NIN-KNAW Summer School on Procedural Learning

NIN-KNAW Summer School on Procedural Learning 26-28 th June 2017 Amsterdam, the Netherlands Programme

NIN-KNAW Summer School 2017 | 1

SPONSORS

We thank the following sponsors for their support:

2 | NIN-KNAW Summer School 2017

GENERAL INFORMATION Organised by NIN-KNAW: Aleksandra Badura, [email protected] Beerend Winkelman, [email protected] Chris de Zeeuw, [email protected] Tycho Hoogland, [email protected] For urgent matters: +31 6 396 88034 Special thanks to: Esmeralda Schemmekes, Laura Cardona, Tini Eikelboom, Wilma Verweij, Marco Adrichem, Ernst Rijvordt, Cock Koeleman, Roxanne ter Haar, Heshow Shaweis, Cynthia Geelen, Rachel Koops Venue: Singelkerk Singel 452 Amsterdam Google Maps: https://goo.gl/CKH1HAL

NIN-KNAW Summer School 2017 | 3

CONTENTS

Scientific Programme

P 5-7

Speaker Abstracts

P 8-19

Ariëns Kapper Award

P 20-21

Poster Abstracts

P 22-31

Directions to venue / party

P 32-33

Notes

P 34-37

4 | NIN-KNAW Summer School 2017

DAY 1 Badge pick-up + coffee/tea

08:00-08:50

Welcome

08:50-09:00

Pieter Roelfsema Deep learning in the brain: the role of cortical feedback connections in the plasticity of sensory representations

Alexandre Pouget

09:00-10:00 10:00-11:00

Learning, uncertainty and confidence

Break

11:00-11:20

Fritjof Helmchen

11:20-12:20

Mesoscale brain dynamics during a sensory discrimination task

Lunch

12:20-13:20

David Ostry

13:20-14:20

The boundary between motor and perceptual learning

Kathleen Cullen The rapid updating of internal models of voluntary self-motion in the primate cerebellum: Implications for perception and action

14:20-15:20

Break

15:20-15:45

Dorret Boomsma Human individual differences: Twin and genetic association studies

15:45-16:45

Drinks / Poster Session

16:45-18:30

Speakers dinner

19:00-21:00 NIN-KNAW Summer School 2017 | 5

DAY 2 Coffee/tea Roy Sillitoe Genetic dissection of cerebellar inhibitory interneuron function

Megan Carey

08:30-09:00 09:00-10:00

Behavioral state modulation of associative learning in mouse cerebellum

10:00-11:00

Break

11:00-11:20

Adam Hantman Neural circuits of dexterity

Lunch Daniel Huber Cortical circuit dynamics during neuroprosthetic learning

John Krakauer

11:20-12:20 12:20-13:20 13:20-14:20

How should we think about motor skill?

14:20-15:20

Break

15:20-15:45

Ariëns Kappers Award

15:45-16:00

Rui Costa

Ariëns Kappers lecture Generating and shaping novel action repertoires

16:00-17:00

Reception

17:00-18:00

Party (Tolhuis)

21:00-01:00

6 | NIN-KNAW Summer School 2017

DAY 3 Coffee/tea

08:30-09:00

Tom Otis

09:00-10:00

Cerebellar circuit mechanisms of normal movement & ataxia

Sam Wang

10:00-11:00

The Cerebellum, Sensitive Periods, and Autism

Break

11:00-11:20

Amy Bastian

11:20-12:20

Learning and relearning movement

Lunch

12:20-13:20

Simon Fisher Understanding the impact of FOXP2 gene disruptions in humans, mice and songbirds

13:20-14:20

Young talent presentation PhD

14:20-14:50

Young talent presentation Postdoc

14:50-15:20

Closing remarks

15:20-15:40

NIN-KNAW Summer School 2017 | 7

SPEAKER ABSTRACTS Deep learning in the brain: the role of cortical feedback connections in the plasticity of sensory representations Pieter R Roelfsema, Netherlands Institute for Neuroscience, Amsterdam, The Netherlands

H

ow does our brain fine tune perceptual representations when we learn about new sensory stimuli? How does a neuron in a sensory area of cortex know whether it should increase or decrease the strength of a particular synapse?

