WWW.DISSERTATION.XLIBX.INFO
FREE ELECTRONIC LIBRARY - Dissertations, online materials
 
<< HOME
CONTACTS



Pages:   || 2 |

«Aaron Gruber1,2, Peter Dayan3, Boris Gutkin3, and Sara Solla2,4 Biomedical Engineering1, Physiology2, Gatsby Computational Neuroscience Unit3 ...»

-- [ Page 1 ] --

Dopamine modulation in basal

ganglia-cortical network implements

saliency-based gating of working memory

Aaron Gruber1,2, Peter Dayan3, Boris Gutkin3, and Sara Solla2,4

Biomedical Engineering1, Physiology2, Gatsby Computational Neuroscience

Unit3 University College London,

and Physics and Astronomy4

London, UK

Northwestern University, Chicago, IL

{ dayan, boris }@gatsby.ucl.ac.uk

{ a-gruber1, solla }@northwestern.edu

Abstract

The sustained neural activity in prefrontal cortex (PFC) underlying working memory (WM) is under under control of dopamine (DA) both through direct modulatory effect on pyramidal neurons and also through the control of basal ganglia (BG) input to PFC.

We construct a network model of BG-PFC interaction to investigate the combined influence of DA on WM formation through both pathways. The cortical component is a line attractor which maintains memory of input through a persistent ’bump’ of activity localized to a subgroup of PFC units. DA increases the contrast of these units leading to increased robustness of the encoded quantity, similar to results of other models of cortical DA effects on WM. The activation of medium spiny neurons (MSN) of the BG provides an additional input to the cortex, which increases the robustness of the memory to noise and also gates access to working memory by triggering input driven switches of the bump location.

Our previously reported biophysically-grounded model of MSNs shows that these neurons become bistable in elevated dopamine, which accounts for both the suppressive and enhancing effects of DA. DA-induced bistability provides a novel mechanism for the network to selectively gate input to WM based on salience: inputs that elicit DA response are strongly encoded in WM through enhanced MSN activation, but subsequent inputs are blocked from switching the WM while DA is elevated through a suppression of MSN activity to these inputs. The combined actions of DA in the BG and PFC lead to WM that is selective for salient input and that has increased robustness to distraction.

1 Introduction Experimental evidence indicates that the maintenance of information in WM is mediated by persistent neural activity in the PFC. Modulation of this activity either through neuromodulation of cortical response properties or alteration of the input to PFC provides mechanisms for controlling WM. Many computational models of DA effects on WM focus on direct neuromodualtion in the cortex,which has been shown experimentally to influence WM. Unfortunately, the multifaceted effects of DA modulation on the electrophysiology of neurons are not fully understood, complicating the construction of computational models. A recent cellular model including many known effects of DA on ionic conductances indicates that DA modulation of pyramidal neurons causes the pattern of network activity at a fixed point attractor to become more robust to noise and input -driven switching of attractor states [7], consistent with the properties of spiking models investigating effects of DA on NMDA and AMPA-mediated transmission and is effects on fixed point attractors [2]. Interestingly, many different styles of models, form high level models designed to investigate human psycobehavioral data [?], to biophysically motivated cellular models [?] and intermediate models [?], broadly share a common phenomenology: DA reduces the efficacy of input for switching WM representation, and in many models, also enhances interaction between active units. Our model indicates that both input from BG and its modulation by DA provide additional functionality to WM.

In addition to modulatory inputs, both excitatory and inhibitory input to PFC also control WM formation. A prominent input to PFC comes from the BG. Although the BG has played a prominent role in models of motor control, this pathway has received very limited attention in models of WM. One group [3] has constructed network models for cognitive tasks that require working memory has ascribed a gating function to the BG in the absence of DA in which the BG disinhibits selective PFC units allowing activation of those units, an idea that has been long-standing in the motor control community. Our goal in this work is to investigate how a biologically-grounded model of BG gates information to PFC, and how the complex effect of DA on MSNs will affect this gating function. We choose to model the WM demand for the memory guided saccade task that is commonly used to study neural activity involved in WM. In this task, subjects must fixate a centrally located spot while a briefly illuminated visual target located in a circular position around a fixation point is presented. Following a delay period of a few seconds, subjects must saccade to the remembered target location. We use a ’bump’ network (line attractor) similar to those used by others [1, 4] to model the working memory of the target angle on the circle. An advantage of this task and model formulation is that the wealth of single unit data from animals performing this task can be used to construct the model and to test its validity. One intriguing finding is that the activity of striatal MSNs show a profound reward dependence [6]. We have shown in a previous model that this response modulation could be mediated by DA [8], providing a route for controlling WM formation in addition to direct modulatory effects in the PFC. We show here that the additional gating imparted by the BG and its modulation by DA work in a complementary manor with DA modulation in the cortex to gate in and enhance WM of salient information.





