1. Neuroscience
Download icon

Inhibition gates supralinear Ca2+ signaling in Purkinje cell dendrites during practiced movements

  1. Michael A Gaffield
  2. Matthew J M Rowan
  3. Samantha B Amat
  4. Hirokazu Hirai
  5. Jason M Christie  Is a corresponding author
  1. Max Planck Florida Institute for Neuroscience, United States
  2. Gunma University Graduate School of Medicine, Japan
Research Article
  • Cited 0
  • Views 870
  • Annotations
Cite this article as: eLife 2018;7:e36246 doi: 10.7554/eLife.36246

Abstract

Motor learning involves neural circuit modifications in the cerebellar cortex, likely through re-weighting of parallel fiber inputs onto Purkinje cells (PCs). Climbing fibers instruct these synaptic modifications when they excite PCs in conjunction with parallel fiber activity, a pairing that enhances climbing fiber-evoked Ca2+ signaling in PC dendrites. In vivo, climbing fibers spike continuously, including during movements when parallel fibers are simultaneously conveying sensorimotor information to PCs. Whether parallel fiber activity enhances climbing fiber Ca2+ signaling during motor behaviors is unknown. In mice, we found that inhibitory molecular layer interneurons (MLIs), activated by parallel fibers during practiced movements, suppressed parallel fiber enhancement of climbing fiber Ca2+ signaling in PCs. Similar results were obtained in acute slices for brief parallel fiber stimuli. Interestingly, more prolonged parallel fiber excitation revealed latent supralinear Ca2+ signaling. Therefore, the balance of parallel fiber and MLI input onto PCs regulates concomitant climbing fiber Ca2+ signaling.

https://doi.org/10.7554/eLife.36246.001

Introduction

Neural circuits that support motor learning must respond to, and adjust for, stimuli relevant for encoding adaptation. With equal importance, these circuits must also prevent network alterations during stimuli unrelated or unnecessary for behavioral modification. In the cerebellar cortex, synaptic re-weighting of parallel fiber input onto Purkinje cells (PCs) is thought to provide the basis for many types of motor learning (Ito and Kano, 1982; Ito et al., 1982; Kano et al., 1992; Hansel and Linden, 2000). Climbing fibers instruct synaptic alterations by evoking dendritic Ca2+ spikes in PCs (Kano et al., 1992; Hansel and Linden, 2000; Coesmans et al., 2004) in response to adaptive stimuli (Simpson and Alley, 1974; Gilbert and Thach, 1977; Medina and Lisberger, 2008). However, climbing fiber activity also occurs regularly at 1–2 Hz (Mukamel et al., 2009; Ozden et al., 2012; De Gruijl et al., 2014). Therefore, PCs must distinguish which climbing fiber signals are relevant for inducing adaptation, as well as prevent or gate inappropriate circuit alterations by climbing fibers when plasticity is unwarranted or apt to produce an incorrect alteration of behavior (Kimpo et al., 2014). The integrative features of PC dendrites offer solutions as to how PCs accomplish this task (Najafi and Medina, 2013).

When activated in conjunction with parallel fibers in ex vivo preparations, climbing fiber-triggered Ca2+ signals in PC dendrites are enhanced in a supralinear manner (Wang et al., 2000; Otsu et al., 2014; Piochon et al., 2016). Supralinear signaling achieves a threshold level of intracellular Ca2+ elevation that triggers long-term depression (LTD) at parallel fiber-PC synapses (Finch et al., 2012). Whether climbing fiber Ca2+ signals in PC dendrites are augmented by preceding parallel fiber activity in vivo is unclear. Certainly, PCs receive sensorimotor information transmitted by granule cells during the execution of movements (Ozden et al., 2012; Wilms and Häusser, 2015; Chen et al., 2017). Furthermore, enhanced PC Ca2+ signals are observed in response to externally produced sensory stimuli (Najafi et al., 2014a; Najafi et al., 2014b). However, if self-generated parallel fiber activity is sufficient to enhance ongoing climbing fiber Ca2+ signals, then PCs would continuously undergo plasticity despite conditions where learning provides no benefit to motor outcomes. To counteract direct parallel fiber excitation of PCs, parallel fibers also excite molecular layer interneurons (MLIs) driving feed-forward inhibition that can attenuate parallel fiber excitatory postsynaptic potentials (EPSPs) (Brunel et al., 2004; Mittmann et al., 2005). MLIs can also directly reduce climbing fiber-evoked responses (Callaway et al., 1995; Kitamura and Häusser, 2011) and impair LTD at parallel fiber-PC synapses (Ekerot and Kano, 1985). Consequently, the balance of excitatory and inhibitory input onto PCs may determine the level of dendritic Ca2+ signaling and, ultimately, the extent of climbing fiber-dependent learning.

We examined climbing fiber-evoked Ca2+ signals in PC dendrites in vivo to determine how these responses are regulated by local circuit activity in the cerebellar cortex of awake behaving mice. During the performance of practiced movements, we found that parallel fiber and climbing fiber co-activity failed to produce supralinear Ca2+ signals. However, disinhibiting the molecular layer through chemogenetic suppression of MLI activity enhanced the amplitude of climbing fiber-evoked Ca2+ signals in PCs, specifically during movements when both parallel fibers and MLIs were activated. Quantitative ex-vivo measurements in PC dendrites confirmed that MLI-mediated feed-forward inhibition limits the ability of parallel fiber excitation to produce supralinearity. Bi-directional optogenetic actuation of MLI activity during parallel fiber stimuli altered associative parallel fiber-climbing fiber Ca2+ signaling, dependent on the level of MLI output. Our results show that climbing fiber Ca2+ signals in PCs are regulated by the counterbalance of MLI-mediated inhibition with parallel fiber-evoked EPSPs.

Results

Climbing fiber Ca2+ signals in PCs are unresponsive to behavior-induced parallel fiber activity

We used two-photon laser scanning microscopy (2pLSM) to measure climbing fiber-evoked Ca2+ activity in PC dendrites of lobule Crus II in head-fixed mice conditioned to lick for water from a port when cued by an audible tone (Figure 1A). After practice, expert mice reliably produced orofacial movements on command (Figure 1B; see also Gaffield and Christie, 2017). Climbing fiber-evoked Ca2+ events were continuously apparent in PCs transduced with the genetically encoded Ca2+ indicator GCaMP6f using an AAV vector under control of a truncated version of the PC-specific Pcp2 promoter (Chen et al., 2013; Nitta et al., 2017). During the initiation of licking, the frequency of Ca2+ events increased nearly five-fold, indicating engagement of the lateral cerebellum during water consumption (Gaffield et al., 2016). This behavior also elicited the activity of granule cell parallel fibers, as determined in separate Ca2+ imaging experiments from GCaMP6f-transduced neurons in the same area of Crus II (Figure 1—figure supplement 1A–D). As with MLIs in this region (Gaffield and Christie, 2017), average parallel fiber Ca2+ activity closely tracked lick rate (Figure 1—figure supplement 1E), indicating encoding of licking-related kinematics in their population as observed in other cerebellar lobules (Jelitai et al., 2016; Chen et al., 2017; Giovannucci et al., 2017; Knogler et al., 2017).

Figure 1 with 3 supplements see all
Climbing fiber-evoked Ca2+ signals in PCs during behavior.

(A) Head-fixed mice were trained to lick water from a port, cued by an audible tone, using the procedure shown. (B) Across-trial distribution of lick probability, aligned to the delivery of water, for a trained mouse. (C) Continuous record of Ca2+ activity in a PC dendrite. Expanded area shows algorithmically identified climbing fiber-evoked events (blue tick marks). Isolated Ca2+ events, indicated by the checkmark, were collected for analysis. (D) Average of climbing fiber-evoked Ca2+ events in PC dendrites occurring during water consumption (blue) or in the absence of licking (black). Measurements were obtained from 11 to 51 PCs in each of 11 mice; 211 cells total. (E) Genetic targeting of PCs and MLI using AAVs with GCaMP6f under control of the Pcp2 promoter and Cre-dependent RCaMP2 in Kit::Cre mice. In vivo images are from an infected area of Crus II. (F) The average frequency of climbing fiber-evoked Ca2+ events in PC dendrites (11 to 19 PCs in each of 5 mice; 82 cells total) plotted against the response in MLIs, acquired simultaneously in a subset of recordings (3 mice). (G) Trial-averaged measurement of MLI activity during cued licking. The peak amplitudes of Ca2+ events in PCs, plotted below, that correspond to three different phases of MLI activation during the task (4 to 19 PCs in each of 6 mice; 79 cells total; p=0.99, ANOVA test).

https://doi.org/10.7554/eLife.36246.002

To examine for an influence of parallel fibers on climbing fiber-evoked Ca2+ signaling in PCs, we collected well-isolated individual dendritic Ca2+ events from PCs (Figure 1C). Events were categorized whether they occurred during water consumption when granule cells were active or in the absence of licking movements when granule cells were relatively inactive. Isolated events comprised the majority of all identified Ca2+ responses in PC dendrites, whether or not the animal was actively engaged in water consumption (Figure 1—figure supplement 2A). Surprisingly, we found no difference in the average amplitudes of isolated Ca2+ events across PCs between these two behavioral states (ΔF/F = 0.223 ± 0.008 and 0.223 ± 0.011, licking and non-licking, respectively; p=0.99; N = 11 mice; paired Student’s t-test; Figure 1D). Similar results were also obtained for all other PC Ca2+ events; this included overlapping responses occurring closely in time (Figure 1—figure supplement 2B and C). The lack of behavioral-dependent differences in PC Ca2+ event amplitudes was unlikely attributable to GCaMP6f biophysics. While this Ca2+ indicator is nonlinear (Chen et al., 2013), it has been shown to be sensitive enough to report slight alteration of PC Ca2+ signals evoked by unexpected sensory stimuli (Najafi et al., 2014b). Thus, climbing fiber-evoked Ca2+ signaling in PC dendrites appeared resistant to licking-related co-activity of granule cells. This also suggests that PC dendritic Ca2+ responses are not always subject to enhancement in the context of behavior (Najafi et al., 2014a; Najafi et al., 2014b) and that the influence of parallel fiber excitation on the integrated dendritic response to climbing fiber input may be subject to regulation.

In addition to PCs, granule cells also excite MLIs. Therefore, it was not surprising that, using in vivo Ca2+ imaging, we found that the licking-induced activation of MLIs showed a close correspondence to that of parallel fibers, measured simultaneously, during licking (Figure 1—figure supplement 1B). This suggests a potential regulatory counterbalance of MLI-mediated feed-forward inhibition onto PCs in superimposition with their direct excitation by granule cells. If so, activation of MLIs during licking could alter conjunctive parallel fiber-climbing fiber Ca2+ signaling in PC dendrites. To examine the relationship between MLIs and climbing fiber-evoked responses in PCs, we used a dual-color imaging approach to simultaneously measure Ca2+ activity in both cell types. By performing experiments in Kit::Cre mice, a driver line with high specificity for MLIs (Amat et al., 2017; these mice may have a low abundance of Golgi cell targeting), we could transduce these cells with an AAV containing Cre-dependent RCaMP2 (Inoue et al., 2015), a red genetically encoded Ca2+ indicator spectrally separable from green GCaMP6f expressed in PCs (Figure 1E). During bouts of cued licking, climbing fibers evoke Ca2+ signals in PCs at the same time that the ensemble of surrounding MLIs was activated by parallel fibers (Figure 1F). However, the amplitude of climbing fiber-evoked Ca2+ events in PCs did not co-vary with the activity level of MLIs (Figure 1G and Figure 1—figure supplement 3) that increased and decreased in proportion to adjustments in lick rate during water consumption (Gaffield and Christie, 2017). In conclusion, the amplitude of climbing fiber-evoked Ca2+ responses in PC dendrites appeared unaffected by movement when both parallel fibers and MLIs were active.

Disinhibition enhances climbing fiber-evoked Ca2+ signals in PCs during parallel fiber activity

To determine whether MLI-mediated inhibition influences climbing fiber-evoked Ca2+ signals in PC dendrites, we chemogenetically suppressed the activity of MLIs and imaged Ca2+ responses in PCs during cued bouts of licking (Figure 2A). For this approach, we injected AAV containing Cre-dependent hM4d (Armbruster et al., 2007) into left Crus II of Kit::Cre mice to transduce MLIs and co-expressed GCaMP6f selectively in PCs using AAVs under control of the truncated Pcp2 promoter (Figure 2B). Intraperitoneal injection of clozapine-N-oxide (CNO), the cognate agonist of hM4d, led to an increase in the amplitude of isolated climbing fiber-evoked dendritic Ca2+ events in PCs relative to control measurements obtained in separate sessions (Figure 2C and F). Similar results were also apparent for all other non-isolated Ca2+ events as well (Figure 2—figure supplement 1A and B). Importantly, the effect of this chemogenetic manipulation was conditional, occurring only when MLIs were active. Whereas climbing fiber-evoked Ca2+ events were enhanced during the consumption of water, events occurring in the absence of licking movements were unaffected by disinhibition of the molecular layer (Figure 2D and F). In a separate set of experiments performed on a cohort of Kit::Cre mice lacking hM4d expression in MLIs, CNO administration had no effect on Ca2+ event amplitudes during water consumption (Figure 2E and F). This result rules out the possibility that an off-target influence of the drug accounted for the alteration of dendritic Ca2+ signaling.

Figure 2 with 4 supplements see all
MLIs suppress climbing fiber-evoked dendritic Ca2+ signals in PCs during licking.

