Purkinje cell outputs selectively inhibit a subset of unipolar brush cells in the input layer of the cerebellar cortex

  1. Chong Guo
  2. Stephanie Rudolph
  3. Morgan E Neuwirth
  4. Wade G Regehr  Is a corresponding author
  1. Department of Neurobiology, Harvard Medical School, United States

Abstract

Circuitry of the cerebellar cortex is regionally and functionally specialized. Unipolar brush cells (UBCs), and Purkinje cell (PC) synapses made by axon collaterals in the granular layer, are both enriched in areas that control balance and eye movement. Here, we find a link between these specializations in mice: PCs preferentially inhibit metabotropic glutamate receptor type 1 (mGluR1)-expressing UBCs that respond to mossy fiber (MF) inputs with long lasting increases in firing, but PCs do not inhibit mGluR1-lacking UBCs. PCs inhibit about 29% of mGluR1-expressing UBCs by activating GABAA receptors (GABAARs) and inhibit almost all mGluR1-expressing UBCs by activating GABAB receptors (GABABRs). PC to UBC synapses allow PC output to regulate the input layer of the cerebellar cortex in diverse ways. Based on optogenetic studies and a small number of paired recordings, GABAAR-mediated feedback is fast and unreliable. GABABR-mediated inhibition is slower and is sufficiently large to strongly influence the input-output transformations of mGluR1-expressing UBCs.

Introduction

Although different lobules of the cerebellar cortex are engaged in diverse motor and non-motor behaviors (Ito, 1998; Kim and Thompson, 1997; Massion, 1992; Murdoch, 2010; Schmahmann and Sherman, 1998; Siuda et al., 2014; Van Overwalle et al., 2014; Villanueva, 2012), a basic circuit motif is repeated within each functional compartment. Mossy fibers provide excitatory inputs to granule cells (GrCs) (Delvendahl and Hallermann, 2016; DiGregorio et al., 2002; Huang et al., 2013; Sotelo, 2008), which in turn excite Purkinje cells (PCs) that provide the output of the cerebellar cortex. In addition, Golgi cells (GoCs) inhibit GrCs, and molecular layer interneurons inhibit PCs (Eccles, 2013). Regional synaptic and cellular specializations exist beyond this basic circuit motif, presumably to meet the computational demands associated with specific behaviors (Cerminara et al., 2015; Diño et al., 1999; Guo et al., 2016; Kozareva et al., 2021; Sekerková et al., 2014; Suvrathan et al., 2016).

One of the most obvious specializations is that the density of unipolar brush cells (UBCs) exhibits large regional variation (Diño et al., 1999; Sekerková et al., 2014). UBCs are excitatory interneurons located in the granular layer that are usually innervated by a single MF (Mugnaini and Floris, 1994). UBCs contribute to temporal processing by converting short-lived MF signals into long-lasting changes in UBC activity (Kennedy et al., 2014; Kinney et al., 1997; Kreko-Pierce et al., 2020; Locatelli et al., 2013; Mugnaini and Floris, 1994; Mugnaini et al., 2011; Rossi et al., 1995; van Dorp and De Zeeuw, 2014). MFs excite UBCs that express metabotropic glutamate receptor type I (mGluR1), suppress firing in UBCs where mGluR2 is prominent, and evoke more complex responses in other UBCs that express both mGluR1 and mGluR2 (Borges-Merjane and Trussell, 2015; Guo et al., 2021; Kinney et al., 1997; Knoflach and Kemp, 1998; Rossi et al., 1995; Russo et al., 2008; van Dorp and De Zeeuw, 2014; Zampini et al., 2016). UBCs are often disregarded in descriptions of the circuitry of the cerebellar cortex. However, UBCs are present in all areas and are exceptionally dense in some regions, notably those involved in vestibular function (Diño et al., 1999; Takács et al., 1999). UBCs are more prevalent and widely distributed in larger mammals including humans (Diño et al., 1999; Munoz, 1990), suggesting a special role in the complex cerebellar computations relating to higher cognition.

PC feedback via collaterals that are restricted to parasagittal planes is another noteworthy regional specialization (Witter et al., 2016). In addition to sending an axon to the deep cerebellar nuclei, each PC axon has a collateral that inhibits PCs (Bernard and Axelrad, 1993; Bernard et al., 1993; Bornschein et al., 2013; de Solages et al., 2008; Orduz and Llano, 2007; Watt et al., 2009; Witter et al., 2016) and several types of inhibitory interneurons in the cerebellar cortex (Crook et al., 2007; Hirono et al., 2012; Witter et al., 2016). In some regions, PCs make extensive contacts within the granular layer and inhibit GrCs (Guo et al., 2016). This feedback is mediated primarily by GABAA receptors (GABAARs) and has a prominent slow component that suggests the involvement of the extrasynaptic high-affinity GABAARs. These PC to GrC synapses allow the output of the cerebellar cortex to provide slow negative feedback to inhibit the input layer. Like UBCs, these synapses are also most prominent in regions involved in vestibular function.

The regional overlap of high densities of UBCs and granular layer PC synapses raises the possibility that PCs might inhibit UBCs and thereby allow PC feedback to refine temporal processing. Here, we examine whether PCs inhibit UBCs. We find that PCs phasically inhibit a subset of UBCs by activating GABAARs, and that UBCs with prominent mGluR1 signaling are preferentially targeted. PCs also provide long-lasting inhibition to most mGluR1-expressing UBCs by activating GABAB receptors (GABABRs). In these ways, PC inhibition of mGluR1-expressing UBCs provides direct feedback from the output layer to input layer to influence the first stage of cerebellar processing.

Results

Regional distribution of PC collaterals and mGluR1+ UBCs

High-power confocal imaging of vermal and floccular slices was used to assess the regional density of PC collateral synapses and UBCs, and to detect putative synaptic connections between PCs and UBCs. PC presynaptic boutons were labeled in Pcp2-Cre×synaptophysin-tdTomato mice (n=2). Weaker labeling was evident in PC dendrites and somata (red, Figure 1A). Inhibitory synapses were labeled with vesicular GABA transporter (VGAT) immunofluorescence (green, Figure 1A). PC synapses were automatically identified with a custom deep neural network using VGAT and synaptophysin-tdTomato immunofluorescence (Figure 1—figure supplement 1, see Materials and methods). For visualization, annotated PC synapses were colored in cyan, and the PC and molecular layers are shown in gray based on the synaptophysin-tdTomato signal (Figure 1B). In the same slices, mGluR1 immunofluorescence labeled UBCs and PCs. mGluR1 labeling of UBC brushes is intense and distinctive (Borges-Merjane and Trussell, 2015; McDonough et al., 2020; Nunzi et al., 2002). In our images, mGluR1 labeling of UBCs dominates the signals in both the high and low magnification of the granular layer. For display purposes, mGluR1 labeling in the PC and molecular layers was masked to highlight UBCs in the granular layer (magenta, Figure 1C-K).

Figure 1 with 1 supplement see all
PC collateral synapses colocalized with mGluR1 UBCs in the vestibular lobules.

(A) Maximal intensity projection of a 15-µm confocal z-stack images of PC collateral synapses co-labeled by synaptophysin-tdTomato (red) and VGAT (green). (B) Annotated PC collateral synapses (cyan) and the synaptophysin-tdTomato signal (gray). (C) A vermal slice of cerebellum analyzed as in (B) with synapses labeled in cyan, synaptophysin-tdTomato in gray, and the mGluR1 labeled UBCs dendritic brush in magenta. (D) Same analysis on a coronal slice of a floccular lobe. (E–I) Zoomed in view of selective lobules showing variable degrees of colocalization between PC synapses and UBCs. (J) Expanded view of lobule X. (K) Two-sample UBCs showing clear examples of collateral synapses onto the brush and the soma. mGluR1, metabotropic glutamate receptor type 1; PC, Purkinje cell; UBC, unipolar brush cell; VGAT, vesicular GABA transporter.

The densities of UBCs and granular layer PC synapses exhibited considerable regional variability. The highest densities of both UBCs and granular layer PC synapses were found in vestibular regions (lobules IX, X, FL, and PFL; see Figure 1C, D and F-I) and in the oculomotor vermis (lobules VIb and VII, Figure 1C). Much lower densities were observed in the anterior cerebellar cortex (lobules I–V and VIa; see Figure 1C,E) and lobule VIII. In lobule X, there was a particularly high density of PC collateral synapses and UBCs (Figure 1J). High-power images of lobule X suggest that PC synapses directly terminate on the brushes and the cell bodies of many UBCs (Figure 1K).

PC collaterals provide GABAAR-mediated inhibition to mGluR1+ UBCs

To determine whether PCs directly inhibit UBCs, we optically stimulated PC axons in Pcp2-Cre×ChR2 mice while recording from UBCs (Figure 2A). We performed these experiments in lobule X of acute slices in the presence of AMPA, NMDA, glycine, and GABABR blockers and a high chloride internal solution. UBCs exhibit diverse responses to MF activation, ranging from excitation mediated by AMPARs and mGluR1, to inhibition mediated by mGluR2, with intermediate cells having more diverse responses mediated by a combination of mGluR1 and mGluR2 (Guo et al., 2021). Electrophysiological studies have shown that fast inhibition in UBCs is mediated by a combination of glycine and GABAARs (Dugué et al., 2005; Rousseau et al., 2012), but that mGluR1-lacking UBCs lack GABAARs (Rousseau et al., 2012). Consistent with these observations, molecular profiling of UBCs using single-cell RNA sequencing (scRNASeq) (Kozareva et al., 2021) indicate that α1, β2, β3, and γ2 subunits of GABAARs have a gradient of expression levels and are selectively enriched in mGluR1-positive UBCs (Figure 2—figure supplement 1B), but that glycine receptors are broadly expressed by UBCs (Figure 2—figure supplement 1C). This opens the possibility that the molecular identity of UBCs could determine whether they receive PC feedback. We therefore characterized UBC responses to brief puffs of glutamate (Figure 2B,C left) prior to measuring optically evoked synaptic responses (Figure 2B,C right). Glutamate (1mM, 50ms) evoked a continuous range of metabotropic responses across UBCs with variable degrees of excitation versus inhibition, as is apparent in the sorted heatmap of glutamate-evoked response (Figure 2D). Optical stimulation evoked currents in a subpopulation of these cells (Figure 2B right, Figure 2D asterisk marks, 17/41 cells) but not in others (Figure 2C, right, 24/41 cells). GABAAR antagonist gabazine eliminated these light-evoked synaptic currents (Figure 2E). We compared the average responses to a glutamate puff for UBCs that had light-evoked GABAA currents with those that did not. Cumulative histograms for glutamate-evoked charge transfer (Figure 2F), and the average glutamate-evoked current for UBCs with or without PC feedback (Figure 2G), show that PC GABAA feedback preferentially targets UBCs that are excited by glutamatergic inputs.

