O'Mara / Tsanov The Connected Hippocampus

1. Auflage 2015
ISBN: 978-0-444-63550-1
Verlag: Elsevier Science & Techn.
Format: EPUB
Kopierschutz: Adobe DRM (»Systemvoraussetzungen)

E-Book, Englisch, Band Volume 219, 292 Seiten

Reihe: Progress in Brain Research

This volume of Progress in Brain Research focuses on the Connected Hippocampus. - This well-established international series examines major areas of basic and clinical research within neuroscience, as well as emerging subfields

Autoren/Hrsg.

O'Mara, Shane

Tsanov, Marian

Leseproben

Chapter 1 If I had a million neurons
Potential tests of cortico-hippocampal theories
Michael E. Hasselmo1    Department of Psychological and Brain Sciences, Center for Memory and Brain, Center for Systems Neuroscience, Graduate Program for Neuroscience, Boston University, Boston, MA, USA
1 Corresponding author: Tel.: [phone]; Fax: [phone] email address: [e-mail] Abstract
Considerable excitement surrounds new initiatives to develop techniques for simultaneous recording of large populations of neurons in cortical structures. This chapter focuses on the potential value of large-scale simultaneous recording for advancing research on current issues in the function of cortical circuits, including the interaction of the hippocampus with cortical and subcortical structures. The review describes specific research questions that could be answered using large-scale population recording, including questions about the circuit dynamics underlying coding of dimensions of space and time for episodic memory, the role of GABAergic and cholinergic innervation from the medial septum, the functional role of spatial representations coded by grid cells, boundary cells, head direction cells, and place cells, and the fact that many models require cells coding movement direction. Keywords Entorhinal cortex Stellate cells Medial septum Time coding Spatial coding Oscillatory interference Population recording 1 Introduction
The title of this chapter has a number of inspirations. The title was partly inspired by a song entitled “If I had a million dollars” by the Canadian band Barenaked Ladies, who humorously sing about the things they would do with a million dollars. This inspiration explains the title, which is not referring to the author having only a million neurons in his own brain, but to the usefulness of data from a million individual neurons recorded simultaneously in a behaving animal. This inspiration also explains the ambitious focus on a million neurons. Obviously, research can benefit tremendously from techniques for recording up to a thousand neurons (Dombeck et al., 2010; Gee et al., 2014; Ghosh et al., 2011; Heys et al., 2014; Sheffield and Dombeck, 2014), and further benefits will also arise from recording tens of thousands of neurons or hundreds of thousands of neurons. The scientific inspiration for the title comes as a response to a surprising comment that I have heard from other scientists over the years. This comment takes different forms, but the common gist is that recordings of thousands or millions of neurons would not be any more useful than data from current technology. I find this comment surprising because it seems obvious how expanding the numbers of neurons would be useful. But I have heard the comment multiple times, even from researchers who were instrumental in developing techniques for the current state of the art for multiple single-unit recording in behaving animals. So I want to take the opportunity to answer the question in the context of my own area of research. This chapter is also inspired by the announcement of the federal BRAIN initiative (Brain Research through Advancing Innovative Neurotechnology). One component of this initiative proposes support for recording of activity in large populations of neurons, showing that many scientists recognize the importance of this type of data. But I think the field can benefit from explicit examples of questions that can be answered if we had large populations of neurons in a well-structured data set obtained from an awake, behaving rat with well-described behavior. Answering this question not only supports the idea of funding innovative neurotechnology but also provides a framework for presenting some of the interesting current questions in the field. As long as I am moving beyond current technology in terms of the number of recorded neurons, I will also assume additional highly desirable features about the data. I will assume that the spiking activity of neurons can be observed at a high temporal resolution, such as that obtained with multiple single-unit recording. This contrasts with the slower time course of activation data obtained from current techniques for calcium imaging in large populations of neurons. I will assume the data are recorded simultaneously over at least 10 min in an awake, behaving rat actively moving around its environment. I will assume the data include tracking the head direction and movement direction of the behaving rat in space and time. I will assume that we can record in multiple different anatomical regions, and, in some cases, that we can identify the individual molecular identity of the neurons in the population. I will not initially make any assumptions about knowledge of the connectivity of the neurons, though connectome data would be useful when coupled with data on physiology and molecular identity of neurons and the behavior of the animal. 2 Cortical Coding of Space