(A) Following an audible cue, head-fixed mice licked from a port during in vivo imaging with 2pLSM. In some sessions, CNO was administered by intraperitoneal injection prior to the start of the task. (B) Image of fixed tissue showing PCs transduced with GCaMP6f and MLIs expressing HA-tagged hM4d (inset is a magnified view of labeled MLI somata). For this set of experiments, PCs were transduced with AAV containing the recombinase FLPo under control of the PC-specific Pcp2 promoter in combination with an FLPo-dependent AAV containing GCaMP6f to generate high-level expression of the Ca2+ indicator. (C–E) Average dendritic Ca2+ events in PCs recorded in control or in sessions with CNO (11 to 35 PCs each from 7 mice, 100 and 156 cells total; control and CNO, respectively). Isolated Ca2+ events were sorted based on whether or not they occurred in correspondence with licking. A separate group of sham mice lacked expression of hM4d were also tested (10 to 51 PCs each from 4 mice, 111 and 100 cells total; control and CNO, respectively). Difference signals are shown below in red. (F) Summary plot of the effect of chemogenetic MLI activity suppression on the amplitude of isolated dendritic Ca2+ events in PCs. Black, group average (mean ± SEM); gray, individual mice (N = 7). Asterisk indicates a significant difference from all other conditions (p<0.01; ANOVA with Tukey’s post-hoc multiple comparison test; all other comparisons were insignificant). (G) Top plot, the difference in trial-averaged PC dendritic Ca2+ activity between sessions in control and in CNO. Note the correspondence of the difference signal in PCs and the licking-evoked activation of MLIs measured simultaneously in a subset of experiments. Inset: the mean integrated Ca2+ activity during water consumption (lick) and in the absence of licking (no lick). Asterisk indicates a significant difference (p=0.025; paired Student’s t-test). Bottom plot, the probability of climbing fiber-evoked Ca2+ events in PCs was unchanged by molecular layer disinhibition, relative to that in control (11 to 35 PCs each from 6 mice, 145 and 97 cells total).

https://doi.org/10.7554/eLife.36246.009

A complementary analysis showed that molecular layer disinhibition by chemogenetics increased trial-averaged Ca2+ activity in PCs (Figure 2—figure supplement 2A and B). This unbiased measure reflects the integration of all dendritic Ca2+ during licking bouts independent of any event detection. The enhancement of trial-averaged dendritic Ca2+ activity was time-locked to the licking-evoked activation of MLIs, measured simultaneously in a subset of these experiments (Figure 2G). The effect of disinhibition on trial-averaged PC Ca2+ activity is likely attributable to an increased amplitude of climbing fiber events because chemogenetic suppression of MLI activity did not affect the rate of Ca2+ events in PCs (1.55 ± 0.19 Hz and 1.51 ± 0.47 Hz for control and CNO, respectively; p=0.51, paired Student’s t-test; Figure 2G) and climbing fiber inputs are known to produce the majority of Ca2+ elevation in dendrites in response to in vivo excitation (Mukamel et al., 2009; Ozden et al., 2009; Najafi et al., 2014b).

Our previous work showed that bilateral chemogenetic suppression of MLIs in Crus II slows the rate of licking, indicating an influence of MLIs on motor output (Gaffield and Christie, 2017). However, in these current experiments, we limited expression of hM4d to MLIs of left Crus II which, upon suppression by CNO administration, failed to significantly affect average licking dynamics (rate, pattern, rhythmicity, and time from cue presentation to lick initiation; Figure 2—figure supplement 3A–C, Figure 2—video 1). Although this result argues against the possibility that chemogenetic-induced behavioral perturbations accounted for alteration of climbing fiber-evoked responses in PCs, it may be that unquantified orofacial movements, such as lateral tongue displacement, were altered. We attempted to control for this more carefully by also examining PC Ca2+ activity in a subset of hM4d-expressing mice that had closely matched licking rates before and during molecular layer disinhibition. Even in these animals, isolated Ca2+ event amplitudes were increased with MLI activity suppression (Figure 2—figure supplement 3D). We conclude that alteration of climbing fiber-evoked Ca2+ signaling in PCs with disinhibition resulted from the influence of MLIs on PC dendritic integration. In summary, these results show that MLI-mediated inhibition recruited during motor behavior (Jelitai et al., 2016; Astorga et al., 2017; Gaffield and Christie, 2017) suppresses climbing fiber-evoked Ca2+ signaling in PCs dendrites. Hence, the integration of information transfer from the inferior olive to the cerebellar cortex, encoded in the Ca2+ activity of PCs will depend, in part, on the output of MLIs.

Disinhibition affects climbing fiber-evoked Ca2+ signaling throughout PC dendrites

We determined the extent to which MLI-mediated inhibition affects Ca2+ signaling across the activated ensemble of PCs. For this analysis, we used movement-related difference images of averaged PC dendritic activity obtained from sessions performed in the disinhibited condition normalized to that obtained in control sessions for each mouse (Figure 3A). More than 70% of identified PC dendrites exceeded a threshold level of altered activity indicative of enhancement during licking movement with MLI activity suppression (Figure 3B). Therefore, MLIs influence Ca2+ signaling in the majority of activated PCs likely owing to the widespread and coherent activation of these interneurons in lobule Crus II during orofacial motor behavior (Astorga et al., 2017; Gaffield and Christie, 2017).

MLIs broadly influence climbing fiber-evoked Ca2+ signaling in PCs.

(A) The across-session change in trial-averaged PC Ca2+ activity, colored-coded based on the extent to which chemogenetic disinhibition affected responses during movement (MDM ratio; see Materials and methods). Each algorithmically identified PC dendrite is numbered. (B) Cumulative probability histograms of the effect of disinhibition on trial averaged Ca2+ activity in identified PCs (2 to 15 PCs each from seven mice, 50 cells total). Dotted line demarcates a threshold level of change indicative of enhancement with disinhibition (MDM ratio >1). (C) Histogram of average Ca2+ activity measurements, determined for each dendritic pixel across the PC population, during licking (L) divided by that observed in the absence of licking (No L), in both control sessions (black) as well as after administration of CNO (gray). For all mice (N = 7), the distributions were significantly different (p<0.0001, Kolmogorov-Smirnov test). The inset shows the summary of ROC analysis on these distributions, obtained for each mouse, where area under the curve was used to calculate the percentage of pixels that showed an effect with chemogenetic disinhibition (gray, individual mice; black, the mean ± SEM). (D) In the image, two segments of a dendrite from distinct branches of the same PC are outlined. Measurements of Ca2+ activity in these segments are shown superimposed in the traces below. (E) Comparison of the amplitudes of simultaneous, inter-branch Ca2+ events for many climbing fiber-evoked responses for the two branches shown in example in panel D. Events were sorted based on whether they occurred during water consumption (black) or in the absence of licking (gray). Unity is marked by the dashed line. (F,G) Cumulative probability of the inter-branch variance of climbing fiber Ca2+ event amplitudes (see Materials and methods). Events were sorted depending on their correspondence with water consumption (black) or the absence of licking (gray). Distributions were not different in control sessions (N = 74 pairs; range: 4 to 20 pairs each from seven mice; p=0.57, Kolmogorov-Smirnov test) but were in the disinhibited condition with CNO (N = 114 pairs; range 12 to 28 pairs each from seven mice; p=0.0013, Kolmogorov-Smirnov test). (H) The effect of molecular layer disinhibition on average inter-branch variability of Ca2+ event amplitudes in PCs. In control sessions, the variance was similar whether or not events occurred during licking movements (p=0.50; paired Student’s t-test). In contrast, a modest, but significant drop in variability occurred during movement with disinhibition (p=0.029; paired Student’s t-test). Black, group average (mean ± SEM); gray, individual mice.

https://doi.org/10.7554/eLife.36246.018

A similar analysis was used to measure for subcellular effects of MLI-mediated inhibition on dendritic Ca2+ signals in PCs. First, we determined the average change in Ca2+ activity due to movements during cued water consumption relative to baseline measurements in the absence of licking for each pixel in all identified dendrites. This was calculated for the same PCs in both control sessions as well as during chemogenetic activity suppression of MLIs. There was a clear shift to larger values in the disinhibited condition (Figure 3C). Next, from these distributions, we used receiver operating characteristic (ROC) curves to estimate that ~66% of the fractional area of all individual dendrites showed enhanced Ca2+ signaling with MLI activity suppression (Figure 3C, inset). This indicates that enhancement occurred throughout individual dendrites suggesting that MLIs produce a widespread influence on PC Ca2+ signaling.

PCs receive input from many MLIs that make distributed synaptic contacts onto their dendrites (Palay and Chan-Palay, 1974; Kim et al., 2014). By producing localized inhibitory effects on climbing fiber-evoked dendritic spiking, MLIs can contribute to increased variability of climbing fiber-evoked Ca2+ signals in the arbors of individual PCs (Callaway et al., 1995; Kitamura and Häusser, 2011). Comparisons of climbing fiber Ca2+ activity in two distinct branches of the same PC dendrite showed that, although evoked responses occurred reliably at both locations, their amplitudes could differ slightly (Figure 3D). This held true whether or not the animal was consuming water (Figure 3E,F and H). However, after MLI activity suppression by intraperitoneal injection of CNO, there was a modest decrease in the inter-branch variability during licking movements (Figure 3G) that was statistically different (Figure 3H). This points to a potential non-uniform effect of MLI-mediated inhibition on the PC response to climbing fiber excitation during task engagement. In conclusion, MLIs broadly influence climbing fiber-evoked Ca2+ signaling in PC dendrites during a practiced motor behavior.

Enhanced PC Ca2+ signaling is not attributable to an alteration in climbing fiber activity

It remains possible that the enhancement of Ca2+ signaling in PCs following molecular layer disinhibition reflects an increase in presynaptic climbing fiber activity which is translated into a larger postsynaptic response. This is because burst firing of olivary projection neurons (Crill, 1970), conveyed to the cerebellar cortex by climbing fibers, promotes increased PC dendritic spiking and larger amplitude Ca2+ signals (Mathy et al., 2009; Kitamura and Häusser, 2011). To assess this possibility, we directly measured the activity of climbing fibers using Ca2+ imaging. We injected AAV containing GCaMP6f under control of the CaMKIIα promoter into the inferior olive transducing excitatory projection neurons (Mathews et al., 2012) as evidenced by transgene expression in climbing fiber axons in the cerebellar cortex (Figure 4A). In recordings from awake animals, we used automated routines (Hyvärinen, 1999) to group like-responding pixels from images obtained in the molecular layer. The resulting segments comprised individual, sagittally-aligned climbing fibers (Figure 4B).

Disinhibition does not affect presynaptic climbing fiber activity.

(A) AAVs containing genetically encoded activity reporters and effectors were injected in the inferior olive and lobule Crus II of Kit::Cre mice, respectively. Image from fixed tissue showing GCaMP6f expression in climbing fibers and HA-tagged hM4d in MLIs. (B) In the image, individual climbing fibers were identified using automated segmentation routines. Traces show activity measurements from color-coded climbing fibers. (C) Ca2+ event rates in PC dendrites and climbing fibers, measured in separate cohorts of mice (11 to 19 PCs and 2 to 6 climbing fibers each from 7 mice, 100 and 29 total, respectively). Black circles, mean ± SEM; gray circles, measurements from individual mice (p=0.92, Student’s t-test). (D) Distribution of Ca2+ event amplitudes for an individual climbing fiber, all climbing fibers in a single mouse (N = 6), and for all mice (2 to 12 climbing fibers each from 7 mice, 36 fibers total). Data were normalized to facilitate comparisons across climbing fibers. (E) The frequency of Ca2+ events in climbing fibers during cued licking (average of 3 mice). (F) Average of isolated Ca2+ events in climbing fibers collected either during the consumption of water (blue) or in the absence of licking (black). Measurements obtained from 4 to 12 climbing fibers each from 5 mice, 38 fibers total. (G) Ca2+ events recorded in climbing fibers both in control and during sessions with chemogenetic MLI activity suppression. Events were collected only during periods of water consumption (4 to 9 climbing fibers each from 5 mice, 26 fibers total). The difference signal is shown in red.

https://doi.org/10.7554/eLife.36246.019

Ca2+ events occurred regularly in climbing fibers with mean rates comparable to the frequency of dendritic events in PCs, measured in separate animals (Figure 4B and C), and within the range of previously published PC event rates (Mukamel et al., 2009; Ozden et al., 2012; De Gruijl et al., 2014). Therefore, in awake mice, climbing fiber activity reliably drives dendritic spiking in their postsynaptic targets. The amplitudes of Ca2+ events in individual climbing fibers showed considerable variation during ongoing activity (Figure 4B). Sorted events, collected across mice, had a non-normal distribution (p<0.0001, Shapiro-Wilk Test; N = 3579 events; 30 climbing fibers; 7 mice). Instead, amplitude distributions skewed towards larger values suggestive of a multimodal composition (Figure 4D). Our interpretation of these observations is that Ca2+ events in climbing fibers are generated by discrete, high-frequency (100–400 Hz) bursts of firing and that the variance in their amplitudes reflects heterogeneity in the number of action potentials contained in the burst (Mathy et al., 2009).

The frequency of Ca2+ events in climbing fibers increased at licking onset (Figure 4E). This indicates that licking initiation is signaled in their activity and generates the corresponding uptick in dendritic Ca2+ events in PCs during the same period of the task (see Figure 1F). To evaluate whether burst firing in climbing fibers encodes licking-related information, we compared Ca2+ events evoked during water consumption to those occurring in the absence of licking (Figure 4F). The lack of significant differences between the amplitudes of mean events in these conditions (p=0.89; paired Student’s t-test) indicates that the spike content of presynaptic bursts in climbing fibers varies independent of practiced movements. Because these measurements were performed in Kit::Cre mice expressing hM4d in MLIs, we disinhibited the molecular layer by intraperitoneal injection of CNO (Figure 4A) and assessed for differences in climbing fiber Ca2+ events. Chemogenetic suppression of MLI activity had no effect on Ca2+ event rates (change of 15.5 ± 16.2% from control, p=0.47, paired Student’s t-test) nor did it affect the amplitude of events evoked during licking when MLIs are normally activated (p=0.66, paired Student’s t-test; Figure 4G). This result rules out that enhancement of PC dendritic Ca2+ signaling with molecular layer disinhibition is due to a change in presynaptic climbing fiber activity. Furthermore, it argues against the possibility that unresolved, closely-spaced Ca2+ events in postsynaptic PCs, erroneously categorized as individual responses, accounted for the amplitude change with MLI activity suppression. Otherwise, this would have also been reflected as a corresponding increase in Ca2+ event amplitudes in presynaptic climbing fibers.

MLIs suppress supralinear climbing fiber-evoked Ca2+ signaling in PC dendrites

The amplitude of climbing fiber-evoked Ca2+ signals in PC dendrites can be enhanced if climbing fibers are stimulated in conjunction with preceding parallel fiber activity (Wang et al., 2000). This supralinearity may reflect a change in dendritic excitability that facilitates the propagation of climbing fiber-evoked Ca2+ spikes into spiny branchlets yielding additional Ca2+ entry (Otsu et al., 2014; but see Wang et al., 2000). Parallel fibers also activate MLIs, driving rapid feed-forward inhibition that attenuates parallel fiber excitation of PCs (Brunel et al., 2004; Mittmann et al., 2005). We reasoned that, by reducing dendritic excitability, feed-forward inhibition could diminish the ability of parallel fibers to enhance subsequent climbing fiber-evoked Ca2+ responses and thus provide a mechanism to explain in vivo gating of non-linear dendritic Ca2+ signaling in PCs during licking movements.