Figure 2 with 1 supplement see all
PC collaterals preferentially inhibit mGluR1+ UBCs via fast GABAAR-mediated feedback.

(A) Schematic showing a mossy fiber (MF) and a PC collateral innervating a UBC. A whole-cell electrode was used to voltage clamp the UBC, and responses were measure for either a glutamate puff applied with a nearby electrode or to optical stimulation of PC collaterals. (B) Whole-cell recordings of glutamate-evoked currents (left) and to optically evoked PC inhibition (right) are shown for cell #1 of (D). (C) As in (B), but for cell #41 of (D). (D) Summary of cells in which responses to glutamate puffs were measured, with red corresponding to an excitatory inward current and blue corresponding to an inhibitory outward current. Cells with light-evoked inhibitory synaptic current are indicated with a black dot (n=41). (E) Light-evoked synaptic currents were recorded in baseline (black) and after the addition of the GABAA receptor antagonist gabazine (gray). (F) Normalized cumulative distributions of charge evoked by a glutamate puff for cells that either had light-evoked responses (black) or that did not (gray). (G) Average currents evoked by a glutamate puff for cells that had PC feedback (black) and that did not have PC feedback (gray). (H) Schematic as in (A), but for application of the mGluR1 agonist DHPG. DHPG-evoked increases in firing were measured with a cell-attached electrode, and then optically evoked synaptic currents were measured with a whole-cell voltage-clamp recording. (I) DHPG-evoked increases in firing (left) and light-evoked synaptic currents (right) are shown for cell #35 in (K). (J) Same as (I) but for cell #45 in (K). (K) Summary of cells in which firing evoked by DHPG puffs was quantified. Cells with light-evoked inhibitory synaptic current are indicated with a black dot (n=60). (L) Normalized cumulative distributions of charge evoked by a DHPG puff for cells that either had light-evoked responses (black) or that did not (gray). (M) Average currents evoked by a DHPG puff for cells that had PC feedback (black) and that did not have PC feedback (gray). (N) Scatter plot of IPSC amplitudes as a function of the peak firing rates for the cells in (K). DHPG, (S)-3,5-Dihydroxyphenylglycine; IPSC, inhibitory post-synaptic current; mGluR1, metabotropic glutamate receptor type 1; PC, Purkinje cell; VGAT, vesicular GABA transporter.

To further characterize the properties of UBCs that receive PC feedback, we categorized UBCs based on their responses to the mGluR1 agonist (S)-3,5-Dihydroxyphenylglycine (DHPG) prior to characterizing light-evoked synaptic responses (Figure 2H). DHPG (100mM, 50 ms) puffs evoked spiking in a subset of UBCs (Figure 2I,J, left, K 45/60). After recording the DHPG-evoked increase in firing with a cell-attached electrode, we obtained a whole-cell configuration and measured light-evoked responses. 13/45 DHPG-responsive UBCs also had significant light-evoked responses (Figure 2I,J, right, K). We then grouped UBCs by the presence (Figure 2I, right and asterisk in K) or absence of PC feedback (Figure 2J, right). Cumulative DHPG-evoked peak firing in UBCs with PC feedback is shifted toward greater response than those without feedback (Figure 2L). Furthermore, the average DHPG-evoked instantaneous firing in UBCs with PC feedback is also slightly bigger than those without (Figure 2M). Cumulative histograms for DHPG-evoked firing rate increases in UBCs (Figure 2L), and the average DHPG-evoked firing rate increases for UBCs with or without PC feedback (Figure 2M), show that PC GABAAR-mediated feedback preferentially targets UBCs with larger mGluR1-mediated excitation. There is considerable scatter in the amplitudes of the average GABAA responses evoked by optical stimulation (Figure 2N).

We also directly recorded from connected PC-UBC pairs (Figure 3, n=2). An on-cell recording allowed us to noninvasively monitor PC spiking while recording synaptic responses in a UBC. Many UBC inhibitory post-synaptic currents (IPSC)s were timed to PC spikes (Figure 3A, top), as is readily appreciated in spike-triggered averages of UBC IPSCs (Figure 3B). The IPSC latency was 1.5±0.4 ms (Figure 3C,D,F) and the distribution of response amplitude was approximately Gaussian (Figure 3E). While the response potency is large (Figure 3F), there was a high failure rate of approximately 85% (Figure 3F).

Paired recordings of PC to UBC synaptic connections.

(A) Schematic of the paired recording configuration. (B) A representative pair showing a cell-attached PC recording (top) and whole-cell voltage-clamp recording of a synaptically connected UBC (bottom). (C) Same two cells showing time-aligned PC spikes (top) and associated IPSCs in UBCs (bottom), with successes in black, failures in gray, and average response from success trials in red. (D) Histogram of IPSC latency (bin size=0.1 ms). (E) Histogram of IPSC amplitude (bin size=4 pA) and Gaussian fit over success trials (red). (F) Failure rate (left), latency (middle), and potency (left) PC to UBC synapse (n=2 cells). IPSC, inhibitory post-synaptic current; PC, Purkinje cell; UBC, unipolar brush cell.

GABABR-mediated slow PC inhibition in mGluR1+ UBCs

One of our goals was to determine if PCs also activate GABABRs and thereby tonically inhibit UBCs. Previous studies established that only mGluR1-positive UBCs contain GABABRs (Kim et al., 2012). This is consistent with scRNASeq analysis of UBCs (Figure 2—figure supplement 1A and D). GABABRs comprises GB1 and GB2 subunits (Gassmann and Bettler, 2012) that are encoded by Gabbr1 and Gabbr2. UBCs exhibit a gradient of expression of Gabbr2 that is similar to Grm1 (mGluR1), and Gabbr1 has a less pronounced gradient of expression (Figure 2—figure supplement 1D). The differences in the expression patterns of Gabbr2 and Gabbr1 are intriguing, because it is thought that GABABRs are heterodimers consisting of GABAB1 and GABAB2 receptors (Frangaj and Fan, 2018). Nonetheless, these findings are consistent with mGluR1+ UBCs preferentially expressing GABABRs, and we therefore restricted our recordings to cells that responded to DHPG. We found that in whole-cell recordings the currents evoked by GABABR agonists washed out within minutes (Figure 4—figure supplement 1A and B, gray). This washout was accompanied by an increase in the leak current (Figure 4—figure supplement 1C). In contrast, in perforated patch recordings, GABAB responses were large and stable, and leak currents were also stable (Figure 4—figure supplement 1, black). These experiments established that perforated patch recordings facilitate the measurement of PC-evoked GABAB responses.

We examined the PC-evoked GABAB responses in UBCs with an approach similar to that used to record GABAA responses, except that we blocked GABAARs, used perforated patch recordings, and elevated PC firing for 2 s to activate the GABABRs on a longer time scale (Figure 4A). We stimulated the entire lobule X with graded light levels to approximately double and triple PC firing rates (Figure 4B–D). Optical stimulation evoked outward currents in all UBCs (Figure 4E–G, 13/13). GABABR-mediated inhibition was slow (Figure 4F, rise time constant=583±6 ms, decay time constant=533±5 ms). The GABABR antagonist CGP eliminated these light-evoked responses (Figure 4K). The average inhibition of 3.8±2.2 pA was sufficiently large to strongly suppress or entirely shut down spontaneous UBC firing (Figure 4H–J in 9/10 spontaneous firing UBCs). These findings suggest that PC firing rates dynamically regulate the activation of GABABRs, and by extension the spontaneous firing of most mGluR1+ UBCs. As with the direct GABAAR-mediated response, it is unclear whether PCs inhibit all mGluR1+ UBCs, or if we are unable to optogenetically activate PC inputs onto a small fraction of UBCs for technical reasons, such as severed and unhealthy PC axons in our brain slice preparation. With this GABABR-dependent mechanism, PCs inhibit a much larger fraction of mGluR1+ UBCs (90–100%, Figure 4G, 13/13 DHPG responding cells and Figure 4J, 9/10 spontaneous firing), compared to GABAAR-dependent inhibition (29%, Figure 2K, 13/45 DHPG-responding cells).

Figure 4 with 1 supplement see all
PC collaterals provide slow GABAB receptor-mediated inhibition onto lobule X UBCs.

(A) Schematic of full-field optical stimulation for modulating PC firing rate in lobule X. (B) Cell-attached recording of spontaneous PC spikes for no (top), low intensity (middle), and high intensity (bottom) full-field optical stimulations. (C) Average instantaneous firing rate of PCs with no (gray), low (light blue), and high (blue) intensity stimulations (n=13). (D) Summary of PC firing rate with no (gray), low (light blue), and high (blue) intensity light-stimulations of PCs (n=13). (E) Voltage-clamp recordings in UBC revealed a slow inhibitory response that is modulated by PC firing rate, same optical stimulation conditions as in (B). (F) Average evoked current of UBCs during stimulations as in (E) (n=13). (G) Summary of inhibitory current amplitude in UBCs with no (gray), low (light blue), and high (blue) intensity light-stimulations of PCs (n=13). (H) Full-field activation of PCs suppressed the frequency of UBC firing. (I) Average instantaneous firing rate of UBCs with no (gray), low (light blue), and high (blue) intensity stimulations (n=13). (J) Summary of firing rate changes in spontaneous firing UBCs with no (gray), low (light blue), and high (blue) intensity light-stimulations of PCs (n=10). (K) PC evoked slow outward currents in UBC in the presence of GABAAR blockers under voltage-clamp before (black) and after blocking GABABR (gray) (n=1). PC, Purkinje cell; UBC, unipolar brush cell.