In computational neuroscience, this question is known as the “credit assignment problem”. At first sight a neuron seems to only have access to information about its own activity as well as activity of neurons from which it receives input. Its influence on the behavior of the organism is often indirect and depends on multiple multi-synaptic pathways. I will review the factors that are instrumental for optimizing the contribution of synapses to the overall network performance. One factor is a feedback signal from the response selection stage, which relies on the activity of NMDA receptors and is thought to produce eligibility traces that gate plasticity for a subset of the synapses. Tagged synapses will be held responsible for the outcome of an action. The second factor is the reward prediction error, which is the difference between the amount of reward that was predicted and the amount that was obtained by the individual. The reward prediction error is signaled by the release of neuromodulators, such as dopamine and acetylcholine, which modify the tagged synapses. The combined action of these synaptic eligibility traces and the reward prediction error changes synapses to promote future actions leading to more reward. The resulting learning is powerful and can explain how learning occurs in deep neurobiological networks. 8 | NIN-KNAW Summer School 2017

Learning, uncertainty and confidence Alexandre Pouget, University of Geneva, Switzerland

C

lassic theories of neural computation rely on the idea that synapses represent synaptic weights, that is, scalars representing the strength of the connection. Within this framework, learning is thought to implement of form of gradient descent in which the weights are slowly adjusted to optimize behavioral performance or reward rate. Given a finite training set, however, it is impossible to know with certainty what the value of the weights should be for a particular task. It is more efficient to infer instead the posterior distribution over the weights given the training set.

I will propose a probabilistic framework for learning and computation in neural networks in which synapses represent probability distributions over synaptic weights given the training set, and the goal of computation is to infer the posterior distribution over the output given the input pattern, the weights and their respective uncertainties. Within this framework, learning should be modulated by both the certainty about the weights as well as the certainty of the output of the network (or confidence). In particularly, learning should be stronger for weights whose uncertainty is larger. I will show that these new learning rules lead to more robust performance on a variety of problems and I will present preliminary experimental evidence suggesting that perceptual learning in olfactory discrimination tasks might indeed be modulated by weight uncertainty and confidence.

NIN-KNAW Summer School 2017 | 9

Mesoscale brain dynamics during a sensory discrimination task Fritjof Helmchen, University of Zurich, Switzerland

T

hrough the combination of in vivo optical imaging and chronic expression of genetically encoded calcium indicators it is now feasible to directly ‘watch’ neuronal population dynamics in the neocortex of awake, head-restrained mice during specific behaviors.

Here, I will exemplify such measurements of mesoscale brain dynamics for a whisker-based texture discrimination task using calcium imaging with either wide-field camera imaging, multi-fiber photometry, or large-area two-photon microscopy. We apply viral-labeling strategies as well as various transgenic mouse lines with distinct expression patterns of sensitive calcium indicators. We are especially interested in the activation patterns across neocortical and subcortical areas during sensory discrimination and how mesoscale signal flow is established during task learning. I will highlight results from applying the various optical imaging tools that are now available for analyzing brain dynamics in behaving mice and will also discuss future prospects.

10 | NIN-KNAW Summer School 2017

The rapid updating of internal models of voluntary self-motion in the primate cerebellum: Implications for perception and action Kathleen E Cullen, Johns Hopkins University, Baltimore, USA

A

prevailing view is that the cerebellum is the site of a forward model that predicts the expected sensory consequences of self-generated action. During motor learning, a mismatch between this prediction and the actual sensory consequences of voluntary behaviour are thought to be vital for updating both the forward model and motor program. However, where and how the brain compares the two was unknown.

In this talk, I will focus on a relatively simple sensory-motor projection from the deep cerebellar nuclei (rostral fastigial nucleus) to the vestibular nuclei, reticular formation, and spinal cord, which is involved in generating reflexes that ensure accurate posture and balance. Trial by trial analysis of these neurons in a motor learning task revealed selective encoding of unexpected self-motion (vestibular information). The implications of these findings are then considered in relation to the calibration of vestibule-spinal reflexes, as well as the higher order processing of vestibular information. Finally, I will discuss the results of a parallel line of our current work demonstrating how non-cerebellar plasticity in direct brainstem pathways further contributes to rapidly shaping motor performances in vivo.

NIN-KNAW Summer School 2017 | 11

The boundary between motor and perceptual learning David J Ostry, McGill University, Montreal Canada

T

here is accumulating evidence that perceptual and motor learning do not occur in isolation. Rather motor learning affects perceptual judgements and changes sensory brain areas. Similarly, perceptual learning changes movements and motor areas of the brain.

In this talk, I will summarize recent work in support of these claims drawing on data from human arm movement and associated fMRI studies of resting-state connectivity. I will also present the results of new work on speech motor learning in which we use a robotic device to selectively alter somatosensory input during speech. This technique is combined with a resting-state neuroimaging analysis based on partial correlation that identifies from among the set of areas that encode learning those whose functional connectivity is both strengthened with learning and cannot be attributed to activity in other parts of the speech network. When we remove the signal attributable to other brain regions, we observe novel roles for sensory and motor systems in learning. Specifically, we find that it is largely sensory areas that survive the partial correlation test in the context of speech motor learning. That is, once the contribution to connectivity that is attributable to other parts of the speech motor network is removed, the remaining connectivity changes that are associated with improvements in motor performance are not in motor areas of the brain but in auditory and somatosensory cortex among others.