To quantitatively assess this possibility, we measured Ca2+ activity in individual PC dendritic branches using 2pLSM in acute cerebellar slices from mature Kit::Cre mice. Experiments were performed in the absence of synaptic blockers while rapidly and reversibly suppressing MLI firing with high temporal precision using the anion-fluxing channelrhodopsin GtACR2 (Govorunova et al., 2015), transduced in MLIs by Cre-dependent AAV (Figure 5A and Figure 5—figure supplement 1A). In whole-cell PC recordings, electrical stimulation of parallel fibers adjacent to the PC dendrite evoked depolarizing post-synaptic potentials (PSPs) that were prolonged when MLI activity was optogenetically prevented with a pulse of blue light coincident with the parallel fiber stimulus (Figure 5B and C and Figure 5—figure supplement 1B and C). The effect of optogenetic MLI activity suppression on the parallel fiber-evoked PSP integral was indistinguishable from that produced following pharmacological block of GABAA receptor-mediated transmission, indicating that GtACR2 completely prevented feed-forward inhibition (Figure 5—figure supplement 1D).

Figure 5 with 2 supplements see all
Feed-forward inhibition attenuates supralinear Ca2+ signaling in PC dendrites.

(A) In acute slices, parallel fibers were stimulated in conjunction with climbing fibers during whole-cell patch recording from PCs in lobule Crus II. Ca2+ imaging was performed using 2pLSM in PC spiny dendrites as shown in the fluorescence image. (B) Evoked responses in a PC either to the parallel fiber tetanus (3 pulses at 100 Hz), the climbing fiber stimulus, or their conjunctive pairing (50 ms interval). (C) In alternate trials, MLIs expressing GtACR2 were photoinhibited using wide-field illumination with blue light (λ = 461 nm; 40 ms; 6.6 mW/mm2), coincident with the parallel fiber tetanus. Average parallel fiber-evoked PSPs in control and during optogenetic suppression of feed-forward inhibition are enlarged on the right. (D) Left, the average climbing fiber-evoked Ca2+ signal in a PC dendrite (location demarcated by the yellow arrowhead in the morphological image in panel A) produced following conjunctive stimulation with parallel fibers. The estimated summed response, shown in gray, for parallel fiber and climbing fiber transients evoked in isolation on separate trials (Ca2+ activity traces shown above). Right, climbing fiber-evoked Ca2+ signals from the same dendritic location but with feed-forward inhibition suppressed by optogenetics during the parallel fiber tetanus. (E) The change in amplitude of climbing fiber-evoked Ca2+ signals with conjunctive stimulation of parallel fibers, measured in the same PC dendrite, in trials either with or without optogenetic suppression of MLI-mediated feed-forward inhibition. Data are normalized to the estimated, summed response of parallel fibers and climbing fibers for each condition. Each point is a measurement from a different dendritic branch (3 to 10 sites for each of 5 PCs, 31 sites total) with unity demarcated by the dashed line. (F) Somatic complex spikes evoked by the climbing fiber stimulus, both in isolation as well as in conjunction with parallel fiber activation. Responses in the same PC in control and with MLIs photo-inhibited during the parallel fiber tetanus.

https://doi.org/10.7554/eLife.36246.021

Electrical stimulation of climbing fibers evoked complex spikes in the PC soma (Figure 5B) and accompanying Ca2+ transients in their spiny dendrites, resolved by including the Ca2+ indicator Fluo-5F in the patch pipette (Figure 5D). Pairing a brief parallel fiber tetanus in conjunction with the climbing fiber stimulus (50 ms interval) produced dendritic Ca2+ responses that were no different than the expected summed combination of the individual parallel fiber and climbing fiber transients, computed from separately evoked responses on alternate trials (100.5 ± 2.1% of sum; N = 31, p=0.78; paired Student’s t-test; Figure 5D). Changing the timing of the preceding parallel fiber tetanus, relative to the conjunctive climbing fiber stimulus, failed to uncover supralinear Ca2+ signaling (range: 25–100 ms; Figure 5—figure supplement 2A). However, when we optogenetically suppressed feed-forward inhibition during the parallel fiber tetanus in interleaved trials, the amplitude of the climbing fiber-evoked Ca2+ transient was greater than the estimated summed response (109.9 ± 2.4% of sum; N = 31; p=0.0003; paired Student’s t-test; Figure 5D). Thus, supralinear Ca2+ signaling in PCs could be uncovered with molecular layer disinhibition.

On average, enhancement of climbing fiber-evoked Ca2+ signaling by parallel fibers in the absence of MLI feed-forward inhibition was greatest for short intervals and decayed to non-significance for long intervals (Figure 5—figure supplement 2B). Thus, supralinear Ca2+ signaling was dependent on the timing of parallel fiber and climbing fiber activity (Wang et al., 2000; Brenowitz and Regehr, 2005). Although, even at short intervals, the effect of disinhibition on supralinear Ca2+ signaling varied to some extent across dendrites (Figure 5E). This suggests that the temporal sensitivity of parallel fiber-climbing fiber interactions may be set not only by the functional region of the cerebellar cortex (Suvrathan et al., 2016) but also locally at the level of individual synapses. The number of spikelets in the somatic complex spike burst was unaffected by suppression of feed-forward inhibition (4.2 ± 0.2 and 4.5 ± 0.3 spikelets; control and with disinhibition, respectively; N = 6; p=0.17; Student’s t-test; Figure 5F) pointing to the compartmentalized influence of MLI activity on climbing fiber signaling in PC dendrites (Callaway et al., 1995). Together, these results indicate that, with feed-forward inhibition intact, brief parallel fiber activation failed to enhance climbing fiber-evoked Ca2+ signals likely due to the attenuating influence of MLI-mediated inhibition on the parallel fiber EPSP. However molecular layer disinhibition revealed the latent ability of parallel fibers to enhance climbing fiber-evoked dendritic Ca2+ signaling, similar to our in vivo findings.

Activity-dependent recovery of supralinear PC dendritic Ca2+ signaling by parallel fibers

Granule cells encode sensorimotor information conveyed through the mossy fiber pathway, with their activation level dependent on self-produced and external stimuli. This includes enhanced firing in response to the multimodal integration of many mossy fiber input streams (Ishikawa et al., 2015; Giovannucci et al., 2017). To further examine if associative climbing fiber-evoked Ca2+ signaling in PC dendrites is sensitive to the level of parallel fiber activity, we increased the number of stimuli in the parallel fiber tetanus (Figure 6A and B). With a more prolonged tetanus, conjunctive climbing fiber-evoked Ca2+ signals were of greater amplitude than the estimated summed responses of parallel fibers and climbing fibers alone (Figure 6B and C; also observed in a matched subset of observations at the same dendritic site: 3 PF stimuli, ΔF/F 101.1 ± 2.5% of expected linear sum; 9 PF stimuli, ΔF/F 114.7 ± 3.6%; p=0.005; N = 14; paired Student’s t-test). Such supralinear Ca2+ signaling was apparent in the majority of PC dendrites examined, in contrast to that observed in separate PC recordings using a tetanus with fewer parallel fiber stimuli (Figure 6C and D). Thus, with a sufficient level of parallel fiber activation, the resulting direct excitation of PCs can overwhelm feed-forward inhibition to recover supralinear Ca2+ signaling. This indicates that the balance of dendritic excitation and inhibition through the mossy fiber pathway is a critical determinate of conjunctive parallel fiber-climbing fiber Ca2+ signaling in PCs.

Figure 6 with 1 supplement see all
Activity-dependent recovery of supralinear climbing fiber Ca2+ signaling is sensitive to MLI inhibition.

(A) Acute slice recording configuration. In a subset of experiments, bReaChES was expressed in MLIs by Cre-dependent AAV in Kit::Cre mice. (B) Comparison of average climbing fiber-evoked Ca2+ signals in two different PCs with the conjunctive parallel fiber tetanus including either 3 or 9 stimuli (100 Hz). The summed response of the parallel fiber and climbing fiber transients, evoked in isolation on separate trials, is shown in gray. (C) Across-cell comparison shows that increasing the number of stimuli in the parallel fiber tetanus results in a supralinear enhancement of climbing fiber Ca2+ signals in PC spiny dendrites. Individual dendritic recording sites are indicated in gray (N = 41 and 42 dendritic sites from 6 and 4 cells; 3 and 9 stimuli, respectively) with mean data in black (±SEM; p<0.0001; Student’s t-test). (D) In the cumulative probability histogram, supralinear climbing fiber Ca2+ signaling was observed in a majority of PC dendrites when stimulated with a longer lasting parallel fiber tetanus. (E) Climbing fiber-evoked Ca2+ signals at the same PC dendritic site. In interleaved trials, optogenetic activation of MLIs (λ = 596 nm; 40 ms; 0.93 mW/mm2) occurred during the parallel fiber tetanus. (F) Relationship between the change in amplitude of climbing fiber-evoked Ca2+ signals with conjunctive activity of parallel fibers, in trials either with or without optogenetic activation of MLIs during the parallel fiber tetanus (3 to 7 dendritic sites from 4 cells, 17 sites total). Data are normalized to the estimated, summed response of the parallel fiber and climbing fiber transients for each condition recorded at the same PC dendritic location. Dashed line is unity. (G) The effect of varying optogenetic MLI-mediated inhibition on the ability of parallel fibers to produce supralinear climbing fiber Ca2+ signals in PC dendrites. Mean data (black symbols ± SEM) are from matched comparisons at the same dendritic recording site (4 to 5 sites from 2 cells, 9 dendrites total; p<0.05; Repeated measures 1-way ANOVA with Tukey’s post hoc multiple comparison test). Gray symbols are individual measurements.

https://doi.org/10.7554/eLife.36246.025

We tested this possibility more rigorously in acute cerebellar slices prepared from Kit::Cre mice infected with a Cre-dependent AAV vector containing the excitatory red-shifted channelrhodopsin bReaChES (Rajasethupathy et al., 2015). This allowed us to optogenetically increase the inhibitory output of transduced MLIs during the prolonged parallel fiber tetanus (Figure 6A). Supralinear Ca2+ signaling in PC dendrites evoked by prolonged parallel fiber and climbing fiber conjunctive stimuli (ΔF/F 121.5 ± 5.3% of the expected linear sum of both inputs) was abolished in alternate trials when MLIs were optogenetically activated coincident with the parallel fiber tetanus (98.0 ± 3.5% of summed response; p=0.004; N = 17; paired Student’s t-test; Figure 6E and F; Figure 6—figure supplement 1A and B). In a subset of experiments, MLI activity was systemically varied during the same recordings using different photostimulus intensities in alternating trials. Supralinear climbing fiber Ca2+ signals in PC dendrites were reduced, dependent on the activity level of MLIs during the parallel fiber tetanus (Figure 6G). This supports the hypothesis that the relative activity of parallel fibers and MLIs not only determines whether coincident parallel fiber and climbing fiber activity produces supralinear Ca2+ signaling in PC dendrites, but also the magnitude of the enhancement as well.

Discussion

Using a combination of Ca2+ imaging and genetically encoded effectors of activity, we find that inhibitory MLIs exert a profound regulatory influence on climbing fiber-evoked Ca2+ signaling in PC dendrites. In awake mice engaged in a routine motor task that activates parallel fibers, MLIs enforce normalization of dendritic climbing fiber-evoked Ca2+ signals matching those occurring spontaneously during quiescence when parallel fibers are inactive. In ex vivo recordings, short bursts of parallel fiber stimuli fail to evoke supralinear climbing fiber Ca2+ signals in PCs dendrites due to MLI-mediated feed-forward inhibition that attenuates parallel fiber EPSPs. Thus, during the performance of practiced movements, recruitment of inhibition from MLIs gates climbing fiber-evoked Ca2+ signaling that might otherwise induce plasticity and, therefore, is well positioned to constrain learning in the absence of motor performance errors.

Encoding of climbing fiber-evoked Ca2+ signals in PCs

Parallel fiber excitation drives simple spiking in PCs while convergent inhibition from MLIs influences the frequency and regularity of these responses (Häusser and Clark, 1997; Mittmann et al., 2005; Dizon and Khodakhah, 2011). In addition, climbing fiber-mediated activation of dendritic voltage-gated Ca2+ channels produces local, regenerative spikes in PC dendrites that are subject to amplification by preceding parallel fiber activity (Wang et al., 2000; Schmolesky et al., 2002; Otsu et al., 2014). Our observations indicate that, in behaving mice, climbing fiber-evoked Ca2+ signaling in PC dendrites is resistant to enhancement by co-active parallel fibers due to inhibition from MLIs. Therefore, in addition to influencing the pattern and timing of PC simple spiking during movements (Jelitai et al., 2016; Chen et al., 2017), MLIs also participate in encoding the Ca2+ response to climbing fiber excitation by locally gating the dynamics of non-linear dendritic signaling. We propose that by performing these non-mutually exclusive operations, MLIs are well-placed to influence motor control (Heiney et al., 2014) as well as motor learning (Jörntell et al., 2010).

Parallel fiber-evoked EPSPs inactivate subthreshold, voltage-gated K+ currents (ISA) in PC dendrites that can facilitate the initiation and propagation of subsequent climbing fiber-evoked Ca2+ spikes into spiny branchlets, thereby enhancing intracellular Ca2+ entry (Otsu et al., 2014). In our slice recordings, we observed that MLI feed-forward inhibition diminished the ability of parallel fibers to enhance climbing fiber-evoked Ca2+ responses. This effect is likely attributable to the marked reduction in parallel fiber-triggered excitatory potentials by feed-forward inhibition (Brunel et al., 2004; Mittmann et al., 2005). Attenuated PSPs are expected to be less efficacious in generating depolarization-induced K+ channel inactivation. However, in our acute slice experiments, the sensitivity of climbing fiber-evoked Ca2+ signaling to enhancement by parallel fibers could be restored by longer lasting parallel fiber stimuli. Although, this effect could be offset, dependent on the level of MLI activity. Thus, gating of non-linear dendritic signaling in PCs is not necessarily an ‘all-or-none’ phenomenon as classically described for direct climbing fiber excitation (Eccles et al., 1966). Rather, the average amplitude of Ca2+ signals in PC dendrites elicited by climbing fibers will depend on the net balance of preceding parallel fiber excitation and MLI-mediated inhibition, with the accumulation and recovery of ISA inactivation evolving with the summating PSPs. Our analysis did not distinguish whether, at the single trial level, the balance of PF-mediated excitation and inhibition influences the probability or the amplitude of a supralinear Ca2+ event in PC dendrites, though both are apt to affect plasticity induction.