GABABR-mediated PC-UBC inhibition temporally sharpens UBC response

In addition to suppressing spontaneous activity, PC firing and GABABR activation could control the excitability of UBCs and their response to depolarization. To assess the effect of small tonic hyperpolarizing currents on UBC excitability, we examined how such currents influence UBC responses to current steps using perforated patch recordings. In spontaneously firing UBCs, small depolarizing current steps evoked firing that declined to a somewhat reduced steady-state response that was proportional to the current injected (Figure 5A,B and C, gray left). A 5-pA tonic hyperpolarizing current, drastically altered these responses, with current steps evoking a much larger transient response and reduced steady-state firing (Figure 5A,B and C, black right). A summary of average spiking responses to different current step amplitudes of all UBCs without and with tonic hyperpolarization is shown in Figure 5B and C. In all cells, hyperpolarization increased the difference between peak and steady-state firing rates (Figure 5C, n=4). We also directly examined the effects of PC firing on UBC excitability. In these experiments, we blocked GABAARs to isolate the effect of GABABRs on spiking. Full-field optical excitation of PCs activated GABABR-mediated inhibition and produced qualitatively similar effects on spiking as hyperpolarizing current injections (Figure 5D, blue, right). Optical activation of PC feedback through GABABR hyperpolarized UBCs (Figure 5D, blue, right), and current steps evoked larger initial responses and smaller steady-state responses (Figure 5E,F, blue vs. gray). These counterintuitive effects on spiking, in which a hyperpolarization increases initial responses, are consistent with previously described properties of UBCs (Perez-Reyes, 2003). They suggest that the effects of PC inhibition of UBCs may be more complex than simply suppressing spiking in UBCs.

PC-UBC feedback via GABABR sharpens temporal response to current injections.

(A) Sample perforated current-clamp recording of UBC spiking response to an 8 pA current step (500 ms), without (gray, left) and with a hyperpolarizing tonic current (black, right). (B) Average instantaneous firing rate to current steps of (0, 2, 4, 6, and 8 pA, from light to dark shade) without (gray, left) and with a hyperpolarizing tonic current (black, right) (n=4). (C) Summary of peak firing rate (triangle marker, mean ± SEM) and steady-state firing rate (circle marker, mean ± SEM) for the two conditions as in (B) (n=4). (D) Same summary plot as in (A) for control (gray, left) versus optical PC feedback (blue, right). (E) Same summary plot as in (B) for control (gray, left) versus optical PC feedback (blue, right) (n=5). (F) Same summary plot as in (C) for control (gray, left) versus optical PC feedback (blue, right) (n=5). (G) In a sample UBC pressure-application of a mGluR1 agonist (DHPG 100 µM, 10 ms) evoked persistent spiking for a few seconds. (H) Pressure-applied GABABR agonist (baclofen 250 μM, 100 ms) 2 s following mGluR1 agonist application readily reduced the firing rate of persistent spiking response. (I) Same as in (B) but only 1 s following mGluR1 agonist application. (J) Average instantaneous firing rate of the sample UBC under mGluR1 agonist application only (black), mGluR1+ GABABR agonist applied 1 s apart (dark gray) or 2 s apart (light gray). PC, Purkinje cell; UBC, unipolar brush cell.

MF activity elevates firing in the input layer of the cerebellar cortex that in turn alters PC spiking. The PC to UBC feedback described here has the potential to allow PC output to dynamically regulate the firing of mGluR1-expressing UBCs. We have previously shown that a brief MF burst can activate mGluR1 to evoke increases in firing that last for more than 10 s in some UBCs (Guo et al., 2021). We briefly applied DHPG to crudely mimic MF-evoked activation of mGluR1-expressing UBCs, and baclofen to mimic GABABR activation arising from elevated PC firing (Figure 5G,J). Brief application of baclofen following DHPG reduced the number of evoked spikes (Figure 5H–J). Thus, activation of GABABRs, as could occur when PC activity is elevated, can strongly suppress mGluR1-activated UBC firing (see Discussion).

Discussion

The PC to UBC synapses described here link two previously described regional specializations of the cerebellar cortex: the high densities of UBCs and the prevalence of PC synapses within the granular layer. PC to UBC synapses provide a new way for PC outputs to dynamically regulate the temporal transformations by UBC populations within the input layer.

Target specificity of GABAA and GABAB PC to UBC feedback

PCs inhibit a subset of UBCs. Optical stimulation of PC terminals evoked GABAAR-mediated responses in 29% (30/101) of UBCs. Light-evoked PC GABAAR-mediated inhibition was much more common in cells in which glutamate evoked a large excitatory current but was also observed in three cases where glutamate evoked inhibitory currents (Figure 2D). This may reflect the graded nature of metabotropic glutamate receptor signaling in UBCs in which outward mGluR1-mediated currents and inward mGluR2-mediated currents can be present in the same cell, and a large mGluR2 component could obscure a small mGluR1 component in some cells (Guo et al., 2021; Kozareva et al., 2021). When we used DHPG to identify mGluR1+ cells, we found that PCs did not inhibit any mGluR1 UBCs (0/15), (Figure 2K). This is in keeping with the observations that mGluR1 UBCs do not respond to either GABAAR or GABABR agonists (Kim et al., 2012; Rousseau et al., 2012), that they do not express GABAAR or GABABR subunits (Rousseau et al., 2012Figure 2—figure supplement 1), and that PCs do not release glycine (Tanaka and Ezure, 2004). In contrast to PCs that only inhibit a subset of UBCs, glycinergic GoCs can inhibit all UBCs, because ionotropic glycine receptors and glycinergic IPSCs are present in all UBCs (Dugué et al., 2005; Rousseau et al., 2012).

Our findings provide new insights into the previous immunohistochemical characterization of the inhibitory synapses onto UBCs that suggested approximately 50% co-release GABA and glycine and the rest only release GABA (Dugué et al., 2005; Rousseau et al., 2012). Previously it was thought that these exclusively GABAergic synapses were made by non-glycinergic GoCs. Our findings suggest that PCs likely constitute most of these synapses.

PCs inhibit UBCs by activating both GABAARs and GABABRs, and light-evoked GABAB currents are observed in a larger percentage of mGluR1+ UBCs (>90%) than are light-evoked GABAA currents (29%). This difference in prevalence likely reflects differential sensitivities of GABAARs and GABABRs. Activation of GABAARs on UBCs requires high GABA levels (Möhler, 2006), and thus requires a direct PC synapse for effective activation. GABABRs are activated by much lower levels of GABA (Galvez et al., 2000; Kaupmann et al., 1998). Consequently, a direct PC input is not required and GABA released from PC synapses onto neighboring cells can pool and spillover to activate GABABRs. This also suggests that the combined activity of many PCs could efficiently activate GABABRs on UBCs and control spontaneous activity.

It is important to note that we have focused on PC feedback to UBCs in lobule X. Both UBCs and PC collaterals are also present throughout the cerebellar cortex, albeit at a lower density. It remains to be seen whether PC to UBC feedback occurs in other regions, and if it does, whether it has the same target dependence.

Comparisons to properties of PC-GrC feedback

Within the input layer, PCs also directly inhibit GrCs, and there are similarities and differences in the inhibition of these two targets. The failure rate and potency of the fast component of inhibition are similar for PC synapses onto UBCs and GrCs (Guo et al., 2016). Therefore, we hypothesize that the overall effect of fast and stochastic GABAAR PC inhibition of UBCs and GrCs is to lower the gain of the input layer. The slow GABAB feedback to UBCs occurs in concert with a slow inhibition of GrCs that is mediated by a very different mechanism (Guo et al., 2016). PCs slowly inhibit GrCs primarily by activating special high-affinity α6 and δ subunit-containing GABAARs that are specialized to respond to low levels of extracellular GABA (Brickley et al., 1996; Hamann et al., 2002; Kaneda et al., 1995; Wall and Usowicz, 1997). It appears that α6/δ containing GABAARs in GrCs and GABABRs in UBCs can both respond to the low GABA levels that reflect the population-averaged PC firing rates, and in both cases, the kinetics of the responses are slow (312 ms in GrCs vs. 580 ms in UBCs). The major differences between slow PC inhibition of UBCs and GrCs are the state change that can occur in UBCs, and the target dependence of the UBC inhibition, neither of which have been described for PC inhibition of GrCs (Guo et al., 2016).

Potential functional consequences of slow PC to UBC inhibition

Slow PC-UBC inhibition represents the average synaptic contribution from multiple PCs, and when the firing rates of these PCs are elevated UBCs are hyperpolarized. Increases in PC firing lead to slow increases in inhibition, and following decreases in PC firing rate this inhibition gradually decreases. Importantly, this can alter the state of the UBC. As has been shown previously, hyperpolarization can relieve the inactivation of T-type calcium channels and make neurons burst upon depolarization (Perez-Reyes, 2003). Concordantly, we observed rebound effects in some UBCs when PC firing returned to baseline (Figure 4H, last row and Figure 4I) as well as steeper initial input/output relationship for depolarizing current steps with PC inhibition (Figure 5E,F). It is likely that this component also allows PCs to constitutively regulate the properties of UBCs (although we have not addressed this possibility in our brain slice experiments where many PC axons and connections to UBCs are not intact). We conclude that PC firing may be able to alter UBC-mediated transformations in the input layer of the cerebellum. Slow inhibition could also allow PCs to dynamically regulate the excitability of mGluR1+ UBCs. MF inputs to the cerebellum can modulate the firing of PCs, and these firing rate changes will feedback to mGluR1+ UBCs to alter their firing. This could be particularly important for UBCs that fire for more than 10 s in response to a brief MF activation in the absence of inhibition (Guo et al., 2021).

