A temporal basis for predic icting the sensory consequences of - - PowerPoint PPT Presentation

Ann Ann Ken ennedy, Gr Greg eg Wayn yne, Patric ick Kaif aifosh, Kar arin ina Al Alviñ iña, La Larry ry F F Abb Abbott & & Na Nathanie iel l B B Sa Sawtell ll (20 (2014) )

A temporal basis for predic icting the sensory consequences of motor commands in in an ele lectric ic fis ish. . Nat. . Neurosci. 19 (3): 416-22.

NEUROSCIENCE JOURNAL CLUB FOR UNDERGRADUATES BIONB 4110, Cornell University

Presented by May-Kate Skoulos and Carl Hopkins

Ann Kennedy

Columbia University 2009 - 2014 (expected)

• Ph.D. Neurobiology and Behavior
• Representation and learning in cerebellum-like

structures

• Advisor: Larry Abbott

Johns Hopkins University 2005 - 2008

• B.S. Biomedical Engineering, Computational Biology

focus

• B.A. Biology
• Minor in Applied Math

Finalist, Intel International Science & Engineering Fair 2005

Greg Wayne

• PhD candidate at Columbia
• Research Interests:Artificial

Intelligence, Neuroscience, Machine Learning, Control Systems Engineering, Computational Neuroscience, Control Theory, Reinforcement Learning, Systems Neuroscience,Theoretical Neuroscience, Computer Science, Cognitive Science, and Artificial Neural Networks

Karina Alviña

• Undergraduate

Patrick Kaifosh

Research interests Neural circuit mechanisms of memory. Education/Training: Ph.D. Candidate Neurobiology and Behavior Program, Columbia University Thesis Advisors: Attila Losonczy, M.D./Ph.D.; L.F. Abbott, Ph.D. Bachelors in 2010 B.Sc., Mathematics & Physics, University of Toronto

• Dr. Larry F Abbott
• LARRY ABBOTT, PH.D.
• William Bloor Professor, Neuroscience, Physiology &

Cellular Biophysics, Biological Sciences Co-Director, Center for Theoretical Neuroscience Member, The Kavli Institute for Brain Science

• AREA OF RESEARCH
• Neurobiology of Learning and Memory,

Cognitive/Systems Neuroscience, Theoretical Neuroscience

• SPECIALIZATION
• Computational modeling and mathematical analysis of

neurons and neural networks.

• RESEARCH THEME
• COMPUTATIONAL AND MATHEMATICAL ANALYSIS OF

NEURONS AND NEURAL NETWORKS.

• Dr. Nathaniel B. Sawtell

Assistant Professor, Neuroscience Member, The Kavli Institute for Brain Science Background: PhD with Mark Bear at Brown Post doc with Curt Bell (Oregon) on e-fish SPECIALIZATION Functions of cerebellum-like structures and the cerebellum. Mechanisms through which past experience and sensorimotor context affect sensory processing RESEARCH THEME Unraveling the functions of neural circuits is a fundamental challenge for neuroscience.

Nate Sawtell, Karina Scalise ,Armen Enikolopov, Shobit Singla

Nature Neuroscience

The journal of choice for 3 out of 13 presentations in BIONB 4110 Impact factor = 15.2 in 2012 Nature Neuroscience is editorially independent, and its editors make their own decisions, independent of the other Nature

• journals. If a paper is rejected from one

Nature journal, the authors can use an

automated manuscript transfer service to submit the paper to another Nature journal via a link sent to them by the editor handling the manuscript. Authors should note that referees' comments (including any confidential comments to the editor) and identities are transferred to the editor of the second journal along with the manuscript.

Like the other Nature titles, Nature Neuroscience has no external editorial board. Instead, all editorial decisions are made by a team of full-time professional editors. For information on their research backgrounds and scientific interests, see About the Editors.

Feedforward control in the CNS is usually associated with motor systems. Can it also have a sensory function?

Wikipedia feedforward Motor command Motor output

Feedforward control is usually associated with motor

• systems. Can it also have a sensory function?

1) Feedforward control in motor systems takes place in the cerebellum. 2) Motor output is often calibrated by having a direct motor control signal accompanied by an indirect feed-forward prediction signal which is adjusted and refined in the cerebellum by supervised learning.

Could a sensory system be made better if the brain were able to make predictions about the sensory consequences of our actions and use them to anticipate and perhaps subtract expected sensations that are simply a consequence of our motor acts? Give an example where a human sensory system could be improved by feed-forward (predictive) control.

Introduction to this system

Weakly electric mormyrid fish emit brief EOD pulses for communication and active electrolocation.

