Lectures

Barry Richmond

"Prospects for understanding information processing in the brain"

It is clear that no one neuron, or even a few neurons can give rise to functionality or flexibility that is seen with large-brained animals. Thus, if there is to be a clear and comprehensive understanding of brain function, it is essential to learn how the responses of single neurons are combined giving rise to higher functions in the brain.

The study of single neuronal responses has been both exciting and frustrating. It is exciting in that we, as a field, have become better and more sophisticated in determining selectivity, tuning and receptive fields of neurons. In the study of neuronal codes we have learned that spike trains have stochastic components riding on top of slower, more deterministic fluctuations of excitation and inhibition. In both experiments and theory we see that these slow fluctuations or ‘waves’ recur in some relatively stereotyped manner, but the spike trains themselves often fluctuate tremendously from one trial to the next, althought in some instances, when the external drive is strong, some spike trains can be remarkably reproducible. All of these responses are quite satisfying when studying functionality early (early sensory, especially vision), or late (motor) in processing.

When studying higher functions, functions depending upon flexibility and decisions, however, the results are less satisfying. A large component of the frustration is the lack of information about how neuronal responses are combined to give rise to high-level cognitive processes. With single neuron recording, sometimes we even end up with conundrum that we can activate single neurons in some brain area in a particular task, but when the area is removed, the animals continue to perform the task without difficulty.

For imaging and field potential studies, the assumption needs to be made that correlated activity across neurons is a reasonable cartoon for describing total signal. If, however, the most interesting components of neuronal activity are the parts that are independent of their neighbors, the cartoon, while helping localization of function, will give little help in understanding how information processing occurs.

How do we go on? A major limitation is that there is little knowledge from whole vertebrates about how information is transformed across synaptic layers. When responses from many incoming neurons impinge on targets it is not clear what the outputs will be. Without this knowledge it is difficult to interpret what information processing has occurred.

An approach that is heavily pursued, but frustrating, is recording from more than one neuron (sometimes quite a few) to determine what the relations are among the responses. This does not guarantee that these neurons are connected, nor even that they are part of a single ensemble. The results are often used to try to infer connectional structure and ensemble membership. Despite the difficulty for interpreting the results of these multiple single neuronal recordings, this remains an important approach.

There are two approaches on the horizon to aid in studying ensemble memberships and information flow from input to output. They both use imaging. In the first approach, Ca++ imaging offers the opportunity to watch small to moderate populations close to the
cortical surface. With this approach, some of the coordination of activity across the population might be observable. In the second approach, we have been using Mn++ imaging to carry out tract tracing in individual monkeys. With this approach the Mn++ transport density can be taken as surrogate for the connection density from one brain region to its efferent target. The density profile of the Mn++ in the efferent target sites will allow targeting of recording sites and interpretation of simultaneous recording data based on the connection density profile.

Another strategy for examining the transformation from input to output is to interfere with a particular mechanism, e.g., a single receptor. This would be particular interesting in brain regions where the transmitter and/or receptor can be associated with a single pathway, or a particular function. There are several efforts under way to do this using technology from molecular biology. These would offer great opportunities for moving beyond identifying correlations to identifying causal mechanisms. We have, for example, had some success with DNA plasmid expression vectors targeting D2 and NMDA receptors. These seem to be temporary, with a lifetime of several weeks. We are getting data that suggest this approach can be generalized to other targets. We, and others, are also investigating other methods using viral vectors, which cause permanent incorporation of the genetic material. Some of these are built to express all of the time, and others are engineered so that expression can be contingent on particular agents, either local or systemic, depending on the system.

Finally, new behavioral methods are needed. Currently studying high level behavior in monkeys has had a severe penalty in training time. We have been pursuing behavioral?@approaches that seem more natural in that the animals seem to learn the tasks more easily. It appears that several aspects of behavior such as categorization, extra-dimensional?@shifting, and rule switching can be learned quickly if the circumstances are appropriate.

Daeyeol Lee

"Prefrontal cortex and economic decision making"

Economic theories provide normative solutions to various problems of decision making, including optimal decision-making strategies during social interactions. Although such formal theories often fail to predict the real behaviors of people and other animals, they still provide a useful vantage point necessary to understand the cognitive constraints and neurobiological mechanisms of decision making. Our experiments have investigated how information about the animal's previous choices and their outcomes during interactive decision making are processed and represented in the fronto-parietal network and medial frontal cortex. The results showed remarkable similarities in the types and time courses of signals represented in various cortical areas. In the dorsolateral prefrontal cortex, we have also found that the expected outcomes of alternative choices are encoded according to their temporally discounted values, suggesting that the information about the magnitude and immediacy of reward might be combined in the prefrontal cortex. Finally, neurons in the medial frontal cortex often encoded the signals related to the gains and losses of the animal's choices. These neurophysiological investigations in non-human primates provide valuable information to understand the neural basis of economic decision making.

