{"title": "A four neuron circuit accounts for change sensitive inhibition in salamander retina", "book": "Advances in Neural Information Processing Systems", "page_first": 384, "page_last": 390, "abstract": null, "full_text": "A four neuron circuit accounts for change sensitive \n\ninhibition in salamander retina \n\nJeffrey L. Teeters \nLawrence Livennore Lab \nPO Box 808, L-426 \nLivennore CA 94550 \n\nFrank H. Eeckman \nLawrence Livennore Lab \nPO Box 808, L-270 \nLivennore CA 94550 \n\nFrank S. Werblin \nUC-Berkeley \nRoom 145, LSA \nBerkeley CA 94720 \n\nAbstract \n\nIn salamander retina, the response of On-Off ganglion cells to a central \nflash is reduced by movement in the receptive field surround. Through \ncomputer simulation of a 2-D model which takes into account their \nanatomical and physiological properties, we show that interactions \nbetween four neuron types (two bipolar and two amacrine) may be \nresponsible for the generation and lateral conductance of this change \nsensitive inhibition. The model shows that the four neuron circuit can \naccount for previously observed movement sensitive reductions in \nganglion cell sensitivity and allows visualization and prediction of the \nspatio-temporal pattern of activity in change sensitive retinal cells. \n\n1 INTRODUCTION \n\nIn the salamander retina. the response of transient (On-Off) ganglion cells to a central \nflash is reduced by movement in the receptive field surround (Werblin. 1972; Werblin & \nCopenhagen. 1974) as illustrated in Fig 1. This phenomenon requires the detection of \nchange in the surround and the lateral transmission of this change sensitive inhibition to \nthe ganglion cell dendrites. Wunk & Werblin (1979) showed that all ganglion cells \nreceive change-sensitive inhibition. and Barnes & Werblin (1987) implicated a change(cid:173)\nsensitive amacrine cell with widely distributed processes. The change-sensitivity of these \namacrine cells has been traced in part to a truncation of synaptic release from the bipolar \ntenninals that presumably drive them (Maguire et al., 1989). The transient response of \nthese amacrine cells, mediated by voltage gated currents (Barnes & Werblin, 1986; Eliasof \net al., 1987) also contributes to this change sensitivity. \n\nThese and other experiments suggest that interactions between four neuron types underlie \nboth the change detection and the lateral transmission of inhibition (Werblin et al., 1988; \nMaguire et al., 1989). To test this hypothesis and make predictions that could be \ncompared with later experiments we have constructed a computational model of the four \nneuron circuit and incorporated it into an overall model of the retina. This model allows \nus to simulate the effect of inhibition generated by the four neuron circuit on ganglion \ncells. \n\n384 \n\n\fStimulus: \n\ncentral test spot \n\nt \n\nWindmill with 1 \nI \n\nI +1(] \n~@:U( Resting level \n\nNormaJ+ \n\nGanglion Cell Response: \n\nStationary \nwindmill \n\nSpinning \nI---t windmill \n1 second \n\n1 \n1 \n------ ---------I \n\nT::::t:~\"\"\" \"~\"] \n\nFigure 1: Change-Sensitive Inhibition. Data is from Werblin (1972). \n\n2 IMPLEMENTING THE HYPOTHETICAL CIRCUIT \nThe proposed change-sensitive circuit (Werblin et al.. 1988; Maguire et al .\u2022 1989) is \nreproduced in Figure 2. This is meant to describe a very local region of the retina where \nthe receptive fields of the two bipolar cells are spatially overlapping. When a visual \ntarget enters this receptive field. the bipolar cells are both depolarized. The sustained \nbipolar cell activates the narrow field amacrine cell that. in tum feeds back to the synaptic \nterminal of the transient bipolar cell to truncate transmitter release after a brief (ca. 100 \nmsec) delay. Because the signal reaching the wide field amacrine cell is truncated after \nabout 100 msec. the wide field amacrine cell will receive excitation when the target enters \nthe recepti ve field. but will not continue to respond in the presence of the target. \nThe spatial profiles of synaptic input and output for the cell types