sign in sign up
<<
Home , Purkinje cell

Report

Purkinje Collaterals Enable Output Signals from the Cerebellar Cortex to Feed Back to Purkinje Cells and

Highlights Authors d In adults, all Purkinje cells have collaterals confined to a Laurens Witter, Stephanie Rudolph, parasagittal plane R. Todd Pressler, Safiya I. Lahlaf, Wade G. Regehr d Collaterals contact Purkinje cells, Lugaro cells, and molecular layer interneurons Correspondence [email protected] d Collaterals allow cerebellar output to regulate processing in parasagittal zones In Brief d Collaterals could regulate firing rate and synchrony in the Witter et al. show that in the juvenile and adult cerebellum, Purkinje cells have collaterals that provide inhibitory feedback to neighboring Purkinje cells and interneurons. These collaterals could allow cerebellar output to feed back and regulate firing in parasagittal zones.

Witter et al., 2016, 91, 1–8 July 20, 2016 ª 2016 Elsevier Inc. http://dx.doi.org/10.1016/j.neuron.2016.05.037 Please cite this article in press as: Witter et al., Purkinje Cell Collaterals Enable Output Signals from the Cerebellar Cortex to Feed Back to Purkinje Cells and Interneurons, Neuron (2016), http://dx.doi.org/10.1016/j.neuron.2016.05.037 Neuron Report

Purkinje Cell Collaterals Enable Output Signals from the Cerebellar Cortex to Feed Back to Purkinje Cells and Interneurons

Laurens Witter,1,2 Stephanie Rudolph,1,2 R. Todd Pressler,1,3 Safiya I. Lahlaf,1 and Wade G. Regehr1,* 1Department of Neurobiology, Harvard Medical School, 220 Longwood Avenue, Boston, MA 02115, USA 2Co-first author 3Current address: Case Western Reserve University, Cleveland, OH 44106, USA *Correspondence: [email protected] http://dx.doi.org/10.1016/j.neuron.2016.05.037

SUMMARY PC collaterals would allow the output of the cerebellar cortex to influence processing within the cerebellar cortex. PC collat- Purkinje cells (PCs) provide the sole output from the erals are known to inhibit Lugaro cells (LCs), which in turn inhibit cerebellar cortex. Although PCs are well character- Golgi cells (GoCs) (Crook et al., 2007; Hirono et al., 2012); PC in- ized on many levels, surprisingly little is known about hibition of LCs would increase GoC inhibition of grCs and pro- their axon collaterals and their target within vide net negative feedback to the grC layer. In contrast, inhibition the cerebellar cortex. It has been proposed that PC of GoCs, which inhibit grCs (Crowley et al., 2009; Palay and collaterals transiently control circuit assembly in Chan-Palay, 1974), would make the grC input layer more excit- able. PC-to-PC could serve a number of roles in early development, but it is thought that PC-to-PC adults. Elevated PC activity could feed back and suppress the connections are subsequently pruned. Here, we output of the cerebellar cortex and thereby control the gain of find that all PCs have collaterals in young, juvenile, PCs. Theoretical studies suggest that mutual PC-to-PC inhibi- and adult mice. Collaterals are restricted to the para- tion could allow PCs to generate prolonged responses (Maex sagittal plane, and most synapses are located in and Steuber, 2013). In addition, it has been proposed that PC close proximity to PCs. Using optogenetics and elec- collaterals could promote local synchronous firing (de Solages trophysiology, we find that in juveniles and adults, et al., 2008), which could be important for the ability of PCs to PCs make synapses onto other PCs, molecular layer regulate the activity of their targets in the interneurons, and Lugaro cells, but not onto Golgi (Gauck and Jaeger, 2000; Person and Raman, 2012). Inhibition cells. These findings establish that PC output can of molecular layer interneurons (MLIs), which in turn inhibit feed back and regulate numerous circuit elements PCs, could indirectly influence PC excitability and PC synchrony. It is therefore important to identify the targets of PC collaterals. within the cerebellar cortex and is well suited to Despite the many potential roles of PC collaterals, we know contribute to processing in parasagittal zones. very little about collaterals in adult animals. Anatomical character- ization of PC collaterals has established that PCs contact other INTRODUCTION PCs in young animals (Chan-Palay, 1971; Watt et al., 2009)and suggested that they inhibit LCs, MLIs, and GoCs (Crook et al., The cerebellum is involved in diverse motor and non-motor behav- 2007; Ha´ mori and Szenta´ gothai, 1968). In post-natal day 9 (P9) iors (Ito, 2008; Wang et al., 2014), and consequently cerebellar mice, PCs have prominent collaterals that provide the primary dysfunction can contribute to disorders such as (Manto source of inhibition to other PCs and can mediate traveling waves and Marmolino, 2009) and spectrum disorders (Baudouin of activity (Watt et al., 2009). Several studies reported that some et al., 2012; Piochon et al., 2014; Tsai et al., 2012). A large portion PCs have collaterals in juveniles (P30) and adults (RP90) (Bernard of cerebellar research has focused on a particular circuit element, et al., 1993; Bishop, 1982; Crook et al., 2007; Hawkes and Le- the Purkinje cell (PC), whose provide the sole output of the clerc, 1989; Larramendi and Lemkey-Johnston, 1970; O’Donog- cerebellar cortex. Although much is known about the intrinsic hue and Bishop, 1990), but their prevalence and targets remain properties of PCs and their synaptic inputs from parallel fibers, largely unknown. Recent work has found functional PC-to-LC climbing fibers, and inhibitory interneurons, remarkably little is connections in P18–P25 animals (Hirono et al., 2012), but connec- known about the prominence and function of PC axon collaterals tions to other types or PCs have not been described, and their synaptic connections within the cerebellar cortex (Ber- and it was hypothesized that PC-to-PC connections are only nard and Axelrad, 1993; Bernard et al., 1993; Bornschein et al., functional in young animals (Watt et al., 2009). 2013; Watt et al., 2009). Prominent collaterals would be a major Here, using a combination of anatomical, optogenetic, and deviation from the feedforward circuitry of the cerebellar cortex, electrophysiological approaches to examine PC collaterals, we with its main flow of signals from mossy fibers (MFs) through find that all PCs have collaterals that are confined to narrow par- granule cells (grCs) to PCs (Eccles et al., 1967; Marr, 1969). asagittal zones. These collaterals onto essentially all