Electric fish (Gnathonemus petersii) Family Mormyridae

electric organ

EOD pulse worm 0.5 ms Electrolocation (object) ELECTRORECEPTION: 1) locates objects (sensed as a change in local current density) 2) communicates with other e-fish 3) senses prey (uncontrolled electric fields from invertebrates and other living things) Current from EO EODs (compressed time scale) 1s E-field from EOD Passive electrolocation of prey electroreceptors communication

Introduction

Weakly electric mormyrid fish emit brief EOD pulses for communication and active electrolocation. However, the fish’s own EOD also affects passive electroreceptors tuned to detect external fields.

15

Mormyrid electric fish produce an ‘electro-motor’ command, and receive a electrosensory response as a consequence.

16

Bell, CC (1982) Ampullary receptor at rest (no EOD) Response to worm EOD Response to EOD nerve

brain

Introduction

Weakly electric mormyrid fish emit brief EOD pulses for communication and active electrolocation. However, the fish’s own EOD also affects passive electroreceptors tuned to detect external fields. Previous studies have shown that such interference, a ringing pattern

• f activation that may persist for as long as the interval between

EODs…….

Post stimulus time histogram responses of ampullary afferent to EODs

Spikes per bin for many EODs (arbitrary units) A ringing pattern of activation that may persist for as long as the interval between EODs…….

Introduction

Weakly electric mormyrid fish emit brief EOD pulses for communication and active electrolocation. However, the fish’s own EOD also affects passive electroreceptors tuned to detect external fields. Previous studies have shown that such interference, a ringing pattern

• f activation that may persist for as long as the interval between

EODs, is cancelled out in medium ganglion cells through the generation

• f motor corollary discharge responses that are temporally

specific negative images of the sensory consequences of the EOD.

20

Medium Ganglion Cell Plasticity

EOD command

Introduction

Weakly electric mormyrid fish emit brief EOD pulses for communication and active electrolocation. However, the fish’s own EOD also affects passive electroreceptors tuned to detect external fields. Previous studies have shown that such interference, a ringing pattern

• f activation that may persist for as long as the interval between

EODs, is cancelled out in medium ganglion cells through the generation

• f motor corollary discharge responses that are temporally

specific negative images of the sensory consequences of the EOD.

3) The paper talks about a corollary discharge of the electric organ discharge signal. What exactly is a corollary discharge? Can you site an example of a corollary discharge in a mammalian or avian nervous system?

http://www.urllabs.com/wad/ia/corollar/corollar.htm

The Basic Architecture of Mormyrid ELL

1. Bell, C. C., Han, V. and Sawtell, N. B. (2008). Cerebellum-like structures and their implications for cerebellar function. Annu Rev Neurosci 31, 1-24.

Neuroscientists have been attracted to the puzzle of the Cerebellum ever since Cajal. The orderly structure, the size of the cerebellum and the regularity of the neural elements suggest it might be possible to understand how it functions like a machine.

Ramón y Cajal 1894

Brains of vertebrates color coded by brain area. Cerebellum in orange

The main players in cerebellar function are: Granule cells: which receive input from mossy fibers, and send their outputs to Purkinje cells via parallel fibers Purkinje cells: the “Principal Cell” of the Cerebellum Climbing fibers: bring “error signals” from the inferior olive to ONE Purkinje cell.

Masao Ito 2008 (Nature Reviews- Neuroscience) Basic circuitry of cerebellum

28

29

ELL CEREBELLUM Medium ganglion cells Granule cells Golgi cells Unipolar brush cells Sensory afferents Mossy fibers Parallel fibers Molecular layer 2) The electric fish electro sensory lateral line lobe (ELL) is a cerebellum-like structure. Which of the structures mentioned in Fig. 1b or in the associated text are found in both the ELL and cerebellum, and what are the names typically used in cerebellum?

Recording from electrosensory lobe in medulla (equivalent of cochlear nuclei for auditory system)

silenced EOD (curarized)

Medium ganglion cells in ELL

4) What is a the benefit of being able to form a negative image of an expected sensory response locked to a motor act?

Granule cell to medium ganglion cell synapses undergo Anti ti-Hebbia ian pla lasti ticity: that is synaptic strength increases for each presynaptic action potential (i.e. non-associative potentiation), and decreased when a post synaptic action potential occurred shortly after a presynaptic action potential (associative depression)

Other spike-timing dependent learning rules

(A) Antisymmetric Hebbian learning rule consistent with Markram et al. (1997), Zhang et al. (1998) and Bi and Poo (1998). A second and later LTD component (dashed line) has been reported in Nishiyama et al.(2000). (B) (B) Antisymmetric Hebbian learning rule consistent with Feldman (2000). (C) (C) Symmetric Hebbian learning rule at the neuromuscular junction (Dan and Poo 1998). (D) (D) Anti-Hebbianlearning rule that is consistent with data presented in Bell et al. (1997). The associative LTP component (dashed) is not statistically significant in vitro, but has been

• bserved in vivo (Bell et al. 1993).

