a favorable CNS safety profile. The Journal of pharmacology and 513 experimental therapeutics 277: 968-981, 1996. 514 Kennedy A, Wayne G, Kaifosh P, Alvina K, Abbott LF, and Sawtell NB. A temporal basis for predicting 515 the sensory consequences of motor commands in an electric fish. Nat Neurosci 17: 416-422, 2014. 516 Kruglikov I, and Rudy B. Perisomatic GABA release and thalamocortical integration onto neocortical 517 excitatory cells are regulated by neuromodulators. Neuron 58: 911-924, 2008. 518 Marr D. A theory of cerebellar cortex. J Physiol 202: 437-470, 1969. 519 Matias S, Lottem E, Dugue GP, and Mainen ZF. Activity patterns of serotonin neurons underlying 520 cognitive flexibility. Elife 6: 2017. 521 Medina JF, Garcia KS, Nores WL, Taylor NM, and Mauk MD. Timing mechanisms in the cerebellum: 522 testing predictions of a large-scale computer simulation. The Journal of : the official journal 523 of the Society for Neuroscience 20: 5516-5525, 2000. 524 Mitchell SJ, and Silver RA. Shunting inhibition modulates neuronal gain during synaptic excitation. 525 Neuron 38: 433-445, 2003. 526 Oostland M, Buijink MR, Teunisse GM, von Oerthel L, Smidt MP, and van Hooft JA. Distinct temporal 527 expression of 5-HT(1A) and 5-HT(2A) receptors on cerebellar granule cells in mice. Cerebellum 13: 491- 528 500, 2014. 529 Ozden I, Dombeck DA, Hoogland TM, Tank DW, and Wang SS. Widespread state-dependent shifts in 530 cerebellar activity in locomoting mice. PLoS One 7: e42650, 2012. 531 Puia G, Costa E, and Vicini S. Functional diversity of GABA-activated Cl- currents in Purkinje versus 532 granule neurons in rat cerebellar slices. Neuron 12: 117-126, 1994. 533 Rossi DJ, and Hamann M. Spillover-mediated transmission at inhibitory synapses promoted by high 534 affinity alpha6 subunit GABA(A) receptors and glomerular geometry. Neuron 20: 783-795, 1998. 535 Rossi DJ, Hamann M, and Attwell D. Multiple modes of GABAergic inhibition of rat cerebellar granule 536 cells. J Physiol 548: 97-110, 2003. 537 Sorenson SM, Kehne, J.H., Fadayel, G.M., Humphreys, T.M., Ketteler, H.J., Sullivan, C.K., Taylor, V.L., 538 Schmidt, C.J. Characterization of the 5-HT2 receptor antagonist MDL 100907 as a putative atypical 539 antipsychotic: behavioral, electrophysiological and neurochemical studies. The Journal of pharmacology 540 and experimental therapeutics 266: 684-691, 1993. bioRxiv preprint doi: https://doi.org/10.1101/411017; this version posted September 6, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

541 Sullivan Hanley NR, and Hensler JG. Mechanisms of ligand-induced desensitization of the 5- 542 hydroxytryptamine2A receptor. The Journal of pharmacology and experimental therapeutics 300: 468- 543 477, 2002. 544 Takeuchi Y, Kimura, H., Matsuura, T., Yonezawa, T., Sano, Y. Distribution of serotonergic neurons in the 545 : a peroxidase-antiperoxidase study with anti-serotonin antibodies. J Histochem 546 Cytochem 31: 181-185, 1983. 547 Tang ZQ, and Trussell LO. Serotonergic Modulation of Sensory Representation in a Central Multisensory 548 Circuit Is Pathway Specific. Cell Rep 20: 1844-1854, 2017. 549 Tang ZQ, and Trussell LO. Serotonergic regulation of excitability of principal cells of the dorsal cochlear 550 nucleus. J Neurosci 35: 4540-4551, 2015. 551 van Welie I, and du Lac S. Bidirectional control of BK channel open probability by CAMKII and PKC in 552 medial vestibular nucleus neurons. J Neurophysiol 105: 1651-1659, 2011. 553 Villalobos C, Beique JC, Gingrich JA, and Andrade R. Serotonergic regulation of calcium-activated 554 potassium currents in rodent prefrontal cortex. Eur J Neurosci 22: 1120-1126, 2005. 555 Wall MJ, and Usowicz MM. Development of action potential-dependent and independent spontaneous 556 GABAA receptor-mediated currents in granule cells of postnatal rat cerebellum. Eur J Neurosci 9: 533- 557 548, 1997. 558 bioRxiv preprint doi: https://doi.org/10.1101/411017; this version posted September 6, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