Neuron 91, 1–8, July 20, 2016 ª 2016 Elsevier Inc. 1 Please cite this article in press as: Witter et al., Purkinje Cell Collaterals Enable Output Signals from the Cerebellar Cortex to Feed Back to Purkinje Cells and Interneurons, Neuron (2016), http://dx.doi.org/10.1016/j.neuron.2016.05.037

Figure 1. Purkinje Cell Collaterals Are Prominent in Mice Ranging from P12 to P90 (A) Schematic of the circuitry of the cerebellar cortex. Mossy fiber (MF), (grC), Purkinje cell (PC), climbing fiber (CF), parallel fiber (PF), (GoC), molecular layer interneuron (MLI), and Lugaro cell (LC). (B) Approach used to label PCs with GFP in P0 mice. (C) Example fluorescent image of a PC labeled with GFP from a P30 mouse. (D–F) Example reconstructions of PCs are shown for P12, P30, and P90 mice. Dotted lines are the lower edges of the molecular layer, cell bodies and main axons are indicated in gray, and axon collateral is shown in green. For each cell, a sagittal and transverse view is shown. (G) Summary of properties of PC collaterals (*p < 0.05). All bars indicate mean ± SEM.

PCs and LCs, and onto about a third of MLIs. Our findings show neurons in an acute slice by whole-cell recording with a dye-filled that PC collaterals are prominent in the mature cerebellum, electrode. Using this approach, however, collaterals are often where they allow the output of the cerebellar cortex to exert feed- severed, and axon health can be compromised. We instead back control over processing in the cerebellar cortex. labeled a small number of neurons in the intact animal (Figure 1C) by driving GFP expression with AAV vectors injected into the RESULTS lateral ventricles of Pcp2-Cre mice at P0 (Figure 1B). Intact PCs were identified within single parasagittal slices, allowing im- In order to evaluate the presence and properties of PC collat- aging of the entire PC collateral. erals, an approach is warranted that allows visualization of the An example PC from a P30 animal shows a typical labeled PC entire axonal arbor. A common approach is to fluorescently label with an axon that extends into the after giving off

2 Neuron 91, 1–8, July 20, 2016 Please cite this article in press as: Witter et al., Purkinje Cell Collaterals Enable Output Signals from the Cerebellar Cortex to Feed Back to Purkinje Cells and Interneurons, Neuron (2016), http://dx.doi.org/10.1016/j.neuron.2016.05.037

Figure 2. Purkinje Cell Collaterals and the Location of Their Synaptic Boutons (A) Reconstruction of a PC collateral labeled with GFP (green) and reconstructed presynaptic bou- ton locations (synaptophysin-tdTomato, red). (B) Detailed 3D isometric view of the same collat- eral as in (A). Cell bodies of contacted PCs are shown in white. (C) Zoom-in and 90 rotations of two collateral locations in close apposition to PCs. (D–F) Same as (A)–(C) for a different collateral reconstruction. (G) Overview of all detected puncta in relation to their position to the . Puncta positions are summarized in histograms along the baso-apical (top) and across the layers of the cerebellar cortex (right). The zero point on the x axis is aligned to the soma center, while the zero point on the y axis is aligned to the bottom the molecular layer (top of the soma). (H) Puncta were convolved with an SD = 1 mm Gaussian to obtain a heat map of all detected puncta. (I) Line plot of synaptic contacts in the baso-apical direction for 14 reconstructed collaterals. All bars indicate mean ± SEM.

