Invertebrate Neurobiology: Polymorphic neural networks

EVE MARDER EVE MARDER

INVERTEBRATE NEUROBIOLOGY

INVERTEBRATE NEUROBIOLOGY

Polymorphic neural networks Recent work on small invertebrate nervous systems provides new insights into the way in which neurons are organized into functional networks to generate behavior. Brains consist of thousands, millions or even billions of neurons that are massively interconnected by synaptic contacts. What is the relationship between the anatomically defined networks in the brain and the functional circuits that produce behavior? Getting [1,2] introduced the term 'polymorphic network' to indicate that anatomically defined networks can have multiple functional modes of operation. Getting highlighted the distinction between anatomical connectivity - the pattern of monosynaptic connections among neurons - and functional connectivity - the effect that neurons may have on one another via a myriad of polysynaptic pathways (Fig. 1). Recent work on several different invertebrate systems illustrates the utility of this distinction in attempts to understand how activity patterns that produce meaningful behavior are transmitted through massively interconnected neural networks.

Synaptic strength modulation during behavior Getting's now classical studies were aimed at understanding the organization of the central-patterngenerating network producing the escape swim movement of the mollusc Tritonia. In Tritonia, as in many animals, the central-pattern-generating network is a group of interneurons that rhythmically drives motor neurons activated in rhythmic movements. In a new study of this network, Katz and colleagues [3] have found that the strengths of several of the synaptic connections within the 'central-pattern-generating' circuitry and between this circuitry and the motor neurons are increased by the prior stimulation of one of the neurons - the dorsal swim interneuron C (DSI-C) - within the central-pattern-generating network, a phenomenon known as heterosynaptic facilitation (Fig. 2).

Fig. 1. Two neurons, A and B, are connected functionally by two pathways: a monosynaptic, excitatory pathway, and a polysynaptic, inhibitory pathway via interneuron . The relative strength of the two functional pathways plays a major role in the animal's switch between two discrete behaviors. The traces show the activity in neurons A and B when the polysynaptic (left) and monosynaptic (right) pathways are active. (Modified from [11 .) 752

© Current Biology 1994, Vol 4 No 8

DISPATCH The Tritonia DSIs contain serotonin [3,4], and serotonin mimics the effects of DSI-C stimulation on synapse strength. That serotonin and serotonergic neurons can produce heterosynaptic facilitation has previously been demonstrated in Aplysia [5]. But the new study [3] differs from most earlier work demonstrating modulation of central-pattern-generating networks, because in the Tritonia system one of the elements of the pattern-generating circuit is responsible for heterosynaptic facilitation. The authors argue that, during ongoing rhythmic behavior, the circuit 'intrinsically' up-modulates its own synaptic strengths, so that activity in the network participates in modifications of synaptic strength important for network function. The same conclusion comes from a recent demonstration that modulatory projection neurons in the stomatogastric nervous system of crabs receive synaptic input from the pattern-generating neurons that they modulate [6]. Many computational models of the brain have been formulated in which learning is based on changes in synaptic strength. The work from Tritonia provides a beautiful example in which synaptic strength is also dynamically modulated during ongoing behavior. Moreover, it shows that if one wants to understand network function in terms of synaptic strengths among constituent neurons, it is necessary to determine these under the same modulatory conditions that prevail during the expression of the behavior. Good measurements of synaptic strengths are best made in the absence of ongoing synaptic activity. Unfortunately, measurements made on quiescent preparations under defined stimulus conditions may not adequately reflect the strengths of synapses during behavior. Multifunctional neurons If networks can function in several modes, it follows that individual neurons may participate in producing several behaviors. A recent paper by Wu et al. [7] deals with this and other issues that are now surfacing in work on small nervous systems. Wu et al. [7] used optical recordings to monitor most of the neurons of the Aplysia abdominal ganglion during several types of behavior, including the classical gill-withdrawal reflex, spontaneous gill movements and respiratory pumping. By monitoring almost all of the neurons in the ganglion, the authors were able to address a number of questions. First, how many neurons were activated during a given behavior? Second, how reproducible were the population and single-cell responses to a repeated stimulus? And third, how many neurons were activated in multiple behaviors? Not surprisingly, Wu et al. [7] found that a large number of neurons in the ganglion were activated during gill movements. Presumably these include the motor neurons directly responsible for the movement, the interneurons that control those motor neurons and other neurons that are 'kept apprised' of the activity in the gill system. Wu et al. [7] further found that many of the same neurons were activated during all three of the behaviors that involve gill movements. Without knowing what fraction of the