(E) (E) Symmetric anti-Hebbian learning rule (Egger et al. 1999). (F) Theoretical antisymmetric anti-Hebbian learning rule without non-associative potentiation

Firing pattern of granule cells

Command discharge PCA

Preeinential nucleus command

Extracellular records

• f Mossy fibers in EGp

Intracellular records from UBC and Golgi

Unipolar Brush Cells in Cerebellum

Cell body axon Discovered in the granule cell layer of the cerebellar

• cortex. Also in the cochlear

nuclei. Long suspected UBCs might involve creation of positive feedback loops that could delay or extend excitatory signals being input to granule cells. Recent data suggests long firing to constant depolarization under appropriate conditions (H current + TRP channel based).

Granule cells, Golgi cells and UBC’s in

Intracellular recording from UBC reveals mechanism

• f delayed and diverse responses in granule cells.

Biocytin filled UBC Rebound firing Same cell fires a late command discharge. Made more bursty by hyperpolarization Single UBC where hyperpolarization yielded trains of spikes Pause type UBC showed a natural

• hyperpolarize. Followed by bursting

Intracellular recordings of subthreshold corollary discharges from 170 granule cells in response to command, arranged by latency and type

Subthreshold responses of granule cells, fit and model

Intracellular recordings of subthreshold corollary discharges from 170 granule cells in response to command, arranged by latency and type

Subthreshold responses of granule cells, fit and model

Model granule cells, with combinations of inputs match both the firing pattern and the subthreshold responses of 4 recorded granule cells

The high quality of the fit suggestes that granule cell input can be understood as the sum of 1 – 3 excitatory inputs without the influence of inhibitory input from Golgi cells.

Granule cell to medium ganglion cell synapses undergo Anti ti-Hebbia ian pla lasti ticity: that is synaptic strength increases for each presynaptic action potential (i.e. non-associative potentiation), and decreased when a post synaptic action potential occurred shortly after a presynaptic action potential (associative depression)

Using the observed timing pattern of mossy fiber cell command related discharges, combined in one two or three inputs to granule cells, combined with the antihebbian learning rule, scaled up by 20,000 inputs, the model was able to cancel the responses to a large number of stimulus waveforms. B) Stimulus that produces a burst pause, generates a negative image after 100 trials. d) In the absence of UBCs, the rate of negative image formation is creatly slowed.

5) The authors of this paper were attempting to explain how spike inputs from parallel fibers onto the dendrites of medium ganglion cells might persist for more than 200 milliseconds following the EOD command – the typical time that negative images are formed. What is their conclusion for how this long, persistent discharge is sustained?

Summary and conclusions

3. Though input coding and plasticity are the critical elements for the functioning of many neural circuits, including other cerebellum-like circuits and the cerebellum itself, there are few cases in which these elements are understood so thoroughly. 4. The function of EGp circuitry demonstrated here closely parallels longstanding, but still untested, expansion- recoding schemes posited for the granular layer of the mammalian cerebellum. 5. Though the notion that motor corollary discharge signals could be used to predict and cancel the sensory consequences of an animal’s own behavior has a long history32,33, there are few cases in which such functions have been characterized at the level of neural circuits34.

• 32. Sperry, R.W. Neural basis of the spontaneous optokinetic response produced by visual inversion. J. Comp.
• Physiol. Psychol. 43, 482–489 (1950).
• 33. von Holst, E. & Mittelstaedt, H. The reafference principle. Naturwissenschaften 37, 464–476 (1950).
• 34. Crapse, T.B. & Sommer, M.A. Corollary discharge across the animal kingdom. Nat. Rev. Neurosci. 9, 587–600

(2008). 1) Using intracellular recordings and modeling of granule cells in mormyrid fish, the authors provided a relatively complete description of granule cell recoding, far more complete than that available in other systems. 2) The remarkably close agreement between recorded and model granule cells (Fig. 3) strongly suggests that the simple rules they used to transform mossy fiber inputs into granule cell responses, i.e., summation of randomly selected excitatory inputs, are essentially correct and complete.

BioNB424 Neuroethology 49

Campylomormyrus numenius (Boulenger)

Learning Goals

Outcome 1: Students will learn how to read, analyze, understand, and orally summarize a recent scientific paper in the subject areas of behavior and neuroscience. Outcome 2: Students will learn about the structure of a typical research paper in neuroscience or behavior. Outcome 3: Students will learn strategies for making effective oral presentations using slides and graphics and demonstrate creativity for engaging audience participation and discussion. Outcome 4: Students will learn about the human and social dimensions of scientific research by learning about the authors, the institutions at which they work, the journals where the work is published, and the funding sources for research. Outcome 5: Students will learn about electronic resources for searching and abstracting recently published papers in the scientific literature, and they will learn about various on-line support resources including encyclopedias, citation analysis services, review articles, news stories, and on-line resources. Outcome 6: Students will learn to work in groups.