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Strain, strain background (Mus musculus)C57BL/6Charles River
Strain, strain background
(M. musculus)
B6.Cg-Tg(Pcp2-cre)3555Jdhu/JJackson LaboratoryStrain#010536
Strain, strain background
(M. musculus)
ChR2-EYFP (Ai32)Jackson LaboratoryStrain#024109
Strain, strain background
(M. musculus)
Synaptophysin-tdTomato (Ai34D)Jackon LaboratoryStrain#021570
Chemical compound, drugNBQX disodium saltAbcamAb120046
Chemical compound, drug(R)-CPPAbcamAb120159
Chemical compound, drugStrychnine hydrochlorideAbcamAb120416
Chemical compound, drugSR95531 (Gabazine)AbcamAb120042
Chemical compound, drugL-GlutamateAbcamAb120049
Chemical compound, drug(RS)-3,5-DHPGTocrisCat. no. 0342/1
Chemical compound, drug(R,S)-BaclofenAbcamAb120149
Chemical compound, drugCGP 55845 hydrochlorideTocrisCat. no. 1248
AntibodiesGuinea pig anti-VGAT (Guinea pig polyclonal)Synaptic Systems1310041 μg/mL stock, 1:200
AntibodiesMouse anti-rat mGluR1 (Mouse monoclonal)BD PharmingenCat. no. 5563310.5 mg/mL stock, 1:800
AntibodiesGoat anti-Guinea pig Alexa Fluor 488 (Goat polyclonal)AbcamAb1501851:500
AntibodiesGoat anti-mouse Alexa Fluor 647 (Goat poly clonal)InvitrogenA327281:500
Software, algorithmIgor Pro 6Wavemetricshttps://www.wavemetrics.com/
Software, algorithmMafPCCourtesy of M.A. Xu-Friedmanhttps://www.xufriedman.org/mafpc
Software, algorithmMATLAB (R2017a)MathWorkshttps://www.mathworks.com/products/matlab.html
Software, algorithmImageJNIHhttps://imagej.nih.gov/ij/
Software, algorithmPythonPython Software Foundationhttps://www.python.org/
Software, algorithmPytorchOpensourcehttps://www.pytorch.org/

Immunohistochemistry and confocal imaging

Request a detailed protocol

Pcp2-Cre (Jackson Laboratory, 010536) × synaptophysin-tdTomato (Ai34D, Jackson Laboratory, 012570) P40 mice were anesthetized with intraperitoneal injections of ketamine/xylazine/acepromazine mixture at 100, 10, and 3 mg/kg, and transcardially perfused with 4% PFA in phosphate-buffered saline solution. The brain was removed from the skull and post-fixed overnight in 4% PFA. The cerebellar vermis and flocculus were dissected out and embedded in 6% low melting agarose before slicing. Sagittal vermal slices and coronal floccular slices at 50 μm were obtained from two animals using Leica VT1000S vibratome (Leica Biosystems, Buffalo Groves, IL). For immunolabeling of inhibitory synapses and the UBCs, slices were permeabilized (0.4% Triton X-100 and phosphate-buffered saline [PBS]) for 30 min and blocked for 1 hr at room temperature (0.4% Triton X-100, 4% normal goat serum) and incubated overnight at 4°C with primary antibodies (Guinea pig anti-VGAT, Synaptic Systems 131004, 1 μg/mL, 1:200 and mouse anti-mGluR1, BD Pharmingen, 0.5 mg/mL, 1:800 in 0.2% Triton X-100%, and 2% normal goat serum). The following day, the slices were washed three times for 10 min in PBS and incubated with secondary antibodies overnight at 4°C (Goat anti-Guinea pig-Alexa Fluor 488, Abcam ab150185, 1:500 and Goat anti-mouse Alexa Fluor 647, Invitrogen A32728, 1:500). Slices were mounted with a #1.5 coverslip using ProLong Diamond Antifade Mountant without DAPI (Thermo Fisher Scientific, Waltham, MA).

Z-stack images of entire vermal and floccular slices were obtained on the Olympus FV3000 confocal microscope using the Multi-Area Time-Lapse Software module. A 60× oil immersion objective with 1.42 NA was used and the x, y, z resolutions were 0.212, 0.212, and 0.906 μm, respectively. Each z-step was 1 μm and the entire stack was 20 μm in thickness.

Image analysis

Request a detailed protocol

PC synapses were detected using a modified deep neural network that is based on the U-Net (Ronneberger et al., 2015). A total of two brains were analyzed. The network was fed in all three imaging channels including the mGluR1 labeling. The rationale being that while colocalization of synaptophysin and VGAT was sufficient for synapse detection, mGluR1 provided additional contextual information for annotation in the molecular layer as it densely labeled PC dendrites. The network was trained using soft-dice loss on 100 annotated images (a single slice in z, 50×50 μm2 in width and height) and validated with an 80/20 train-test-split. Our human annotator criteria were that the inhibitory PC synapses should be synaptophysin-TdTomato and VGAT positive puncta with clear morphological features that are present across at least two to three consecutive z-slices. After training the network, we verified that it works consistently across different cerebellar layers and regions using out-of-sample validation. We found that using an ensemble of 10 independently trained networks qualitatively improved the annotation by reducing the amount of variability in synapse identification. The true positive rate of synapse detection was at 95.5% and the false-positive rate was 10.1%. We note however that the false positive rate was likely inflated as the human annotator sometimes missed difficult-to-spot synapses (Figure 1—figure supplement 1A–B, yellow arrows). For illustrating the UBCs, the mGluR1 signals in the molecular layer were manually cropped in the final composite image. The neural network approach was necessary to generate high-quality overviews of synaptic density maps over large regions.

Slice preparation for electrophysiology

Request a detailed protocol

Adult (P30–P40) C57BL/6 or Pcp2-Cre×ChR2-EYFP (Ai32, Jackson Laboratory, 024109) mice were first anesthetized with an intraperitoneal injection of ketamine/xylazine/acepromazine mixture at 100, 10, and 3 mg/kg, and transcardially perfused with an ice-cold choline slicing solution consisting (in mM): 110 choline Cl, 7 MgSO4, 2.5 KCl, 1.2 NaH2PO4, 0.5 CaCl2, 11.6 Na-ascorbate, 2.4 Na-pyruvate, 25 NaHCO3, and 25 glucose equilibrated with 95% O2 and 5% CO2. The cerebellum was dissected, mounted against an agar block, and submerged in the choline solution during slicing. Sagittal slices of the vermis were obtained using a Leica VT1200S vibratome and allowed to recover for 30 min at 33°C in the artificial cerebral spinal fluid (ACSF) consisting of (in mM): 125 NaCl, 26 NaHCO3, 1.25 NaH2PO4, 2.5 KCl, 1 MgCl2, 1.5 CaCl2, and 25 glucose (pH 7.4, osmolarity 315) equilibrated with 95% O2 and 5% CO2. The incubation chamber was then removed from the warm water bath and kept at room temperature for recording for up to 6 hr.

Electrophysiology

Request a detailed protocol

Recordings were performed at 34–36°C in ASCF containing 5 μM NBQX, 2 μM R-CPP, and 1 μM strychnine set to a flow rate of 2 mL/min. Visually guided recording of UBCs was obtained under a 60× objective with differential interference contrast imaging on an Olympus BX51WI microscope. Identity of the UBCs was also verified with fluorescent dye after recording (Alexa Fluor 594, 100 μM). Borosilicate patch pipette (3–5 MΩ) containing either a KCl internal (in mM: 140 KCl, 4 NaCl, 0.5 CaCl2, 10 HEPES, 4 MgATP, 0.3 NaGTP, 5 EGTA, and 2 QX-314, pH adjusted to 7.2 with KOH) or a K-methanesulfonate internal (in mM: 122 K-methanesulfonate, 9 NaCl, 9 HEPES, 0.036 CaCl2, 1.62 MgCl2, 4 MgATP, 0.3 Tris-GTP, 14 Tris-creatine phosphate, and 0.18 EGTA, pH7.4) was used for whole-cell recordings of GABAAR- or GABABR-mediated current, respectively. A junction potential of −8 mV was corrected for the K-Methansulfonate internal during recording. For cell-attached UBC recordings, ACSF was used as the internal solution. For perforated-patch recording, an internal containing (in mM) was used: 100 K-methanesulfonate, 13 NaCl, 2 MgCl2, 10 EGTA, 1 CaCl2, 10 HEPES, 0.1 Lucifer Yellow, and 0.25 amphotericin B, pH 7.2 with KOH. A junction potential of +7 mV was corrected online. For recordings of the GABABR-mediated current, 10 μM SR95531/gabazine was included in the bath to isolate the metabotropic inhibition. For paired PC-UBC recordings, a whole-cell UBC recording with KCl was obtained first. Then 10–20 PCs in lobule X were screened with cell-attached recordings using a large pipette (1–2 MΩ) containing ACSF.

Pharmacology

Request a detailed protocol

For identification of UBC subtypes in either whole-cell or perforated recordings, ACSF containing glutamate (1 mM, 50 ms) or DHPG (100 µM, 20 ms) was pressure-applied at 5 psi through a borosilicate pipette with a Picospritzer III (Parker Hannifin, Hollis, NH). For the sequential mGluR1 and GABABR activation experiment (Figure 5G–J), concentrations and durations of pressure applications are 100 µM, 50 ms for DHPG, and 250 µM, 100 ms for baclofen. The same baclofen concentration and duration were used for testing the stability of GABABR-mediated current (Figure 4—figure supplement 1). For wash-in experiments, solution exchange was controlled via ValveLink8.2 Controller (AutoMate Scientific, Berkeley, CA). Gabazine (SR95531, 5 μM) was used for blocking GABAAR and CGP (1 μM) was used for blocking GABABR.