12 | NIN-KNAW Summer School 2017

Human individual differences: Twin and genetic association studies Dorret Boomsma, Netherlands Twin Register, VU Univ. Amsterdam

T

he classical twin study is a powerful heuristic in neuroscience, psychiatry and behavioral research. Its application has contributed to the awareness that variation in almost every human trait is influenced our genome and that covariation among traits often reflects genetic pleiotropy. For many traits, estimates of heritability and genetic correlations among traits derived from twin studies encouraged the search for causal genetic variants through genome-wide association studies (GWAS).

Here, I consider twin and GWA studies in the current era of molecular genetics. GWAS have successfully identify tens of thousands of genetic variants associated with complex human traits and led to large repositories of genome wide association results, providing opportunities for further multivariate approaches and Mendelian randomization studies, which contribute to the study of causality in human data. The twin design continues to be of importance through for example the application of the discordant monozygotic twin design. Twins offer unique opportunities to research that extend beyond the estimation of heritability. For heritable traits, the comparison of discordant monozygotic twins represents a powerful improvement over the traditional case–control study to search for disease-associated biological marks, for example in epigenetics and gene expression studies.

NIN-KNAW Summer School 2017 | 13

Cerebellar circuit mechanisms of normal movement & ataxia Tom Otis, Synapses and Circuits Section, pRED, Basel, Switzerland

I

n the first part of the talk I will describe published and unpublished optogenetic experiments that provide insight into how Purkinje neurons and deep cerebellar nucleus neurons shape movements. We have found that robust forms of associative learning can be generated by pairing optogenetic circuit manipulation with sensory stimuli.

This artificial learning is quite similar in properties to natural associative learning in exhibiting extinction and savings. Different configurations of training lead to learned responses with distinct properties, suggesting interesting mechanisms by which consolidation and extinction might occur at a circuit level. In the second part of the talk I will present collaborative data exploring circuit and cellular mechanisms underlying cerebellar dysfunction in various models of spinocerebellar ataxia type 2. I will also present some promising treatment strategies for this devastating neurodegenerative disease.

14 | NIN-KNAW Summer School 2017

Genetic dissection of cerebellar inhibitory interneuron function Roy Sillitoe, Baylor College of Medicine, USA

T

he cerebellar cortex contains at least seven distinguishable classes of interneurons. Among these are the excitatory granule cells and unipolar brush cells, and the inhibitory Golgi cells, Lugaro cells, and candelabrum cells, in addition to the developmentally related stellate and basket cells. Although it is well accepted that stellate and basket cells arise from a common progenitor pool, it is not clear whether these two cell types constitute unique functional classes or if they contribute differentially to cerebellar behaviors.

To investigate these problems, we devised a genetic inducible approach to silence GABAergic neurotransmission selectively at stellate versus basket cell interneuron synapses in the mouse. The flexibility of this approach allows us to integrate it with optogenetics as well as with genetically controlled diptheria toxin killing of select neurons. With these tools, we ask how stellate and basket cells influence circuits during development compared to adulthood, and also how cerebellar function is impacted when their synapses are silenced over the long-term, manipulated acutely, or when the whole interneuron is removed from the circuit altogether. We then analyze the consequences on cerebellar function and behavior using a combination of in vivo electrophysiology, EMG, gene expression patterns, anatomy, electron microscopy, behavioral assays and deep brain stimulation. Our results suggest that stellate and basket cells play discrete functional roles. The data raise the possibility that interneuron defects partly determine cerebellar disease pathophysiology.

NIN-KNAW Summer School 2017 | 15

Neural circuits of dexterity Adam Hantman, Janelia Research Campus, Ashburn, Virginia, USA

D

exterous movements serve the major functions of the brain, perception and manipulation of the world. Considering the range of possible actions and the complexity of musculoskeletal arrangements, control of the hand is an amazing achievement of the nervous system.

Dexterous behavior involves understanding objects in the world, developing appropriate plans, converting those plans into appropriate motor commands, and adaptively reacting to feedback. The myriad of these underlying operations is likely performed by a diverse set of neural circuits. By combining anatomy, physiology, and specific (genetic and temporal) manipulations, we hope to identify and understand the neural elements responsible for dexterous motor control. Currently, we focus on the role of the cortico-cerebellar loop in a skilled reach-grab-eat task in the rodent.