group (Figures 1D–1G, S1, and S2), but collateral was similar between different ages, with only the total length of the collateral being significantly longer in P12 animals than in P30 and P90 ani- mals (Figure 1G; ANOVA, p < 0.05). There was also a trend towards smaller width and thickness, and the number of branches in older animals (Figure 1G). Collaterals were present towards both the apex and the base of the lobule (the direction the primary axon takes as it leaves the cortex and progresses to- wards the deep nuclei), but there was a slight directional bias, such that more col- laterals were present in the direction of the base, which became somewhat more pronounced in older animals (Fig- ure 1G). Overal, the qualitative impres- sion is that collaterals are similar in P12, several collateral branches (Figure 1C). PC collaterals in lobules P30, and P90 animals, but they are slightly more complex in II–IX of the cerebellar cortex were reconstructed from different P12 animals than in juveniles and adults. age animals (P12, n = 13; P30, n = 28; and P90, n = 9; Figures In order to provide more insight into potential synapses made S1 and S2, available online). Extensive collaterals were observed by PC collaterals, we combined single-cell labeling with a trans- in P12 animals (Figure 1D). There were large differences in collat- genic approach to label PC synapses (Figure 2). Pcp2-Cre ani- eral properties within each age group. Collateral branches and mals crossed with conditional synaptophysin-tdTomato mice re- the total width of the arbor varied considerably, but all collaterals sulted in faint labeling of PC bodies, , and axons, and were confined to a narrow sagittal plane, as shown in the side intense labeling of presynaptic boutons (Figures S2A–S2F). By views in Figures 1D–1F. Most branches reached the bottom of labeling individual PCs in Pcp2-Cre 3 synaptophysin-tdTomato the PC layer; some extended to the top of the PC layer or into mice, we were able to identify all presynaptic boutons of a given the molecular layer (Figure 1D, right). In P30 and P90 animals, PC, as shown for two reconstructed cells (Figures 2A–2F). We re- PC collaterals showed considerable variability within each age constructed 14 cells from lobules III–VII (Figure S2G) in P30 mice.

Neuron 91, 1–8, July 20, 2016 3 Please cite this article in press as: Witter et al., Purkinje Cell Collaterals Enable Output Signals from the Cerebellar Cortex to Feed Back to Purkinje Cells and Interneurons, Neuron (2016), http://dx.doi.org/10.1016/j.neuron.2016.05.037

Figure 3. Identification of a Cre Line that Allows Expression Restricted to Purkinje Cells (A) Schematic of the experimental approach to evaluate conditional expression of ChR2 in two different Cre lines. A Purkinje cell was recorded in whole-cell configuration while synaptic inputs were either evoked with light or by electrical stimulation. (B and C) Representative experiments showing the effects of the CB1R agonist WIN (5 mM) on optically and electrically evoked IPSCs. (B) Pcp2- Cre Mpin and (C) Pcp2-Cre Jdhu line. (D) Summary of the experiments in (B) and (C). (E) Example photocurrents measured in PCs (top traces) and MLIs (bottom traces) for the Mpin line (left) and the Jdhu line (right). (F and G) Summary of photocurrents measured in PCs (F) and MLIs (G). (H and I) Confocal images of the cerebellar cortex from the Pcp2-Cre Mpin 3 tdTomato (top left) and Pcp2-Cre Jdhu 3 tdTomato mice (bottom left) and summary of the percentage of MLIs expressing tdTomato in the two different Cre lines (right; also see Figure S3). All bars indicate mean ± SEM.

There is considerable variability in the location of synaptic con- the presence of inhibitors of excitatory and glycinergic synaptic tacts in the direction of the apex and base of the lobules (Figures transmission, we alternated optical stimulation just below the PC 2G–2I). The total spatial range over which synaptic contacts layer with electrical stimulation of the upper molecular layer, pri- were observed was comparable to the total width of the collat- marily activating MLI-to-PC synapses (Figure 3A). We found that eral (220 ± 35 mm versus 210 ± 20 mm). Most synapses were for the Mpin line, WIN strongly attenuated IPSCs evoked by near the PC layer (Figures 2G and 2H). PC somata close to the either light or electrical stimulation (Figures 3B and 3D). In GFP-labeled collateral were reconstructed, and their distance contrast, WIN did not affect optically evoked IPSCs in the Jdhu to the nearest synaptic contact was determined. Our analysis line, but attenuated the electrically evoked responses (Figures shows that PC collaterals come within 1 mm of 7.1 ± 1.1 other 3C and 3D). These experiments indicate that optically evoked PCs (Figures 2C and 2F). IPSCs in the Jdhu line are the result of PC-to-PC connections, We next tested whether PC collaterals make functional synap- but in the Mpin line MLI-to-PC synapses also contribute. ses onto PCs, LCs, MLIs, and GoCs. We used optogenetics We also tested for photocurrents in MLIs and never recorded to selectively activate PCs and record light-evoked inhibitory photocurrents in the Jdhu line (0 of 55), but 21% of the MLIs (6 of postsynaptic currents (IPSCs) in target neurons. This approach 29) had photocurrents in the Mpin line (Figures 3E and 3G). requires crossing a conditional channelrhodopsin-2 (ChR2) A cross of each Cre line with a conditional tdTomato line labeled mouse with a Cre line selective for PCs. We characterized two PCs in both lines (Figures 3H and 3I), but some cell bodies in the candidate Cre lines, the Pcp2-Cre Mpin (Barski et al., 2000) molecular layer were tdTomato positive in the Mpin line (Fig- and the Pcp2-Cre Jdhu (Zhang et al., 2004) lines, for expression ure 3H, white arrow head), but not in the Jdhu line (Figure 3I). selectivity. Our primary concern was possible non-specific Cre tdTomato expression was apparent in 3% ± 0.2% (N = 3 animals, expression (Lammel et al., 2015; Stuber et al., 2015). 738 neurons) of MLIs in Mpin-tdTomato mice, but not in MLIs in Mice conditionally expressing ChR2 were crossed with either Jdhu-tdTomato mice (N = 3 animals, 705 neurons; Figures S3, the Mpin or Jdhu Cre lines, and acute slices were cut. In both 3H, and 3I). In Jdhu mice, photocurrents were never observed lines, large photocurrents were evoked in all PCs, indicating in GoCs (n = 19), LCs (n = 12), or grCs (n = 69). To sample that PCs robustly express ChR2 (Figures 3E and 3F). We also from a larger population of grCs, we recorded from MLIs with found that brief pulses (0.5 ms) of blue light evoked IPSCs in only inhibitory transmission blocked while stimulating with light PCs of both lines, but it was unclear whether these IPSCs origi- and never observed any excitatory inputs (n = 28). nated from PCs or from MLIs that non-specifically expressed These experiments highlight the importance of assessing the ChR2. To distinguish between these possibilities, we used the specificity of Cre lines. The observation that for the Mpin line, type 1 receptor (CB1R) agonist WIN 55,212-2 a fluorescent reporter was detected in 3% of MLIs, whereas a (Figures 3A–3D). It is known that MLIs express CB1Rs and photocurrent was recorded in 21% of MLIs, suggests that func- MLI-to-PC synapses are strongly attenuated by WIN (Kreitzer tional testing is important for optogenetic studies that rely on a and Regehr, 2001), while PCs do not express CB1Rs. Therefore, Cre line. We conclude the Jdhu line is suitable for our studies, synapses made by PCs should be insensitive to WIN. In but the Mpin line is not.