Optogenetics

Request a detailed protocol

Slices from Pcp2-Cre×ChR2-EYFP were kept in the dark before recording. A laser source (MBL-III-473-50 mW, Optoengine) was fiber coupled to the excitation path of the microscope. For over bouton stimulation, brief (0.5 ms) high intensity (160 mW/mm2) light pulses were delivered under the 60× objective and focused down to a 50-μm spot. For the optical modulation of PC spontaneous firing, much lower intensities (10 or 25 μW/mm2) of light were delivered through the 10× objective over the entire lobule X (~1 mm spot) for 2 s. To ensure the quality of the slice, we recorded many PCs before the experiment to check for spontaneous activity and to ensure that these low-intensity stimulations did not result in PCs bursting. Only healthy slices were used for subsequent UBC recordings.

Data acquisition and analysis

Request a detailed protocol

Electrophysiology experiments were performed with a MultiClamp 700B amplifier (Molecular Device) and controlled by mafPC (Matthew Xu-Friedman, SUNY Buffalo) in Igor Pro 7 (WaveMetrics). Data were filtered at 4 kHz in MultiClamp and digitized at 50 kHz by InstruTECH ITC-18 (Heka Instrument Inc). Analysis of DHPG evoked spiking in UBC was done via peak detection in MATLAB. sIPSC timing was determined by time of 5% to peak threshold crossing. The timing of the PC spike for determining the latency of synaptic transmission was determined by the first derivative of the action potential. The time constant for the slow GABABR-mediated current was obtained using a single exponential fit. Data are reported as mean ± SEM.

Data availability

The RNAseq data for cerebellar UBCs can be visualized through (https://singlecell.broadinstitute.org/single_cell/study/SCP795/). Raw and processed data that support the findings of this study have been deposited in GEO under accession number GSE165371 and in at the Neuroscience Multi-omics (NeMO) Archive (https://nemoarchive.org/).

The following data sets were generated
    1. Guo C
    2. Rudolph S
    (2021) GitHub
    ID PC-UBC_elife2021. Purkinje cell outputs selectively inhibit a subset of unipolar brush cells in the input layer of the cerebellar cortex.
The following previously published data sets were used

References

    1. Guo C
    2. Huson V
    3. Macosko E
    4. Regehr WG
    (2021)
    Graded heterogeneity of metabotropic signaling underlies a continuum of cell-intrinsic temporal responses
    Nature Communications In press.
    1. Kozareva V
    2. Martin C
    3. Osorno T
    4. Rudolph S
    5. Guo C
    6. Vanderburg C
    7. Nadaf N
    8. Regev A
    9. Regehr WG
    10. Macosko E
    (2021)
    A transcriptomic atlas of the mouse cerebellum reveals regional specializations and novel cell types
    Nature In press.
  1. Conference
    1. Ronneberger O
    2. Fischer P
    3. Brox T
    (2015)
    U-Net: convolutional networks for biomedical image segmentation
    International Conference on Medical Image Computing and Computer-Assisted Intervention.

Decision letter

  1. Jennifer L Raymond
    Reviewing Editor; Stanford University School of Medicine, United States
  2. Kenton J Swartz
    Senior Editor; National Institute of Neurological Disorders and Stroke, National Institutes of Health, United States
  3. Jennifer L Raymond
    Reviewer; Stanford University School of Medicine, United States
  4. Philippe Isope
    Reviewer; CNRS, France

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

This paper expands our understanding of the cerebellar microarchitecture, and, in particular, the level of recurrence in this circuit. A combination of electrophysiological, pharmacological and optogenetic approaches demonstrate a new inhibitory feedback connection from Purkinje cells onto a subset of unipolar brush cells. This is an important discovery for the cerebellar field, and should also be of interest to readers studying feedback in neural circuits.

Decision letter after peer review:

Thank you for submitting your article "Purkinje cell outputs selectively inhibit a subset of unipolar brush cells in the input layer of the cerebellar cortex" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Jennifer L Raymond as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Kenton Swartz as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Philippe Isope (Reviewer #2).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

The finding that PCs inhibit mGluR1 positive UBCs is important in its own right and well established by the evidence in Figure 2 and 4.

1) The finding of feedback from the PCs to the UBCs opens up exciting new research directions. The reviewers appreciated the authors' efforts to gain insight about the signal processing function of this new feedback connection, however, they agreed that the specific claims about the function of the fast and slow inhibitory feedback were not well supported by the evidence provided, and hence should be clearly identified as highly speculative and the conclusions significantly tempered. Additional evidence to support the claims about the specific signal processing functions of the fast and slow feedback would be highly desirable, but is not essential. See specific suggestions below.

2) Some clarifications are needed regarding the histology:

– What is the reader supposed to take away from the convolutional neural network analysis of Purkinje cell synapses onto UBCs? It seems that two brains were annotated by a human expert, and then the network was trained on those two brains and could reproduce similar results on held-out data, as shown in Figure 1-Supp 1. What criteria did the human annotator use in marking synapses? Was the neural network approach used for anything beyond what the human annotator did?

– Is it absolutely certain that there is no mGluR1 expression in Lugaro cells and Golgi cells? (See Knöpfel and Grandes, 2002 and Grandes et al. 1994.) The high magnification images in Figure 1K do show the characteristic morphology of UBCs, but there is a possibility that low mag. images of distribution in lobules could include these other cell types in spite of the difference in size, in case they express mGluR1. This possibility could change the interpretation of these results, so please clarify.

– There seem to be PC synapses on the cell body of UBCs as well as on the brush (Figure 1K), although the synapses on the brush are highlighted for mention. Please clarify.

Specific, non-essential suggestions for the authors to consider, should they decide to strengthen the evidence about the function of the fast and slow feedback.

– There is the implicit assumption that regular PC firing hardly affects mGluR1 positive UBC discharge, but it would be interesting to address or at least discuss constitutive PC inhibition. For example, this could be assessed in UBCs that don't express mGluR2, by applying an mGluR2 agonist to inhibit Golgi cell discharge, then GABAA+GABAB blockers in the bath.

– Tonic UBC hyperpolarization by steady PC excitation in figure 5D-F was not fully quantified. For example, is there different behaviors at different light intensities as in Figure 4? In figure 5D-F, a bigger burst at the beginning of the depolarization step was observed, but the steady state current seems unchanged. This is an important point, as rebound discharge is an important physiological feature of UBCs (Diana MA et al., J Neurosci. 2007) which may modify the current knowledge about ON and OFF UBC computation. Noticeably, there is a clear rebound discharge in Figure 4I but not quantified. The interplay between T-type channel and GABAB hyperpolarization may be addressed or discussed more.

– Is there any combined effect of GABAA and GABAB signaling during trains of stimulation? Probability of release at the PC-UBC connection appeared low, but we expect a buildup of GABAA mediated current during trains and a synergistic effect with GABAB signaling on UBC firing rate. However, in the current form, experiments were performed either with GABAA or GABAB antagonist in the bath. Experiments with neither GABAA nor GABAB blockers would likely be informative.

– Can you draw any conclusions on the number of PC inputs to a UBC from the paired recordings? How many PCs are required to inhibit UBCs via GABAA or GABAB receptors? This question is important as groups of PCs can be synchronized or PCs can fire in burst at high frequencies, suggesting that the low probability of release of GABA may act as a high pass filter for GABAA signaling. While paired recordings (Figure 3) are challenging, loose-cell attached recordings on PCs (Barbour and Isope, J Neurosci. Methods, 2000) or focalized light illumination in ChR2 expressing PCs maybe used instead of cell-attached.

– The study focused on mGluR1 positive cells, but showed that Gabbr1 is also expressed in other UBCs. Therefore, GABAB inhibition could be relevant for mGluR1 negative UBCs as well. If the authors have these data, it would be an interesting extension.

– A discussion on the physiological consequences of a selective inhibition of mGluR1 positive vs mGluR2 UBCs would strengthen the study. Can we extrapolate that "ON" UBCs are inhibited by PCs while "OFF" cells are not?

– Quantification of the GABAA current recorded in figure 2 would be informative as it relates to glutamate induced current or DHPG induced increase in firing rate.

– It has been thought that UBCs serve to prolong GrC firing after brief mossy fiber inputs. The negative feedback described here would serve the opposite purpose. Thus, it is important to understand the time course of the negative feedback relative to mossy fiber input.

– The major conclusions of this paper relate to feedback regulation of the input layer. But the data presented here don't directly show a feedback loop, i.e., that the GrCs that feedforward mossy fiber information to the Purkinje cells are the ones that are regulated by the UBCs that these PCs project to. Given the observation of slow GABAB-dependent feedback, the role of direct feedback vs. feedback onto UBCs without a direct connection is important.

– In relation to the previous question, it remains unclear how modulating UBC input would affect Purkinje cells.

– The authors previously showed that PC-GrC feedback is also heterogeneously distributed, and will serve to inhibit the input layer. Since PCs directly provide both fast and slow feedback to GrCs in these regions, the PC-UBC connection demonstrated here suggests additional negative feedback via UBCs to GrCs. Therefore, the effectiveness of second level of negative feedback regulation of PCs to the input layer remains unclear.

– Until the current finding of PC-UBC feedback, inhibitory inputs to UBCs were thought to be limited to those from Golgi cells. The relative time course of Golgi cell vs. PC feedback inhibition in controlling UBCs is worth consideration.

– It would be important to know whether PC feedback to UBCs is in the same parasagittal plane.

– As an additional example that relates to the same overall question about how this feedback loop will actually affect neural signals via mossy fibers, the authors also show that the faster inhibitory mGluR2/3-mediated current comes first, and the mGluR1-mediated excitation comes later (Guo et al., Biorxiv, 2020). When would PC feedback come, in relation to this sequence?