Emmanuel Procyk

"Cognitive control and performance monitoring: modulation of prefrontal cortico-cortical interactions"

Rapid adaptation of behavior during exploration and exploitation necessitates the coordination of different cognitive processes that allow the regulation, verification, and adjustment of performance. The mechanism by which neural networks compute information and organize during flexible behaviour is a core issue in cognitive neuroscience.

Adaptive behaviours particularly require the appropriate functioning of some prefrontal areas, the striatum, and aminergic systems. These structures independently participate in working memory and performance monitoring. Brain imaging has shown that dorsolateral prefrontal and the anterior cingulate cortices (DLPF and ACC) are co-activated during numerous cognitive tasks. Single-unit recordings showed that the neural coding of intended actions and expected outcomes are modulated during trial and error learning in both cortical areas, and that both reveal shift in activity at the transition between exploration and exploitation periods. The dynamics and timing of these modulations are comparable in both areas, although the basic information processed in each is very different. Whereas tonic DLPF activity clearly shows properties that have been associated with cognitive control, ACC activity reveals expectation of outcome, encoding of task values, and computations of prediction errors. In both areas, activity patterns have properties similar to those observed in mesencephalic structures that play a key role in learning, but global changes in activity also fit perfectly with theoretical accounts of cortical noradrenergic modulations.

The real-time interactions between these structures as well as the origin of their modulation are yet to be described. I will present here past and current experiments as well as projects that explore the dynamical interactions between prefrontal cortical areas and their modulation during behavioral adaptation.

Okihide Hikosaka

"Neural mechanisms of reward-oriented behavior: Beyond dopamine neurons"

Obtaining a reward facilitates your action that has led to the reward. Dopamine is thought to be a key factor in such reinforcement learning, and a likely site of the dopamine action is the basal ganglia. Using a visually guided saccade task with positionally biased reward outcomes we have provided several lines of evidence supporting this hypothesis. A likely scheme is that a burst of spikes in dopamine neurons in response to the visual stimulus predicting a large reward facilitates the signal transmission in the caudate nucleus which then leads to a selective disinhibition of saccadic neurons in the superior colliculus.

However, a robust effect of biased rewards was the suppression of saccades when a small reward was expected rather than the facilitation of saccades when a large reward was expected. In other words, obtaining a small (or smaller-than-expected) reward suppresses your action. While this fact has been well recognized by psychologists studying operant conditioning and has been implemented in more recent reinforcement learning theories, neuroscientific research has largely been focused on the facilitatory effects of large rewards, not the suppressive effects of small rewards.  
One possibility is that dopamine neurons are instrumental in the suppressive effects as well. They decrease or cease firing in response to a sensory stimulus predicting a small reward. What then induces the inhibition of dopamine neurons? We recently found evidence that a small epithalamic structure called the lateral habenula inhibits dopamine neurons when a small reward is expected. A majority of lateral habenula neurons were excited by a visual stimulus that predicted a small reward and were inhibited by a stimulus that predicted a large reward. This activity pattern was opposite to that of dopamine neurons. Furthermore, weak electrical stimulation inside the lateral habenula elicited strong inhibitions in dopamine neurons. These results suggest that the lateral habenula guides the basal ganglia to suppress less rewarding saccadic eye movements by inhibiting dopamine neurons.

This discovery may provide a clue beyond dopamine neurons to understanding the neural mechanisms of reward-contingent learning. It also raises many questions. Is the inhibition of dopamine neurons, which may be caused by the lateral habenula, sufficient to suppress the action leading to a small reward? Is the lateral habenula simply a relay station of signals that eventually reach dopamine neurons? Or, is there a functional dichotomy such that lateral habenula neurons encode aversive signals while dopamine neurons encode rewarding signals? The lateral habenula is located at the crossroads of the limbic system, basal ganglia, and monoaminergic (dopaminergic and serotonergic) system. Future studies disentangling the complex crossroads may provide answers to these questions.

