My work builds on the electrophysiological analyses of molluscan nervous systems that my university has specialised in for more than three decades (Benjamin & Rose, 1979; Benjamin & Kemenes, 2008). The use of relatively simple invertebrate systems as models for understanding brains in general goes back to the characterization of the action potential (Hodgkin & Huxley, 1939; 1952), through to the molecular mechanisms of synaptic memory formation (Kandel, 2001) and the study of dynamic neural circuits (Selverston & Ayers, 2006; Briggman & Kristan, 2008; Elliot & Suswein, 2002). Whereas the vertebrate brain, for the most part, remains impossibly complex, invertebrate nervous systems are small (20.000 neurons in the molluscan brain - 100 billion neurons in the human brain) and have large, relatively accessible neurons that survive and remain functional in vitro. Moreover, the location, connectivity and function of invertebrate neurons are virtually identical among members of a species, much unlike the individualistic vertebrate brain.
"In the mammalian brain, the precise relationship between the dynamics of individual neurons and functional networks remains extremely complex. The main reason for this is a lack of knowledge of the detailed cell-to-cell connectivity patterns and the biophysical properties of the individual neurons and their synaptic connections. Attempts to understand brain dynamics by large-scale modeling have been attempted frequently but without knowledge of the detailed parameters such as the number and kind of synaptic connections, the results have been disappointing (e.g., Foldiak and Young 1995). Moreover, each physiological synapse may result from numerous anatomical synapses that may have complex spatial geometries in neuronal branches. The numerically less complex microcircuits of invertebrates have neurons and synapses which are identifiable from animal to animal. Therefore, a much more detailed understanding of neural circuit dynamics is possible." - Selverston & Ayers (2006)
Unlike many other sciences, neuroscience lacks what Gerald Edelman calls a "global brain theory", a conceptual and mathematical foundation for understanding brain and behaviour. In this context, the aim of invertebrate neuroscience is the development of principles and computer models that describe the operations of biological neural networks in general. The tacit assumption here is that vertebrate brains work like their spineless cousins, and though this has proved largely accurate on the molecular and cellular level, it has not been conclusively established on the poorly understood level of neuronal networks, although there are tantalizing similarities (Yuste et al., 2005; Grillner 2006).
Of particular interest, at least in our lab, is the generation, by these networks, of the complex patterns of electrical activation that underlie the adaptive behaviour of living organisms. Invertebrates show a striking degree of adaptive variability in the way they behave, which through processes of sensory feedback, neuromodulation and reinforcement learning allows them to behave intelligently in a constantly changing environment (Horn et al., 2004; Selverston & Ayers, 2006), an ability currently unavailable to computer software and robotics. We are also concerned with reinforcement learning on longer time-scales, and with the competitive and cooperative interactions between neural networks that allow organisms to select different behaviours in different situations.
We study these processes primarily through a detailed analysis of a network of neurons that control the feeding musculature in the freshwater gastropod Lymnaea stagnalis. The network comprises some 500 neurons in two almost identical ganglia (shown below, black dots are extracellular electrodes).
The network occasionally generates a complex pattern of activation (below), which in vivo drives sequential muscle contractions that extend a tongue-like structure, scoop food into the mouth cavity and swallow it (a video of this behaviour can be found here). Remarkably, the neural network continues to generate the appropriate sequential activation of motor neurons for hours even when the brain is isolated and monitored with microelectrodes. Each cycle of activity is then called a "fictive" feeding cycle. Below are five such cycles, recorded on 14 of the extracellular electrodes in the figure above (note the cycle-to-cycle variability).
A number of methods have been developed for experimental initiation of feeding cycles, including intracellular activation of cerebral command-neurons, dopamine perfusion, and perfusion of sensory tissues with gustatory stimuli.
I'll return to this work in future posts, because there is an enormous number of questions and avenues for research here, and I know some of my readers work on similar projects. There are also a number of direct links between this research and reinforcement learning, motivation, dopamine, the iPlant, and the overarching goal of developing a qualitative, easily communicable understanding of the living human brain.