Ariëns Kappers lecture

Generating and shaping novel action repertoires

Rui Costa, Champalimaud Centre for the Unknown, Lisbon, Portugal

T

his lecture will be preceded by a speech by NIN-KNAW director Pieter Roelfsema and presentation of the Ariëns Kappers medal (p 20) to Rui Costa.

16 | NIN-KNAW Summer School 2017

Cortical circuit dynamics during neuroprosthetic learning Daniel Huber, University of Geneva, Switzerland

N

euronal motor commands, whether generating real or neuroprosthetic movements, are shaped by ongoing sensory feedback from the displacement being produced. I will present how we explore the use of cortical microstimulation to provide artificial feedback during operant conditioning of cortical neurons.

Combining two-photon imaging and real-time optogenetic stimulation we train mice to activate a single neuron in motor cortex (M1), while continuous feedback of its activity level is provided by proportionally stimulating somatosensory cortex. We found that this artificial signal is necessary to rapidly learn to increase the conditioned activity, detect correct performance and maintain the learned behavior. Population imaging in M1 revealed that learning-related activity changes are observed in the conditioned cell only, which highlights the functional potential of individual neurons in the neocortex. Our experiments demonstrate the capacity of animals to use an artificially-induced cortical channel in a behaviorally relevant way and reveal the remarkable speed and specificity at which this can occur.

How should we think about motor skill? John Krakauer, Johns Hopkins, Baltimore, USA

M

otor skill is a vague term that is used interchangeably with motor adaptation, motor learning, and motor ability. Here a framework will be provided that attempts to define motor skill more carefully and dissect it into components. Experimental results will be shown that hopefully fit into this framework and illuminate a way forward.

NIN-KNAW Summer School 2017 | 17

Behavioral state modulation of associative learning in mouse cerebellum Megan Carey, Champalimaud Centre for the Unknown, Lisbon, Portugal

D

elay eyeblink conditioning is a relatively simple form of cerebellum-dependent associative learning. Recent work has demonstrated, however, that neither the learned behavior nor its underlying neural circuitry are as simple as once thought.

I will present our recent findings that locomotor activity modulates delay eyeblink conditioning through mechanisms that act on the mossy fiber pathway within the cerebellar cortex. These results suggest a novel role for behavioral state modulation in associative learning and provide a potential mechanism through which engaging in movement can improve an individual’s ability to learn.

The Cerebellum, Sensitive Periods, and Autism Sam S.-H. Wang, Princeton University, New Jersey, USA

A

lthough the cerebellum is mainly known as a sensory/motor structure, its influence and connections reach many regions known for cognitive function, emotion, and reward. I will present evidence for the idea that the cerebellum acts during sensitive periods to shape the developing brain. This hypothesis can explain a wide range of observations in autism, and may illuminate how the brain’s wiring is shaped by early-life sensory experience.

18 | NIN-KNAW Summer School 2017

Learning and relearning movement Amy Bastian, Johns Hopkins University in Baltimore, Maryland

H

uman motor learning depends on a suite of brain mechanisms that are driven by different signals and operate on timescales ranging from minutes to years. Understanding these processes requires identifying how new movement patterns are normally acquired, retained, and generalized, as well as the effects of distinct brain lesions. The lecture focuses on normal and abnormal motor learning and how we can use this information to improve rehabilitation for individuals with neurological damage.

Understanding the impact of FOXP2 gene disruptions in humans, mice and songbirds Simon Fisher, Donders Institute for Brain, Cognition and Behaviour, the Netherlands

I

n humans, rare mutations of the FOXP2 gene cause problems with learning to sequence mouth movements during speech, accompanied by wide-ranging deficits in language skills. FOXP2 encodes a regulatory protein which modulates expression of other genes. It is evolutionarily ancient, found in similar form in diverse vertebrates, where it helps regulate development and function of corresponding neuronal subpopulations. Intriguingly, expression of the gene is enriched in corticostriatal and corticocerebellar circuits known to be involved in sensorimotor integration and motor-skill learning. In my talk I will describe how in-depth studies of FOXP2 disruptions in humans, mice and songbirds are yielding novel insights at multiple levels, from molecules and cells to circuits and behaviour. NIN-KNAW Summer School 2017 | 19

C.U. ARIËNS KAPPERS MEDAL This year the Ariëns Kappers Medal is awarded to Rui Costa for his exceptional contribution to the field of neuroscience, in particular to his role in clarifying the importance of the basal ganglia in action sequencing.