– There is known to be a heterogeneous distribution of UBCs, as shown in this study too. Is there also preferential localization of mGluR1-positive UBCs? Figure 1 shows only mGluR1-positive UBCs. Since the presence of mGluR1 and the response to glutamate are related to PC feedback, it is important to know whether the proportion of UBCs that do not get PC feedback even in these UBC-rich regions is different from elsewhere, or not. This would have important implications for circuit processing in vestibular-input-related vs. other regions

– A significant minority of DHPG-responsive cells don't have PC feedback (although the absence of an optogenetically-induced response is not definitive proof of lack of PC inputs). Notwithstanding this caveat, I think this significant minority should also be discussed.

– In the PC-UBC paired recordings, were there UBC responses that were not timed to PC spikes?

– The entire study is motivated by describing specialization of different regions of the cerebellum, but the physiology was only performed in lobule X of vermis. Although there are fewer UBCs in other parts of the cerebellum, do the same patterns of feedback from the Purkinje cells apply?

Summary:

The major strength of this study is the straightforward and convincing demonstration of a previously unknown synaptic connection from Purkinje cells (PCs) to mGluR1-positive unipolar brush cells (UBCs) in lobule X of the cerebellum. UBCs are interneurons of the cerebellar input layer, which preferentially distributed in some lobules of the cerebellum; especially those involved in vestibular function. The authors utilize a combination of electrophysiological recordings from UBCs in the acute rodent cerebellar slice, optogenetic activation of PCs, and pharmacological manipulations to establish that PCs inhibit UBCs in cerebellar lobule X through both fast synaptic transmission through GABAA receptors, as well as long-lasting inhibition via GABAB receptors. Since PCs constitute the output stage of the cerebellar cortex, the inhibitory signaling described in this study identifies a new feedback control pathway to the input layer, which constitutes a computational specialization of lobule X.

Another strength of this study is the determination of how molecular heterogeneity corresponds to circuit connectivity and variable responses to synaptic input. There is heterogeneity in the molecular profiles of UBCs, and expression of mGluR1 is a key marker for a sub-population of UBCs. The authors demonstrate that mGluR1-expressing UBCs are preferentially targeted by PC feedback. – It is noteworthy that GABAA-mediated PC feedback affects a majority of, but not all, mGluR1-containing UBCs, but GAABB-mediated PC feedback affected all UBCs tested. This finding suggests tonic control of the UBC by PCs even without a direct PC connection.

The results raise the interesting question of the signal processing and computational function of this newly identified inhibitory feedback pathway, which should inspire future research. The authors offer some ideas along with limited evidence, however, the reviewers felt that additional research would be required to establish the authors' claims about the specific roles of PC inhibition of UBCs via GABAA and via GABAB, especially in the context of the full complement of signaling pathways in the intact cerebellar circuit.

The claim that "phasic GABAAR-mediated PC-UBC synapses convey noisy and unreliable

output to input layer feedback" is based on recordings from two pairs of synaptically recorded PCs and UBCs. The reviewers fully appreciate how technically challenging these experiments are, nevertheless it is difficult to conclude much from n=2 observations. Moreover, even if the high (85%) synaptic failure rate were to be replicated in a large number of PC-UBC pairs, it is not clear that this should be interpreted as "noise." Purkinje cells fire at high rates of 50-100 Hz, and one would expect a buildup of GABAA mediated current during spike trains and potentially a synergistic effect with GABAB signaling on UBC firing rate.

Likewise, the pharmacological approach used in Figure 5G-J lacks the temporal resolution to establish the claim that "Slow GABABR-mediated inhibition allows elevated PC activity to sharpen the input-output transformation of UBCs."

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Purkinje cell outputs selectively inhibit a subset of unipolar brush cells in the input layer of the cerebellar cortex" for further consideration by eLife. Your revised article has been evaluated by Kenton Swartz as the Senior Editor, and a Reviewing Editor.

The two main conclusions from the original review still apply in their entirety:

1) The finding that PCs inhibit mGluR1 positive UBCs is important in its own right and well established by the evidence provided.

2) The specific claims about the function of the fast and slow inhibitory feedback were not well supported, in particular the following two claims, which are made in the last two lines of the abstract "GABAAR-mediated feedback is fast, unreliable, noisy, and may help linearize input-output curves and decrease gain. Slow GABABR-mediated inhibition could allow elevated PC activity to sharpen the input-output transformation of UBCs, and allow dynamic inhibitory feedback of mGluR1-expressing UBCs."

During the review of the original submission, the reviewers and editor agreed there were two reasonable paths forward to publication: i) provide more evidence for the unsubstantiated claims, or ii) simply back off on the unsubstantiated claims, and focus on the main finding of the new connection from PCs to mGluR1+ UBCs. Inexplicably, the authors did not pursue either of these options, but rather resubmitted the original manuscript almost verbatim, complete with the two unsubstantiated claims at the end of the abstract.

The editor apologizes if placement of the key concerns of the reviewers at the end of the decision letter caused the authors to overlook these central concerns, which left the reviewers completely unconvinced of the functional claims:

The claim that "phasic GABAAR-mediated PC-UBC synapses convey noisy and unreliable

output to input layer feedback" is based on recordings from two pairs of synaptically recorded PCs and UBCs. The reviewers fully appreciate how technically challenging these experiments are, nevertheless it is difficult to conclude much from n=2 observations. Moreover, even if the high (85%) synaptic failure rate were to be replicated in a large number of PC-UBC pairs, it is not clear that this should be interpreted as "noise." Purkinje cells fire at high rates of 50-100 Hz, and one would expect a buildup of GABAA mediated current during spike trains and potentially a synergistic effect with GABAB signaling on UBC firing rate.

Likewise, the pharmacological approach used in Figure 5G-J lacks the temporal resolution to establish the claim that "Slow GABABR-mediated inhibition allows elevated PC activity to sharpen the input-output transformation of UBCs."

The authors provided a lengthy rebuttal of the suggestions the reviewers had offered about how the unsubstantiated claims might be strengthened, however, the second option of removing those claims still seems like a reasonable path forward, if the authors are willing to substantially revise the manuscript.

https://doi.org/10.7554/eLife.68802.sa1

Author response

Essential revisions:

The finding that PCs inhibit mGluR1 positive UBCs is important in its own right and well established by the evidence in Figure 2 and 4.

1) The finding of feedback from the PCs to the UBCs opens up exciting new research directions. The reviewers appreciated the authors' efforts to gain insight about the signal processing function of this new feedback connection, however, they agreed that the specific claims about the function of the fast and slow inhibitory feedback were not well supported by the evidence provided, and hence should be clearly identified as highly speculative and the conclusions significantly tempered. Additional evidence to support the claims about the specific signal processing functions of the fast and slow feedback would be highly desirable, but is not essential. See specific suggestions below.

We have addressed the specific issues raised below.

2) Some clarifications are needed regarding the histology:

– What is the reader supposed to take away from the convolutional neural network analysis of Purkinje cell synapses onto UBCs? It seems that two brains were annotated by a human expert, and then the network was trained on those two brains and could reproduce similar results on held-out data, as shown in Figure 1-Supp 1. What criteria did the human annotator use in marking synapses? Was the neural network approach used for anything beyond what the human annotator did?

We have clarified the explanation in both the results and methods section. We emphasize that we did not manually annotate two brains. The expert annotator only worked on 100 images, each a single 2D slice of 50x50 micron. This is less than 0.001% of the parasagittal vermis volume in Figure 1C for reference. Our human annotator criteria were that the inhibitory PC synapses should be synaptophysin-TdTomato and VGAT positive puncta with clear morphological features that are present across at least two to three consecutive z-slices. After training the network, we verified that it works consistently across different cerebellar layers and regions using out of sample validation. The neural network approach was necessary to generate high-quality overviews of synaptic density maps over large regions. (pages 12-13).

– Is it absolutely certain that there is no mGluR1 expression in Lugaro cells and Golgi cells? (See Knöpfel and Grandes, 2002 and Grandes et al. 1994.) The high magnification images in Figure 1K do show the characteristic morphology of UBCs, but there is a possibility that low mag. images of distribution in lobules could include these other cell types in spite of the difference in size, in case they express mGluR1. This possibility could change the interpretation of these results, so please clarify.

The brush expresses extremely high level of mGluR1 and dominates the signals in both the high and low magnification image of UBCs. mGluR1 labeling is used in many previous papers to identify UBCs (i.e. Borges-Merjane and Trussell, 2015, Nuzi et al., 2002, McDonough et al., 2020). Please note that the entire image was obtained at high resolution, which allows us to inspect the morphology of the labelled structures. The distinctive, intense mGluR1 immunofluorescence pattern we observe is characteristic of UBC brush labelling, and is not consistent with Golgi cell or Lugaro cell labelling. (page 4)

– There seem to be PC synapses on the cell body of UBCs as well as on the brush (Figure 1K), although the synapses on the brush are highlighted for mention. Please clarify.

We agree that synapses are near both the brush and the cell body in these images, and we have modified the legend and text (page 4).

Specific, non-essential suggestions for the authors to consider, should they decide to strengthen the evidence about the function of the fast and slow feedback.

– There is the implicit assumption that regular PC firing hardly affects mGluR1 positive UBC discharge, but it would be interesting to address or at least discuss constitutive PC inhibition. For example, this could be assessed in UBCs that don't express mGluR2, by applying an mGluR2 agonist to inhibit Golgi cell discharge, then GABAA+GABAB blockers in the bath.

Constitutive PC inhibition of UBCs is an interesting possibility, and we have added to the discussion to deal with this issue. UBCs can receive GABAR-mediated inhibition from GoCs and PCs. However, it is easy to sever the extensive PC collaterals that extend across lobule X, and GoCs often die in adult slices. In addition, a large percentage of UBCs express mGluR2/3Rs, so the application of an agonist is difficult to interpret. These technical challenges make it difficult to determine the magnitude and fractional contributions of constitutive PC and GoC inhibition onto UBCs.(page 10)

– Tonic UBC hyperpolarization by steady PC excitation in figure 5D-F was not fully quantified. For example, is there different behaviors at different light intensities as in Figure 4? In figure 5D-F, a bigger burst at the beginning of the depolarization step was observed, but the steady state current seems unchanged. This is an important point, as rebound discharge is an important physiological feature of UBCs (Diana MA et al., J Neurosci. 2007) which may modify the current knowledge about ON and OFF UBC computation. Noticeably, there is a clear rebound discharge in Figure 4I but not quantified. The interplay between T-type channel and GABAB hyperpolarization may be addressed or discussed more.

