WIXLIB

8CHAPTER 1. PURKINJE CELL VOLTAGE-CLAMP IN VIVO1.1IntroductionMuch attention has been given recently to HCN1 channels, mediating the hyperpolarization-activated current (Ih ) and mainly responsible for the resetting of the restingmembrane potential after spike-induced hyperpolarization. The HCN1 subunit wasshown to be involved in motor-learning and motor-coordination tasks controlled bythe cerebellar cortex. Previous studies in HCN1 global knockout mice proved thatthe deletion of the channels causes major changes in the firing rates of Purkinje cells(PC) leading to profound deficits in motor learning [23].Current interest in the Nolan Lab is the investigation of how the inferior olivarycerebellar circuit is affected by selective HCN1 deletion and what behavioural outcomes this produces [22]. Preliminary data suggests that HCN1 channels have alarge influence on the initiation of action potentials in IO neurons, upstream of thecerebellum. In vitro recordings show that deletion of HCN1 subunits have two maineffects:- increasing the spontaneous firing rate (this is paradoxical for depolarising channels);- preventing the emergence of spikelets during the long afterdepolarising potential (ADP).The molecular mechanisms underlying these phenomena are currently under study.Recent evidences have demonstrated that IO bursts are faithfully transmittedto the PCs via the climbing fibers (CF) pathway [19]. A successful space-clamprecording of a PC is necessary to distinguish the EPSCs of CF inputs. As eachlarge EPSC in the PC corresponds to a spike originated upstream in the IO [19], byanalysing the statistics of the EPSCs we can infer the statistics of the IO bursts.The goal of the present work is to verify feasibility of an in vivo investigationof the role of HCN1 channels regulating the olivary cerebellar activity by voltageclamping single PCs in anaesthetised mice. The choice of the correct anaestheticis always critical for obtaining unbiased results but, unfortunately, the offer is verylimited. The basic requirement is that the drug does not act directly on the channel

1.2. MATERIALS AND METHODS9under study. In the present case, urethane was selected the most suitable anaestheticbecause other drugs were recently proved to act on HCN channels (e.g. ketamineand isofluorane [3] [4]).1.21.2.1Materials and methodsSurgeryC57BL6 KOPSI mice (age 6-7 weeks) were anaesthetised with isofluorane and putin a stereotaxic frame. A first surgery was performed to attach a head post, neededto fix the head during the recordings. Mice were given post-operative analgesia andwere returned to their cage for at least 24 hours. On the recording day mice wereanaesthetised again with isofluorane, a 500 µm 300 µm craniotomy was made overlobules V of the cerebellar vermis and the dura mater was removed. Mice were thaninjected with urethane (1.6 mg Kg 1 ) and carried to the recording rig.1.2.2In vivo voltage clampWhole cell recordings were made with Multiclamp 700A amplifier (Axon instrument),filtered at 4 kHz and digitised at 10-20 kHz. The pipette contained a Cs basedinternal solution with QX314 (Cs MethSo4 140 mM, NaCl 8 mM, HEPES 10 mM,MgATP 2 mM, Na ATP 2 mM, Na GTP 0.5 mM, EGTA 0.3 mM, QX314 Cl 5mM). The Cs blocks K channels while QX314 blocks Na channel in order to avoiddendritic spikes. Artificial cerebro-spinal fluid (aCSF: NaCl 125 mM, NaHCO3 25mM, NaH2 PO4 1.25 mM, KCl 2.5 mM, CaCl2 1.5 mM, MgCl2 1 mM, Glucose 25mM) was used as external solution. The electrodes (resistance range 5-6 MΩ) werepositioned by the means of a 4-axis manipulator. PCs were univocally identified fortheir typical high capacitance ( 100 pF) and depth from brain surface (190-300µm).