T

he Ariëns Kappers Medal is a scientific honor named after the Dutch neurologist Cornelius Ubbo Ariëns Kappers. From 1909 to 1946, he was the first director of the Netherlands Central Institute for Brain Research (Nederlands Instituut voor Hersenonderzoek) —now the Netherlands Institute for Neuroscience. The medal is awarded by the Royal Netherlands Academy of Arts and Sciences on recommendation of the Netherlands Institute for Neuroscience to people who have made an outstanding contribution to neuroscience.

20 | NIN-KNAW Summer School 2017

Previous recipients of the C.U. Ariëns Kappers Medal: Pasko Rakic (1987) Anders Björklund (1988) Mortimer Mishkin (1989) Robert Y. Moore (1991) Dale Purves (1993) Joseph S. Takahashi (1995) Patricia S. Goldman-Rakic (1996) Dean H. Hamer (1999) Gerald M. Edelman (1999) Vilayanur Ramachandran (1999) Steven Rose (1999) Michael Gazzaniga (1999) Antonio Damasio (1999) Rudolf Nieuwenhuys (2000) Mark H. Tuszynski (2001) Dennis D.M. O’Leary (2003) Clifford B. Saper (2005) James W. Fawcett (2008) Frans B.M. De Waal (2009) Marcus Raichle (2010) György Buzsáki (2014) NIN-KNAW Summer School 2017 | 21

ABSTRACTS

Cerebellum-dependent locomotor adaptation on a split-belt treadmill in mice Dana Darmohray, Megan R. Carey Champalimaud Center for the Unknown, Lisbon, Portugal

L

ocomotor patterns are constantly adapted for changing environments but the neuralmechanisms underlying this basic form of learning are not well understood. Locomotor adaptation has been extensively studied in humans, using a split-belt treadmill that controls the speeds of opposite body sides independently (Reisman et al, 2005). Here, we describe split-belt adaptation in mice, using high-speed videography and 3D whole body tracking (Machado, Darmohray et al, 2015). We find that mice learn to adapt their locomotor patterns to achieve a more symmetrical gait in a way that is remarkably similar to human adaptation: measures of interlimb coordination adapt to become more symmetrical over several minutes of split-belt walking. A stable walking pattern is learned, in which each side of the body matches the speed of its belt, while interlimb and whole-body coordination improves to enable a more symmetrical gait. To investigate the neural substrate of this adaptation, we tested two cerebellar mutants (pcd and reeler) on the split-belt treadmill. While these ataxic mutants were able to respond appropriately to the changes in belt speed, they showed no evidence of learning. Interlimb coordination did not adapt over the course of split-belt walking, and there were no aftereffects observed post-adaptation. Thus, split-belt locomotor adaptation in mice, as in humans, appears to be a cerebellum-dependent form of learning. Our results suggest that the neural mechanisms of locomotor adaptation may be conserved across species, opening up the possibility of using genetic manipulations to dissect the neural circuit mechanisms underlying this form of motor learning. The combined action of these synaptic eligibility traces and the reward prediction error changes synapses to promote future actions leading to more reward. The resulting learning is powerful and can explain how learning occurs in deep neurobiological networks.

22 | NIN-KNAW Summer School 2017

Towards Automated Training of Mice in a Bimodal Attention Task Mahyar Moghimi, Leona Enke and Johannes J. Letzkus Max-Planck Institute for Brain Research, Frankfurt, Germany

T

raining time for difficult cognitive tasks is at present the major bottleneck for investigation of higher brain functions in mice. In addition, experimenter-controlled, manual training suffers from other downsides such as restricting water or food, imposing the time of day and duration of the training onto the animals, and causing handling stress for them. An automated training system that provides the animals with access to training around the clock while monitoring the behavior and allowing for individually tailored training schedules can mitigate the aforementioned problems. Here, we develop a social home cage system for housing up to 6 mice which provides RFID-controlled access to a two alternative forced choice task using audiovisual cues. Voluntary access to the training station led to animals performing many more trials per day compared to manual training, and receiving close to their ad libitum water intake. In preliminary experiments, animals entered the training station on average 10.3 times per day (range: 2 to 22), with an average time in the station of 16:54 minutes (ranging up to 5 hours). The number of entries during the light and dark period of the circadian cycle were not significantly different, while the duration of stays was slightly longer for the dark period. In addition, the animals maintained their initial weight during the experiment (9 days). This system thus provides a promising avenue for acquisition of complex tasks, while enhancing the welfare of the animals by keeping them in a social and more spacious cage, and without water restriction.