involved in the model \nare summarized in Figure 3. The bipolar and narrow field amacrine cell sensitivities \nextend over a region corresponding roughly to their dendritic spread. The wide field \namacrine cell appears to receive input over a local region near the cell body, but delivers \nits inhibitory output over a much wider region corresponding the the full extent (ca. 500 \nmm) of its processes. \n\nFigure 4 shows the electrical circuit model for each cell type. and illustrates the \ninteractions between cells that are implemented in the model. In Figure 4. boxes contain \nthe circuit for each cell and arrows between them represent synaptic interactions thought \n\n.\u00b7\u00b7\u00b7\u00b7NarrowJietd (amacrine .... \n\nTo \n\n_ Ganglion \n\ncells. \n\nFigure 2: Circuitry to be Analyzed \n\n\fBipolar 1\\ (Inpul and oulpull \n\nNarrow fleld~ (Input and output) \n\namacrine / \n\n\" \n\nWide \n\nfield \n\n~Inpull \n\n500 \n\nDistance from 0 cell center (Ilm) \n\n500 \n\nFigure 3: Spatial Profiles of Input Sensitivity and Output Transmission \n\nto occur as determined through experiments in which a neurotransmitter is puffed onto \nbipolar dendrites. Bipolar cells are modeled using two compartments. corresponding to \nthe cell body and axon terminal as suggested in Maguire et at. (1989). Amacrine cells are \nmodeled using only one compartment as in Eliasof et at. (1987). \nEach compartment has a voltage (Vbs. Vbst, Vbtt. Van. Vaw). The cell body for the \nsustained and transient bipolar are assumed to be the same. Batteries in the figure \ncorrespond to excitatory (E+. Ena) or inhibitory reversal potentials (E-. Ek, Eel). \nResistors represent ionic conductances. Circles and arrows through resisters indicate \ntransmitter dependent conductances which are controlled by the voltage of a presynaptic or \nsame cell. Functions relating voltages to conductances are mostly linear with a threshold. \nMore details are given in Teeters et at. (1991). \n\nNeurotransmitter Input \n\nWide field \n\nFi~ure 4: Details of Circuitry \n\n\fA Four Neuron Circuit Accounts for Change Sensitive Inhibition \n\n387 \n\n3 TESTING THE COMPUTATIONAL MODEL \nComputer simulation was used to tune model parameters. and test whether the single cell \nproperties and proposed interactions between cells shown in Figure 4 are consistent with \nthe responses recorded from the neurons during applications of a neurotransmitter puff. \n\nResults are shown in Figure 5. Voltage clamp experiments electrically clamp the cell \nmembrane potential to a constant voltage and determine the current required to maintain \nthe voltage over time. Downward traces indicate that current is flowing into the cell; \nupward traces indicate outward current For simplicity. scales are not shown, but in all \ncases the magnitude of the simulated response is close to that of the observed response. \nThe simulated and observed responses voltage clamps of the wide field amacrine shown in \nthe fourth row vary because there is a sustained outward current observed experimentally \nthat is not apparent in the simulations. This shows that the model is not perfect and is \nsomething that needs further investigation. \n\nThis difference between the model and observed response does not prevent the \nhypothesized function of the circuit from being simulated. This is shown on the bottom \nrow where both the observed and simulated voltage responses from the wide field amacrine \nare transient. \n\n4 SIMULATING INHIBITION TO GANGLION CELLS \nFigure 5 illustrates that we have, to a large degree, succeeded in combining the \ncharacteristics of single cells into a model which can explain many of the observed \nproperties thought to be due to the interaction between these cells in a local region. \n\nExperiment \n\nNeurotransm Itter \nPuff Input \n\nVoltage clamp of \nbipOlar cell body \n\nObserved response \n-.r--\n\"-\n\nSimulated Response \nJ \n\n-\n\nnarrow field amacrine \n\nVoltage clamp of V' E=======:: \n\n-\n\nWide field amacrine \n\nVoltage clamp \n\n~., \n-\" \n\ny-\n\n--v-\n\npIcrotoxin block \n\nVoltage