In our in vivo recordings, the amplitude of climbing fiber-evoked Ca2+ events in PC dendrites did not co-vary with the level of MLI activation during water consumption. This implies that a homeostasis is achieved between the excitatory activity of parallel fibers and inhibition from MLIs to prevent supralinear climbing fiber Ca2+ signaling. Even under conditions where MLI activity decreased during motor behavior (e.g., as the lick rate decreased), excitation and inhibition appeared in balance, as determined by the high correspondence of parallel fiber and MLI activity. This likely limited any enhancement of PC Ca2+ events during task performance. Whether feed-forward inhibition can continuously counteract parallel fiber excitation across a range of behavioral conditions is yet to be determined.

Apart from regulating supralinear enhancement of Ca2+ signals by parallel fibers, MLIs can also reduce the amplitude of climbing fiber-evoked Ca2+ transients through direct electrogenic suppression of PC dendritic spiking (Callaway et al., 1995; Kitamura and Häusser, 2011). We cannot rule out the possibility that this mechanism also affects climbing fiber-evoked Ca2+ signaling during practiced movements. However, direct suppression of dendritic spiking by inhibition would result in smaller amplitude Ca2+ events during movement when MLIs are active compared to spontaneous events occurring in quiescence. We did not observe such an effect in our in vivo measurements. Therefore, we conclude that the predominant role of MLIs on climbing fiber-mediated Ca2+ signaling during practiced motor behavior is through their inhibitory influence on non-linear dendritic operations in PCs.

Climbing fiber Ca2+ signaling and plasticity

Ca2+ signals produced in the PC dendrite by climbing fiber excitation have been subject to intensive investigation because intracellular Ca2+ elevation is a biochemical trigger for inducing synaptic plasticity (Finch et al., 2012; Lamont and Weber, 2012), and synaptic plasticity has been implicated as a neural correlate of motor learning and memory (Marr, 1969; Albus, 1971; Ito and Kano, 1982). When activated alone, climbing fiber-evoked Ca2+ signals are insufficient to induce synaptic re-weighting of parallel fiber-PC synapses. However, amplification of climbing fiber signals by preceding parallel fiber activity can achieve a threshold level of Ca2+ elevation necessary to generate short- and long-term plasticity at parallel fiber inputs (Wang et al., 2000; Brenowitz and Regehr, 2005; Tanaka et al., 2007). Therefore, by preventing parallel fiber enhancement of Ca2+ signaling, MLI-mediated inhibition may suppress the induction of climbing fiber-mediated plasticity despite the interaction of these two inputs.

In the absence of MLI constraint during behavior, parallel fiber enhancement of climbing fiber Ca2+ signaling would occur continuously during self-generated movement. This would result in the continuous induction of climbing fiber-mediated plasticity at co-active parallel fiber inputs. Unconstrained plasticity in the disinhibited cortex may force parallel fiber-PC synapses into a saturated state where motor learning can no longer occur (Nguyen-Vu et al., 2017). Interestingly, genetic deletion of GABAARs from PCs impedes consolidation of motor learning (Wulff et al., 2009), a result that may be attributable to the disinhibition of PCs and indiscriminate plasticity produced by unregulated, supralinear climbing fiber Ca2+ signaling.

Inhibitory control of plasticity by modulation of dendritic excitability may be a ubiquitous function of GABA-releasing interneurons in the brain. Interneurons in the cortex, hippocampus, amygdala, and striatum are known to dynamically regulate dendritic processing in their postsynaptic targets, electrogenically gating synaptic interactions that lead to alterations in circuit function (Paulsen and Moser, 1998; Letzkus et al., 2015). In this report, we extend inhibitory regulation to non-linear Ca2+ signaling in PCs. Although inhibition can elicit branch-and synapse-specific control of Ca2+ signaling (Callaway et al., 1995; Chiu et al., 2013), our results indicate that during behavior, MLIs exert a widespread suppression of supralinear climbing fiber Ca2+ signaling throughout the PC dendrite, perhaps owing to the highly coherent activation of their ensemble during orofacial movements (Astorga et al., 2017; Gaffield and Christie, 2017). Even in experiments where molecular layer disinhibition reduced inter-branch variance of climbing fiber-evoked Ca2+ responses, the effect was small. This suggests that heterogeneity of MLI-mediated inhibition on Ca2+ signaling within PC dendritic arbors is limited during practiced movement. Therefore, we envision that inhibition gates plasticity on a cell-wide scale.

Behavioral significance of MLI-gating of Ca2+ signaling in PCs

A novel aspect of our findings is that Ca2+ responses in PC dendrites are not always augmented by the context of active motor behavior. However, the lack of supralinear Ca2+ signaling during movement in our study does not preclude the possibility that such responses could occur under different behavioral conditions. In our experiments, mice elicited well-rehearsed motor responses that, through prior practice, resulted in highly stereotyped licks trial after trial indicative of few performance errors (Gaffield and Christie, 2017). During these trials, as well as during un-cued licking with free access to water, climbing fiber-evoked Ca2+ activity increases in PCs at the onset of water consumption (Gaffield et al., 2016). Importantly, we have not yet determined the role these signals play in the cued licking task, nor whether they (or the cerebellum in general) are required for the behavior. That the amplitudes of these evoked Ca2+ events are indistinguishable from spontaneous activity suggests that they may not be useful for learning. As the mice are performing a skilled behavior, these movement-evoked Ca2+ events could help with motor memory retention (Medina et al., 2002). During more mistake-prone behaviors, where mice must learn to carefully articulate their movements using the benefit of sensorimotor associations, it may be that parallel fibers contribute to climbing fiber-evoked Ca2+ signaling to reach a threshold necessary for inducing plasticity and modification of motor output. In this way, we do not discount the possibility that the extent of climbing fiber-mediated excitation of PCs, which may vary dependent on the severity of motor errors, plays a role in determining the magnitude of learning outcomes (Yang and Lisberger, 2014).

This hypothesis is consistent with work showing that external sensory cues that can guide associative learning (e.g., classical eyeblink conditioning), evoke graded climbing fiber-evoked Ca2+ signals in PC dendrites (Najafi et al., 2014a; Najafi et al., 2014b). Enhancement of these signals may be driven, in part, from conjunctive activity of parallel fibers (Giovannucci et al., 2017). In this scenario, enhanced climbing fiber-evoked Ca2+ signaling in PCs during learning may arise from alterations in network activity within the cerebellar cortex. Spike bursting of granule cells could shift the balance of parallel fiber excitation and feed-forward inhibition through short-term plasticity (Häusser and Clark, 1997; Grangeray-Vilmint et al., 2018), perhaps unlocking the ability of parallel fibers to generate non-linear Ca2+ signaling in PCs when activated by multimodal streams of sensorimotor information, as occurs during associative learning (Chadderton et al., 2004; Ishikawa et al., 2015; Giovannucci et al., 2017). In addition, MLIs inhibit one another through their structured GABAergic interconnections (Kim et al., 2014; Rieubland et al., 2014). Increased inhibition between MLIs in response to salient learning events would allow parallel fibers to generate supralinear Ca2+ signals in PCs during climbing fiber excitation. Alternatively, sensorimotor signals useful for producing learning could be conveyed to PCs through a separate population of granule cells whose parallel fibers bypasses MLIs (Ekerot and Jörntell, 2001). Future work will help clarify if and how MLI regulation of climbing fiber Ca2+ signaling in PCs is altered during learning.

An important caveat to our study is that we did not specifically monitor granule cell activity in the disinhibited state. Chemogenetic suppression of some Golgi cells, which may be expected considering that Kit::Cre mice are not pristinely selective for MLIs (Amat et al., 2017), might produce enhanced excitability in a subpopulation of granule cells. We cannot rule out a scenario where such enhancement of PF signaling onto PCs - not apparent in licking behavior nor in the responsiveness of MLIs whose activity we monitored - directly contributes to increased PC dendritic Ca2+ activity. Notably, the absence of a disinhibitory effect on the behavior-induced MLI population response following CNO administration may reflect the inability of our GCaMP imaging approach to discern the contribution of spike-driven Ca2+ entry from activity mediated by synaptic sources (e.g., Ca2+ permeable AMPA and NMDA receptors); the latter is expected to be relatively less susceptible to alteration by chemogenetic MLI activity suppression. However, strong hM4d-mediated suppression of MLI neurotransmission, apart from spiking and synaptic activation (Stachniak et al., 2014; Amat et al., 2017), combined with evidence from our ex vivo experiments showing a direct inhibitory influence of MLIs on PF-evoked enhancement of PC dendritic Ca2+ signaling, support our conclusion that MLI activity gates PC responses to climbing fiber excitation in vivo.

In summary, our results emphasize the importance of the PC dendrite as a central locus for encoding the integrated response to climbing fiber input (Najafi and Medina, 2013) and, hence, determining the physiological consequence of olivary signaling in the cerebellar system. Climbing fibers fire in response to motor errors. Climbing fibers also fire during normal movements, activity that may be important for coordinating motor timing (Lang et al., 2017). Context-specific MLI activity might allow climbing fibers to functionally multiplex. Complex spikes in the PC soma could be transmitted to downstream premotor targets and influence motor control apart from Ca2+ signaling in the PC dendrite where plasticity is induced. By gating supralinear climbing fiber-evoked Ca2+ signaling, molecular layer inhibition may prevent unwarranted or unnecessary adaptation during accurately performed movements.

Materials and methods

Key resources table
Reagent type
or resource
DesignationSource or referenceAdditional information
strain, (Mus Musculus)Kit::CreAmat et al., 2017on C57Bl/6
background
transfected constructAAV1-Pcp2.4-GCaMP6fUniversity of North
Carolina
custom
transfected constructAAV1-Pcp2.4-FLPoUniversity of North
Carolina
custom
transfected constructAAV1-CAG-Flex(FRT)
rev-RCaMP2
University of North
Carolina
custom
transfected constructAAV1-CAG-Flex(loxP)
rev-RCaMP2
University of
Pennsylvania
custom
transfected constructAAV1-CAG-Flex(FRT)
rev-GCaMP6f
University of North
Carolina
custom
transfected constructAAV1-CAG-Flex(loxP)
rev-ChR2.HA-2a-hM4d
ViGenecustom
transfected constructAAV1-Syn-GCaMP6fUniversity of
Pennsylvania
AV-1-PV2822
transfected constructAAV1-CaMKIIα-GCaMP6fUniversity of PennsylvaniaAV-1-PV2822
transfected constructAAV1-EF1α-Flex(loxP)
rev-GtACR2.eYFP
ViGenecustom
transfected constructAAV5-EF1α-Flex(loxP)
rev-bReachES-TS-YFP
University of North
Carolina
shelf
antibodyanti HAAbcam#ab9110
software, algorithmPrismGraphPadStatistical analysis
software, algorithmMatlabMathworksImage analysis
software, algorithmImageJNIHImage analysis
software, algorithmbControlCarlos Brody, PrincetonBehavior control
software, algorithmScanImageVidrio TechnologiesMicroscope control

Animals

Animal procedures were conducted using protocol 15–205 approved by the Institutional Animal Care and Use Committee (IACUC) at the Max Planck Florida Institute for Neuroscience. Heterozygous adult Kit::Cre mice (Amat et al., 2017) of both genders were used for all experiments (in vivo and ex vivo:>10 and>7 weeks of age, respectively).

Surgical procedures

As described previously (Gaffield et al., 2016), cranial windows for in vivo imaging in the cerebellum were prepared from mice under isoflurane (1.5–2.0%). Warmth was provided by a heating pad using biofeedback to maintain a stable core body temperature (37°C). Non-responsiveness to intermittent toe pinches confirmed the surgical plane of anesthesia. For this procedure, the skull was exposed through surgical excision of the scalp (subcutaneous injection of lidocaine/bupivacaine provided local anesthesia). A custom-engineered stainless steel head post was then attached onto the dried, exposed bone, centered on the midline of the cranium, using Metabond (Parkell, Edgewood, NY). A small craniotomy (~2 mm square) was cut over the left lateral cerebellum using a scalpel without disturbing the underlying dura mater. The opening was covered with a small glass coverslip (CS-3R, Warner Instruments, Hamden, CT), and cemented in place with Metabond such that the window applied minimal pressure to the brain. Post-operative analgesia (buprenorphine; 0.35 mg/kg) was administered and the animal recovered under supervision until ambulatory.

Prior to placement of the window, adeno-associated viruses (AAVs) were pressure injected into the brain using beveled glass micropipettes. Viruses included: AAV1-Pcp2.4-GCaMP6f, AAV1-Pcp2.4-FLPo, AAV1-CAG-Flex(FRT)rev-RCaMP2, AAV1-CAG-Flex(loxP)rev-RCaMP2, AAV1-CAG-Flex(FRT)rev-GCaMP6f, AAV1-CAG-Flex(loxP)rev-ChR2.HA-2a-hM4d, AAV1-Syn-GCaMP6f (all custom prepared at the University of North Carolina Vector Core Facility, the University of Pennsylvania Vector Core Facility, or ViGene, Rockville, MD). For injections into lobule Crus II, the micropipette generally contained multiple viruses (100–150 nl) in order to avoid repeated penetrations of the same location. Three different injection depths (150–350 µm below the dura) were used to evenly transduce both PCs and MLIs. To transduce granule cells, AAV1-Syn-GCaMP6f was injected 250–350 mm below the dura (Giovannucci et al., 2017); this non-selective approach also resulted in expression of the Ca2+ indicator in MLIs and Golgi cells. For climbing fiber transduction, AAV1-CaMKIIα-GCaMP6f (University of Pennsylvania) was injected into the inferior olive through a surgical opening in the back of the neck with access to the brainstem through the foramen magnum at the following coordinates from lambda: x = 0.3 mm, y = −4.9 mm, z = −4.6 mm at a depth of 3.6 mm using an approach angle of 62°. Total volumes of injections were ~500 nl. All injection rates were ~25 nl/min.