NIN-KNAW Summer School 2017 | 23

Anatomical evidence that multiple striatal regions influence motor cortex in the rat Sho Aoki1,2, Masakazu Igarashi1, Patrice Coulon3, Jeffery R. Wickens1, Tom J.H. Ruigrok2. 1Okinawa Institute of Science and Technology, Japan, 2Erasmus Medical Center, the Netherlands, 3Institut de Neurosciences de la Timone, France

S

triatum receives monosynaptic cortical input and sends multisynaptic output back to the cortex. Dorsolateral striatum (DLS) receives motor cortical input but dorsomedial (DMS) and ventral striatum (VS) do not. However, accumulated evidence suggests the involvement of DMS and VS in action such as goal-directed and Pavlovian approach behaviors. Those findings raise the possibility of output from DMS and VS to motor cortex, despite an absence of motor cortical input. We address this possibility by retrograde transneuronal tracing using rabies virus (RABV). RABV was co-injected with an anterograde monosynaptic tracer (cholera toxin b-subunit, CTb) into the rat motor cortex, which resulted in third-order RABV+ striatal neurons and CTb+ cortico-striatal fibers. DLS showed both RABV and CTb labeling, suggesting a closed-loop between DLS and motor cortex. Surprisingly, a few clusters of RABV+ neurons were observed in DMS, VS, and the tail of striatum. These areas were devoid of CTb labeling but made output connections to motor cortex, suggesting an open-loop structure. In another set of rats, injection of RABV/CTb into prefrontal cortex resulted in RABV labeling in DMS and VS but not in DLS. Since these RABV-labeled regions receive prefrontal input, this suggests a closed-loop structure. These results demonstrate co-existence of closed- and open-loop structures between the striatum and motor cortex, but a closed-loop structure for the striatum and prefrontal cortex. The present study provides an anatomical basis for understanding how the striatum can access motor cortex. The animals maintained their initial weight during the experiment (9 days). This system thus provides a promising avenue for acquisition of complex tasks, while enhancing the welfare of the animals by keeping them in a social and more spacious cage, and without water restriction. 24 | NIN-KNAW Summer School 2017

Impact of inhibitory and excitatory inputs on subthreshold oscillations in the inferior olive Sebastian Loyola1, Tycho Hoogland1,2, Mario Negrello2, Hugo Hoedemaker1, Chris de Zeeuw1,2. 1Netherlands Institute for Neuroscience, Amstedam, the Netherlands, 2 Erasmus Medical Center, Rotterdam, the Netherlands

O

ne of the hallmarks of neurons in the inferior olive (IO) nucleus is the presence of subthreshold oscillations (STOs), which are promoted and synchronized over the network through dendro-dendritic gap junctions. Dendritic elements of olivary neurons are contacted by excitatory afferents, originating in the mesodiencephalic junction (MDJ); and inhibitory afferents from the cerebellar nuclei (CN), (De Zeeuw et al, 1990). This unusual configuration could have important functional implications (Segev and Parnas, 1988), considering that the timing of excitatory and inhibitory inputs can affect the phase and amplitude of STOs and thus the probability to elicit complex spikes. Recent studies have shown that activation of IO NMDA receptors (Turececk et al, 2014) and selective stimulation of inhibitory nucleo-olivary inputs (Lefler et al, 2014) modulates STOs. Here we performed whole cell patch clamp recordings in sagittal slices of the IO. By combining transgenic lines, electrical stimulation and spectrally separable opsins (Klapoetke et al, 2014) we characterized the synaptic responses of these inputs, and their phase responses in oscillating neurons. Excitatory IO inputs evoked tri-phasic synaptic responses consisting of a depolarization, hyperpolarization and rebound depolarization, which persisted in the presence of the GABA receptor (GABAR) blocker picrotoxin. Nucleo-olivary inhibition evoked a GABAR-dependent hyperpolarization followed by a rebound with comparable latency to the rebound evoked by excitatory input. Both inputs affected the oscillation phase, while inhibition was more effective than excitatory input in phase resetting. In some cases, suppression of oscillations was observed when inhibition was followed by excitation in a narrow time window. This suggests that timing of excitatory and inhibitory inputs can be important in gating IO output. NIN-KNAW Summer School 2017 | 25

Pupil responses are indicators of value-based decision-making Joanne C. Van Slooten, Sara Jahfari, Tomas Knapen en Jan Theeuwes Department of Cognitive Psychology, Vrije Universiteit Amsterdam