clamp with ~ -V-\nVoltage response P--\n\n--\"'-\n\nFigure 5: Example Puff Simulations \n\n\f388 \n\nTeeters, Eeckman, and Werblin \n\nThe next step in our analysis is to investigate how this circuit influences the response of \nganglion cells. To do this requires simulating the input to the bipolar dendrites and \nsimulating the ganglion cells which receive the transient inhibition generated by the wide \nfield amacrine. This amounts to a integrated model of an entire patch of retina. including \nreceptors. horizontal cells. the four neuron circuit discussed earlier. and ganglion cells. \nThe manner in which we accomplish this is illustrated in Figure 6. \n\nThe left side of figure 6 shows the model elements. Receptors and horizontal cells are \nmodeled as low pass filters with different time constants and different spatial inputs. The \nganglion cell model receives a transient excitatory input generated phenomenologically by \na thresholded high pass filter from the transient bipolar. Inhibitory input to the ganglion \ncell is implemented as coming from the transient wide field amacrine cells described \npreviously. For simplicity. voltage gated currents and spiking are not implemented in the \nganglion cell model. and only the off bipolar pathways are simulated. \n\nThe right hand of Figure 6 illustrates how the model is implemented spatially. The \ncircuit for each cell type is duplicated across the retina patch in a matrix fonnat. The \nknown spatial properties of each cell. such as the spatial range of transmitter sensitivity \nand release are incorporated into the model. Details are given in Teeters et al. 1991. \n\n5 SIMULATING INHIBITION TO GANGLION CELLS \nTo test if the model can account for the observed reduction in ganglion cell response \nduring movement in the receptive field surround. we simulated the experiment depicted in \nFigure 1. mainly the flashing of a central light during the presence of a stationary and \nspinning windmill. The results are shown in Figure 7. \n\nModel Elements \n\nSpatial Implementation \n\nReceptor \n\nR \u00b7\" \n\nHorizontal Cell \n\nThreshold \nHigh-pass \nfilter \n\n... \n\nwCf). 1: 'b~ .... \n\u00b7 ... 1--\n~E -\n'\u00b7\u00b7 \n\u00b7t:t- -rEcIT r } \n\nang Ion el \n\n............. . \n\n. . \n\n.'. \n\n. \n\n- E\n\n\u2022 \n\u2022 \n\nOn-Off Ganglion cells \n\nFigure 6: Integrated Retinal Model \n\n\fA Four Neuron Circuit Accounts for Change Sensitive Inhibition \n\n389 \n\nRather than displaying a single curve representing the response of a single unit over time, \nFigure 7 shows the simultaneous pattern of activity in an array of neurons spatially \ndistributed across the retina patch at an instant in time (just after a central light spot is \nturned on). The neuron responses are the transient bipolar terminal, the wide field \namacrine neurotransmitter release, and the ganglion cell voltage response. On the left \ncolumn is shown the response to a flashing spot when the windmill is stationary. On the \nright is shown the response to the same flashing spot but with a spinning windmill. \n\nWhen the windmill is stationary, the transient bipolar terminal responds only to the \ncenter flash. Responses to the windmill vanes are suppressed by the narrow field \namacrine cell causing the appearance of four regions of hyperpolarizing responses around \nthe center. The wide field amacrine responds to the central test flash and releases \ntransmitter as shown in the second row. The array of ganglion cells responds to both the \nexcitatory input generated by the spot at the bipolar terminals and the inhibitory input \ngenerated by the wide field amacrines. Because the wide field inhibition has not yet taken \neffect at this point in time, the ganglion cells respond well to the flashing spot. \n\nWhen the windmill is spinning, as is shown on the right hand column, the transient \nbipolar terminals generate a response to the leading edge of the windmill vanes. The wide \nfield amacrine cells receive excitatory input from the transient bipolar terminal responses \nto the vane, and consequently release