4 Neuron 91, 1–8, July 20, 2016 Please cite this article in press as: Witter et al., Purkinje Cell Collaterals Enable Output Signals from the Cerebellar Cortex to Feed Back to Purkinje Cells and Interneurons, Neuron (2016), http://dx.doi.org/10.1016/j.neuron.2016.05.037

Figure 4. Functional Identification of the Synaptic Targets of Purkinje Cell Collaterals (A–D) Example recordings and cell identifica- tion using two-photon imaging. Responses were evoked with a brief pulse of light in the grC layer. Experiments for P30 mice are shown for (A) Pur- kinje cells, (B) molecular layer interneurons, (C) Lugaro cells, and (D) Golgi cells. (E) Summary of the fraction of cells with optically evoked IPSCs (top), and IPSC size (bottom) in each cell type both for juvenile (P30) and for adult animals (P90). (F) Example of a connected PC pair (top). The presynaptic neuron was loaded with Alexa 594 (red), and the postsynaptic neuron with Alexa 488 (green). Magnified image (bottom) shows red boutons in close proximity to the green PC body. (G) Example average response from one pair. Presynaptic spike (top); postsynaptic current (bottom). (H) Example successes (black) and failures (gray). (I) Amplitude histogram of postsynaptic currents. Successes are shown in black; failures in gray. (J and K) Summary of failure rate (J) and average synaptic current and potency (K) of the nine con- nected pairs. All bars indicate mean ± SEM.

cell was hyperpolarized to prevent spon- taneous firing, action potentials were evoked with brief current steps, and We examined synaptic contacts onto different potential PC tar- IPSCs were measured from the postsynaptic PC. Repeated gets using P30 Pcp2-Cre Jdhu 3 ChR2-EYFP mice. Optically stimulation of the presynaptic cell resulted in responses that var- evoked synaptic currents were observed in PCs (Figure 4A; 28 ied in size, with clear successes and failures (Figures 4H and 4I). of 29), MLIs (Figure 4B; 9 of 29), and LCs (Figure 4C; 11 of 12), The average IPSC amplitude was 65 ± 14 pA (n = 9), the potency but not in GoCs (Figure 4D; 0 of 19). Optically evoked IPSCs was 122 ± 29 pA (Figure 4K), the success rate was 44% ± 7%, were blocked by the GABAAR antagonist SR95531 (5 mM; Figures (Figure 4J), and the latency was 0.94 ± 0.07 ms. 4A–4C, gray traces) and were reduced to 7% ± 2.6% (n = 14, PCs), PC collaterals may help to synchronize activity. A recent in vivo 7% ± 3.4% (n = 5, MLIs), and 5% ± 1.1% (n = 9, LCs) of the initial study observed high-frequency oscillations that reflect synchro- IPSC values. There was considerable variability in the amplitude of nous PC firing (de Solages et al., 2008). Our observation that col- optically evoked IPSCs observed in the different cell types, and laterals are prominent and functional in the adult is consistent the average amplitude was 454 ± 71 pA in PCs (n = 28), 192 ± with PC-to-PC connections contributing to synchrony in vivo 55 pA in MLIs (n = 9), and 221 ± 48 pA in LCs (n = 11; Figure 4E). (see Discussion and Figure S4). We therefore studied PC syn- In P90 animals, we found that all PCs (12 of 12) received optically chrony in slices. Although the firing of nearby PCs was evoked IPSCs (530 ± 160 pA, n = 12) and a fraction of MLIs (7 of 25) rarely correlated (Figures S4C and S4D), we found that most received inputs (266 ± 90 pA, n = 7; Figure 4E). healthy PCs had intact dendrites, but severed axons (Fig- Our findings contrast with a previous study where PC-to-PC ure S4A). However, short latency-correlated activity was often synapses were not found using paired recordings in animals observed when the slice orientation preserved PC collaterals older than P22 (Watt et al., 2009). To resolve this discrepancy, (Figures S4B–S4D). Spike-triggered average synaptic currents we performed recordings from pairs of PCs in juvenile animals revealed that inhibition decreases at the time of synchronous to directly test for functional synapses between PCs. Based on firing (Figures S4G and S4H). In many cases, large spike-trig- our anatomical findings, we hypothesized that connections be- gered averages and correlated activity were observed even tween two random PCs would be rare and that severing collat- when no direct connection between the cells was apparent erals during the slicing procedure would further impede finding (although we only tested in one direction; Figures S4E and connected pairs. We therefore fluorescently labeled individual S4F). Correlated activity and spike-triggered averages were