2 Model description

–  –  –

and noise: either all units have very low activity, or a subset of units participate in a ’bump’ of elevated activity shown by the filled dots in Fig 2A. This configuration is capable of encoding a continuous angle by assigning each unit a direction label and translating the bump on the ring of cortical units; the encoded angle is the corresponding labeled direction at the peak of the bump, and can be read out by computing the population vector. The action of DA in the cortex is modeled as increasing the slope of the logistic function of the cortical units, leading to a more narrow activity bump with a higher peak as shown by the open dots in Fig 2A. This leads to an increased stability of the bump location, and therefore encoded angle, as described next.

External excitatory input is capable of moving the bump. A quantitative measure of bump robustness against movement is obtained by providing uniform input to five units for a fixed duration and measuring the resulting bump displacement in the absence of noise. This procedure is illustrated in figure 2B. The activity of the cortical units is displayed in pseudo-color versus time to visualize network dynamics.

Current injection centered at θ† causes a drift of the bum initially centered at (or encoding?) θ0 towards the injection location. The bump ceases to move once the injection is terminated, and its location θ1 is read out 40 ms later. Bump displacement is dependent on the location of the injection relative to the bump location as shown in Fig 2C. The injection is capable of moving the bump completely to the injection location when the injection occurs close to θ0, however, the injection becomes less effective as the distance increases. The incomplete bump movement illustrated in Fig 2B represents the data point in Fig 2C marked by an aserisk.

The intensity and duration of injection affect the sensitivity profile, but the trends remain qualitatively similar. The increases in activation function slope in high DA decreases the sensitivity of the bump to current injection (Fig 2C, dotted line), hence increasing the robustness of the encoded quantity. DA also reduces the diffusion of the bump caused by noise as shown in Fig 3. The trajectories (inset) and variance of the population vector readout for 60 independent realizations are displayed in conditions of low (dashed line) and high (solid line) cortical DA. The increased robustness of the memory encoded by the bump in high DA is consistent with reported results obtained with other line and fixed point attractor models of DA effects on WM.

BG effects on cortex: increased memory robustness and gating actions We now investigate the effects of BG input on cortical bump stability in the absence of dopamine modulation. Tonic input from a single MSN to the cortical units is fig4.pdf fig3.pdf

–  –  –

able to anchor the location of the bump, leading to an increased robustness of the bump to noise induced diffusion as shown in Fig 4A, and to current injection as shown in Fig 4B. The tonic input to cortex has the effect of breaking the symmetry of the line attractor in the cortex and causing a single fixed point occurring at the location of maximal input to the cortex, which corresponds to the preferred direction of the active MSN (but the cortical net can still go to 0 activity, so its still bistable, but with the active state being a fixed point rather than line attractor).

The transition to a single fixed point reduces the maximal deviation of the bump due to current injection, and also has the property of pulling the bump back to the fixed point following a perturbation. Therefore θ1 is computed with short latency (40 ms) after injection offset so that the deviation is reflected in the plot in Fig 4B before reconvergence to the fixed point.

The output of the MSNs provides control over the encoded memory not only by enhancing robustness, but also by providing input that can assist the direct visual input to the cortex in triggering a switch in the attractor location when a visual stimulus is presented. The network activity resulting from the presentation of two different target locations is shown in Fig 5. The top plots show the angle encoded by a gaussian bump of activity in the input layer as a grey bar, superimposed with a thin black line indicating the population vector of the cortical layer. The middle and bottom plots show the activity of BG and PFC layers respectively. The initial input activates one MSN strongly and later two neighboring units, and causes activation of a group of PFC units. When the input is extinguished, the MSNs become inactive and the cortical layer relaxes to its characteristic bump state. In Fig 5A, the presentation of a second input stimulus does not activate MSN units and has only a minimal effect on the bump state. In Fig 5B, the second input activates MSNs and the resulting additional input to the cortex causes a switch of the bump state so that it encodes the new stimulus location. The MSNs are effectively gating the input to the cortex and can control access to working memory.

DA effects in BG: saliency-based gating We now explore the effects of DA on the properties of MSNs and the resulting effects on the gating function of BG. Many lines of evidence indicate that DA modulation has an impact on neuronal activity of MSNs. Data of particular relevance to our fig5.pdf Figure 5: Cortical input to PFC in not sufficient to switch attractor states (A).