Gary Aston-Jones

"The cortex in context: Locus coeruleus, optimal performance, and maximal utility "

Optimal performance on a specific task requires selective attention to filter out extraneous stimuli and behaviors. It is also important, however, to disengage from the current task and find a new behavioral goal if motivation or the stimulus context changes. This tension between selective attention and the search for a new task in an altered context has been characterized as the exploitation-exploration tradeoff, and is critical for adaptive behavior. Our recent studies indicate that this tradeoff is at the heart of noradrenergic locus coeruleus (LC) system function, and indicate a role for LC in maximizing behavioral utility. LC neurons exhibit two patterns or modes of activity in monkeys performing a signal detection task: a phasic mode in which these cells are transiently activated selectively following target cues at latencies preceding behavioral responses, and a tonic mode when these cells have elevated tonic levels of discharge but no phasic responses. These modes correspond to excellent task performance and selective attention (phasic) vs. poorer performance due to apparent scanning and high behavioral flexibility (tonic). Computational modeling reveals how modulated electrotonic coupling among LC neurons could produce these two modes of LC activity, and how these modes may produce the corresponding differences in task performance. Additional recordings during a two-alternative forced choice task reveal that phasic responses of LC neurons are closely linked to behavioral responding rather than to stimulus presentation. Moreover, LC responses do not occur for stimuli that the animal foveates but does not respond to, nor for lever responses that occur in the absence of task stimuli. These findings lead us to propose that phasic LC responses are driven by decision outcome, and that they facilitate behaviors and processes engaged by the decision. We hypothesize that the phasic mode of LC activity promotes selective behavioral responding to task-defined stimuli (selective attention), whereas the tonic mode produces a broader sampling of stimuli in the environment (search for new task, high behavioral flexibility). Both of these modes serve to increase the utility of behavior: The phasic mode increases utility on a short time-scale by facilitating selective responding in the task at hand, whereas the tonic mode increases long time-scale utility by promoting the identification of a new task that is adaptive in a changed context.

Christoph Kayser/ Christopher Petkov

"Imaging neuronal assemblies and connectivity"

Individual techniques often prove unsatisfactory for understanding brain mechanisms at multiple scales. The combination of high-resolution functional and structural imaging with methods from electrophysiology, histology, and neurochemistry promises great insights into neural organization and processing that could not be achieved with any individual technique alone. For example, combinations of technologies allow insights into sensory and perceptual networks, long-range neural interactions, dynamic connectivity as well as learning-related neurochemical changes and plasticity. This talk will touch upon some of our work to address such issues, such as: (a) spatially resolved fMRI of visual, auditory and multisensory processes in the brains of primates, (b) combined physiology and MRI for examining the relationship of electrical activity and BOLD activation, (c) the study of in vivo connectivity using electrical microstimulation combined with fMRI, or manganese-enhanced MRI tractography, and (d) molecular imaging based on targeted and “smart” (Ca+ or Ph sensitive) MR contrast agents. We will discuss insights obtained using such combined approaches using both the visual and auditory systems as model, and how they can help us to uncover ways in which our brain merges information from multiple senses into a coherent percept.

Jason Kerr

"Combining two-photon imaging with electrophysiology in vivo: from the single neuron to the network"

The appeal of in-vivo imaging to any neuroscientist is not hard to understand: neurons are almost impossible to isolate with their complex interactions with surrounding tissue remaining intact. Moreover, their interaction with thousands of interconnected cells leads to the complex network dynamics that most likely underlies neural computation. In-vivo imaging allows the study of both form and function in a reasonably intact preparation, with cellular and sub-cellular resolution over time periods of milliseconds to periods spanning months. Using two-photon imaging in combination with electrophysiological techniques, the limits of what can be achieved in-vivo have been pushed into a terrain previously only accessible in-vitro. This expansion has been due to advances in physical imaging technology and to the design of molecular contrast agents. More recently, two-photon imaging has allowed the measurement of ongoing activity in small ensembles of neurons with single cell and single action-potential resolution. I will describe the physical principles of two-photon imaging and the methods that allow for measurement of activity at the single cell and neuronal population level. Then I will highlight recent work from our lab that uses a combination of imaging and electrophysiological techniques to image neuronal population activity during sensory input.