PCs (see Supplemental Experimental Procedures) and used both suppressed by blocking GABAA receptors. Thus, when fluorescence to identify target PCs. We recorded from nine pairs the slice orientation preserves PC collaterals, we observed of synaptically connected PCs of P31–P39 animals. A typical correlated activity that appears to be a network phenomenon paired recording is shown in Figures 4F and 4G. The presynaptic that likely involves many PCs and possibly MLIs.

Neuron 91, 1–8, July 20, 2016 5 Please cite this article in press as: Witter et al., Purkinje Cell Collaterals Enable Output Signals from the Cerebellar Cortex to Feed Back to Purkinje Cells and Interneurons, Neuron (2016), http://dx.doi.org/10.1016/j.neuron.2016.05.037

DISCUSSION Another possibility suggested by a recent modeling study is that reciprocal inhibition between populations of PCs would Our major finding is that PCs have prominent axon collaterals allow PCs to generate firing-rate changes lasting tens of seconds that contact neighboring PCs and local interneurons in adult (Maex and Steuber, 2013). Here we find that PC-to-PC connec- mice. These findings indicate that PC collaterals are well suited tions are made toward both the apex and the base, and they to route inhibitory feedback to interneurons and PCs in the cere- could provide the requisite reciprocal connections. Thus, the bellar cortex to regulate cerebellar processing, which is counter anatomy of PC collaterals is compatible with a role in generating to the view that cerebellar processing is strictly feedforward. long-lasting signals. It was surprising that all PCs, even in adults, have collaterals PC collaterals could also control spike timing and synchrony. and receive collateral input. Moreover, we observed only a Loosely connected inhibitory networks, such as the one formed mild reduction in the complexity of axon collaterals over time by PCs and MLIs, have also been shown to promote synchrony (P12 versus P30 and P90), which was much less than the overall in other brain areas (Diba et al., 2014; Hu et al., 2011; Lagier et al., variation within age groups. Collaterals were confined to a nar- 2004). Indeed, a previous study described synchronous firing of row sagittal plane but extended hundreds of micrometers within nearby PCs and high-frequency oscillations in the cerebellar cor- that plane. The prominence of collaterals in adults suggests that tex in vivo (de Solages et al., 2008). A mechanism that relied on they are functional in adults and do not only serve a develop- PC-to-PC collaterals was advanced to explain these observa- mental role. The location of synaptic contacts in and near the tions, but the apparent absence of PC-to-PC connections in PC layer, and to a lesser extent in the molecular layer, suggested adults called this mechanism into question (Watt et al., 2009). that apart from the known contacts to LCs in juvenile animals Our findings suggest that the mechanism to explain in vivo syn- (Dieudonne´ and Dumoulin, 2000; Hirono et al., 2012), PC collat- chronous and oscillatory activity advanced by de Solages et al. erals might contact PCs and perhaps MLIs (Figure 2). Functional (2008) is viable. In our in vitro recordings, where even in optimal studies established that essentially all PCs are inhibited by other conditions the PC collateral network is not completely intact, we PCs, both in juveniles (P30) and adults (P90). Paired recordings observed synchronous PC activity that relied on inhibitory neuro- allowed us to measure the properties of synapses between transmission and was more prominent when the network of PC two PCs. By combining the average amplitude of PC-to-PC con- collaterals was preserved (Figure S4). The issue of PC synchrony nections (65 ± 14 pA) with the amplitude of light-evoked IPSCs is important because it has been proposed that synchronously (454 ± 71 pA), we estimate that each PC receives input from firing PCs could entrain firing in the deep cerebellar nuclei if their five to ten other PCs (Figure 4). This is in good agreement with main axons converged onto common cerebellar nuclei neurons our estimates of convergence from our synaptic labeling exper- (De Zeeuw et al., 2011; Gauck and Jaeger, 2000; Person and iments in which each PC forms synaptic contacts near approxi- Raman, 2012). mately six to eight PCs (Figure 2). PC collaterals could also be important when climbing fibers PC axon collateral synapses onto PCs and MLIs could regu- (CFs) are activated. CFs evoke characteristic complex spikes late activity in narrow parasagittal strips, which are likely con- followed by a brief pause in target PCs, whereas neighboring tained within broader zebrin bands that constitute functional PCs are inhibited (Bosman et al., 2010; Schwarz and Welsh, units (Apps and Hawkes, 2009). Both PC collaterals and MLI 2001). This inhibition is thought to arise from spillover activation axons are restricted to narrow parasagittal planes (Gao et al., of MLIs (Szapiro and Barbour, 2007; Coddington et al., 2013), but 2006; Hawkes and Leclerc, 1989). Therefore, PC feedback reg- our findings suggest that PC collaterals could also contribute to ulates cerebellar activity at the output stage in these functionally surround inhibition. CF activation could also help to reset the delimited zones, and could potentially act to regulate the rate or phase of synchronously firing PCs. timing of firing of PCs and MLIs. The cerebellum has long been considered primarily a feedfor- PC collaterals could allow the output of the cerebellar cortex ward circuit where inputs are processed sequentially without to feed back and control the gain of the cerebellar cortex. Gain much feedback (Eccles et al., 1967; Marr, 1969). The ubiquitous control by inhibitory feedback is a common mechanism to main- presence of PC collateral synapses within the cerebellar cortex tain the dynamic range of neural circuits. When principal output indicates that this is not the case. Our findings establish that neurons are excitatory, inhibitory feedback requires interneu- feedback is prominent in the adult cerebellum. Until recently, rons as in the cerebral cortex (Olsen et al., 2012) and hippocam- the only feedback described in the cerebellar system was that pus (Freund and Buzsa´ ki, 1996). When output neurons are of GoC inhibition of local grCs. More recently, it was shown GABAergic, as in the basal ganglia and as described here for that some neurons of the cerebellar nuclei send collaterals the cerebellum, gain control can be achieved by connections back up to the cerebellar cortex contacting grCs or GoCs (Ankri between the output cells (Brown et al., 2014). If PC collaterals et al., 2015; Gao et al., 2016; Houck and Person, 2015). Here we control the firing rate of their targets, then PC-to-PC connec- establish that feedback is prominent and important in the cere- tions allow PC activity to suppress the output of the cerebellar bellum, and establish that the cerebellum is not exclusively a cortex. In contrast, PC-to-MLI synapses would have the oppo- feedforward circuit. site effect and would suppress inhibition of MLIs to PCs, thereby providing positive feedback. The time course and extent of EXPERIMENTAL PROCEDURES feedback on PC firing rates will thus depend on collateral con- nectivity and the balance of direct inhibition and indirect In all figures, bars are mean ± SEM. Experimental Procedures are described in disinhibition. detail in the Supplemental Experimental Procedures. All animals were handled