A B PF stim 1 stim 2

Stim V-clamp control GoC 5-HT 20 pA NBQX 10 ms

pA

GrC MF n.s. n.s. n.s. n.s. C 1.0 0 0 1.5 -50 -50 0.5 -100 PPR 1.0 -100

failure rate -150 potency (pA) 0.0 mean amp (pA) -150

control 5-HT control 5-HT control 5-HT control 5-HT

D PF E -60 5-HT

-70 I-clamp GoC Vm (mV) -80

pA

GrC MF 0 2 4 6 time (min) F control 5-HT G 30 control

20 5-HT +16 pA 10

firing rate (Hz) 0 -4 pA 20 mV 0 10 20 50 ms current (pA)

Figure 1. 5-HT does not alter evoked excitation onto granule cells or granule cell membrane potential directly. A. Schematic of the voltage-clamp recording configuration. B. Whole-cell voltage-clamp recordings from an example granule cell. Averaged EPSCs recorded in control, 5-HT (10 mM) and NBQX (5 mM) at - 70 mV. C. Summary of paired-pulse ratio, failure rate, mean amplitude, and potency for recorded granule cells. Overall mean and SEM (solid bars) are overlaid (n = 10). n.s. = statistically non-significant. D. Schematic of current-clamp recording configuration. E. Average membrane potential of granule cells before and after 5- HT application (n = 11) in the presence of NBQX (5 mM) CPP (5 mM) and SR95531 (5 mM). F. Whole-cell current- clamp recordings from two example granule cells given a series of 2 pA steps in the presence of NBQX (5 mM) CPP (5 mM) and SR95531 (5 mM). G. Average relationship between firing rate and current steps (2 pA steps, input/output relationship) for granule cells with synaptic blockers only (control, n = 13) or synaptic blockers and 5-HT (n = 10). bioRxiv preprint doi: not certifiedbypeerreview)istheauthor/funder.Allrightsreserved.Noreuseallowedwithoutpermission. change in conductance during application 5-HT in control (open, n = 6) and 1 M. μM). (1 time of modulation, peak and 5-HT all experiments were performed in TTX (gray, n = 4). potential (Vm) over time added with to 5-HT control (open, n = 7) and MDL epne idw n oto ad n D. MDL. in and control in window response (gray, n = 4). normalized spike rates across time in control (open, n = 10) and in MDL MDL in and 10) = n (open, control in time across rates spike normalized ) n S951 (5 SR95531 and mM) receptor-specific antagonist, MDL 100907 (500 nM). (5 NBQX during bath application of 5-HT (10 μM), with and without the 5-HT2A 5-HT2A the without and with μM), (10 5-HT of application bath during individual whole-cell,current-clamprecordingsfromGolgicellsbeforeand ceai o te eodn cngrto. configuration. recording the of Schematic Figure 2. 5-HT depolarizes Golgi cells by activating 5-HT2A receptors. MDL MDL (gray, n = 5) as measured by a 5 mV test pulse. MDL MDL and MDL + 5-HT. Membrane potentials in 5-HT were measured at the https://doi.org/10.1101/411017 E Vm (mV) A GrC -50 -60

pA 0 GoC I-clamp .. saitcly o-infiat non-significant statistically = n.s. MF PF 4 2 time (min) F. Summary of mean membrane potential during control, 5-HT, D. Summary of mean normalized spike rates during peak 5-HT B

5-HT MDL control 500 ms 20mV +5-HT 5-HT ; 6 this versionpostedSeptember6,2018. ) rsn i al eodns recordings. all in present mM) C Vm (mV) F -50 -40 -70 -60 spike rate (norm) 3 4 2 1 0 control

* 4 2 5-HT time (min) 5-HT MDL xml tae fo two from traces Example B. umr o normalized of Summary G. umr o membrane of Summary E. n.s. +MDL 5-HT 6 The copyrightholderforthispreprint(whichwas

G D conductance (norm) spike rate (norm) umr of Summary C. 4 0 0 6 2 2 1 5-HT M) mM) CPP (5 5-HT * * +MDL +MDL 5-HT 5-HT A.