The reviewer makes an important point. We now point out the rebound discharge in Figure 4 and discuss in the context of T-type calcium channels and the study of Diana et al. (2007). We discuss how pauses in PC firing could potentially generate rebound firing in UBCs. (page 10)

– Is there any combined effect of GABAA and GABAB signaling during trains of stimulation? Probability of release at the PC-UBC connection appeared low, but we expect a buildup of GABAA mediated current during trains and a synergistic effect with GABAB signaling on UBC firing rate. However, in the current form, experiments were performed either with GABAA or GABAB antagonist in the bath. Experiments with neither GABAA nor GABAB blockers would likely be informative.

We had also considered the possibility of performing experiments in the absence of any GABA receptor antagonists. Ultimately, we felt that it was more informative to explore the contributions of GABAA and GABAB receptors in isolation, and that performing experiments in the absence of GABAR antagonists did not provide a sufficient additional advance to justify the considerable time and effort required.

– Can you draw any conclusions on the number of PC inputs to a UBC from the paired recordings? How many PCs are required to inhibit UBCs via GABAA or GABAB receptors? This question is important as groups of PCs can be synchronized or PCs can fire in burst at high frequencies, suggesting that the low probability of release of GABA may act as a high pass filter for GABAA signaling. While paired recordings (Figure 3) are challenging, loose-cell attached recordings on PCs (Barbour and Isope, J Neurosci. Methods, 2000) or focalized light illumination in ChR2 expressing PCs maybe used instead of cell-attached.

We share the reviewer’s desire to determine the number of PCs that can influence each UBC. However, we do not think that electrophysiological studies will provide a reliable estimate of the number of PCs that inhibit a UBC. This is essentially two questions, because it is likely that direct contacts are required for GABAA responses, GABAB responses do not require direct contacts and can be activated by spillover. In order to estimate input numbers, it is necessary to determine the average size of an individual input and determine the size of the response to all inputs activated simultaneously. We used loose-cell attached recordings to record from PCs, but these experiments were very challenging and the n’s were small (Figure 3). We are less confident about our ability to reliably activate all PC inputs to a UBC. There is the challenge that over bouton stimulation and the use of ChR2 can artificially elevate the initial probability of release. We also think it likely that many PC axons and their associated synapses may be damaged, which would cause us to underestimate the number of inputs.

For these reasons, in future studies (within the next 2 or 3 years) we hope to determine the total number of PC inputs onto a single UBC using serial EM reconstruction, which we feel will provide much more reliable estimates. Estimating the number of PCs that can contribute to the GABAB response is much more challenging because in addition to the challenges outlines for direct GABAA synapses, it is extremely difficult to determine the magnitude of a single PC to UBC GABAB component. It is also highly likely that responses from different inputs will not sum linearly and that there will be a strong frequency dependence to the magnitude. For these reasons we prefer to be cautious and not draw conclusions that go beyond our data regarding the numbers of inputs.

– The study focused on mGluR1 positive cells, but showed that Gabbr1 is also expressed in other UBCs. Therefore, GABAB inhibition could be relevant for mGluR1 negative UBCs as well. If the authors have these data, it would be an interesting extension.

It is generally thought that GABAB receptors are heterodimers consisting of GABAB1 and GABAB2 receptors (Frangaj and Fan 2018). It is therefore of considerable interest that Gabbr2 shows such a pronounced gradient and such low expression levels in mGluR1- cells, but Gabbr1 expression does not exhibit such a gradient. We now comment on this in the text (page 6)

– A discussion on the physiological consequences of a selective inhibition of mGluR1 positive vs mGluR2 UBCs would strengthen the study. Can we extrapolate that "ON" UBCs are inhibited by PCs while "OFF" cells are not?

We have added additional discussion in the possible physiological consequences of PCs selectively inhibiting a subset of UBCs. In this discussion we strive to avoid the convenient “ON: and “OFF” terms that we feel are insufficiently precise. We have demonstrated in a recent manuscript (Guo et al. 2021, Nat Comm in press, and bioRxiv 2021) that the population of UBCs have graded responses, and mGluR1 excitation is prominent in some cells, mGluR2/3 inhibition is prominent in others, but for most mGluR1 and mGluR2/3 are both present. We find that PCs preferentially inhibit UBCs with high mGluR1 expression, and predominantly excitatory MF responses.

– Quantification of the GABAA current recorded in figure 2 would be informative as it relates to glutamate induced current or DHPG induced increase in firing rate.

We have provided an additional scatter plot (Figure 2N) to address this issue.

– It has been thought that UBCs serve to prolong GrC firing after brief mossy fiber inputs. The negative feedback described here would serve the opposite purpose. Thus, it is important to understand the time course of the negative feedback relative to mossy fiber input.

This is not a simple issue. In general, elevated PC inhibition will rapidly suppress firing in UBCs if there is a GABAA component, and will slowly inhibit UBCs if there is a GABAB component. MFs can excite many UBCs for a few hundred milliseconds to tens of seconds, and MFs inhibit many UBCs for hundreds of milliseconds by activating mGluR2/3 receptors. However, it is not clear whether MF evoked population activity in UBCs will increase or decrease PC firing, because UBCs excite granule cells that both directly excite PCs and disynaptically inhibit PCs by activating MLIs. We therefore do not know if the UBCgRCPCUBC loop leads to positive or negative feedback. Hopefully, future in vivo studies will address this issue, but they will be challenging. We feel that it is reasonable and appropriate to restrict this study and the discussion to a basic description of the PC to UBC synapse, and that many of the interesting issues raised by the reviewer are perfect topics for future studies.

– The major conclusions of this paper relate to feedback regulation of the input layer. But the data presented here don't directly show a feedback loop, i.e., that the GrCs that feedforward mossy fiber information to the Purkinje cells are the ones that are regulated by the UBCs that these PCs project to. Given the observation of slow GABAB-dependent feedback, the role of direct feedback vs. feedback onto UBCs without a direct connection is important.

We feel that it is appropriate to indicate that PCs are providing feedback from the output layer to UBCs in the input layer. Also note that each UBC excites many granule cells, and each granule cell excites about half of the local PCs. While it seems extremely likely that many GrCs activated by a UBC will directly excite the PCs that inhibit that UBC, we have not explicitly shown that this is the case.

– In relation to the previous question, it remains unclear how modulating UBC input would affect Purkinje cells.

We echo the sentiment of the reviewer. Indeed, in general the manner in which UBCs influence PC firing is complex and poorly understood, as is the case for extrinsic MFs. UBCs and MFs activate granule cells that in turn excite PCs directly and inhibit PCs through disynaptic inhibition, so the net effect is difficult to predict. We have therefore focused additional discussion on the fact that PC-UBC feedback allows the output of the cerebellar cortex to shape the temporal representation of the input layer.

– The authors previously showed that PC-GrC feedback is also heterogeneously distributed, and will serve to inhibit the input layer. Since PCs directly provide both fast and slow feedback to GrCs in these regions, the PC-UBC connection demonstrated here suggests additional negative feedback via UBCs to GrCs. Therefore, the effectiveness of second level of negative feedback regulation of PCs to the input layer remains unclear.

On page 9 we point out: The major differences in slow inhibition of UBCs and GrCs are the state change of UBCs and the target dependence of the UBC inhibition, neither of which have been described for PC inhibition of GrCs (Guo et al., 2016).

– Until the current finding of PC-UBC feedback, inhibitory inputs to UBCs were thought to be limited to those from Golgi cells. The relative time course of Golgi cell vs. PC feedback inhibition in controlling UBCs is worth consideration.

This is an interesting issue that is part of the larger question of the distinctive functional roles of PC and GC inhibition of UBCs. However, the relative times courses of GC and PC inhibition would begin with determining the activity patterns of these two cell types during behaviors.

– It would be important to know whether PC feedback to UBCs is in the same parasagittal plane.

We reconstructed PC collaterals in a previous paper across various cerebellar regions. The PC feedback appears to be confined in the parasagittal plane (Witter et al., 2016) as we now point out. This will potentially allow the PC-UBC feedback to operate within functional macrozones. (page 2)

– As an additional example that relates to the same overall question about how this feedback loop will actually affect neural signals via mossy fibers, the authors also show that the faster inhibitory mGluR2/3-mediated current comes first, and the mGluR1-mediated excitation comes later (Guo et al., Biorxiv, 2020). When would PC feedback come, in relation to this sequence?

The reviewer raises an interesting question regarding the relative timing of mGluR2 inhibition and PC inhibition. The timing of mGluR2/3 current is diverse across UBCs, ranging from about one hundred milliseconds to a few seconds. There are so many factors contributing to relative time courses of PC inhibition and mGluR2-mediated inhibition (including stimulus, MF activity pattern, continuous UBC temporal variability, the magnitudes of the GABAA and GABAB components, etc.) that we feel it would be premature to discuss this issue (regardless of how interesting it is).

– There is known to be a heterogeneous distribution of UBCs, as shown in this study too. Is there also preferential localization of mGluR1-positive UBCs? Figure 1 shows only mGluR1-positive UBCs. Since the presence of mGluR1 and the response to glutamate are related to PC feedback, it is important to know whether the proportion of UBCs that do not get PC feedback even in these UBC-rich regions is different from elsewhere, or not. This would have important implications for circuit processing in vestibular-input-related vs. other regions

We are interested in investigating the properties of PC to UBC inhibition in other regions of the cerebellar cortex. However, it would be necessary to essentially repeat our study in each region we wish to evaluate this issue functionally. We feel that this is beyond the scope of the current study. We deal with this issue on page 8.

