Extended dynamic clamp: controlling up to four neurons using a single desktop computer and interface

https://doi.org/10.1016/S0165-0270(01)00368-5Get rights and content

Abstract

The dynamic clamp protocol allows an experimenter to simulate the presence of membrane conductances in, and synaptic connections between, biological neurons. Existing protocols and commercial ADC/DAC boards provide ready control in and between ≤2 neurons. Control at >2 sites is desirable when studying neural circuits with serial or ring connectivity. Here, we describe how to extend dynamic clamp control to four neurons and their associated synaptic interactions, using a single IBM-compatible PC, an ADC/DAC interface with two analog outputs, and an additional demultiplexing circuit. A specific C++ program, DYNCLAMP4, implements these procedures in a Windows environment, allowing one to change parameters while the dynamic clamp is running. Computational efficiency is increased by varying the duration of the input–output cycle. The program simulates ≤8 Hodgkin–Huxley-type conductances and ≤18 (chemical and/or electrical) synapses in ≤4 neurons and runs at a minimum update rate of 5 kHz on a 450 MHz CPU. (Increased speed is possible in a two-neuron version that does not need auxiliary circuitry). Using identified neurons of the crustacean stomatogastric ganglion, we illustrate on-line parameter modification and the construction of three-member synaptic rings.

Introduction

The dynamic current clamp protocol is a powerful, computer-based method that allows an experimentalist to inject voltage- and time-dependent currents into a neuron, so as to simulate the presence of specific ionic conductances in the neuronal membrane (Robinson and Kawai, 1993, Sharp et al., 1993a, Sharp et al., 1993b). The computer monitors the membrane potential of the neuron under study by intracellular recording and calculates the current to be injected according to equations supplied by the experimenter. The technique has two main applications. First, one can study the contribution of particular voltage-dependent, ligand-gated, or leak conductances to the electrophysiology of individual cells (Sharp et al., 1993a, Sharp et al., 1993b, Gramoll et al., 1994, Harris-Warrick et al., 1995, Ma and Koester, 1996, Turrigiano et al., 1996, Hughes et al., 1998, Hughes et al., 1999, Bartos et al., 1999, Kinard et al., 1999). Second, one can simulate the action of postsynaptic currents. The simulated synaptic input may be driven by patterns of voltage signals defined by the experimenter (Reyes et al., 1996, Ulrich and Huguenard, 1996, Ulrich and Huguenard, 1997, Bartos et al., 1999, Jaeger and Bower, 1999), or by the membrane potential recorded in a second cell (Sharp et al., 1993a, Sharp et al., 1993b, Sharp et al., 1996, Wilders et al., 1996). In the latter case, the dynamic clamp inserts a simulated synaptic connection between the two neurons; in this way one can begin to build up artificial synaptic circuits with neurons among which natural synapses are blocked or non-existent.

Implementation of the dynamic clamp requires hardware, an ADC/DAC interface so that the computer can acquire voltage signals and generate current commands; and software programs that calculate the current equations and operate the interface. There are two corresponding constraints. First, many commercial ADC/DAC boards possess, at most, two analog output channels, thus limiting ready control to two neurons only. Second, the need to compute the injection of current in real time puts a premium on the speed, efficiency and reliability of the software running in a single processor multitask environment (described in detail later).

The methods described here were motivated by research on the functional architecture of central pattern generator circuits (CPGs) — neural circuits that generate the oscillatory motor output patterns underlying rhythmical movements. In certain CPGs (mainly in invertebrates), it has proved possible to identify and characterize most or all of the component neurons together with their connectivity (for a recent review, see Marder and Calabrese (1996)). Simultaneous intracellular recordings can be made routinely from multiple neurons in a given circuit. Pharmacological blockade and photoablation allow the experimenter to disconnect neurons, thereby ‘dissecting’ a circuit (Miller and Selverston, 1979, Bidaut, 1980, Selverston and Miller, 1980, Peterson, 1983). Further studies have revealed that the intrinsic electrical properties of neurons, as well as the strength of their interconnections, can be greatly modified by the action of neuromodulators (Harris-Warrick and Marder, 1991, Harris-Warrick et al., 1992a, Katz et al., 1994). In this context, the dynamic clamp is an important experimental tool, because it allows one to observe the functional consequences of manipulating cellular properties within existing (natural) circuits, or altering either cellular or synaptic properties within artificial circuits constructed by inserting simulated connections between (earlier unconnected) biological neurons (Sharp et al., 1993a, Sharp et al., 1993b, Sharp et al., 1996). However, constraints upon the technique (mentioned above) have, until now, limited its application to controlling just two neurons and their interconnections. To manipulate existing CPG circuits, or to construct more complex, artificial synaptic circuits, it is crucial to exert control at more than just two sites. With existing hardware and software (see below), experiments with more than two cells would require multiple, synchronized computers and ADC/DAC boards, together with additional electronic circuits to sum current commands coming to each cell from separate computers.