bioRxiv preprint doi: not certifiedbypeerreview)istheauthor/funder.Allrightsreserved.Noreuseallowedwithoutpermission. -55 mV A waveforms between subsequent conditions. view of the afterhyperpolarization (AHP). Bottom, subtractions of AP AP of subtractions Bottom, (AHP). afterhyperpolarization the of view rate across experiments (n = 5). Error bars are S.E.M. potentiation prior to 5-HT application . Bottom, mean normalized firing normalized mean Bottom, . application 5-HT to prior potentiation hyperpolarizing step (-50 pA, 3 min) was applied to induce firing rate firing induce to applied was min) 3 pA, (-50 step hyperpolarizing eodn soig cin oetas rm n xeiet hr a where experiment an from potentials action showing recording Bottom, subtraction between conditions. (5 NBQX 3 cells). Inset shows expanded view of the afterhyperpolarization (AHP). following 5-HT application in the subset of cells with stable AP height (n = vrgd og cl ato ptnil A) aeom i cnrl and control in waveforms (AP) potential action cell Golgi averaged clddb yeplrzto-nue rn aeptnito. potentiation. rate firing hyperpolarization-induced by occluded 5H-eitd nrae n og cl ectblt i not is excitability cell Golgi in increase 5-HT-mediated 3. Figure AP waveforms (n = 4 cells) for each condition. Inset shows expanded shows Inset condition. each for cells) 4 = (n waveforms AP M) included in all recordings. B. Top, (5mM)includedinallrecordings. ExampleGolgicell SR95531

D Vm (mV) https://doi.org/10.1101/411017 0 4 2 5-HT -control 5-HT 5-HT control 2 mV 2 mV 5 ms

spike rate (norm) B 0 5 ; 0 control this versionpostedSeptember6,2018. hyp 10 hyp time (min) post-hyp 20 1s 20 mV 5-HT 5-HT 30 -55 mV M) mM) CPP (5 C D Vm (mV) The copyrightholderforthispreprint(whichwas C. Top, averaged 0 4 2 5-HT -post-hyp 5-HT post-hyp -control 5-HT post-hyp control mM) and Top,A. 2 mV 2 mV 5 ms

bioRxiv preprint doi: not certifiedbypeerreview)istheauthor/funder.Allrightsreserved.Noreuseallowedwithoutpermission. https://doi.org/10.1101/411017 D C A V-clamp S.E.M. n.s. = statistically non-significant. individual granule cells. Solid black lines are mean and current quantified at the time of peak 5-HT effect for effect 5-HT peak of time the at quantified current n = 9) and MDL (gray, n = 7). Right, mean holding mean Right, 7). = n (gray, MDL and 9) = n normalized holding current across time in control (open, lc lns r ma ad ... S.E.M. and mean are lines black of peak 5-HT effect for individual granule cells. Solid cells. granule individual for effect 5-HT peak of 5). Right, frequency mean quantified sIPSC at the time across time in control (open, n = 13) and MDL (gray, n = niioy otsnpi ptnil ISs frequency (IPSCs) potential post-synaptic inhibitory (right). (right). eetratgns D 097(ide r receptor antagonist MDL 100907 (middle) or SR95531 application of 5-HT (left), with and without the 5-HT2A (left),withandwithout the5-HT2A application of5-HT potential for excitation (0mV) before and during bath during and before (0mV) excitation for potential eodns rm gaue el hl a te reversal the at held cells granule 3 from recordings ofiuain configuration. eedn manner. dependent tonic holding current on granule cells in a 5-HT2AR- a in cells granule on current holding tonic Figure 4 Iholding (norm) sIPSC freq (norm) 10

1 2 0 5 GrC 0 0

pA C. . 5-HT increasesspontaneousinhibitionand 5-HT . GoC ; 4 2 4 2 Left, Mean normalized spontaneous spontaneous normalized Mean Left, this versionpostedSeptember6,2018. time (min) time (min) MF PF Example whole-cell, voltage-clamp whole-cell, Example B. 5-HT 5-HT B 5H MDL+5HT +5-HT ceai o te recording the of Schematic A. 6 6

I (norm) sIPSC freq (norm) MDL 15

holding 10 5 3 0 0 2 1 5-HT 5-HT * * D. +MDL 5-HT +MDL 5-HT The copyrightholderforthispreprint(whichwas 5-HT SR95531+ SR95531 et mean Left, n.s. ..n.s. n.s. 500 ms 20pA +SR 5-HT 5-HT 5-HT 5-HT +SR