W

hat can we learn about how people reach and evaluate value-based decisions by looking at their pupil responses? To answer this question, we investigated the mapping between the cognitive processes underlying value-based decision-making and fluctuations in pupil dilation (under constant illumination). In the present study, we used a probabilistic selection task (Frank, 2004) and fit participants’ choices with a Bayesian hierarchical reinforcement learning model (Frank et al., 2007; Jahfari & Theeuwes, 2016). This allowed us to quantify learning and choice strategies (sensitivity to losses and gains, and explore/exploit tendency) on an individual basis. Furthermore, this quantification delivered single-trial estimates of variables such as value expectation and its violations (reward prediction errors). Prior to choice, we observed that a dilatory pupil response was predicted by the value of the chosen stimulus, but not the value of the stimulus that was not chosen. Furthermore, the peak amplitude of this expected value-related pupil response was positively modulated in individuals that employed a more exploitatory choice strategy. At the time of feedback (after choice) we observed a biphasic pupil response, where early dilation was negatively correlated with the expected value of the chosen stimulus, while late constriction correlated positively with signed reward prediction errors. These findings show strong, specific relationships between per-trial pupil responses and reinforcement learning variables that describe value-based learning and decision-making. Our findings potentially benefit research on Parkinson’s disease, as pupil responses might be used to assess therapeutic efficacy by providing non-invasive access to the affected reinforcement learning process.

26 | NIN-KNAW Summer School 2017

Adaptation of Whisker Movements Requires Cerebellar Potentiation Vincenzo Romano1, Licia De Propris1,2, Laurens Bosman1, Pascal Warnaar1, Michiel ten Brinke1, Sander Lindeman1, Chiheng Ju1, Arthiha Velauthapillai1, Jochen Spanke1, Mario Negrello1, Egidio D’Angelo2,3 and Chris De Zeeuw1,4, 1Erasmus Medical Center, Rotterdam, the Netherlands, 2University of Pavia, Italy, 3Brain Connectivity Center, Pavia, Italy, 4Netherlands Institute for Neuroscience, Amsterdam, the Netherlands

T

he ability to rapidly adapt movement patterns while exploring the world is critical for survival, yet how it comes about is unclear. We developed a sensory-driven whisker movement adaptation protocol and show with trial-by-trial analyses that long-term enhancements and accelerations of whisker behaviour require increased and accelerated simple spike activity of Purkinje cells. Blocking potentiation specifically in Purkinje cells prevents both adaptation of whisker movements and concomitant changes in simple spike patterns. The ability to express sensory-induced simple spike plasticity is set by the temporal windows in which complex spike activity is low. Our data indicate that enhancement of simple spike activity guides long-term storage of novel whisker motor patterns during learning by presenting instruction signals to the cerebellar nuclei, whereas synchronized complex spike activity orchestrates acute stereotypic behaviour, during which the simple spikes merely relay efference copy signals consistent with those needed for a fully trained internal model.

NIN-KNAW Summer School 2017 | 27

In vivo whole-cell recordings of cerebellar nuclei neurons in awake head-fixed mice during Pavlovian eyeblink conditioning Robin Broersen1,2, Cathrin Canto1, Chris de Zeeuw1,2. 1Netherlands Institute for Neuroscience, Amstedam, the Netherlands, 2Erasmus Medical Center Rotterdam, the Netherlands

T

he cerebellar nuclei form a central part of the cerebellar circuit, integrating multiple types of input, while they provide the output to downstream extra-cerebellar areas. Although the extracellular electrophysiological responses of cerebellar nuclei neurons have been studied during natural behaviour, insight in their subthreshold membrane potential dynamics during cerebellar-dependent learning and behaviour remains largely lacking. The goal of the study was to differentiate between the various inputs that converge onto cerebellar nuclei neurons during Pavlovlian eyeblink conditioning. Therefore we characterized the strength and timing of the various inputs during this task. We obtained in vivo whole-cell recordings from anterior interpositus neurons in awake head-fixed mice, while presenting conditioned and unconditioned stimuli. We found that both types of stimuli evoked strong changes in membrane potential and that a subset of neurons showed distinctive subthreshold activity patterns that correlated with both learned and spontaneous eyelid movements. As such our results show unequivocally for the first time how cerebellar nuclei neurons integrate information such that they lead to a well-timed motor output.

28 | NIN-KNAW Summer School 2017

Low intensity repetitive transcranial magnetic stimulation modulates skilled motor learning in adult mice Alexander Tang1,2, William Bennett3+, Claire Hadrill3, Jessica Collins3, Barbora Fulopova3,Karen Wills3, Rohan Puri3, Michael Garry3, Mark Hinder3, Jeffery Summers3,4, Alison J Canty3 and Jennifer Rodger1, 1 University of Western Australia, Perth, Australia, 2OIST, Okinawa, Japan, 3University of Tasmania, Hobart, Australia, 4 Univeristy of Liverpool, Liverpool, United Kingdom