Kazuto Kobayashi

"Application of immunotoxin cell targeting for primate brain research"

Mechanisms underlying brain functions are dependent on interactions among a variety of neurons that are consisted of a complex brain network. An understanding of these functions has progressed with development of techniques that manipulate the activity of specific neural types with particular identities in the network. Our research group previously developed a transgenic mouse technology for conditional ablation of specific neuronal types based on the specificity of recombinant immunotoxins. By using this immunotoxin cell targeting, it is possible to eliminate the neurons that are genetically engineered to express a target molecule of the immunotoxins in transgenic mice. This technology makes it possible to eliminate specific neuronal types that can be distinguished by different genetic markers from the neural circuitry. In addition, recent studies propose new molecular genetic technologies that permit conditional suppression of the activity of specific neuronal types.

Application of the molecular genetic techniques that manipulate the activity of specific neuronal types is useful for the progress of the study on the neural mechanisms underlying brain functions of the monkey. For this purpose, one powerful technique is the use of recombinant virus vectors that retrogradely transport from the injection sites in the brain through the axonal fibers. These vectors make it possible to deliver the transgene into neurons that innervate the injected brain regions. Introduction of agents that influence the neural activity into one of these upstream regions enables to manipulate the activity of specific neural pathways that innervate the target regions.  Recently, our research group developed a modified lentivirus vector that possesses an efficient retrograde transport activity. This vector system will provide a useful approach to elucidate the mechanisms underlying operation and regulation of the neural network in the monkey brain.

Naotaka Fujii

"Living in reality: Social neurophysiology"

In conventional electrophysiological approaches, the experimental paradigms were optimized to address a specific question. All of relevant parameters used in the experiments were controlled strictly. Therefore, the experimental environment was highly artificial and events in the environment are probabilistically predictable. The methods worked for solving bottom-up questions, but were not adequate for solving top-down questions.

Social brain function is one of the unsolved questions that require top-down approach, because the function is implemented by combinations of multiple levels of networks from neural networks to social networks in and between the brains so that reality has to be introduced to the experimental design to tackle the issue. It is commonly agreed that society rules our behaviors. One action which was valid few second ago is not guaranteed to be valid now. Even though there is no visible sign of rules in the space, it undoubtedly exists and forces us to follow. The social rules are unstable and continuously changing depending on behaviors of people who are sharing the space so that updating the internal representation of the social environment is essential for selecting socially correct behavior. For revealing a mechanism of the social brain function, conventional approach is not sufficient because we can not control the social environment. Although we can not control the social environment, we can monitor and record the environment. To reveal the social brain function, I developed Multi-dimensional Recording (MDR) system. MDR was designed to record simultaneously any biological and environmental information as much as possible. It consists of chronic multi-electrodes recording system, motion capture system and multi-video recording system. The method allows subjects making natural behavior that may reflect social rules defined by the environment. While recording behaviors and environment, neuronal activity is recorded from multiple areas of the brain of multiple subjects. After recording these data, we can reconstruct multitiered network from single cell activity to real small society on the computer. Using MDR, we can ask questions about social brain functions by interacting with subjects and by modulating the environment.

In this lecture, I will introduce a detail of MDR technique and show preliminary results about how neurons recorded from multiple sites and multiple subjects were contributing to the social selection of behaviors.

Tadashi Isa

"How the CNS overcomes its partial damage"

Neuro-rehabilitation is based on the concept that training recruits the remaining neuronal systems to compensate for partial injury of the CNS.  However, the neuronal basis of such take-over mechanism is poorly understood.  To assess the neuronal mechanism of functional compensation, we need a longitudinal study using an animal model with defined lesion of the particular neuronal system and apply quantitative behavioral evaluation.  Recently, it has been found that dexterous finger movements can be restored within a few weeks to 1-3 months after making lesion of the direct cortico-motoneuronal (CM) connection via the lateral corticospinal tract (l-CST) at the border between C4 and C5 segments of the spinal cord, which is rostral to the segments where motoneurons of hand muscles are located (“C4/C5 l-CST lesion”) in macaque monkeys.  This result has suggested that an indirect corticomotoneuronal pathways, mediated by subcortical or spinal interneuronal systems, can mediate commands for the control of precision grip from the motor cortex to digit motoneurons in primates (3).  By combining brain imaging with PET and pharmacological reversible inactivation techniques we show that recovery involves the bilateral primary motor cortex (M1) at the early recovery stage, and contralesional M1 (co-M1) and bilateral premotor cortex at the late stage.  Such change in activation pattern of frontal motor related areas represents the adaptive strategy for functional compensation after spinal cord injury.