Behavior task

Mice were head restrained in a previously described behavior apparatus (Gaffield et al., 2016; Gaffield and Christie, 2017) by attaching the surgically implanted post to a solid rod that provided stiff resistance to movement while the animal sat in a metal tube (diameter 25.4 mm). Water was delivered through a gavage needle with the end positioned about 2.5 mm from the mouth. All behavior timing was controlled using bControl (Brody Lab; Princeton). Licks were detected using a low-current circuit whereby contact of the tongue with the port produced a small electrical signal that was recorded digitally and could be used to register lick timing against time-series images acquired with 2pLSM. Lick timing was used to calculate lick probabilities within a given time bin. We determined lick rate as the inverse of the average inter-lick interval with the exception of the adjusted lick rate (a measure accounting for the intervals before and after each lick), which was calculated as published previously (Gaffield and Christie, 2017). Videography was used to ensure consistent placement of the lick port across all sessions for each mouse.

To provide motivation for participation in the behavior task, mice were maintained on a water-restricted diet (1 ml of water/day) and monitored daily for health. In the first few sessions, mice were familiarized with the apparatus and head restraint, and were trained to lick from the water delivery port by providing free access to water. After mastery, demonstrated by consistent licking, and judged by the uptake of water under high-speed videography, a tone cue (6 kHz) was added signaling water availability. After many trials (100–200 trials/day; 3–5 days), mice learned to lick at the onset of the tone and refrain from licking after fully consuming the dispensed water droplet. Mice typically completed >150 trails until sate. Tone response time was the interval between the end of the tone cue and the first subsequent lick. Licking movement epochs were defined based on whether a lick had been detected within 0.25 s (roughly two lick cycles). Epochs of non-licking were classified based on the absence of licking for >0.75 s.

For experiments in mice expressing the engineered receptor hM4d in MLIs, animals received an intraperitoneal injection of CNO prior to start of the task (45 min; 5 mg/kg; Tocris Bioscience, Bristol, UK; stock solutions were made by dissolving CNO in DMSO to 50 mM). Control measurements in the absence of CNO were made from the same mice on alternate days (>48 hr between sessions).

In vivo Ca2+ imaging

We used 2pLSM to image in vivo Ca2+ activity in neurons of the lateral cerebellum. The microscope was purpose built and included resonant scan mirrors (CRS 8K, Cambridge Technologies, Bedford, MA), a low power objective (16X, 0.8 NA water, Olympus, Center Valley, PA), and high-sensitivity PMTs (H10770PA-40, Hamamatsu, Bridgewater, NJ) for both red and green channels producing high-resolution (512 pixels x 512 pixels), high frame rate images (30 frames/s). This rate is faster than the response time of the Ca2+ indicators (Chen et al., 2013). Imaging was controlled using ScanImage 2015 software (Vidrio Technologies, Ashburn, VA). In the emission pathway, a 700 nm shortpass filter limited stray excitation light from reaching the PMT detectors. A 570 nm dichroic split the emission light into red and green channels. The green channel also included a 525/50 filter. GCaMP6f was excited at 900 nm (Chameleon Vision S, Coherent, Santa Clara, CA) with <50 mW of power at the objective except for climbing fibers where ~ 100 mW was required. RCaMP2 was excited at 1070 nm (Fidelity 2, Coherent) with <60 mW of power. The same region of cerebellum was imaged across sessions and generally included most of the same PC dendrites although not all dendrites could be automatically re-identified (Gaffield et al., 2016).

Acute brain slice recording

Parasagittal slices of the lateral cerebellum containing lobule Crus II were prepared from mature Kit::Cre mice. Animals were anesthetized with ketamine and xylazine by intraperitoneal injection (100 mg/kg and 10 mg/kg, respectively) and, after opening the chest, were perfused with ice-cold (~4°C) saline through the heart. The cerebellum was then removed by rapid dissection and mounted on an agar block and sectioned in thin slices (200 μm) using a vibraslicer (VT1200S, Leica Biosystems, Buffalo Grove, IL). Sectioning was performed in an ice-cold solution containing (in mM) 87 NaCl, 25 NaHO3, 2.5 KCl, 1.25 NaH2PO4, 7 MgCl2, 0.5 CaCl2, 10 glucose, and 75 sucrose that was continuously bubbled with carbogen gas (95% O2/5% CO2). Once cut, slices were immediately transferred to a holding chamber containing (in mM) 128 NaCl, 26.2 NaHO3, 2.5 KCl, 1 NaH2PO4, 1.5 CaCl2, 1.5 MgCl2 and 11 glucose, maintained at 34°C for 30 min and then at room temperature (~23°C) thereafter. For experiments, acute slices were placed in a recording chamber under a microscope and continuously perfused with an oxygenated saline solution of identical composition to that used for holding after slicing. The concentration of Ca2+ was kept low (1.5 mM) to reflect a more physiological-like condition (Jones and Keep, 1988) and the solution maintained at a near-body temperature (34–36°C) using an inline heater (TC-344; Warner Instruments). Except where noted, all experiments were performed in the absence of drugs so that both excitatory and inhibitory synaptic transmission were unaffected.

Neurons in Crus II were visually targeted for recording using IR contrast imaging with an upright video microscope (BX51WI; Olympus) and a QIClick CCD Camera (Q-Imaging, Surrey, BC, Canada). PCs and MLIs were easily distinguished based on their location and morphology. Recording pipettes were pulled from thin-walled borosilicate glass (PG 52–165;World Precision Instruments, Sarasota, FL) and filled with a solution containing (in mM) 124 potassium gluconate, 2 KCl, 9 HEPES, 4 MgCl2, 4 NaATP, 3 L-Ascorbic Acid, and 0.5 NaGTP (pH = 7.25). For PC recordings, the Ca2+ indicator dye Fluo-5F (200 μM; Life Technologies, Carlsbad, CA) as well as the volume indicator Alexa 594 (60 μM; Life Technologies) were also included in the pipette solution. Cell-attached recordings from MLIs were achieved by forming a loose seal with the patch electrode (~400 MΩ). This prevented dialyzing the cell and changing the intracellular concentration of Cl- and, hence, the reversal potential for currents generated by GtACR2.

We used a Multiclamp 700B amplifier (Molecular Devices, Sunnyvale, CA) for electrophysiological recordings. Signals were filtered online at 10 kHz and digitized at 20 kHz with a Digidata 1440 A-D converter (Molecular Devices). For whole-cell recordings, the membrane potential of PCs was maintained −75 mV using constant current injections whereas MLIs were maintained at −70 mV. Current offset was not used during cell-attached recordings of MLIs, allowing these cells to fire spontaneously under their own control, at an unperturbed resting membrane potential. Pipette capacitance was neutralized online and series resistance adjusted using the bridge balance circuitry of the amplifier. Liquid junctional potentials, calculated to be 10 mV, were corrected offline. Climbing fibers were stimulated using bi-polar glass electrodes placed near the axon hillock of the targeted PC. Brief electrical pulses (20 μs; 0.1–1.0 V) were delivered using a stimulation isolation unit (Model DS2A; Digitimer, Ft. Lauderdale, FL). Parallel fibers were also stimulated electrically using an electrode placed in the molecular layer adjacent to the dendritic recording site. The intensity of the stimulus was adjusted to produce PSPs, recorded at the PC soma, of similar amplitude across recordings (0.1–1.0 mV). For conjunctive stimulation experiments, parallel fibers were stimulated in bursts at 100 Hz. Climbing fiber were stimulated 50 ms after the end of parallel fiber tetanus, or at a varying interval where noted. Trials occurred at a relatively low frequency (0.125 Hz). In interleaved trials, either parallel fibers or climbing fibers were stimulated in isolation. This also occurred for trials that included optogenetic actuation of MLIs. Light pulses for optogenetic actuation or inactivation of MLIs activity started 10 ms prior to the beginning of the parallel fiber tetanus and lasted for the duration of the electrical stimulus.

For 2pLSM imaging in slices, we used a commercial scan head (Ultima; Bruker, Billerica, MA) fitted on top of an upright microscope (BX51-WI, Olympus; Tokyo, Japan). The scan head directed laser light (λ = 810 nm) from a mode-locking Ti:sapphire laser (Chameleon Ultra II; Coherent) through a scan lens and pair of galvanometer mirrors (Cambridge Technologies) onto the back aperture of a high-power objective (60X; 1.0 NA). To image Ca2+ activity in PCs, indicators dyes were allowed to dialyze for >30 min before starting recordings. Inclusion of the red volume dye allowed for identification of dendrites and spines for subsequent Ca2+ activity measurements. However, fluorescence in the red channel was not collected during neural activity measurements because of the interference of light used for the optogenetic stimuli. Ca2+ transients were recorded in PC spines using line scans (500 Hz). The PMT used to collect green Ca2+-indicator fluorescence was shuttered during blue-light optogenetic stimuli to prevent damage. Ca2+ activity was therefore not measured during this period. To facilitate comparison between conditions, a comparable region of the control response was blanked.

For our slice experiments, AAV1-EF1α-Flex(loxP)rev-GtACR2.eYFP (prepared by ViGene) and AAV5-EF1α-Flex(loxP)rev-bReachES-TS-YFP (University of North Carolina) were injected into the lateral cerebellum of Kit::Cre mice using a surgical procedure identical to that described above except that the size of the craniotomy was reduced (~0.5 mm diameter). Acute slices were prepared from these mice 7 to 14 days after surgery. GtACR2 and bReaChES were activated using blue and amber light, respectively, delivered from separate LEDs (M470L3 and M590L3; Thorlabs, Newton, NJ). The emission of the LEDS was combined with a dichroic (T570plxr; Chroma, Bellows Falls, VT) and directed, unfiltered (λ = 461 ± 20 nm and λ = 596 ± 16 nm), into the back epi-port of the microscope. This light was combined into the 2P excitation pathway using a second dichroic (700dcxru; Chroma). LEDs were modulated by separate current controllers (LEDD1B; Thorlabs) using digital commands out of the A-D converter and under computer control from the electrophysiology software (Clampex v10; Molecular Devices).

Post-hoc histology and confocal imaging

Transgene expression was confirmed by visual inspection of tissue from paraformaldehyde-perfused mice. Following the completion of Ca2+ imaging sessions, mice were anesthetized by intraperitoneal injection of ketamine/xylazine (100 mg/kg and 10 mg/kg, respectively) and the chest cavity opened, exposing the heart. The heart was accessed by a needle and the animal was perfused at 2 ml/min with a 0.1 M phosphate-buffered (PB) solution followed by paraformaldehyde (4% by volume in PB) until the perfusate exiting an opening from the pulmonary artery ran clear. The cerebellum was removed by dissection and sliced into 80 µm sections in cold PB. When necessary, HA immunostaining was used to confirm hM4d expression. Samples were first incubated with anti-HA antibody (#ab9110, Abcam, Cambridge, UK), followed by an Alexa 633 secondary antibody (Thermo Fisher Scientific, Waltham, MA). DAPI (D1306, Thermo Fisher Scientific) counterstaining was used to identify cell locations in some cases. Images were collected on a confocal microscope (LSM 780 Axio Imager 2; Zeiss, Oberkochen, Germany) using 488 nm excitation and 493–598 nm emission for GCaMP6f, 633 nm excitation and 638–747 nm emission for Alexa 633, 405 nm excitation and 410–507 nm emission for DAPI, and 514 nm excitation and 519–620 nm emission for YFP.

Image analysis

Image analysis of in vivo data was performed blind to the experimental condition. Time-series images of Ca2+ activity in neurons expressing genetically encoded Ca2+ indicators were aligned using a least-squares algorithm. For dual-color imaging, the translation coordinates from PC images were also used to co-register the corresponding MLI images. Individual PC dendrites or individual climbing fibers were segmented using an independent component analysis algorithm (Hyvärinen, 1999; Gaffield et al., 2016). Individual MLIs or parallel fibers were identified using hand-drawn ROIs from averaged images. Ca2+ events in both PCs and climbing fibers were identified using an inference algorithm (Vogelstein et al., 2010); events were identified whether or not they occurred during the decay of other events. However, to avoid potential uncertainties associated with GCaMP6f non-linearity (Chen et al., 2013), we performed an initial analysis on well-isolated (non-overlapping) events. For inclusion in this analysis, events must have occurred at least 500 ms from proceeding or following events. This time window allowed for the full decay of climbing fiber-evoked PC Ca2+ responses (τ is approximately 150 ms; Gaffield et al., 2016). The relative prevalence of isolated events was calculated by counting the number of isolated events, along with the number of times two events occurred within 500 ms, but were isolated by 500 ms from any other events, and the number of times three events occurred within 1000 ms, but were isolated by 500 ms from any other events. A subsequent analysis was performed on overlapping events comprising of two distinct events. In this case, consecutive events were selected that occurred within 150–200 ms of each other; all other events not meeting this criteria were rejected. In a final analysis, we also simply measured the peaks of all algorithmically identified Ca2+ responses.

For in vivo measurements of Ca2+ events in PCs and climbing fibers, ΔF/F was calculated using a baseline fluorescence period immediately prior to an identified event (~200 ms). For overlapping responses, this baseline period was during the decay of preceding events. For trial-averaged PC Ca2+ activity measurements, we calculated ΔF/F for all PC dendrite ROIs using the smallest GCaMP6f-fluorescence values obtained during recordings as the baseline. The average of responses in control was subtracted from that measured in CNO for each mouse before generating an overall average. In bouton measurements from parallel fibers, ΔF/F values were corrected by subtracting the neuropil signal from an area immediately adjacent to each fiber. The MLI-dependent movement (MDM) ratio was defined as the following equation:

(ΔF/FCNOΔF/FCtrl)movement/(ΔF/FCNOΔF/FCtrl)no movement

ROC analysis involved generating true positive and false positive rates for each threshold in the distribution of fluorescence values from all dendritic pixels reporting Ca2+ activity. This was used to generate a ROC curve. The percent of pixels showing a CNO effect was estimated from the area under the ROC curve (Najafi et al., 2014b). Analysis of inter-branch climbing fiber Ca2+ activity in PCs was performed by selecting two equally sized segments (average area = 190 ± 4 µm2; centers separated by 98 ± 4 µm) from a single dendrite then comparing the amplitudes for each identified event occurring simultaneously at each location. Similar to that of previous work (Kitamura and Häusser, 2011), dendritic variability was defined as:

2|A1A2|A1+A2

Where A1 and A2 are the peak fluorescence amplitudes for branch 1 and branch 2, respectively.