Fig. 2. Heterosynaptic facilitation of a synapse - from cerebral neuron 2 (C2) to dorsal swim interneuron B (DSI-B) - within the Tritonia swim system by one of the neurons (DSI-C) of the central pattern generator. (Modified from [31.) monitored population are motor neurons, it is somewhat difficult to assess how many of the interneurons that control these gill movements during sensory-evoked, spontaneous and respiratory associated gill movements are shared. Nonetheless, it is likely that a significant population of interneurons are employed during all three movements. Wu et al. [7] were also able to assess the trialto-trial reproducibility of the single-cell and population responses to the same stimulus. Although there is a great deal of similarity from trial to trial, if one looks at the responses of individual neurons, or at the number of neurons that are recruited, one sees considerable variation. How do we assess the significance of variability of this sort, in Aplysia or monkey brain recordings? This is an important issue for understanding both simple and complex nervous systems, and the relatively simple Aplysia nervous system is a good one in which to investigate the issue. As many investigators are now employing multiple electrodes to record from large and complex nervous systems during behavior [8], it becomes crucial to understand the reproducibility and reliability of singlecell and population responses during repeated executions of the same behavior. A priori, it is difficult to know, for any nervous system and any behavior, how precise the neural codes that govern behavior must be for that behavior to be coherent. The Aplysia system, consisting of relatively few neurons, poses the challenge we face of extracting meaning from many channels of data during

753

754

Current Biology 1994, Vol 4 No 8 behavior. When is variation 'noise' and when is it a significant component contributing to behavioral flexibility? The paper by Wu et al. [7] reinforces on a larger scale the inference from recent work on the stomatogastric nervous system that many neurons participate in the generation of multiple behaviors [9]. In an elegant new study, Meyrand et al. [10] have explored the global reorganization of the stomatogastric neural networks by the rhythmic discharge of a pair of identified interneurons. These neurons, the PS cells, synapse onto a large number of neurons previously thought to be part of several separate networks. However, Meyrand et al. [10] have now shown that sensory inputs that activate the PS neurons result in the temporary formation of a new functional network, consisting of neurons 'borrowed' from four different pattern generating circuits. The study of Meyrand et al. [10], together with previous work on the stomatogastric nervous system, shows that in different modulatory environments the same neurons are combined into different functional networks [9,10]. Indeed, these studies indicate that the functional neuronal circuitry is a meaningful concept only if the modulatory environment is specified. The extreme statement of this position [10] is that the modulatory environmnent 'constructs' the functional network for the behavior. Thus, Getting's concept of a 'polymorphic' network captures well the flexibility of networks that are subject to modulation at multiple sites.

References: 1. Getting PA, Dekin MS: Tritonia swimming: a model system for integration within rhythmic motor systems. In Model Neural Networks and Behavior. Edited by Selverston Al. New York: Plenum Press; 1985:3-20. 2. Getting PA: Emerging principles governing the operation of neural networks. Annu Rev Neurosci 1989, 12:185-204. 3. Katz PS, Getting PA, Frost WN: Dynamic neuromodulation of synaptic strength intrinsic to a central pattern generator circuit. Nature 1994, 367:729-731. 4. McClellan AD, Brown GD, Getting PA: Modulation of swimming in Tritonia: excitatory and inhibitory effects of serotonin. J Comp PhysiolA 1994, 174:257-266. 5. Mackey SL, Kandel ER, Hawkins RD: Identified serotonergic neurons LCB1 and RCB1 in the cerebral ganglia of Aplysia produce presynaptic facilitation of siphon sensory neurons. J Neurosci 1989, 9:4227-4235. 6. Nusbaum MP, Weimann JM, Golowasch J, Marder E: Presynaptic control of modulatory fibers by their neural network targets. J Neurosci 1992, 12:2706-2714. 7. Wu J,Cohen LB, Falk CX: Neuronal activity during different behaviors in Aplysia: a distributed organization? Science 1994, 263:820-822. 8. Wilson MA, McNaughton B: Dynamics of the hippocampal ensemble code for space. Science 1993, 261:1055-1058. 9. Marder E, Weimann JM: Modulatory control of multiple task processing in the stomatogastric nervous system. In Neurobiology of Motor Programme Selection. Edited by Kien J, McCrohan C, Winlow B. New York: Pergamon Press; 1992:3-19. 10. Meyrand P, Simmers J, Moulins M: Dynamic construction of a neural network from multiple pattern generators in the lobster stomatogastric nervous system. J Neurosci 1994, 14:630-644. 11. Getting PA, Dekin MS: Mechanisms of pattern generation underlying swimming in Tritonia. Gating of a central pattern generator. I Neurophysiol 1985, 53:466-480.

Eve Marder, Volen Center for Complex Systems and Biology Department, Brandeis University, Waltham, Massachusetts 02254, USA.