6 Neuron 91, 1–8, July 20, 2016 Please cite this article in press as: Witter et al., Purkinje Cell Collaterals Enable Output Signals from the Cerebellar Cortex to Feed Back to Purkinje Cells and Interneurons, Neuron (2016), http://dx.doi.org/10.1016/j.neuron.2016.05.037

in accordance with federal guidelines and protocols approved by Harvard Brown, J., Pan, W.-X., and Dudman, J.T. (2014). The inhibitory microcircuit of University. the substantia nigra provides feedback gain control of the basal ganglia output. eLife 3, e02397. SUPPLEMENTAL INFORMATION Chan-Palay, V. (1971). The recurrent collaterals of Purkinje cell axons: a corre- lated study of the rat’s cerebellar cortex with electron microscopy and the Supplemental Information includes Supplemental Experimental Procedures Golgi method. Z. Anat. Entwicklungsgesch. 134, 200–234. and four figures and can be found with this article online at http://dx.doi.org/ Coddington, L.T., Rudolph, S., Vande Lune, P., Overstreet-Wadiche, L., and 10.1016/j.neuron.2016.05.037. Wadiche, J.I. (2013). Spillover-mediated feedforward inhibition functionally segregates interneuron activity. Neuron 78, 1050–1062. AUTHOR CONTRIBUTIONS Crook, J.D., Hendrickson, A., Erickson, A., Possin, D., and Robinson, F.R. (2007). Purkinje cell axon collaterals terminate on Cat-301+ neurons in L.W. and S.R. contributed equally. R.T.P. and W.G.R. conceived the experi- Macaca monkey cerebellum. Neuroscience 149, 834–844. ments. L.W., S.R., and R.T.P. conducted electrophysiological experiments. Crowley, J.J., Fioravante, D., and Regehr, W.G. (2009). Dynamics of fast and L.W., S.R., and S.I.L. did anatomical studies. L.W., S.R., R.T.P., and S.I.L. con- slow inhibition from cerebellar golgi cells allow flexible control of synaptic inte- ducted analysis. L.W., S.R., and W.G.R. wrote the manuscript with contribu- gration. Neuron 63, 843–853. tions from all authors. de Solages, C., Szapiro, G., Brunel, N., Hakim, V., Isope, P., Buisseret, P., ´ ACKNOWLEDGMENTS Rousseau, C., Barbour, B., and Lena, C. (2008). High-frequency organization and synchrony of activity in the purkinje cell layer of the cerebellum. Neuron 58, 775–788. This work was supported by NIH R01NS032405 and R01NS092707; the Nancy Lurie Marks, Goldenson, Lefler, and Khodadad Foundations (W.G.R); a Gold- De Zeeuw, C.I., Hoebeek, F.E., Bosman, L.W.J., Schonewille, M., Witter, L., enson and a Mahoney Fellowship to L.W.; a Brooks fellowship and an NIH and Koekkoek, S.K. (2011). Spatiotemporal firing patterns in the cerebellum. F32NS087708 to S.R.; and NEI F32EY020718 to R.T.P. We thank the Image Nat. Rev. Neurosci. 12, 327–344. and Data Analysis Core (IDAC) and H.L. Elliott for help with image analysis, Diba, K., Amarasingham, A., Mizuseki, K., and Buzsa´ ki, G. (2014). Millisecond the Neurobiology Imaging Facility (NINDS P30 Core Center Grant timescale synchrony among hippocampal neurons. J. Neurosci. 34, 14984– NS072030), C. Hull and the W.G.R. lab for comments on the manuscript, 14994. and K. McDaniels for mouse colony maintenance. Dieudonne´ , S., and Dumoulin, A. (2000). Serotonin-driven long-range inhibi- tory connections in the cerebellar cortex. J. Neurosci. 20, 1837–1848. Received: October 26, 2015 Revised: March 30, 2016 Eccles, J., Ito, M., and Szenta´ gothai, J. (1967). The Cerebellum as a Neuronal Accepted: May 23, 2016 Machine (Springer Berlin Heidelberg). Published: June 23, 2016 Freund, T.F., and Buzsa´ ki, G. (1996). Interneurons of the hippocampus. Hippocampus 6, 347–470. REFERENCES Gao, W., Chen, G., Reinert, K.C., and Ebner, T.J. (2006). Cerebellar cortical molecular layer inhibition is organized in parasagittal zones. J. Neurosci. 