For ex vivo Ca2+ imaging experiments, fluorescence changes in PC dendrites were quantified as ΔF/F (average ~10 trials per condition). The peak climbing fiber-evoked Ca2+ transient was determined from an exponential fit of the fluorescence decay immediately following the electrical stimulus.

All image analysis was performed with Matlab (Mathworks, Natick, MA) or ImageJ (NIH). AxoGraph (Axograph) was used to analyze electrophysiological data. Additional calculations and plotting was performed with Excel (Microsoft, Redmond, WA) and Prism (GraphPad, La Jolla, CA). In figures, error bars indicate SEM. For data contained within the figure set where individual measurements are not already shown please see Figure 1—source data 1, Figure 1—figure supplement 1—source data 1, Figure 1—figure supplement 2—source data 1, Figure 2—source data 1, Figure 2—figure supplement 1—source data 1, Figure 2—figure supplement 2—source data 1, Figure 2—figure supplement 3—source data 1, Figure 4—source data 1, and Figure 5—figure supplement 2—source data 1.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
    Cerebellar Cortex: Cytology and Organization
    1. SL Palay
    2. V Chan-Palay
    (1974)
    Springer.
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74

Decision letter

  1. Jennifer L Raymond
    Reviewing Editor; Stanford School of Medicine, United States
  2. Eve Marder
    Senior Editor; Brandeis University, United States
  3. Martijn Schonewille
    Reviewer; Erasmus MC, Netherlands
  4. Jennifer L Raymond
    Reviewer; Stanford School of Medicine, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Inhibition gates instructional Ca2+ signaling in Purkinje cell dendrites during movement" for consideration by eLife. Your article has been reviewed by three peer reviewers, including Jennifer L Raymond as the Reviewing Editor and Reviewer #3, and the evaluation has been overseen by Eve Marder as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Martijn Schonewille (Reviewer #1).

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

Summary:

This manuscript by Gaffield and colleagues described an elegant series of experiments investigating the effects of excitatory and inhibitory synaptic inputs to the Purkinje cells climbing fiber-evoked calcium signals. Historically, the climbing fibers were thought to elicit a reliable, all-or-none (binary) calcium response in Purkinje cells, which triggered associative LTD in coactive parallel fibers. However, a number of recent in vitro and in vivo studies have demonstrated that the climbing fiber-elicited calcium response is graded, and that variations in the amplitude of the calcium response can modify the probability that associative plasticity is induced. The current manuscript extends this work by identifying inhibition from the molecular layer interneurons (MLIs) as a signal that can regulate the calcium response elicited by the climbing fibers.

The authors employ a powerful combination of complementary in vivo and in vitro calcium imaging, optogenetic and pharmacogenetic manipulation, and slice physiology. They show that the amplitude of calcium transients during a well-learned behavior (licking), which increases both MLI activity and the rate of these transients (Figure 1), is similar to calcium transients in the absence of movement. Chemogenetic inactivation of MLIs, however, unmasks a widespread, movement-specific increase in transient amplitude (Figure 2, Figure 3). This increase appears to be postsynaptic, as direct imaging of climbing fibers fails to reveal an influence of the ligand on presynaptic calcium preceding or during movement (Figure 4). Optogenetic suppression of MLIs in acute slices enhances parallel fiber (PF) postsynaptic potentials at the Purkinje cell soma and induces a supralinear dendritic calcium response to paired parallel and climbing fiber stimulation. Finally, the authors demonstrate that long trains of PF stimulation can induce a similar supralinear calcium response, but that this interaction is erased by optogenetic activation of MLIs.

The work is of high potential impact. The results extend our understanding of events that can gate or modulate the climbing fiber's impact on their Purkinje cell targets, which is thought to provide instructive signals that guide the induction of cerebellum-dependent learning. The experiments are well-designed, the claims are clearly explained, and the data support the claims. However, the reviewers agreed that some issues should be addressed by the authors.

Essential revisions:

1) Because long trains of PF stimulation can induce a supralinear PF+CF effect on calcium, the authors should rule out the possibility that pharmo- and opto-genetic perturbations drives an increase in PF activity (e.g. by increasing the frequency of some possibly unobserved movement, or via inhibition of Golgi cells, which also have some expression in the c-kit line). The most direct way to achieve this would be to perform calcium imaging in granule cells and check that activity is similar between treatment and sham conditions. It may also be helpful to provide a more thorough analysis of the effects of the MLI manipulations on behavior. Is enhanced calcium still observed with CNO if the analysis is limited to CNO and control trials with the behavior as closely matched as possible (subsample Figure 2C data to control for licking rate)?

2) More thorough analysis of the data, and more complete presentation of the results are needed throughout the manuscript.

a) In the analysis of the in vivo calcium events, the authors focus on the unitary calcium responses, and exclude the multipeaked responses from most pf the analyses. The description of how the unitary events were identified is scanty (subsection “Image Analysis”). The authors should provide more information about how they did the analysis and provide an analysis of the robustness of the results to their specific choices for the analysis. What does the distribution of all (single and multipeaked) calcium event amplitudes look like? What is the distribution of single, double, triple peaked events? Does the latter differ during cued licking behavior vs. spontaneous? The example trace in Figure 1C suggests there is surprisingly large variation in the size of unitary events (a ~3-fold difference in the peak with the green check mark vs. the first peak in the multiple event)-this makes me wonder how many of the unitary events selected for analysis are actually composed of two smaller calcium events occurring close together in time. What is the temporal resolution for discriminating one vs. two closely spaced events? A more careful consideration of how the choice of analysis could affect the results and the interpretation of the results is needed.

b) The authors show in Figure 3C that ~63% of the dendritic area shows enhanced Ca2+ signaling after MLI disinhibition. If I understand correctly the figure also indicates that the other ~37% has a substantial decrease of the Ca2+. This should be addressed, also in relation to the chance of random changes in Ca2+ signaling (is the 67% statistically significant from random 50%). Similarly, the remark about the 'arbor-wide' influence should be re-worded to more careful terms.

c) For Figure 5D and Figure 6, it would be very helpful if the evoked PF and CF transients were also shown individually, before the summed transient is shown. Also, please state clearly if the summed response for the suppressed inhibition is based on the PF with or without inhibition.

d) Figure 6G and associated text suggest that the MLIs exert a graded effect on the climbing fiber-elicited calcium response. Is there truly a graded effect on amplitude, or might it be an effect on the probability of an additive vs. supralinear event of a more uniform amplitude? Additional analysis could distinguish these two equally interesting possibilities.

e) The point of Figure 1 is that the amplitude of the calcium events is the same during movement or no movement. However, in Figure 2F, this does not seem to be the case for the hM4d animals, even in the absence of CNO. If the calcium sensing is nonlinear, this could potentially influence what is measured with CNO. This possibility should be considered.

f) There are some places where individual examples are shown, but no group data, for example, Figure 4E.

g) There are some places where results are presented without statistics. For example, is there is a significant difference between movement vs. no movement and between CNO vs. control in Figure 3FG? Also, the error bars on the CNO-control PC data for simultaneous PC/MLI imaging (Figure 2G, red trace) are very large. Please provide the same statistical analyses for these data as were used for the data in Figure 2F to justify the claim that the difference is positive. Also, please provide the same analysis for the climbing fiber imaging data in Figure 4G, where the difference appears to be negative, but is asserted to not differ from zero.

h) To support the claim that peak PC ΔF/F does not co-vary with MLI activity (Figure 1G), please plot compute the correlation between the peak amplitude and MLI activity for all dendrites and complex spike event, and display a scatterplot of these values.

i) A video of mouse behavior and two-color imaging (Figure 1) would be useful.

j) Figure 5D shows that MLI inactivation induces a supralinear interaction between PF and CF stimulation on dendritic calcium. Is the same true of the somatic membrane potential? Please overlay plots of the PF+CF and PF+CF (sum) for control and light conditions (similarly to Figure 5D) in Figure 5F, and make a similar plot for the ephys data in the long pulse train experiments (Figure 6E).

k) Was there any correlation between calcium transient amplitude and behavior? For instance, were transients larger on trials with longer or mistimed licking bouts? Similarly, did the correlation between PC calcium and MLI calcium change with any aspect of behavior? Although the authors have addressed some of these questions in previous work, it would be helpful to have this information for the current manipulations and data.

l) In general, the figure legends and text describing the figures should be edited to be more precise and make it easier for the reader to understand what is in the figure. For example, in Figure 2G and associated text, I am guessing that the "totality of all calcium responses" includes the multipeaked as well as the unitary responses, but after considerable effort, I am still not sure. Another example, it is not clear whether the averaged calcium activity in Figure 1F is from one mouse or all mice. Please carefully review all figure legends to make sure that the reader is provided with sufficient information about what is in each figure panel.

m) Please provide time units for the event probability plots to allow comparison across figures, and with previously published climbing fiber firing rates. I would expect the probabilities in Figure 1E and Figure 4E to be similar, but they differ by twofold. How does this compare with the typical climbing fiber rate of 1 Hz?

3) The current results would have more impact if more effectively framed in the context of what is already known about variations in climbing fiber-triggered responses in the Purkinje cells, how the current work extends what is known, the conclusions that can be drawn, and any caveats.

a) Previous work has associated the calcium transients with plasticity, however, the reviewers thought there was too much emphasis on plasticity (using "instructional Ca2+ signaling" in the Title and last sentence of the Abstract and "circuit modifications" in the first sentence of the Abstract), given that plasticity is not tested in the in vitro or in vivo experiments.

b) Najafi et al., have reported that climbing fiber-associated calcium responses are enhanced when they occur during a learning task as compared with the spontaneous calcium events, which are presumably due to spontaneous climbing fiber spiking. In contrast, the current results indicate that the calcium responses during the well-learned lick task are the same as the spontaneous calcium events. The authors cite Najafi, but don't directly compare the two results so that the reader can effectively appreciate the new finding that the calcium responses are not always enhanced in the behavioral context. This would better frame the current Results section than the current statement in the Introduction "whether CF Ca2+ signals in PC dendrites are augmented by preceding PF activity in vivo is unclear" and would explain why the results in Figure 1D are described as "Surprising".

c) More discussion of the cued licking task, and the role of the Ca2+ signal in Purkinje cells during the movement in the lick task is needed. From the example mouse data in Figure 1B, it is difficult to tell how much of the behavioral response and neural activity is learned (a conditioned response) versus an unconditioned response to the water — it looks like the majority of the neural and behavioral response could be unconditioned. Which components of the lick task are cerebellum dependent-acquisition of learning? expression of the condition response? performance of unconditioned licks? Discussion of these issues would help to clarify the potential role of the calicum transients in the behavior.

d) In Figure 5 and Figure 6, might the lack of supralinear calcium responses result from the use of a suboptimal parallel fiber-climbing fiber pairing interval?

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

Thank you for submitting your article "Inhibition gates supralinear Ca2+ signaling in Purkinje cell dendrites during practiced movements" for consideration by eLife. Your article has been reviewed by three peer reviewers, including Jennifer L Raymond as the Reviewing Editor and Reviewer #3, and the evaluation has been overseen by Eve Marder as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Martijn Schonewille (Reviewer #1).

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

Summary:

This manuscript provides evidence that the molecular layer interneurons in the cerebellar cortex modulate the climbing fiber-triggered calcium responses in Purkinje cells. Historically, the climbing fibers were thought to elicit a reliable, all-or-none (binary) calcium response in Purkinje cells, which triggered associative LTD in coactive parallel fibers. However, a number of recent in vitro and in vivo studies have demonstrated that the climbing fiber-elicited calcium response is graded, and that variations in the amplitude of the calcium response can modify the probability that associative plasticity is induced. The current manuscript extends this work by identifying inhibition from the molecular layer interneurons (MLIs) as a signal that can regulate the calcium response elicited by the climbing fibers. The authors employ a powerful combination of complementary in vivo and in vitro calcium imaging, optogenetic and pharmacogenetic manipulation, and slice physiology. In response to the previous review, the authors have conducted a number of additional analyses, which increase the rigor and strengthen the conclusions. In addition, they included new data from experiments imaging the parallel fiber axons during their cued licking task. Overall, this study represents a significant advance in understanding how the cerebellum implements learning, in particular, the events governing the induction of associative LTD at the parallel fiber-Purkinje cell synapse. The reviewers did not think additional experiments were necessary for this to make a valuable contribution, however, they felt that some additional revisions of the text were needed to better acknowledge and discuss some of the limitations of the current study that were raised in the previous review, and which were not fully addressed by the additional analyses and experiments provided in the revised manuscript.

Essential revisions:

1) The manuscript provides convergent evidence that inhibition from the molecular layer interneurons (MLIs) can gate supralinear calcium responses to combined parallel fiber and climbing fiber activity. This is likely correct. However, it is also difficult to rule out the possibility that in vivo, the chemogenetic suppression of MLIs could have resulted in changes (increases) in granule cell/parallel activity, which could contribute to the enhanced calcium responses. The authors need to be more forthcoming and explicit in the manuscript about acknowledging this possibility and the limited extent to which their evidence addresses it.

Their argument that argument that parallel fiber activity is not altered by suppression manipulation of MLI activity hinges on:

a) new data showing that in the absence of MLI manipulation, calcium imaging-based measurement of parallel fiber activity closely tracks MLI activity (Figure Figure 1—figure supplement 1) and an assertion in their point-by-point response that "unilateral chemogenetic disinhibition of MLIs in Crus II did not affect the population response of MLIs (data not shown). There are some concerns with this argument. First, just because the activity of two populations of neurons is similar under one measurement condition, it does not mean that it will be similar under all conditions, such as experimental manipulation of inhibition. Second, the whole point of the chemogenetic manipulation of MLIs is to suppress their activity, so why does it not affect the population response? Are we missing something? Third, the Kit promoter has some expression in Golgi cells, which directly inhibit the granule cells-this needs to be acknowledged.

b) Lack of change in the behavior. The additional data provided in Figure 2—figure supplement 3 is helpful, but does not fully address the possibility of other, unmeasured behavior differences (lateral deviation of the tongue, other orofacial movements), which is why the reviewers had asked for sample videos with and without chemogenetic suppression. Also, the lack of behavioral difference does not rule out a difference in the granule cells that does not affect the behavior, but could contribute to calcium transients in the Purkinje cells.

In the absence of more direct recordings from the granule cells or parallel fibers comparing responses in the presence or absence of the disinhibition manipulation, we would be satisfied with a more thorough acknowledgement and discussion of the above caveats in the manuscript.