Input driven MSN activation triggers switching (B), gating information into WM model comes from single unit recordings of MSNs in monkeys performing a memory guided saccade task. The responses of many MSNs to visual targets were of higher frequency and longer duration when the targets indicated reward rather than an audible tone for correct task performance. A minority of MSNs were suppressed in the reward condition. Visual stimuli indicating reward are known to trigger DA response. Our model of MSNs accounts for both types of response modulation through DA-induced bistability. We identified two consequences of DA-induced bistability that can have a significant impact on the response and potentially the function of MSNs: firstly, the threshold for reaching the active ’up’ state is increased by DA, but secondly, units that reach the up state have higher frequency activity that can be extended in duration due to hysteresis.

The dual enhancing/suppressing nature of DA modulation on MSNs has significant consequences on the network’s response to inputs, as shown in Fig 6. The top 4 panels show the input angle superimposed with the encoded angle, DA signal γ, BG ∗ activity, and PFC activity respectively. The sequence of inputs is θA, θB, θA, θB, ∗ where θA is a conditioned stimulus that triggers DA release. The MSNs activate following the first two inputs and allow the input to be encoded in working memory.

∗ The presentation of θA begins to activate the same set of MSNs as θA but the units become bistable allowing only one unit to become active, consequently switching the ∗ bump state to encode θA. Now however, the presentation of θB does not elicit MSN activation because the total excitatory input is below the threshold for reaching the ∗ up state while MSNs are bistable, and the bump state continues to encode θA. DA modulation in the BG has imparted salience selectivity to the gating function of fig6.pdf Figure 6: DA locks the activation of MSNs, preventing a switch to stimuli following DA release and preserving the conditioned stimulus in WM.

the BG by inducing bistability. In addition, the robustness of the bump during the delay period is enhanced by the combined effect of DA through both increasing the slope of cortical unit activation function and by sustaining MSN input during the delay (Fig 6, bottom panel; dashed line compared to solid line).

4 Conclusions

Our model represents one of the first efforts to link DA modulation in both cortex and BG and the resulting influence of BG input to cortex on WM formation.

Our model indicates that both input from BG and its modulation by DA complements the cortical effects of DA to provide more sophisticated control over WM and addresses some of the fundamental drawbacks inherent to line attractor models.



Pages:   || 2 |


Similar works:

«Models of Gemini Surfactants∗ Haim Diamant1 and David Andelman2 1 School of Chemistry and 2 School of Physics & Astronomy, Beverly and Raymond Sackler Faculty of Exact Sciences, Tel Aviv University, Tel Aviv 69978, Israel arXiv:cond-mat/0302184v1 [cond-mat.soft] 10 Feb 2003 I. INTRODUCTION Gemini surfactants are composed of two monomeric surfactant molecules linked by a spacer chain. They constitute a new class of amphiphilic molecules having its own distinct behavior. Since their first...»

«Few-Body Systems 0, 52–57 (2003) FewBody Systems c by Springer-Verlag 2003 Printed in Austria Quasilinear Approximation and WKB∗ R. Krivec1, V.B. Mandelzweig2,∗∗, F. Tabakin3 1 J. Stefan Institute, PO Box 3000, 1001 Ljubljana, Slovenia 2 Racah Institute of Physics, The Hebrew University, Jerusalem 91904, Israel 3 Department of Physics and Astronomy, University of Pittsburgh, Pittsburgh, PA 15260, USA Abstract. Quasilinear solutions of the radial Schr¨dinger equation for different...»

«Kochhar, Rajesh & Narlikar, Jayant (1995) Astronomy in India: A Perspective (New Delhi: Indian National Science Academy), Chap.1 1 Historical perspective India, as can be expected from an ancient culture, has a long astronomical tradition. The oldest astronomical text in India is the Vedanga Jyotisha (astronomy as part of the Vedas), one part of which is attributed to Lagadha. It is dated about 1400 BC on the basis of the statement in the book that the winter solstice took place at the star...»

«Website; www.tauranga-astro.150m.com Correspondence to Secretary; 14 Hazel Terrace, Tauranga Issue No. 26 – March 2010 The Moon, taken through Andrew Walker’s 12-inch Telescope in his backyard observatory. 29 Brown Street, Tauranga, P.O. Box 1147 Phone 07 578 0059, Fax 07 578 2692 A visit to the CERN Large Hadron Collider and other interesting highlights of my July 2009 tour of Europe By Ursula Macfarlane To treat myself for my 50th birthday I impulsively decided to go on a trip to Europe....»