– A significant minority of DHPG-responsive cells don't have PC feedback (although the absence of an optogenetically-induced response is not definitive proof of lack of PC inputs). Notwithstanding this caveat, I think this significant minority should also be discussed.

The reviewer rightly points out the very important caveat that “the absence of an optogenetically-induced response is not definitive proof of lack of PC inputs”. We have added a sentence dealing with this issue (page 6)

– In the PC-UBC paired recordings, were there UBC responses that were not timed to PC spikes?

Yes, there were IPSCs not time locked to PC spikes. This could come from either other PCs or GoCs. We did not use mGluR2 agonist to suppress GoC input in these recordings because many UBCs express mGluR2 receptors (including many mGluR1+ UBCs).

– The entire study is motivated by describing specialization of different regions of the cerebellum, but the physiology was only performed in lobule X of vermis. Although there are fewer UBCs in other parts of the cerebellum, do the same patterns of feedback from the Purkinje cells apply?

This is an interesting question that we are unable to answer definitively. Images of other regions such as lobule IX, the paraflocculus and the flocculus (Figure 1) suggest that collateral synapses may target UBCs in these regions as well. We are, however, reluctant to draw conclusions based solely on these light level images. It would require a great deal of effort to address the issue of PC inhibition of UBCs in other regions. We would need to repeat our experiments of Figures 2 and 4 for each of the other regions under consideration. We now briefly discuss this issue (page 9).

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

The two main conclusions from the original review still apply in their entirety:

1) The finding that PCs inhibit mGluR1 positive UBCs is important in its own right and well established by the evidence provided.

2) The specific claims about the function of the fast and slow inhibitory feedback were not well supported, in particular the following two claims, which are made in the last two lines of the abstract "GABAAR-mediated feedback is fast, unreliable, noisy, and may help linearize input-output curves and decrease gain. Slow GABABR-mediated inhibition could allow elevated PC activity to sharpen the input-output transformation of UBCs, and allow dynamic inhibitory feedback of mGluR1-expressing UBCs."

During the review of the original submission, the reviewers and editor agreed there were two reasonable paths forward to publication: i) provide more evidence for the unsubstantiated claims, or ii) simply back off on the unsubstantiated claims, and focus on the main finding of the new connection from PCs to mGluR1+ UBCs. Inexplicably, the authors did not pursue either of these options, but rather resubmitted the original manuscript almost verbatim, complete with the two unsubstantiated claims at the end of the abstract.

The editor apologizes if placement of the key concerns of the reviewers at the end of the decision letter caused the authors to overlook these central concerns, which left the reviewers completely unconvinced of the functional claims:

The claim that "phasic GABAAR-mediated PC-UBC synapses convey noisy and unreliable

output to input layer feedback" is based on recordings from two pairs of synaptically recorded PCs and UBCs. The reviewers fully appreciate how technically challenging these experiments are, nevertheless it is difficult to conclude much from n=2 observations. Moreover, even if the high (85%) synaptic failure rate were to be replicated in a large number of PC-UBC pairs, it is not clear that this should be interpreted as "noise." Purkinje cells fire at high rates of 50-100 Hz, and one would expect a buildup of GABAA mediated current during spike trains and potentially a synergistic effect with GABAB signaling on UBC firing rate.

Likewise, the pharmacological approach used in Figure 5G-J lacks the temporal resolution to establish the claim that "Slow GABABR-mediated inhibition allows elevated PC activity to sharpen the input-output transformation of UBCs."

The authors provided a lengthy rebuttal of the suggestions the reviewers had offered about how the unsubstantiated claims might be strengthened, however, the second option of removing those claims still seems like a reasonable path forward, if the authors are willing to substantially revise the manuscript.

Your additional clarification has been most helpful. We now have a much better idea of the issues and the remedies. We have made a real effort to remove anything that you would consider to be speculative from the abstract, the end of the introduction and the results. We have also eliminated the discussion of the GABAA component’s functional role, as per your suggestion.

The one area where I would like to push back a bit concerns Figure 5. This is the figure where we try to provide insight into what this feedback might be doing to PCs. Dr. Raymond has expressed reservations about this figure. Nonetheless, we feel that this figure adds to the paper and furthers our understanding of what PC to UBC feedback might be doing. In our point-by-point response we explain why we feel that these experiments and the associated discussion should be included. Of course, if you still feel that including Figure 5 and the associated discussion is a deal breaker, we will remove it. But we request that you allow us to retain it, because we feel that it makes the paper better.

https://doi.org/10.7554/eLife.68802.sa2

Article and author information

Author details

  1. Chong Guo

    Department of Neurobiology, Harvard Medical School, Boston, United States
    Present address
    Department of Brain and Cognitive Sciences and McGovern Institute for Brain Research, Massachusetts Institute of Technology, Cambridge, United States
    Contribution
    Conceptualization, Data curation, Software, Formal analysis, Investigation, Visualization, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-1230-5333
  2. Stephanie Rudolph

    Department of Neurobiology, Harvard Medical School, Boston, United States
    Present address
    Albert Einstein College of Medicine, New York, United States
    Contribution
    Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0388-7762
  3. Morgan E Neuwirth

    Department of Neurobiology, Harvard Medical School, Boston, United States
    Contribution
    Software, Investigation
    Competing interests
    No competing interests declared
  4. Wade G Regehr

    Department of Neurobiology, Harvard Medical School, Boston, United States
    Contribution
    Conceptualization, Resources, Funding acquisition, Writing - original draft, Writing - review and editing
    For correspondence
    wade_regehr@hms.harvard.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3485-8094

Funding

National Institute of Neurological Disorders and Stroke (R35NS097284)

  • Wade G Regehr

Stuart H.Q. & Victoria Quan Fellowship

  • Chong Guo

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

Imaging was performed in the Vision Core and NINDS P30 Core Center (NS072030) to the Neurobiology Imaging Center at Harvard Medical School with the help of M El-Rifai (performed RNA scope) and M Ocana. The authors thank Laurens Witter for his insights in the early stages of this project, and Vincent Huson for feedback on the manuscript.

Ethics

Animal experimentation: Animal experimentation: All experiments were conducted in accordance with federal guidelines and protocols (#1493) approved by the Harvard Medical Area Standing Committee on Animals.

Senior Editor

  1. Kenton J Swartz, National Institute of Neurological Disorders and Stroke, National Institutes of Health, United States

Reviewing Editor

  1. Jennifer L Raymond, Stanford University School of Medicine, United States

Reviewers

  1. Jennifer L Raymond, Stanford University School of Medicine, United States
  2. Philippe Isope, CNRS, France

Version history

  1. Preprint posted: March 26, 2021 (view preprint)
  2. Received: March 26, 2021
  3. Accepted: July 23, 2021
  4. Version of Record published: August 9, 2021 (version 1)

Copyright

© 2021, Guo et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 863
    Page views
  • 136
    Downloads
  • 8
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Chong Guo
  2. Stephanie Rudolph
  3. Morgan E Neuwirth
  4. Wade G Regehr
(2021)
Purkinje cell outputs selectively inhibit a subset of unipolar brush cells in the input layer of the cerebellar cortex
eLife 10:e68802.
https://doi.org/10.7554/eLife.68802

Further reading

    1. Computational and Systems Biology
    2. Neuroscience
    Huu Hoang, Shinichiro Tsutsumi ... Keisuke Toyama
    Research Article Updated

    Cerebellar climbing fibers convey diverse signals, but how they are organized in the compartmental structure of the cerebellar cortex during learning remains largely unclear. We analyzed a large amount of coordinate-localized two-photon imaging data from cerebellar Crus II in mice undergoing ‘Go/No-go’ reinforcement learning. Tensor component analysis revealed that a majority of climbing fiber inputs to Purkinje cells were reduced to only four functional components, corresponding to accurate timing control of motor initiation related to a Go cue, cognitive error-based learning, reward processing, and inhibition of erroneous behaviors after a No-go cue. Changes in neural activities during learning of the first two components were correlated with corresponding changes in timing control and error learning across animals, indirectly suggesting causal relationships. Spatial distribution of these components coincided well with boundaries of Aldolase-C/zebrin II expression in Purkinje cells, whereas several components are mixed in single neurons. Synchronization within individual components was bidirectionally regulated according to specific task contexts and learning stages. These findings suggest that, in close collaborations with other brain regions including the inferior olive nucleus, the cerebellum, based on anatomical compartments, reduces dimensions of the learning space by dynamically organizing multiple functional components, a feature that may inspire new-generation AI designs.

    1. Neuroscience
    Max L Sterling, Ruben Teunisse, Bernhard Englitz
    Tools and Resources Updated

    Ultrasonic vocalizations (USVs) fulfill an important role in communication and navigation in many species. Because of their social and affective significance, rodent USVs are increasingly used as a behavioral measure in neurodevelopmental and neurolinguistic research. Reliably attributing USVs to their emitter during close interactions has emerged as a difficult, key challenge. If addressed, all subsequent analyses gain substantial confidence. We present a hybrid ultrasonic tracking system, Hybrid Vocalization Localizer (HyVL), that synergistically integrates a high-resolution acoustic camera with high-quality ultrasonic microphones. HyVL is the first to achieve millimeter precision (~3.4–4.8 mm, 91% assigned) in localizing USVs, ~3× better than other systems, approaching the physical limits (mouse snout ~10 mm). We analyze mouse courtship interactions and demonstrate that males and females vocalize in starkly different relative spatial positions, and that the fraction of female vocalizations has likely been overestimated previously due to imprecise localization. Further, we find that when two male mice interact with one female, one of the males takes a dominant role in the interaction both in terms of the vocalization rate and the location relative to the female. HyVL substantially improves the precision with which social communication between rodents can be studied. It is also affordable, open-source, easy to set up, can be integrated with existing setups, and reduces the required number of experiments and animals.