Theoretical Framework - Principles of Cortical Self-Organization
The mammalian brain is a highly sophisticated self-organizing system, which learns by transforming its experience into changes in circuitry that improve the animalís chances of survival. The same principles that guide day to day learning also allow the cortex to compensate for damage either to peripheral sensory structures or within the central nervous system. Although studies of long-term potentiation and depression have demonstrated that plasticity mechanisms are dependent on correlation-based rules, we still do not understand the principles that govern how sensory experience alters the distributed responses of thousands of cortical neurons in a behaviorally useful manner.
The anatomy of cortical connections strongly suggests that both bottom-up and top-down information shape perception and guide cortical plasticity. The sensory saltation illusion provides a robust example of the degree to which our experiences are shaped by attention and expectation (Nature, 1995). Other experiments from Merzenich and colleagues have established that attention is required for sensory input to drive reorganization of cortical maps. These results suggest that attention gates cortical plasticity mechanisms, allowing correlation-based rules to operate preferentially on stimuli which are relevant to the animal. It has been proposed that the brain uses ascending neuromodulatory projections, such as the central cholinergic system, to differentiate important stimuli from among the tens of thousands of behaviorally irrelevant stimuli encountered each day. I have developed a simple and robust plasticity paradigm using electrical activation of the cholinergic nucleus basalis and confirmed that release of acetylcholine paired with sensory stimuli is sufficient to generate enduring reorganizations of cortical circuitry.
This powerful new paradigm will serve as the basis for a series of experiments exploring the basic principles of plasticity that shape the cortical representation of stimulus features with the explicit goal of assembling these principles into a functional general theory of cortical self-organization.
Nucleus basalis (NB) neurons located in the basal forebrain provide almost all of the cholinergic input to the cortex. In my experiments, activation of NB, via a chronic stimulating electrode, is repeatedly paired with the presentation of an auditory stimulus to adult rats that are awake and unrestrained. After four weeks of such pairing, a detailed map of the response properties of primary auditory cortex neurons is reconstructed from up to one hundred microelectrode penetrations. The reorganizations that result are among the largest ever recorded in primary sensory cortex. Importantly, the plasticity observed is specific to the stimulus paired with NB stimulation. For example, when 9 kHz tones are paired, the region of the A1 map representing this frequency is expanded as neurons that previously responded to other frequencies shift their responses toward 9 kHz (Science, 1998). In contrast, stimuli presented without NB activation do not result in cortical reorganization. These results confirm the hypothesis that the NB functions to demark significant stimuli allowing cortical plasticity mechanisms to operate specifically on important events.
My goal is to use NB activation to investigate the principles of self-organization that specify how connection strengths and network dynamics are modified to result in plasticity that is behaviorally useful. Stimulation of NB provides a means of enabling cortical plasticity, and the auditory system allows a freedom of stimulus generation that will facilitate the investigation of the rules that guide plasticity in response to stimuli located within a continuous multi-dimensional feature space.
The first principle I investigated relates changes in receptive field size to specific qualities of the stimulus paired with acetylcholine release. Recanzone and colleagues showed that in monkeys cortical receptive field size decreases after practicing a task requiring discrimination of location on the receptor surface (cochlea or skin), and increases following training on a task requiring detection of changes in stimulus modulation rate. In my simplified preparation, this differential plasticity can be mimicked, without behavioral training, by changing the statistics of the sensory input paired with NB stimulation. Receptive field sizes are increased when amplitude modulated stimuli are paired with acetylcholine release and decreased when different tone frequencies are paired. By pairing six different types of stimuli, I have shown that receptive fields are altered as a continuous function of spatial variability and temporal modulation of stimuli. These results demonstrate that simple rules operate in the cortex to generate useful changes in circuitry based on the statistics of sensory stimuli marked by NB activity.
My second class of experiments examined representational plasticity of time-varying stimulus features. The maximum following rate of cortical neurons can be significantly increased after repeated pairing of NB activation with stimuli modulated at 15 Hz, and significantly decreased after pairing with 5 Hz stimuli. Interestingly, my first attempt to generate temporal plasticity, by pairing 9 kHz tones modulated at 15 Hz, failed. Although this procedure resulted in a large reorganization of the cortical map of frequency, the maximum following rate was unaltered. In contrast, when the tone frequency was randomized while maintaining the 15 Hz modulation rate, dramatic plasticity of A1 temporal responses occurred without any map reorganization. Thus, variability of one feature can profoundly impact plasticity of another.
Determining which features of a stimulus are behaviorally important is a difficult problem that has been largely ignored in the plasticity literature. In tasks involving simple tonal stimuli, it seems obvious that experience with 9 kHz should increase the 9 kHz region of the map, but how does the cortex know that tone frequency is important and not duration, intensity, modulation rate, bandwidth, or any other stimulus feature? It would not only be inefficient to simultaneously adjust the cortical tuning for every stimulus feature, it would create representations that did not generalize well. There is no evidence that primary sensory cortex has specific information about task goals, so the cortex probably uses information contained in the input itself to make an educated guess about how to improve performance.
Variability shapes behavioral generalization functions in humans and animals. My results demonstrate that input variability can serve as an important cue for the cortex about which feature(s) of the stimulus contain information. This result further substantiates the hypothesis that NB activity marks important stimuli and allows simple cortical rules to improve the representation of features likely to be useful based on the statistics of input.
These studies represent my initial investigations of the principles of cortical self-organization using NB stimulation. I have three major goals for my continuing research. The first is to develop a general theory of cortical plasticity by extending my investigations of the plasticity rules that operate in relation to simple features for which auditory cortex neurons are tuned. These features include duration, intensity, bandwidth, FM direction and rate. I will also examine how plasticity rules operate on conjunctions of these features, such as harmonic relationships and sequenced stimuli. Finally, I will determine how well the principles investigated with simple stimuli apply to complex spatiotemporal stimuli, such as human speech. I believe that this incremental approach, exploiting both the power of cholinergic modulation and the flexibility of auditory stimuli, will yield a more complete understanding of cortical plasticity.
My second goal is to relate cortical plasticity to changes in behavior. To date it has not been possible to demonstrate that cortical plasticity is sufficient to improve performance. Previous studies have correlated cortical plasticity with behavior, but could not establish causality. An important advantage of my model is that plasticity can be generated independently of behavior. I will use NB stimulation to generate cortical reorganizations and quantify the consequences on behavioral performance. Initial experiments will focus on improving behavior by exaggerating the distinction between the internal representations of features required for task performance (e.g. 15 vs. 30 Hz AM discrimination). Such preparation may speed learning by facilitating relevant distinctions. Subsequently, I will use NB stimulation to degrade the distinction between representations and explore how inappropriate plasticity can affect performance. These studies should clarify the functional principles relating cortical plasticity to behavioral performance.
The third goal of my experiments will be to use these principles to develop models to study the role of plasticity in the genesis and remeditation of CNS pathology. Determining how plasticity contributes to the stability and instability of cortical representations will provide insight into a number of disorders in which aberrant plasticity processes have been implicated, such as tinnitus, epilepsy, Alzheimerís disease, and dyslexia. I will also develop animal models to test the feasibility of using cholinergic modulation to accelerate functional recovery from head-trauma and stroke. Studies of cortical plasticity have already proven useful in designing treatment strategies for a number of neurological disorders, and a more complete understanding will facilitate the application of neuroscience principles in clinical settings.