bioRxiv preprint doi: not certifiedbypeerreview)istheauthor/funder.Allrightsreserved.Noreuseallowedwithoutpermission. https://doi.org/10.1101/411017 D population. n.s. = statistically non-significant. Solid lines reflect mean and S.E.M. across the across S.E.M. and mean reflect lines Solid n pa rsos fr niiul og cells. Golgi individual for response peak and Right, mean spike rates measured at baseline at measured rates spike mean Right, (CPP (5 μM), NBQX (5 μM), SR95531 (5 μM)). SR95531 (5 μM), NBQX (5 μM), (CPP bak n 3 snpi tasiso blockers transmission synaptic 3) = n (black, cell spike rate with (open, n = 10) and without and 10) = n (open, with rate spike cell et ma woecl, urn-lm Golgi current-clamp whole-cell, mean Left, D. reflect mean and S.E.M. across the population. the across S.E.M. and mean reflect response for individual Golgi cells. Solid lines Solid cells. Golgi individual for response IS rates sIPSC application of (10 5-HT C. Left, mean frequency IPSC before and during mV) of before IPSCs and after 5-HT application. Example Golgi cell voltage-clamp recording (0 recording voltage-clamp cell Golgi Example ceai o te eodn cngrto. configuration. recording the of Schematic f5H-nue pk aeehneet. . enhancement rate spike 5-HT-induced of sIPSCs on Golgi cells, reducing the magnitude the reducing cells, Golgi on sIPSCs 5H icess h feuny of frequency the increases 5-HT 5. Figure C A spike rate (norm) sIPSC freq (Hz) 10 20 3 0 2 1 0 0 GrC ; 5-HT 5-HT

this versionpostedSeptember6,2018.

pA 2 2 GoC time (min) time (min) V-clamp esrd t aeie n peak and baseline at measured MF PF SR95531 4 4 B μM; n = 5). 5-HT control SR95531 6 6

sIPSC freq (Hz) spike rate (norm) 10 20 6 0 0 4 2 blockers synaptic Right, mean control The copyrightholderforthispreprint(whichwas n.s. *** 400 ms 50pA aCSF 5-HT B. A.

bioRxiv preprint doi: not certifiedbypeerreview)istheauthor/funder.Allrightsreserved.Noreuseallowedwithoutpermission. https://doi.org/10.1101/411017 C F E A = statistically non-significant. and S.E.M. across the population. n.s. cells (n = 18). Solid lines reflect mean measured from individual granule granule individual from measured latency, P P R and failure rate rate failure and R P P latency, br. fibers. to electrical stimulation of the mossy the of stimulation electrical to granule cell held at 0 mV in response in mV 0 at held cell granule evoked IPSCs from an example example an from IPSCs evoked the recording configuration. nogauecls cells. granule onto le eoe fe-owr inhibition feed-forward evoked alter 5-HT does not significantly 5-HT Figure 6. V-clamp

PPR mean amp (pA) ; this versionpostedSeptember6,2018. GrC 25 50 3 0 0 2 1 control control pA Mean IPSC amplitude, IPSC C-F.Mean n.s. n.s. GoC MF 5-HT 5-HT PF Stim D B Schematic of Schematic A. failure rate latency (ms) stim 0.5 1.0 0.0 15 10 5 0 control control B. Mean n.s. n.s. 10 ms 10 pA 5-HT 5-HT 5-HT control NBQX The copyrightholderforthispreprint(whichwas

bioRxiv preprint doi: https://doi.org/10.1101/411017; this version posted September 6, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

stim: 1 2 3 stim: 1 2 3 A B C 0 PF control control spike + 5-HT + 5-HT 50 + MDL + MDL

cell-attached Stim trial 100 + gabazine GoC + gabazine

GrC + NBQX pA 150 MF 20 pA + NBQX 5 ms 20 ms

stim: 1 2 3 n.s. n.s.

D E * control *

n.s. n.s. 5-HT * 2 *** 6 n.s. 2 n.s. MDL ** 2 gabazine NBQX 4 1 1 1 2 S.D. (ms) latency (ms) spike probability mean spike number

(norm to max control) 0 0 0 0 L 0 20 40 60 80 T T 5-H MDL 5-H MD 5-HT MDL time (ms) control NBQX control control gabazine gabazine gabazine

Figure 7. 5-HT reduces mossy fiber-evoked spike probability of granule cells in a 5-HT2A- dependent manner. A. Schematic of the recording configuration. B. Example traces from a cell-attached granule cell recording during electrical stimulation of mossy fibers and sequential bath application of 5-HT, MDL 100907, gabazine and NBQX. C. Spike raster from the same example cell in B. D. Mean spike probability for granule cells across conditions, (n = 4). E. Summary of mean spike number (left), mean first spike latency (middle) and spike jitter (right). n.s. = statistically non-significant.