2) Figure 1—figure supplement 1B and 1E show that PF activity is well correlated with licking behavior after the cue. However, it is not very clear in Figure 1—figure supplement 1D. In particular, fluorescence signals begin to increase mostly before licking starts in the B3 bouton. It is therefore important to show PF and MLI activity not only when the cue induces licking but also fails to induce licking as described above. In addition, please indicate in Figure 1—figure supplement 1D when the cue was presented. Otherwise, readers cannot tell which licking bouts are learned behavior.

3) In the first round of review, one of the reviewers asked more discussion of the cued licking task, for example, which components of the lick task are cerebellum dependent-acquisition of learning. The authors address the comment, but their rebuttal letter has addressed the comment better than the manuscript itself. It is important to let readers know that it is currently unclear (1) the role of climbing fibers in the cued licking task, and (2) which aspects of this learning task are regulated by the cerebellum. These limitations do not diminish the value of this study.

https://doi.org/10.7554/eLife.36246.029

Author response

Essential revisions:

1) Because long trains of PF stimulation can induce a supralinear PF+CF effect on calcium, the authors should rule out the possibility that pharmo- and opto-genetic perturbations drives an increase in PF activity (e.g. by increasing the frequency of some possibly unobserved movement, or via inhibition of Golgi cells, which also have some expression in the c-kit line). The most direct way to achieve this would be to perform calcium imaging in granule cells and check that activity is similar between treatment and sham conditions. It may also be helpful to provide a more thorough analysis of the effects of the MLI manipulations on behavior. Is enhanced calcium still observed with CNO if the analysis is limited to CNO and control trials with the behavior as closely matched as possible (subsample Figure 2C data to control for licking rate)?

We have collected a new set of data using GCaMP6f to measure the activity of parallel fiber axons during our cued-licking task. Parallel fibers were robustly activated during the consumption of water and were relatively inactive in the absence of licking (Figure 1—figure supplement 1). Parallel fiber activity also closely tracked licking kinematics. Interestingly, the movement-induced activation of MLIs was essentially an exact match of that of the average response of the parallel fibers in the same region. This indicates that MLI activity measurements can be used to infer the overall activity of the granule cell population. Unilateral chemogenetic disinhibition of MLIs in Crus II did not affect the population response of MLIs, measured using our Ca2+ imaging approach. Thus, by inference, disinhibition did not affect the overall level of granule cell activity. This rules out the possibility that enhanced Ca2+ signaling in PCs was the result of a CNO-induced change in the presynaptic activity level of granule cells.

Addressing the specific question of parallel fiber signaling onto Purkinje cell dendrites with and without MLI block is not technically feasible in awake behaving animals at this time. We are therefore left with relying on the reduced preparation experiments where we show that shifting the balance between parallel fiber excitation and MLI inhibition of PCs can likewise affect PC Ca2+ activity. These points are now emphasized in the revision.

We apologize for not being forward in our reporting of potential motor changes with chemogenetic suppression of MLI activity. We actually went out of our way to identify conditions that were not conducive to affect behavior as this could complicate the interpretation of our results. We now present our analysis showing the absence of changes in average licking performance following unilateral suppression of MLI activity in the small area of left Crus II under our imaging window (Figure 2—figure supplement 3). We do not think this result contradicts our previous work where we reported a small, but significant reduction in lick rate with MLI activity suppression by hM4d activation (Gaffield et al., 2017). This is because in this prior work, we inhibited MLI activity bilaterally in large swaths of both left and right Crus II, a much more robust perturbation of the MLI population controlling orofacial function. At the reviewers’ request, we examined PC Ca2+ activity in a subset of mice showing the smallest deviations in lick rate with chemogenetic disinhibition, compared to control. MLI activity suppression continued to enhance climbing fiber-evoked Ca2+ events in PCs from these mice (Figure 2—figure supplement 3). Moreover, limited behavioral differences indicate that granule cell activity (which tracks behavior) is also unlikely to change with chemogenetic inhibition of MLIs, which further supports our conclusion that the observed alteration of PC Ca2+ signaling by chemogenetics is not due to activity changes in granule cells.

2) More thorough analysis of the data, and more complete presentation of the results are needed throughout the manuscript.

We have made extensive new additions to the manuscript including many new supplemental figures to address each of the reviewers’ concerns.

a) In the analysis of the in vivo calcium events, the authors focus on the unitary calcium responses, and exclude the multipeaked responses from most pf the analyses. The description of how the unitary events were identified is scanty (subsection “Image Analysis”). The authors should provide more information about how they did the analysis and provide an analysis of the robustness of the results to their specific choices for the analysis. What does the distribution of all (single and multipeaked) calcium event amplitudes look like? What is the distribution of single, double, triple peaked events? Does the latter differ during cued licking behavior vs. spontaneous? The example trace in Figure 1C suggests there is surprisingly large variation in the size of unitary events (a ~3-fold difference in the peak with the green check mark vs. the first peak in the multiple event)-this makes me wonder how many of the unitary events selected for analysis are actually composed of two smaller calcium events occurring close together in time. What is the temporal resolution for discriminating one vs. two closely spaced events? A more careful consideration of how the choice of analysis could affect the results and the interpretation of the results is needed.

We apologize for the inadequacy of our description of unitary event selection for analysis of PC dendritic Ca2+ signals. Specifically, we used a previously published inference algorithm (Vogelstein et al., 2010) to identify all climbing fiber-evoked Ca2+ events in PCs, whether or not they occurred during the decay of preceding events. The reviewers refer to these overlapping events as “multi-peaked”. We are hesitant to use this terminology as it seems to imply some sort of physiological relevance rather than simply reflecting that climbing fiber-evoked events can occur in succession and the decay time of GCaMP6f is relatively slow (τ is approx. 150 ms for PCs [Gaffield et al., 2016]). We rejected overlapping dendritic Ca2+ events from our initial analysis because of possible uncertainties of GCaMP6f nonlinearity. Therefore, our criteria for inclusion was that an event occurred at least 500 ms from proceeding or following events, a window that allows for GCaMP6f fluorescence to decay back to baseline levels (as shown in Figure 1D). We have now included a better description of our selection criteria and rationale in the Materials and methods section.

In response to the reviewers urging, we now report the distribution of single vs. overlapping events. Isolated, single events comprise the majority of the response population; this makes sense because the average Ca2+ event rate in PCs is low, about 1.5 Hz. We found that these distributions don’t change during motor behavior (Figure 1—figure supplement 2).

A major finding of our report is that with chemogenetic suppression of MLI activity, Ca2+ event amplitudes increased during movement. This is a robust result. New analysis shows that this holds true not only for discrete single events, but also for all other events including overlapping responses (Figure 2—figure supplement 1). Regarding the discrimination of overlapping Ca2+ events, our imaging rate is 30 Hz (~33 ms); this defines a theoretical temporal resolution by which we could distinguish two closely spaced dendritic events. That said, we don’t believe that unresolved, closely-spaced responses appearing as single events accounted for the difference in response amplitude with chemogenetic disinhibition. If so, this would have been reflected in the presynaptic activity of climbing fibers. Because Ca2+ event amplitudes (and frequency) did not change in climbing fibers with molecular layer disinhibition, this argues strongly against this possibility. We now state this in the text.

Like the reviewers, we are fascinated by the variability in the amplitudes of PC dendritic Ca2+ events. Identifying the mechanistic underpinnings of this variability is a major goal of ours. That similar variability is apparent in the amplitudes of Ca2+ events in climbing fibers indicates the PC dendrites are simply integrating the level of presynaptic activity from these inputs. We hypothesize that MLI-mediated gating of dendritic nonlinearities allow for the faithful conversion of presynaptic climbing fiber activity into postsynaptic Ca2+ signals in PCs. However, this investigation is beyond the scope of the current report.

b) The authors show in Figure 3C that ~63% of the dendritic area shows enhanced Ca2+ signaling after MLI disinhibition. If I understand correctly the figure also indicates that the other ~37% has a substantial decrease of the Ca2+. This should be addressed, also in relation to the chance of random changes in Ca2+ signaling (is the 67% statistically significant from random 50%). Similarly, the remark about the 'arbor-wide' influence should be re-worded to more careful terms.

To address this concern, we have changed our analysis approach and now include use of receiver operating characteristic (ROC) curves to diagnose the probability of disinhibition-induced shifts in Ca2+ activity across all pixels of identified PC dendrites (i.e., area under the curve). This analysis indicates a widespread effect (Figure 3C). In addition, we have re-worded the Results section, taking caution about the use of the phrase ‘arbor-wide influence’.

c) For Figure 5D and Figure 6, it would be very helpful if the evoked PF and CF transients were also shown individually, before the summed transient is shown. Also, please state clearly if the summed response for the suppressed inhibition is based on the PF with or without inhibition.

We now show both the individual, parallel fiber- and climbing fiber-evoked Ca2+ transients in Figure 5D (the parallel fiber response is always quite small or negligible for all recordings). This should help readers understand our approach for quantifying supralinearity in dendritic PC Ca2+ signaling more easily. The same method applied for dendritic Ca2+ activity measurements in Figure 6. We now state, in the accompanying legend of Figure 5, that the summed response for the suppressed condition is based on a PF stimulus without feed-forward inhibition (MLI activity during the parallel fiber tetanus was eliminated by GtACR2).

d) Figure 6G and associated text suggest that the MLIs exert a graded effect on the climbing fiber-elicited calcium response. Is there truly a graded effect on amplitude, or might it be an effect on the probability of an additive vs. supralinear event of a more uniform amplitude? Additional analysis could distinguish these two equally interesting possibilities.

Single-trial analysis of climbing fiber-evoked Ca2+ transients could distinguish these possibilities. However, the feasibility of such analysis has not been established for PCs, at least to our knowledge. This is because single trial analysis is made difficult due of the low signal-to-noise level of climbing fiber-evoked responses. We could change our recording parameters in an attempt to help improve the quality of evoked Ca2+ signals (e.g., increasing the concentration of external Ca2+ or the number of conjunctive climbing fiber stimuli, etc.). This would be very laborious and time consuming. In the end, parsing which mechanism is responsible for the observed effect would not change the overall interpretation of our results as both possibilities are interesting and relevant. Therefore, we have removed the term “graded” from the manuscript in favor of simply stating that the effect of parallel fiber excitation on climbing fiber-evoked Ca2+ signaling in PCs was dependent on the level of MLI activity.

e) The point of Figure 1 is that the amplitude of the calcium events is the same during movement or no movement. However, in Figure 2F, this does not seem to be the case for the hM4d animals, even in the absence of CNO. If the calcium sensing is nonlinear, this could potentially influence what is measured with CNO. This possibility should be considered.

The reviewers point to a trend towards a larger Ca2+ response with movement, compared to events in the absence of movement, in hM4D-expressing mice. However this difference is not significant. We now state this in legend of Figure 2. The major concern with uncertainties regarding GCaMP6f nonlinearity is whether there is a faithful reporting of the relative change in amplitudes (including the absence of change due to saturation). This is compounded for responses that occur in the decay of others because Ca2+ unbinding isn’t yet complete. This is a major reason why we avoided including such events in our initial analysis (we now include such events in a second analysis at the reviewers’ request). Heeding this concern, we were also hesitant about making significant conclusions based on how much change we observed in the disinhibited condition. Thus, we don’t believe that enhancement of Ca2+ signaling by GCaMP6f can be explained by a nonlinearity, although the absence of effect may warrant caution. In response to the reviewers, we now state in the text that GCaMP6f is known to be nonlinear but, based on previous reports, is sensitive enough to report small changes in climbing fiber-evoked PC Ca2+ signaling to different behavioral stimuli. This point is quite moot later in the manuscript when we show that enhanced Ca2+ signaling can be uncovered with chemogenetic-induced, molecular layer disinhibition.

f) There are some places where individual examples are shown, but no group data, for example, Figure 4E.

We have changed Figure 4E to include group data; we can find no other instances where group data are not reported either in the text or in a separate Figure panel.

g) There are some places where results are presented without statistics. For example, is there is a significant difference between movement vs. no movement and between CNO vs. control in Figure 3FG? Also, the error bars on the CNO-control PC data for simultaneous PC/MLI imaging (Figure 2G, red trace) are very large. Please provide the same statistical analyses for these data as were used for the data in Figure 2F to justify the claim that the difference is positive. Also, please provide the same analysis for the climbing fiber imaging data in Figure 4G, where the difference appears to be negative, but is asserted to not differ from zero.

The statistical comparisons for the data shown in Figure 3F and 3G are presented in Figure 3H (in this comparison, mean variability of PC Ca2+ activity was significantly different following chemogenetic suppression of MLIs). We apologize for not pointing out this figure panel in our original submission. This has now been corrected.

For activity measurements in climbing fibers, we compared the peak amplitudes of averaged Ca2+ events collected in control and during disinhibition of the molecular layer (Figure 4G). The difference trace, like those shown in Figure 2C-2E, are for the reader’s convenience. The statistical test for this comparison is reported in the text; it was non-significant (P = 0.66; Student’s t-test).

Regarding our results examining trial-averaged PC Ca2+ activity during cued licking, we show the subtracted difference between responses measured in control and with molecular layer disinhibition (Figure 2G, top panel, red trace). We chose to present the subtracted difference because it facilitated comparison to the activity plot of MLIs (i.e., the change in PC activity with disinhibition closely corresponded to the time period when MLIs were normally activated by the behavior). This was also for the reader’s convenience but, as requested, we performed a statistical analysis of this dataset (inset of Figure 2G). Based on the reviewers’ comments, we believe presenting the data as we did may have caused confusion. Therefore, we now show responses of trial-average PC activity in the two conditions used to generate the difference plot (control and with chemogenetic MLI activity suppression) as well as an accompanying statistical analysis of these data (Figure 2—figure supplement 2).

h) To support the claim that peak PC ΔF/F does not co-vary with MLI activity (Figure 1G), please plot compute the correlation between the peak amplitude and MLI activity for all dendrites and complex spike event, and display a scatterplot of these values.

As shown in our analysis of mean activity (Figure 1G), there is no relationship between the amplitude of climbing fiber-evoked dendritic Ca2+ events in PCs and the corresponding level of MLI activation. However, in response to the reviewers, we have provided scatterplots from three representative mice showing all PC dendritic events and the corresponding level of activity in surrounding MLIs (Figure 1—figure supplement 3).No relationship was found.

i) A video of mouse behavior and two-color imaging (Figure 1) would be useful.