If I had data from a million neurons, one top priority would be to analyze how grid cells and place cells are generated. Fundamental questions about the nature of spatial representations in the cortex would be answered through an understanding of the mechanisms of generation of the spatial firing patterns of grid cells. Extensive data from multiple interacting brain regions should be able to elucidate the mechanism of generation of grid cells, and I think it is useful to consider the steps that could be taken with such extensive data. The following sections focus on different aspects of this fundamental question, including the possible rate coding of movement direction, the possible phase coding of movement direction and speed, and the coding of sensory cues and boundaries. The Nobel prize in physiology or medicine in 2014 acknowledged the importance of grid cells and place cells by recognizing O’Keefe for the discovery of place cells (O’Keefe, 1976; O’Keefe and Dostrovsky, 1971) and May-Britt and Moser for the discovery of grid cells (Fyhn et al., 2004; Hafting et al., 2005; Moser and Moser, 2008). Initially, grid cells were proposed as a mechanism for driving place cells (McNaughton et al., 2006; Solstad et al., 2006), but recent data showing loss of grid cells with inactivation of the hippocampus suggest that place cells might be driving grid cells (Bonnevie et al., 2013). In either case, understanding the generation of one of these phenomena is important to understanding the other. The highly regular pattern of grid cell firing gives a sense that they can be accounted for by elegant theoretical principles, and numerous published models address the mechanism of grid cell generation. Grid cell models can be grouped into categories based on some of their component features. One category of models uses attractor dynamics to generate the characteristic firing pattern of grid cells (Bonnevie et al., 2013; Burak and Fiete, 2009; Bush and Burgess, 2014; Couey et al., 2013; Fuhs and Touretzky, 2006; Guanella et al., 2007; McNaughton et al., 2006). Most of the attractor models use circularly symmetric inhibitory connectivity within a large population of grid cells to generate competition between grid cells coding nearby locations. This results in a pattern of neural activity across the population that matches the characteristic hexagonal array of grid cell firing fields (also described as falling on the vertices of tightly packed equilateral triangles). Large-scale recording of cells particularly during first entry to a familiar environment would allow testing of whether the population dynamics appear to settle into an attractor state or whether individual neurons independently code location. As noted by the models, the shared orientation and spacing of the firing fields of grid cells within individual modules (Barry et al., 2007; Stensola et al., 2012) and the shared shifts of firing fields with environmental manipulations (Barry et al., 2007; Stensola et al., 2012; Yoon et al., 2013) already support the existence of attractor dynamics. However, generating a grid-like pattern across a population is not sufficient for modeling individual grid cells. Replicating the changes in firing of an individual grid cell over time requires that the grid-like pattern in the population needs to be shifted in proportion to the behavioral movement of the animal, that is, in proportion to its running velocity. To generate this movement, most attractor models of grid cells explicitly cite a role for experimental data on conjunctive grid-by-head-direction cells (Sargolini et al., 2006). In attractor models of grid cells (Burak and Fiete, 2009; Couey et al., 2013; McNaughton et al., 2006), these grid-by-head-direction cells are proposed to drive adjacent neurons within the population based on the movement of the animal. However, there is a fundamental problem in using grid-by-head-direction cells to represent the movement direction of an animal, as described in Section 2.1. A similar problem occurs for oscillatory interference models of grid cells (Burgess, 2008; Burgess et al., 2007; Hasselmo, 2008) that also require velocity as an input. Data show that the movement direction coding required by these models cannot be provided by cells coding head direction. 2.1 Coding of...