In this paper, we outline methods to increase the computational efficiency in running dynamic clamp programs, and to control and/or connect up to four neurons using a conventional, two-output-channel interface augmented by a simple, analog electronic circuit. We also describe a specific program, DYNCLAMP4, that implements these methods and has the added advantages of operation and user control in Microsoft Windows, on-line parameter adjustment, and straightforward formulations for the (in)activation and kinetics of conductances. The methodology is illustrated by inserting simulated synapses between neurons in the crustacean stomatogastric ganglion (STG) (Harris-Warrick et al., 1992b). We show de novo construction of a ring circuit involving three neurons and five simulated synapses. A preliminary report of the methods and their application has appeared in abstract form (Pinto et al., 2000a). Details and downloadable versions of the software are available via the internet (Pinto, 2000). When we started this work, a program controlling up to two neurons was available commercially (DCLAMP2.0: Dyna-Quest Technologies, Inc; Sudbury, MA). The program ran under DOS in 386 or 486 computers, at a ∼1 kHz update rate. DYNCLAMP4 offers some advantages over this commercial program: as well as being able to control up to four neurons, the program can operate faster (update rate ∼10 kHz for two neurons, ∼5 kHz for four) and runs in a Windows environment. The software is free and its source code is available for modification by the user.

Section snippets

Dynamic clamp cycle

Implementation of the basic dynamic current clamp protocol involves repetition of a three-stage cycle:1) Acquisition. A computer reads the neuronal membrane potential, recorded and amplified by a conventional intracellular electrometer, via an analog to digital converter (ADC).2) Computation. Using the sampled membrane potential together with mathematical models of membrane and/or synaptic conductances, the computer calculates the current, which would be generated by the chosen combination of

Update rate capability of the extended dynamic clamp

The minimum update rate of the dynamic clamp cycle was measured with all 18 synapses and eight Hodgkin–Huxley conductances activated. Under these conditions the processor needs to calculate the effect of all conductances in the currents for each cell, and the performance will be the slowest possible. The minimum update rate, for a Pentium III 450 MHz computer running only the dynamic clamp application under Windows NT 4.0, measured directly from one of the digital gating pulses (D0-D3, see Fig.

Discussion

Since its effective introduction by Sharp et al., 1993a, Sharp et al., 1993b, the dynamic clamp protocol has been a valuable tool for studying the functional roles of specific membrane and synaptic properties in the electrophysiology of neurons in many preparations (Marder, 1998). For research upon circuits of neurons, however, the technique has been limited to studying synaptic pairs (Sharp et al., 1996, Wilders et al., 1996). Our present work extends dynamic clamp capabilities in a number of

Acknowledgements

R.D. Pinto was supported by the Brazilian agency Fundação de Amparo à Pesquisa do Estado de São Paulo-FAPESP, under proc. 98/15124-5. Partial support for this work came from the US Department of Energy, Office of Science, under grants DE-FG03-90ER14138 and DE-FG03-96ER14592, and the US Office of Naval Research under grant N00014-00-1-0181. R.D. Pinto also acknowledges encouragement and discussion from José C. Sartorelli, Ramón Huerta, Pablo Varona, Mauro Copelli, Gregg Stiesberg, and Paco

References (50)

  • C.C. Canavier et al.

    Control of multistability in ring circuits of oscillators

    Biol. Cybern.

    (1999)
  • H. Chiel et al.

    Evolution and analysis of model CPGs for walking: I. Dynamical modules

    J. Comput. Neurosci.

    (1999)
  • J.J. Collins et al.

    Hard-wired central pattern generators for quadripedal locomotion

    Biol. Cybern.

    (1994)
  • A. Destexhe et al.

    An efficient method for computing synaptic conductances based on a kinetic model of receptor binding

    Neural Comput.

    (1994)
  • J.S. Eisen et al.

    Mechanisms underlying pattern generation in lobster stomatogastric ganglion as determined by selective inactivation of identified neurons. III. Synaptic connections of electrically coupled pyloric neurons

    J. Neurophysiol.

    (1982)
  • R.C. Elson et al.

    Mechanisms of gastric rhythm generation in the isolated stomatogastric ganglion of spiny lobsters: bursting pacemaker potentials, synaptic interactions, and muscarinic modulation

    J. Neurophysiol.

    (1992)
  • P.A. Getting

    Emerging principles governing the operation of neural networks

    Ann. Rev. Neurosci.

    (1989)
  • P.A. Getting

    A network oscillator underlying swimming in Tritonia

  • S. Gramoll et al.

    Switching in the activity state of an interneuron that controls coordination of the hearts in the medicinal leech (Hirudo medicinalis)

    J. Exp. Biol.

    (1994)
  • R.M. Harris-Warrick et al.

    Modulation of neural networks for behavior

    Ann. Rev. Neurosci.

    (1991)
  • R.M. Harris-Warrick et al.

    Modulation of stomatogastric networks by identified neurons and transmitters

  • R.M. Harris-Warrick et al.

    Dopamine modulation of two subthreshold currents produces phase shifts in activity of an identified motoneuron

    J. Neurophysiol.

    (1995)
  • A.L. Hodgkin et al.

    A quantitative description of membrane current and its application to conduction and excitation in nerve

    J. Physiol.

    (1952)
  • S.W. Hughes et al.

    All thalamocortical neurones possess a T-type Ca2+ ‘window’ current that enables the expression of bistability-mediated activities

    J. Physiol. (London)

    (1999)
  • Cited by (65)

    • Modulation of respiratory sinus arrhythmia in rats with central pattern generator hardware

      2013, Journal of Neuroscience Methods
      Citation Excerpt :

      Computer models of Hodgkin–Huxley neurons have been interfaced with biological neurons to understand the function of one neuron in a network. This method has been used to drive the half centered oscillator in the leech heart (Olypher et al., 2006; Sorensen et al., 2004) and map the modes of synchronization between pairs of coupled neurons (Elson et al., 1998; Pinto et al., 2001; Varona et al., 2001). Computer models of biological neurons have the advantage of being easy to program and to modify.

    View all citing articles on Scopus
    View full text