inhibitory neurotransmitter over a wide area as \nshown in in the right column. Because inhibition is being continuously generated by the \nspinning windmill, the response of the ganglion cells across the retinal patch has a large \n\nStationary Windmill \n\nSpinning windmill \n\nTransient Bipolar Terminal \n\nWide field \n\nGanglion cell \n\nFig. 7 - Ganglion Cell Inhibition Caused By Spinning Windmill \n\n\fbowl shaped area of hyperpolarization which reduces the ganglion cell response of the \ncells to the central test flash. This is seen by the fact that the height of depolarization in \nthe centrally located ganglion cells is much smaller under conditions of a spinning \nwindmill than if the windmill is stationary. This is consistent with the results found \nexperimentally which are illustrated in Figure 1. Experimental data not yet attained. but \nwhich are predicted by the model simulations illustrated in Figure 7. are the spatial \npatterns of activity generated in the bipolar. amacrine. and ganglion cells in response to \nthe different stimuli. \n\n6 SUMMARY \nUsing computer simulation of a neurophysiologically based model. we demonstrate that \nthe experimental data describing properties of four neurons in the inner retina are \ncompatible with the hypothesis that these neurons are involved in the detection of change \nand the feedforward of change-sensitive inhibition to ganglion cells. First. we build a \ncomputational model of the hypothesized four neuron circuit and determine that the \nproposed interactions between them are sufficient to reproduce many of the observed \nnetwork properties in response to a puff of neurotransmitter. Next. we integrate this \nmodel into a full retina model to simulate their influence on ganglion cell responses. \n\nThe model verifies the consistency of presently available data. and allows formation of \npredictions of neural activity are subject to refutation or verification by new experiments. \nWe are currently recording the spatio-temporal response of ganglion cells to moving \nstimuli so that direct comparisons to these model predictions can be made. \n\nReferences \n\nBarnes. S. and Werblin. F.S. (1986). Gated currents generate single spike activity in \namacrine cells of the tiger salamander. Proc. Natl. Acad. Sci. USA 83: 1509 - 1512. \nBarnes. S. and Werblin. F.S. (1987). Direct excitatory and lateral inhibitory synaptic \ninputs to amacrine cells in the tiger salamander retina. Brain Res. 406: 233 - 237. \nEliasof S .\u2022 Barnes S. and Werblin. F.S. (1987). The interaction of ionic currents \nmediating single spike activity in retinal amacrine cells of the tiger salamander. 1. \nNeurosci. 7: 3512 - 3524. \nMaguire. G .\u2022 Lukasiewicz. P. and Werblin F.S. (1989). Amacrine cell interactions under(cid:173)\nlying the response to change in the tiger salamander retina. 1. Neurosci. 9: 726 - 735. \nTeeters. J.L .\u2022 Eeckman. F.H .\u2022 Werblin F.S. (1991). A computer model to visualize \nchange sensitive responses in the salamander retina. In MA. Arbib and J-P. Ewert (eds.) \nVisuomotor Coordination: Amphibians. Comparisons. Models and Robots. Plenum. \nWerblin. F.S. (1972). Lateral interactions at inner plexiform layer of a vertebrate retina: \nantagonistic response to change. Science. 175: 1008 - 1010. \nWerblin. F.S. and Copenhagen. D.R. (1974). Control of retinal sensitivity. III. Lateral \ninteractions at the inner plexiform layer. 1. Gen. Physiol. 63: 88 - 110. \nWerblin. F.S .\u2022 Maguire. G., Lukasiewicz, P., Eliasof. S .\u2022 and Wu. S. (1988). Neural \ninteractions mediating the detection of motion in the retina of the tiger salamander. Visual \nNeurosci. 1: 317 - 329. \nWunk, D.F. and Werblin, F.S. (1979). Synaptic inputs to ganglion cells in the tiger \nsalamander retina. 1. Gen. Physiol. 73: 265 - 286. \n\n\f", "award": [], "sourceid": 368, "authors": [{"given_name": "Jeffrey", "family_name": "Teeters", "institution": null}, {"given_name": "Frank", "family_name": "Eeckman", "institution": null}, {"given_name": "Frank", "family_name": "Werblin", "institution": null}]}