26, Ankri, L., Husson, Z., Pietrajtis, K., Proville, R., Le´ na, C., Yarom, Y., Dieudonne´ , 8377–8387. S., and Uusisaari, M.Y. (2015). A novel inhibitory nucleo-cortical circuit con- Gao, Z., Proietti-Onori, M., Lin, Z., Ten Brinke, M.M., Boele, H.J., Potters, J.W., trols cerebellar Golgi cell activity. eLife 4, http://dx.doi.org/10.7554/eLife. Ruigrok, T.J., Hoebeek, F.E., and De Zeeuw, C.I. (2016). Excitatory cerebellar 06262. nucleocortical circuit provides internal amplification during associative condi- Apps, R., and Hawkes, R. (2009). Cerebellar cortical organization: a one-map tioning. Neuron 89, 645–657. hypothesis. Nat. Rev. Neurosci. 10, 670–681. Gauck, V., and Jaeger, D. (2000). The control of rate and timing of spikes in the Barski, J.J., Dethleffsen, K., and Meyer, M. (2000). Cre recombinase expres- deep cerebellar nuclei by inhibition. J. Neurosci. 20, 3006–3016. sion in cerebellar Purkinje cells. Genesis 28, 93–98. Ha´ mori, J., and Szenta´ gothai, J. (1968). Identification of synapses formed in Baudouin, S.J., Gaudias, J., Gerharz, S., Hatstatt, L., Zhou, K., Punnakkal, P., the cerebellar cortex by Purkinje axon collaterals: an electron microscope Tanaka, K.F., Spooren, W., Hen, R., De Zeeuw, C.I., et al. (2012). Shared syn- study. Exp. Brain Res. 5, 118–128. aptic pathophysiology in syndromic and nonsyndromic rodent models of autism. Science 338, 128–132. Hawkes, R., and Leclerc, N. (1989). Purkinje cell axon collateral distributions reflect the chemical compartmentation of the rat cerebellar cortex. Brain Bernard, C., and Axelrad, H. (1993). Effects of recurrent collateral inhibition on Res. 476, 279–290. Purkinje cell activity in the immature rat cerebellar cortex—an in vivo electro- physiological study. Brain Res. 626, 234–258. Hirono, M., Saitow, F., Kudo, M., Suzuki, H., Yanagawa, Y., Yamada, M., Nagao, S., Konishi, S., and Obata, K. (2012). Cerebellar globular cells receive Bernard, C., Axelrad, H., and Giraud, B.G. (1993). Effects of collateral inhibition monoaminergic excitation and monosynaptic inhibition from Purkinje cells. in a model of the immature rat cerebellar cortex: multineuron correlations. PLoS ONE 7, e29663. Brain Res. Cogn. Brain Res. 1, 100–122. Houck, B.D., and Person, A.L. (2015). Cerebellar premotor output neurons Bishop, G.A. (1982). The pattern of distribution of the local axonal collaterals of collateralize to innervate the cerebellar cortex. J. Comp. Neurol. 523, 2254– Purkinje cells in the intermediate cortex of the anterior lobe and paramedian 2271. lobule of the cat cerebellum. J. Comp. Neurol. 210, 1–9. Bornschein, G., Arendt, O., Hallermann, S., Brachtendorf, S., Eilers, J., and Hu, H., Ma, Y., and Agmon, A. (2011). Submillisecond firing synchrony be- Schmidt, H. (2013). Paired-pulse facilitation at recurrent Purkinje neuron syn- tween different subtypes of cortical interneurons connected chemically but 31 apses is independent of calbindin and during high-frequency not electrically. J. Neurosci. , 3351–3361. activation. J. Physiol. 591, 3355–3370. Ito, M. (2008). Control of mental activities by internal models in the cerebellum. 9 Bosman, L.W.J., Koekkoek, S.K.E., Shapiro, J., Rijken, B.F.M., Zandstra, F., Nat. Rev. Neurosci. , 304–313. van der Ende, B., Owens, C.B., Potters, J.-W., de Gruijl, J.R., Ruigrok, Kreitzer, A.C., and Regehr, W.G. (2001). Cerebellar depolarization-induced sup- T.J.H., and De Zeeuw, C.I. (2010). Encoding of whisker input by cerebellar pression of inhibition is mediated by endogenous . J. Neurosci. 21, Purkinje cells. J. Physiol. 588, 3757–3783. RC174.