We show a plot of Ca2+ activity measured simultaneously in PCs and MLIs using a dual-color, Ca2+ indicator imaging approach. The average responses are demarcated based on their correspondence to cell type and aligned to the onset of behavior (licking). We are not sure how the addition of a video would help clarify our approach or provide any additional insight into this result to readers.

j) Figure 5D shows that MLI inactivation induces a supralinear interaction between PF and CF stimulation on dendritic calcium. Is the same true of the somatic membrane potential? Please overlay plots of the PF+CF and PF+CF (sum) for control and light conditions (similarly to Figure 5D) in Figure 5F, and make a similar plot for the ephys data in the long pulse train experiments (Figure 6E).

Hyperpolarization of the membrane potential by somatic current injection diminishes climbing fiber-evoked responses in PC dendrites including supralinear Ca2+ signaling (Wang et al., 2000, Kitamura et al., 2011, and Otsu et al., 2014). This is likely mediated through the passive spread of voltage into the dendritic compartment, preventing parallel fiber-mediated inactivation of Kv4 currents (Otsu et al., 2014). It follows then that GABAAR-mediated hyperpolarization, elicited by MLI activity, is sufficient to prevent the supralinear interaction of parallel fibers and climbing fibers in PC dendrites, as observed in our experiments. At the reviewers’ request, traces of complex spikes with and without preceding parallel fiber stimulation are shown superimposed in Figure 5F for both control and the responses with optogenetic suppression of MLIs. For the data presented in Figure 6, these are included as a supplement additional figure (Figure 6—figure supplement 1).

k) Was there any correlation between calcium transient amplitude and behavior? For instance, were transients larger on trials with longer or mistimed licking bouts? Similarly, did the correlation between PC calcium and MLI calcium change with any aspect of behavior? Although the authors have addressed some of these questions in previous work, it would be helpful to have this information for the current manipulations and data.

We did not observe significant changes to average licking behavior with unilateral chemogenetic suppression of MLI activity in left Crus II. This is now reported in the manuscript (Figure 2—figure supplement 3). A more in depth examination of deviant licking, perhaps reflecting motor errors that lead to learning, was not pursued. Regarding motor errors and climbing fiber-evoked signals, it is certainly a future interest of ours to understand the behavioral conditions conducive to producing PC supralinear Ca2+ responses. To address this point carefully, will require more exacting control of the movement task as well as additional tools to measure and manipulate neural activity in the granule cell and molecular layers. Thus, this is beyond the scope of the current study. As the reviewers point out, we have previously reported that the rate of climbing fiber-evoked Ca2+ events in PC dendrites increases at the onset of licking while MLI activity increases and decreases with changes in licking rate during water consumption. This is now discussed in greater detail.

l) In general, the figure legends and text describing the figures should be edited to be more precise and make it easier for the reader to understand what is in the figure. For example, in Figure 2G and associated text, I am guessing that the "totality of all calcium responses" includes the multipeaked as well as the unitary responses, but after considerable effort, I am still not sure. Another example, it is not clear whether the averaged calcium activity in Figure 1F is from one mouse or all mice. Please carefully review all figure legends to make sure that the reader is provided with sufficient information about what is in each figure panel.

The “totality of all calcium responses” was meant to indicate that we simply averaged all Ca2+ activity (including overlapping events spaced closely in time) in all identified PCs across all trials. We have attempted to improve the clarity of our language regarding this dataset and have added a new figure to illustrate this result more directly (Figure 2— figure supplement 2). We now report numbers for Figure 1F in the legend.

m) Please provide time units for the event probability plots to allow comparison across figures, and with previously published climbing fiber firing rates. I would expect the probabilities in Figure 1E and Figure 4E to be similar, but they differ by twofold. How does this compare with the typical climbing fiber rate of 1 Hz?

We changed the plots to include time units (rate/frequency). We show that the average rate of Ca2+ events is the same for both climbing fibers and PCs. The increase in peak Ca2+ event rates at the initiation of licking is also not significant between climbing fibers and PCs (there is large amount of variability in the climbing fiber response); we don’t report this because our preference is to more carefully compare climbing fibers and PCs using simultaneous dual-color imaging of pre- and post-synaptic activity in the same mouse which we hope to publish in the near future. The rates of Ca2+ events detected in PCs in our experiments closely matches that of previously published values. This is now stated in the text with accompanying references. To our knowledge, we are unaware of anyone else who has measured activity rates directly from climbing fibers, so we are proud to say that this is a first!

3) The current results would have more impact if more effectively framed in the context of what is already known about variations in climbing fiber-triggered responses in the Purkinje cells, how the current work extends what is known, the conclusions that can be drawn, and any caveats.

We thank the reviewers for highlighting areas of our manuscript where we could include clarification to increase the impact of our work.

a) Previous work has associated the calcium transients with plasticity, however, the reviewers thought there was too much emphasis on plasticity (using "instructional Ca2+ signaling" in the Title and last sentence of the Abstract and "circuit modifications" in the first sentence of the Abstract), given that plasticity is not tested in the in vitro or in vivo experiments.

As requested, we have removed most of these phrases from the Title and Abstract. However, the first sentence of the Abstract simply describes what the cerebellum does and emphasizes the motivation for understanding dendritic Ca2+ signaling in PCs. Similarly in the Introduction, our use of terms like “circuit alterations” and “plasticity” is necessary to put our experiments in context. Therefore we kept these words in the first two paragraphs while removing them from the third as this was merely speculative.

b) Najafi et al. have reported that climbing fiber-associated calcium responses are enhanced when they occur during a learning task as compared with the spontaneous calcium events, which are presumably due to spontaneous climbing fiber spiking. In contrast, the current results indicate that the calcium responses during the well-learned lick task are the same as the spontaneous calcium events. The authors cite Najafi, but don't directly compare the two results so that the reader can effectively appreciate the new finding that the calcium responses are not always enhanced in the behavioral context. This would better frame the current Results section than the current statement in the Introduction "whether CF Ca2+ signals in PC dendrites are augmented by preceding PF activity in vivo is unclear" and would explain why the results in Figure 1D are described as "Surprising".

We thank the reviewers for pointing this out. We now emphasize the novelty of our findings regarding that lack of behavior-related augmentation of Ca2+ signaling in PCs. A detailed comparison of the work by Najafi et al., contrasting their findings to ours, is also provided.

c) More discussion of the cued licking task, and the role of the Ca2+ signal in Purkinje cells during the movement in the lick task is needed. From the example mouse data in Figure 1B, it is difficult to tell how much of the behavioral response and neural activity is learned (a conditioned response) versus an unconditioned response to the water — it looks like the majority of the neural and behavioral response could be unconditioned. Which components of the lick task are cerebellum dependent-acquisition of learning? expression of the condition response? performance of unconditioned licks? Discussion of these issues would help to clarify the potential role of the calicum transients in the behavior.

We have added to the Discussion section to address some of these points. In summary though, we have not yet identified which aspects of lick-related behavioral refinement are due to this region of the cerebellum nor do we understand the importance of the climbing fiber-evoked responses in PCs, elicited at licking onset, to task performance. It is enticing to speculate on the potential role of these PC Ca2+ signals in guiding learning. But we conjecture that learning is not apt to be occurring during this well-practiced motor behavior. That said, climbing fiber activity has been shown to be important for motor memory retention (Medina et al., 2002). It may be that the climbing fiber-evoked Ca2+ signals that we studied are import for maintenance of some earlier aspect of behavioral adaptation during task training (e.g., timing of lick responses to the cue). Certainly, there is an abundant literature indicating that complex spikes evoked by climbing fibers are important for online motor control. As discussed, MLI-mediated suppression of nonlinear Ca2+ signaling could allow climbing fibers to multitask. Supporting both online motor control through the influence of complex spike bursts at the PC soma separate from the acquisition of plasticity in the dendrite.

d) In Figure 5 and Figure 6, might the lack of supralinear calcium responses result from the use of a suboptimal parallel fiber-climbing fiber pairing interval?

We tried different pairing intervals. However, supralinear Ca2+ responses were not apparent with feed-forward inhibition intact. These new data are now included in the manuscript (Figure—figure supplement 2).

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

[…] Essential revisions:

1) The manuscript provides convergent evidence that inhibition from the molecular layer interneurons (MLIs) can gate supralinear calcium responses to combined parallel fiber and climbing fiber activity. This is likely correct. However, it is also difficult to rule out the possibility that in vivo, the chemogenetic suppression of MLIs could have resulted in changes (increases) in granule cell/parallel activity, which could contribute to the enhanced calcium responses. The authors need to be more forthcoming and explicit in the manuscript about acknowledging this possibility and the limited extent to which their evidence addresses it.

In the revised manuscript, we now attempt to make these points more explicit and state the limitations of our approach/conclusions.

Their argument that argument that parallel fiber activity is not altered by suppression manipulation of MLI activity hinges on:

a) new data showing that in the absence of MLI manipulation, calcium imaging-based measurement of parallel fiber activity closely tracks MLI activity (Figure 1—figure supplement 1) and an assertion in their point-by-point response that "unilateral chemogenetic disinhibition of MLIs in Crus II did not affect the population response of MLIs (data not shown). There are some concerns with this argument. First, just because the activity of two populations of neurons is similar under one measurement condition, it does not mean that it will be similar under all conditions, such as experimental manipulation of inhibition. Second, the whole point of the chemogenetic manipulation of MLIs is to suppress their activity, so why does it not affect the population response? Are we missing something? Third, the Kit promoter has some expression in Golgi cells, which directly inhibit the granule cells-this needs to be acknowledged.

We now state/discuss these caveats. The chemogenetic inhibitor hM4d is thought to work by a presynaptic action (now cited in the discussion, Stachniak et al., 2014). This would block release of GABA from MLIs, but not necessarily affect the calcium activity driven in large part by parallel fiber input. The revision includes specific mention of possible Golgi cell expression of hM4d in our Kit::Cre driver line.

b) Lack of change in the behavior. The additional data provided in Figure 2—figure supplement 3 is helpful, but does not fully address the possibility of other, unmeasured behavior differences (lateral deviation of the tongue, other orofacial movements), which is why the reviewers had asked for sample videos with and without chemogenetic suppression. Also, the lack of behavioral difference does not rule out a difference in the granule cells that does not affect the behavior, but could contribute to calcium transients in the Purkinje cells.

In the absence of more direct recordings from the granule cells or parallel fibers comparing responses in the presence or absence of the disinhibition manipulation, we would be satisfied with a more thorough acknowledgement and discussion of the above caveats in the manuscript.

We now point this out in the Results section. Specifically we state that we cannot rule out behavioral alterations such as those described by the reviewer.

2) Figure 1—figure supplement 1B and 1E show that PF activity is well-correlated with licking behavior after the cue. However, it is not very clear in Figure 1—figure supplement 1D. In particular, fluorescence signals begin to increase mostly before licking starts in the B3 bouton. It is therefore important to show PF and MLI activity not only when the cue induces licking but also fails to induce licking as described above. In addition, please indicate in Figure 1—figure supplement 1D when the cue was presented. Otherwise, readers cannot tell which licking bouts are learned behavior.

The data for PF activity have been added to Figure 1—figure supplement 1. It is essentially a flat line. The result for MLIs is the same, but has been excluded from the plot for clarity. This is stated in the figure legend. The timing of the sound cue is now shown in panel D.

3) In the first round of review, one of the reviewers asked more discussion of the cued licking task, for example, which components of the lick task are cerebellum dependent-acquisition of learning. The authors address the comment, but their rebuttal letter has addressed the comment better than the manuscript itself. It is important to let readers know that it is currently unclear (1) the role of climbing fibers in the cued licking task, and (2) which aspects of this learning task are regulated by the cerebellum. These limitations do not diminish the value of this study.

Provided that the reviewers found our response in the rebuttal informative, we simply moved these sentences into the Discussion section of the revision.

https://doi.org/10.7554/eLife.36246.030

Article and author information

Author details

  1. Michael A Gaffield

    Max Planck Florida Institute for Neuroscience, Jupiter, United States
    Contribution
    Conceptualization, Formal analysis, Investigation, Methodology, Writing—original draft
    Competing interests
    No competing interests declared
  2. Matthew J M Rowan

    Max Planck Florida Institute for Neuroscience, Jupiter, United States
    Contribution
    Formal analysis, Investigation, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  3. Samantha B Amat

    Max Planck Florida Institute for Neuroscience, Jupiter, United States
    Contribution
    Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  4. Hirokazu Hirai

    Gunma University Graduate School of Medicine, Maebashi, Japan
    Contribution
    Resources, Writing—review and editing
    Competing interests
    No competing interests declared
  5. Jason M Christie

    Max Planck Florida Institute for Neuroscience, Jupiter, United States
    Contribution
    Conceptualization, Funding acquisition, Methodology, Writing—original draft, Project administration
    For correspondence
    jason.christie@mpfi.org
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0276-2554

Funding

National Institutes of Health (NS083894)

  • Jason Christie

Max-Planck-Gesellschaft

  • Jason Christie

Max Planck Florida Institute for Neuroscience

  • Jason Christie

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

Acknowledgements

We thank the members of the Christie Lab for their helpful discussion during preparation of this manuscript. We are especially grateful to the GENIE program and the Janelia Research Campus (Drs. Jayaraman, Kerr, Kim, Looger, and Svoboda) for generously making GCaMP6f widely available to researchers and Dr. Bito (University of Tokyo) for use of RCaMP2. This work was supported by the Max Planck Society, the Max Planck Florida Institute for Neuroscience, and National Institutes of Health Grant NS083894 (JMC).

Ethics

Animal experimentation: Animal procedures were conducted using protocol 15-205 approved by the Institutional Animal Care and Use Committee (IACUC) at Max Planck Florida Institute for Neuroscience.

Senior Editor

  1. Eve Marder, Brandeis University, United States

Reviewing Editor

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

Reviewers

  1. Martijn Schonewille, Erasmus MC, Netherlands
  2. Jennifer L Raymond, Stanford School of Medicine, United States

Publication history

  1. Received: February 26, 2018
  2. Accepted: August 16, 2018
  3. Accepted Manuscript published: August 17, 2018 (version 1)
  4. Version of Record published: September 3, 2018 (version 2)

Copyright

© 2018, Gaffield 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

  • 870
    Page views
  • 186
    Downloads
  • 0
    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)

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

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

Further reading

    1. Cell Biology
    2. Neuroscience
    Matthew S Scarnati et al.
    Research Article Updated
    1. Neuroscience
    Qiaoli Huang et al.
    Research Article Updated