R

epetitive transcranial magnetic stimulation (rTMS) is commonly used to alter motor learning in clinical and non-clinical populations. Clinically, rTMS is delivered to targeted regions of the cortex at high intensities (>1T). We have previously shown that even at low intensities, rTMS induces structural and molecular plasticity in the rodent cortex. To determine whether low intensity rTMS (LI-rTMS) alters behavioural plasticity, daily intermittent theta burst LI-rTMS (120mT) or sham was delivered as a priming or consolidating stimulus to mice completing 10 consecutive days of skilled reaching training. Relative to sham, priming LI-rTMS (before each training session), increased baseline skill accuracy (~8%) but did not alter the rate of increase in accuracy over time. In contrast, consolidating LI-rTMS (after each training session), resulted in a small increase in the rate of increase in accuracy (~1.3%) over time but did not alter baseline skill accuracy. Changes in behaviour with LI-rTMS were not accompanied with long lasting changes in brain-derived neurotrophic factor (BDNF) expression or in the expression of plasticity markers at excitatory and inhibitory synapses for either priming or consolidation groups. These results suggest that LI- rTMS can alter specific aspects of skilled motor learning in a manner dependent on the timing of intervention.

NIN-KNAW Summer School 2017 | 29

Function of Neuregulin3-ErbB4 signalling in the GABA circuitry development Catarina Osório1,2,4, Sandra Pascual3, Cristina García-Frigola3, Beatriz Rico1,2 and Oscar Marín1,2 1Centre for Developmental Neurobiology, Institute of Psychiatry, Psychology and Neuroscience, King’s College London, London SE1 1UL, UK 2MRC Centre for Neurodevelopmental Disorders, King’s College London, London SE1 1UL, UK 3Instituto de Neurociencias, Consejo Superior de Investigaciones Cientı´ficas & Universidad Miguel Herna´ ndez, Sant Joan d’Alacant 03550, Spain 4Current address: Department of Neuroscience, Erasmus MC, Rotterdam, The Netherlands

I

n schizophrenia, alteration of specific brain circuits contributes for the disease. It has been reported that the stability between excitatory and inhibitory synapses is altered and GABAergic inhibition mediated by fast-spiking interneurons is compromised in patients. Moreover, genetic studies have shown a strong association between genes encoding for NRG1, NRG3, their receptor ERBB4 and schizophrenia. ErbB4 deletion in vivo reduces the wiring of parvalbumin interneurons by decreasing the number of excitatory synapses and a specific fraction of GABAergic synapses. Besides, overexpression of type-III-Nrg1 significant increases the formation of inhibitory synapses into pyramidal neurons. Since, Nrg3 is exclusively expressed in the developing brain there is a strong possibility that Nrg3-ErbB4 signalling may control the balance between excitatory and inhibitory synapses within the cortex. We hypothesize that Nrg3 might regulate synapse development in the brain. To address this question we performed gain-of-function and loss-of-function experiments to explore the involvement of the Nrg3-ErbB4 signalling in the formation of GABAergic and glutamatergic synapses. By analyzing the effects that reduced Nrg3 expression has in synapse formation and its molecular mechanisms we will understand for the first time the physiological relevance of Nrg3 in forming brain circuitries and neuronal excitability. The study of genes linked to schizophrenia is crucial to comprehend how the disease starts and progresses. Unravelling the molecular and cellular mechanisms underlying psychiatric disorders is necessary to develop future therapeutic interventions. 30 | NIN-KNAW Summer School 2017

NIN-KNAW Summer School 2017 | 31

DIRECTIONS

Singel Kerk

T

he Singel church is located near the tram (lines 1, 2, and 5) stop Koningsplein on the S ingel canal. It takes 12 minutes to get there from the Piet Hein hotel by tram (for line 1 take the stop Eerste Constantijn Huygenstraat, for lines 2 and 5 the stop Van Baerlestraat next to the Conservatorium hotel towards Centraal Station. When traveling from the Rho hotel take lines 1,2 and 5 at the stop Dam towards Koningsplein.

Speakers dinner

A

boat will pick up the speakers in front of the Krijtberg church which is just next to the Singel church. We will depart from the Singel kerk at 18:50.

Tolhuis (party venue)

T

he Tolhuis is located across the IJ river and can be reached the ferries located behind Centraal Station in the direction of Buiksloterweg. The Tolhuis is located just behind Café de Pont after you get off the ferry.

32 | NIN-KNAW Summer School 2017

Tolhuis (Party)

Rho Hotel Singelkerk

Piet Hein Hotel

NIN-KNAW Summer School 2017 | 33

NOTES

34 | NIN-KNAW Summer School 2017

NOTES

NIN-KNAW Summer School 2017 | 35

NOTES

36 | NIN-KNAW Summer School 2017

NOTES

NIN-KNAW Summer School 2017 | 37

SEE YOU SOON!