Neuron 91, 1–8, July 20, 2016 7 Please cite this article in press as: Witter et al., Purkinje Cell Collaterals Enable Output Signals from the Cerebellar Cortex to Feed Back to Purkinje Cells and Interneurons, Neuron (2016), http://dx.doi.org/10.1016/j.neuron.2016.05.037

Lagier, S., Carleton, A., and Lledo, P.-M. (2004). Interplay between local plasticity and motor learning deficits in a copy-number variation mouse model GABAergic interneurons and relay neurons generates g oscillations in the rat of autism. Nat. Commun. 5, 5586. olfactory bulb. J. Neurosci. 24, 4382–4392. Schwarz, C., and Welsh, J.P. (2001). Dynamic modulation of mossy fiber sys- Lammel, S., Steinberg, E.E., Fo¨ ldy, C., Wall, N.R., Beier, K., Luo, L., and tem throughput by inferior olive synchrony: a multielectrode study of cerebellar Malenka, R.C. (2015). Diversity of transgenic mouse models for selective tar- cortex activated by motor cortex. J. Neurophysiol. 86, 2489–2504. geting of midbrain dopamine neurons. Neuron 85, 429–438. Larramendi, L.M., and Lemkey-Johnston, N. (1970). The distribution of recur- Stuber, G.D., Stamatakis, A.M., and Kantak, P.A. (2015). Considerations when rent Purkinje collateral synapses in the mouse cerebellar cortex: an electron using cre-driver rodent lines for studying ventral tegmental area circuitry. 85 microscopic study. J. Comp. Neurol. 138, 451–459. Neuron , 439–445. Maex, R., and Steuber, V. (2013). An integrator circuit in cerebellar cortex. Eur. Szapiro, G., and Barbour, B. (2007). Multiple climbing fibers signal to molecular J. Neurosci. 38, 2917–2932. layer interneurons exclusively via glutamate spillover. Nat. Neurosci. 10, Manto, M., and Marmolino, D. (2009). Cerebellar . Curr. Opin. Neurol. 735–742. 22, 419–429. Tsai, P.T., Hull, C., Chu, Y., Greene-Colozzi, E., Sadowski, A.R., Leech, J.M., 202 Marr, D. (1969). A theory of cerebellar cortex. J. Physiol. , 437–470. Steinberg, J., Crawley, J.N., Regehr, W.G., and Sahin, M. (2012). Autistic-like O’Donoghue, D.L., and Bishop, G.A. (1990). A quantitative analysis of the dis- behaviour and cerebellar dysfunction in Purkinje cell Tsc1 mutant mice. Nature tribution of Purkinje cell axonal collaterals in different zones of the cat’s cere- 488, 647–651. bellum: an intracellular HRP study. Exp. Brain Res. 80, 63–71. Wang, S.S., Kloth, A.D., and Badura, A. (2014). The cerebellum, sensitive pe- Olsen, S.R., Bortone, D.S., Adesnik, H., and Scanziani, M. (2012). Gain control riods, and autism. Neuron 83, 518–532. by layer six in cortical circuits of vision. Nature 483, 47–52. Palay, S.L., and Chan-Palay, V. (1974). Cerebellar Cortex: Cytology and Watt, A.J., Cuntz, H., Mori, M., Nusser, Z., Sjo¨ stro¨ m, P.J., and Ha¨ usser, M. Organization (Springer). (2009). Traveling waves in developing cerebellar cortex mediated by asymmet- rical Purkinje cell connectivity. Nat. Neurosci. 12, 463–473. Person, A.L., and Raman, I.M. (2012). Purkinje neuron synchrony elicits time- locked spiking in the cerebellar nuclei. Nature 481, 502–505. Zhang, X.-M., Ng, A.H.-L., Tanner, J.A., Wu, W.-T., Copeland, N.G., Jenkins, Piochon, C., Kloth, A.D., Grasselli, G., Titley, H.K., Nakayama, H., Hashimoto, N.A., and Huang, J.-D. (2004). Highly restricted expression of Cre recombi- K., Wan, V., Simmons, D.H., Eissa, T., Nakatani, J., et al. (2014). Cerebellar nase in cerebellar Purkinje cells. Genesis 40, 45–51.

8 Neuron 91, 1–8, July 20, 2016