10CHAPTER 1. PURKINJE CELL VOLTAGE-CLAMP IN VIVO1.31.3.1ResultsFeasibility of obtaining an adequate space-clamp of Purkinjecells in vivoAn optimal space-clamp of the PC is required in order to distinguish the number ofEPSCs in the burst of CF inputs. If the cell is not space clamped correctly, spurious currents and potentially also unclamped dendritic spikes appear on the traceand could be misinterpreted as a genuine EPSC. Unfortunately, the PC has a hugedendritic tree that makes the space clamp a demanding task [26]. Moreover, in vivoaccess resistance in whole-cell configuration is typically higher than in vitro and theclamping quality is worse [17].The plots of Fig. 1.1 show two typical CF evoked EPSC of a PC [this voltageclamp recording was made by Ian Duguid]. The holding potential was set at -100mV because a strong hyperpolarisation enhances the probability that proximal anddistal dendrites were at least clamped at -70 mV. This technique reduces the impactof a relative inadequate space-clamp. Nevertheless, the result is suboptimal: theevidence for unclamped currents is the long duration of the EPSC, whose decay timeis ruled by at least two different decay constants. This observation makes unsure theinterpretation of the second peak: it is not possible to determine with certainty if itis a second EPSC or an unclamped dendritic conductance.For comparison, an in vivo voltage clamp recording from a granule cell is shownin Fig. 1.2 [this voltage-clamp recording was made by Paolo Puggioni]. Granule cellsare very small, diameter 10 µm, and therefore easy to space-clamp correctly. Theleft picture shows a typical EPSC recorded with holding potential -70 mV, while theright one represents a typical IPSC with holding potential 20 mV. The signal isclean and the cell is space clamped correctly, as the kinetic of each event indicate anadequate space-clamp.

1.3. RESULTS11Figure 1.1: PC voltage-clamp recording. Two typical CF inputs recorded with holding potential -100 mV. Despite the Cs internal solution and the strong hyperpolarised membranepotential, the space-clamp is not optimal. The evidence for unclamped currents is the longduration of the EPSC, which display a bi-exponential decay reflecting unclamped dendriticconductances. [This recording was made by Ian Duguid]Figure 1.2: Granule cell voltage clamp recording. (Left) Typical EPSC with holding potential-70 mV and (Right) typical IPSC with holding potential 20 mV. The signal is clean andthe cell is space clamped correctly, as proved by the exponential decay.

12CHAPTER 1. PURKINJE CELL VOLTAGE-CLAMP IN VIVO1.3.2The effect of urethane on the inferior olive activity in vitroAt the same time, the effect of the urethane on the activity of IO was tested in vitro[in collaboration with Nolan Lab, this study was conducted by Dr. D. Garden andis reported here due to its relevance for this work].As spontaneous action potential (AP) frequency is low in vitro the IO have beenstimulated by 2 ms depolarising steps. Each plot of Fig. 1.3 (Top) shows 10 consecutive APs before and after application of 20 mM urethane. In control conditions800 pA was sufficient to generate an AP while in 20 mM urethane the amplitudehad to be increased to 1100 pA. Interestingly, the plateau potential and spikelets arecompletely abolished in urethane. Finally, the resting membrane potential in controlconditions is about 5 mV more depolarised.1.4DiscussionAs stated in Sec. 1.1, the goal of the present work is to verify feasibility of an in vivoinvestigation of the role of HCN1 channels regulating the olivary cerebellar activityby voltage-clamping single PCs in anaesthetised mice. The quality of the PC spaceclamp is fundamental to distinguish the number of CF evoked EPSC in each IOburst and test the two predictions based on in vitro data (increased CF activity butsuppression of the bursts).This study showed that the in vivo measurements cannot be made according tothe original plans for two reasons:- the poor quality of the PC space clamp does not allow determination of individual EPSCs in a CF burst;- the urethane anaesthesia reduces the occurrence of CF inputs and abolishesthe bursts.The only possibility left for an in vivo investigation is to switch to awake animals,because it is now proved that both of the two available anaesthetic drugs (ketamineand urethane) affect directly and strongly the CF inputs.

1.4. DISCUSSION13Figure 1.3: Effect of urethane on the IO activity. (Top) 10 consecutive APs before (left) andafter (right) application of 20 mM urethane. As spontaneous action potential (AP) frequencyis low in vitro the IO have been stimulated by 2 ms depolarising steps. In control conditions800 pA was sufficient to generate an AP while in 20 mM urethane the amplitude had to beincreased to 1100 pA. (Middle) The plateau potential and spikelets are completely abolishedin urethane. (Bottom) The resting membrane potential in control conditions is about 5 mVmore depolarised.