«Astronomy Cast Episode 76: Lagrange Points Fraser Cain: Gravity is always pulling you down, but there are places in the solar system where gravity balances out. These are called Lagrange points and space agencies use them as stable places to put spacecraft. Nature is on to them and has already been using them for billions of years. Before we get on to it, let’s talk about pronunciation. I said it Lagr-ahunge points. Is that okay? Dr. Pamela Gay: I have heard it said Lagr-ahunge points,...»

«Vol. 1, N.º 49 (enero-marzo 2016) Organización Barrial Tupac Amaru en San Salvador de Jujuy: ¿Un movimiento social urbano? Fernanda Valeria Torres Universidad Nacional de La Plata/ Consejo Nacional de Investigaciones Científicas y Técnicas (Argentina) Resumen Este trabajo se propone revisitar la categoría Movimientos sociales urbanos enlazando con las concepciones posibles de ciudad que dicha categoría analítica puede asumir. Nos referimos por un lado, a las interpretaciones que...»

«Antonio Marín Márquez: Quejío, Los Palos, Herramientas, 1972-1977, La Cuadra de Sevilla [Investigación realizada para el proyecto A la calle me sali: teatro urbano, corralas de vecinos y polígonos de vivienda organizado por Plataforma Independiente de Estudios Flamencos Modernos y Contemporáneos-PIE.FMC y UNIA arteypensamiento en colaboración con el Centro de Documentación de las Artes Escénicas de Andalucía] Quejío, Los palos y Herramientas 1. El objeto Esta charla aborda los...»

«Laser-induced ultrafast electron emission from a field emission tip B Barwick1, C Corder, J Strohaber, N Chandler-Smith, C Uiterwaal and H Batelaan Department of Physics and Astronomy, University of Nebraska—Lincoln, 116 Brace Laboratory, PO Box 880111, Lincoln, Nebraska 68588-0111, USA E-mail: brettbarwick@gmail.com Abstract. We show that a field emission tip electron source that is triggered with a femtosecond laser pulse can generate electron pulses shorter than the laser pulse duration...»

«Site24x7: Key Mistakes in Data Center Operations Introduction In today’s connected society, the demands placed on data center professionals is at an all-time high due to the need for modern devices to always have connectivity. While investing in quality equipment is one way to avoid downtime, you can greatly improve the performance of your datacenter by working to minimize common ine ciencies within your company. Whether you are a C-Level executive or employee directly handling the operations...»

«SAPPORO: A way to turn your graphics cards into a GRAPE-6 Evghenii Gaburov∗,c,a,b, Stefan Harfsta,b, Simon Portegies Zwarta,b,c a Astronomical Institute “Anton Pannekoek”, University of Amsterdam b Section Computational Science, University of Amsterdam c Leiden Observatory, University of Leiden the Netherlands Abstract We present Sapporo, a library for performing high-precision gravitational N -body simulations on NVIDIA arXiv:0902.4463v2 [astro-ph.IM] 5 Mar 2009 Graphical Processing...»

«On-Sky Tests of an A/R Coated Silicon Grism on board NICS@TNG F. Vitali*a, V. Fogliettib, D. Lorenzettia, E. Ciancic, F. Ghinassid, A. Harutyunyand, S. Antoniuccia, C. Riverold, L. Riverold a INAF Osservatorio Astronomico di Roma, Via Frascati 33, 00040-I Monte Porzio Catone Italy b CNR Istituto di Struttura della Materia, Area della Ricerca Roma 1, Montelibretti, Via Salaria, Km. 29,300, 00016-I Monterotondo Italy c CNRIstituto per la Microelettronica e Microsistemi, Laboratorio MDM, Via C....»

«Sistema de Retiro del Servicio Civil (CSRS) Sistema de Retiro de Empleados Federales (FERS) ¿Está pensando en su retiro? Una visión general de las cosas más importantes que los empleados federales deben hacer cuando planeen su retiro. Oficina de Administración de Personal Federal (OPM) Servicio de Retiro y Seguro RI 83-11SP Revisado en octubre de 1997 La edición de octubre de 1996 aún está vigente. Información adicional sobre el retiro y todas las publicaciones de la Oficina de...»





 
<<  HOME   |    CONTACTS
2016 www.dissertation.xlibx.info - Dissertations, online materials

Materials of this site are available for review, all rights belong to their respective owners.
If you do not agree with the fact that your material is placed on this site, please, email us, we will within 1-2 business days delete him.