14CHAPTER 1. PURKINJE CELL VOLTAGE-CLAMP IN VIVO

Chapter 2Purkinje cell dynamics in vivo:analysis of the spike traintemporal structureAbstractPurkinje cells (PCs), the only output of the cerebellar cortex, show two distincttypes of spiking activity: complex (CSs) and simple (SSs) spikes. Recent functionalcerebellar theories, such as the adaptive-filter model, consider the SSs as the carrierof the information, while the CSs as a teaching signal, responsible of the learningprocess. At the same time, the most recent coding theories propose that the PCinformation is stored in the rate and pauses of the spike train. These two perspectivesare somehow incompatible, as it was shown that CSs induce pauses and alter the SStemporal structure and therefore necessarily carry information. Therefore it will beof interest to examine the reciprocal influence that SSs and CSs have on each otheron a dynamical basis. This study investigates the dynamical temporal structure ofSS and CS trains in PC recordings from urethane-anaesthetised mice. The findingssuggest the conception of two simple models describing (i) the pauses as a possibleintrinsic feature of PC dynamics and (ii) the effect of a CS on the next SS interval.The huge variability of responses to the same inputs implies the presence of different15

16CHAPTER 2. PURKINJE CELL DYNAMICS IN VIVO’states’ in which the PC can operate. This sort of state-dependent computation, inwhich the state of the network is determined by the previous inputs, could potentiallybe described in a liquid-state machine framework.2.1IntroductionPurkinje cells (PCs) are the only output of the cerebellar cortex, known to play afundamental role in motor coordination and learning. Two distinct types of spikingactivity characterise PC recordings. Complex spikes (CSs), generally occurring at lowfrequencies (0.05-2 Hz), are massive depolarising events driven by excitatory inputsfrom a single climbing fiber (CF) coming from the inferior olive (IO) [5]. Single spikes(SSs) generally occur at higher frequency (30-100 Hz) and are intrinsically generatedwithout the need of any synaptic input [2]. Two different mechanisms modulate SSdynamics: excitation from the mossy fiber (MF) - granule cell (GC) - parallel fibers(PF) pathway and inhibition from basket cell interneurons (BC). PCs inhibit deepcerebellar nuclear (DCN) cells which, in turn, inhibit the IO. The IO completes theloop by exciting back the DCN and, as already described, by inducing CSs in PCs.The picture, already very complicated, is not complete yet. Classical theories onthe function of cerebellum (Albus-Ito-Marr [1] [13] [18]) consider the SS as the carrierof the information, while the CS as the responsible of the learning process. In recentstudies, the cerebellum is modelled as a filter whose inputs (coming from the brainstem to the PC through the MF-GC-PF path) are weighted by synapse strengths inthe PCs and then recombined to form the SS output. Moreover, the filter is adaptive,because synapses activated synchronously with a CS (an error teaching signal) aredown-regulated by long term depression mechanisms (LTD), while synapses carryinginputs negatively correlated with CSs are up-regulated by long term potentiation(LTP) [7]. The adaptive-filter model, although mathematically elegant and consistent with many experimental findings, is still not fully convincing.Present theoretical works are investigating the nature of the code adopted by PCsto transfer information. The most up-to-dated theories supported by experimentalevidences suggest that the DCN (target of PCs) is sensible to two parameters: (i)

2.1. INTRODUCTION17the rate of the spikes and (ii) the length of the pauses of PC spike trains. PCsinhibit DCN through GABA synapses: their regular firing controls DCN membranepotential and acts as a rate code. The pauses, when sufficiently synchronised acrossafferent PCs, generate bursts of activity in DCN and are interpreted to be a temporalcoding signal [6].A recent hot and rather controversial topic is related to the observed bi-stabilityin the firing rate of a relevant fraction of PCs. This phenomenon was shown with invitro [25] and in vivo anaesthetised [16] preparations, but its importance was harshlycriticised by objecting that it is rarely observed in awake animals [27]. Recently,awake recordings from cats not engaged in any behavioural task revealed that about50% of the cells show frequent long pauses suggesting the presence of up- and downstates [32]. However, the functional relevance of bistability is not clear.In this view, CSs cannot be considered as a pure teaching signal as in the adaptivefilter cerebellar theory, but necessarily they carry information. As a matter of fact,experimental evidence collected in the last years showed that CSs could affect therate and the pauses of PCs in many ways:- CSs introduce a typical 8-10 ms refractory period eventually followed by apause whose length depends on the number of Ca2 dendritic spikes triggeredby the full blown somatic Na action potential [5];- CSs can help the synchronisation of spike trains between paired PCs (definedas PCs with significant correlation in CF inputs, usually located in the samemicrozone [31]);- CSs can act as a switch between up- and down- states of the PC and thereforeproducing directly a pause or a regular firing period [16] [25].The last case is subtle and tricky from a coding point of view, as the result of theCS action depends on the state of the PC. Therefore, if we assume that the IO codesin a deterministic way, we should admit that it somehow knows the state of the PCand changes it deliberately.

18CHAPTER 2. PURKINJE CELL DYNAMICS IN VIVOThe study of the interaction of CSs with SS trains, the emergence of pauses andthe possible coding consequences requires the application of statistical techniquesthat go beyond the characterisation of overall averaged quantities. Processes havingcompletely different dynamic behaviours could produce the same mean result [9].The present work should be considered as an effort to propose the study of neuraldynamics by applying tools originally developed for the analysis of general dynamicalsystems. The main challenge is that the stochastic component is often dominant andtherefore the detection of regularities and recurrences is quite difficult.A total of 15 cell-attached recordings were made from PCs of urethane anaesthetised mice. The results presented in Sec. 2.3.1 and 2.3.2 give the flavour of thehuge variability characterising both average and dynamical properties of the PCs.Continuous firing, pausing and bistable PC are described in detail and the data areconsistent with published studies conducted on anaesthetised and awake preparations.The second part (Sec. 2.3.3 and 2.3.4) is devoted to the description of two simplemodels explaining (i) a possible intrinsic mechanism of pause generation, (ii) thesynchronisation effect that CSs have on the subsequent SSs.Section 2.3.5 contains the analysis of 2 awake extracellular recordings that areconsistent with the previous findings, although the range of the firing frequencyshown in one of the cells is massive.The discussion (Sec. 2.4) is focused on outlining a possible robust method to cooperate with and possibly explain in a probabilistic framework the striking biologicalvariability of the dynamical properties of the PCs.2.22.2.1Materials and methodsElectrophysiologySurgeryC56BLJ6 mice (age 6-7 weeks) were anaesthetised with isofluorane and put in astereotaxic frame. A first surgery was performed to attach a head post, needed tofix the head during the recordings. Mice were given a post-operative analgesia and

2.2. MATERIALS AND METHODS19were returned to their cage for at least 24 hours. On the recording day mice wereanaesthetised again with isofluorane, a 500 µm 300 µm craniotomy was made overlobules V of the cerebellar vermis and the dura mater was be removed. Mice werethan injected with urethane (1.6 mg Kg 1 ) and carried to the recording rig.In vivo cell attached recordingsCell-attached recordings were made with Multiclamp 700A amplifier (Axon instrument), band-pass filtered at 300 Hz-4 kHz and digitised at 10-20 kHz. The pipettecontained a HEPES based in vivo external solution (NaCl 150 mM, KCl 2.5 mM,HEPES 10 mM, CaCl2 2 mM, CaCl2 1 mM). Artificial cerebrospinal fluid (NaCl 125mM, NaHCO3 25 mM, NaH2 PO4 1.25 mM, KCl 2.5 mM, CaCl2 1.5 mM, MgCl2 1mM, Glucose 25 mM) was used as external solution. The electrodes (resistance range2.5-3.5 MΩ) were positioned by the means of a 4-axis manipulator. PCs were identified for their high frequency rate ( 30 Hz) and depth from brain surface (190-300µm). Although the presence of a CS is a final proof for PC identification, not allPCs show CSs (see Sec. 2.3.1). Therefore the detection of CSs cannot be used as theunique PC identification strategy.2.2.2Data analysisAll the data analysis was performed off-line by using a custom-made code implemented in MATLAB (MathWorks, Natick, MA, USA).Spike detection and sortingThe spikes are detected by setting thresholds at a proper value chosen carefully foreach cell after a visual inspection of peak amplitude and noise level. Spikes are thensorted as CS of SS by checking the signal between 1.5 and 3.5 ms after a spike. If themembrane current is higher than a given threshold, the spike is s

University of Edinburgh School of Informatics - DTC Neuroinformatics Spike dynamics in Purkinje cells: in vivo recordings and computational modelling