- Email: [email protected]
Emergence of temporal-pattern sensitive neurons in the midbrain of weakly electric fish Gymnarchus niloticus
Journal of Physiology - Paris 96 (2002) 531–537 www.elsevier.com/locate/jphysparis
Emergence of temporal-pattern sensitive neurons in the midbrain of weakly electric fish Gymnarchus niloticus Masashi Kawasakia,b,*, Yuan-Xing Guoa a
Department of Biology, Gilmer Hall, University of Virginia, Charlottesville, VA 22903, USA PRESTO, Japan Science and Technology Corporation, Kawaguchi, Saitama 332-0012, Japan
b
Abstract Sensitivity of neurons in the torus semicircularis of a weakly electric fish, Gymnarchus niloticus, to two stimulus parameters that are critical for its behavior the jamming avoidance response was examined. The first parameter is the sign of frequency difference between discharge frequencies of fish’s own electric organ and that of a neighbor’s. The second parameter is the spatial orientation of neighbor’s electric field. Whereas neuronal ambiguity of frequency coding for different orientations of neighbor’s electric field is predicted, unambiguous JAR occurs at the behavioral level. Most neurons in the torus semicircularis showed sensitivity to the sign of frequency difference. Although a small number of neurons showed preference to a consistent sign of the frequency difference, the coding of the sign of frequency differences was found to be ambiguous with a highly variable pattern of responses for different orientations in most of neurons. # 2003 Elsevier Ltd. All rights reserved. Keywords: Electric fish; Combination sensitivity; Jamming avoidance response; Spatio-temporal pattern; In vivo whole-cell recording
1. Introduction In the jamming avoidance response (JAR) of weakly electric fishes, fish shift their frequencies of electric organ discharge (EOD) away from each other to avoid mutual jamming of the electrosensory system [2]. Behavioral studies showed that a fish shifts its own discharge frequency in the direction that increases the frequency difference without a trial-and-error behavior. Thus, the computational task for this behavior is to determine if a neighbor’s EOD frequency is higher or lower than its own. If the frequency difference, Df, between two fish is defined as Df=f2 f1, where f2 is frequency of the EOD of the neighboring fish, and f1 is that of the fish’s own electric organ, the fish’s task is to determine the sign of Df. Computational algorithms and neuronal mechanisms of the jamming avoidance response of weakly electric fishes have been extensively studied in two genera of electric fishes, Eigenmannia and Gymnarchus, which lack common electric fish ancestors [3,7]. Both fish determine the sign of Df by joint evaluation of two stimulus parameters, amplitude and phase modulation in the electrosensory stimulus. Amplitude and phase * Corresponding author. Tel.: +1-434-982-5763. E-mail address: [email protected] (M. Kawasaki). 0928-4257/03/$ - see front matter # 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0928-4257(03)00008-1
modulation are created by summation of EODs from fish’s own electric organ and that from a neighboring fish. The information about the sign of Df cannot be determined by evaluation of amplitude or phase modulation alone since both of them occur at the absolute frequency difference, |Df|. The temporal relation between the time course of amplitude and phase modulation, however, is specific to the sign of Df. Behavioral experiments demonstrated that joint evaluation of amplitude and phase modulation is necessary for JAR [4,6]. In Gymnarchus, neurons in the first electrosensory brain station, the electrosensory lateral line lobe (ELL), process amplitude and phase information independently. These neurons are sensitive to either amplitude modulation or phase modulation and commonly project to the medial dorsal (MD) and lateral anterior (LA) subdivisions of the midbrain structure, the torus semicircularis (hereafter simply ‘torus’) [1,9]. In the present study, neuronal sensitivity to the temporal pattern of amplitude and phase modulation was explored in the common projection areas in the torus. In the JAR, a fish raises or lowers its discharge frequency according to the frequency of a neighbor. The ’correct’ JAR, i.e. the frequency shift of EOD resulting in increase of the frequency difference between two fish,
532
M. Kawasaki, Y.-X. Guo / Journal of Physiology - Paris 96 (2002) 531–537
occurs whatever the spatial orientation of a neighbor with respect to the fish. The phase computation mentioned above, however, is inherently sensitive to the orientation of the electric field created by a neighbor since phase information is detected by phase sensitive neurons in the ELL as the phase difference between two areas on the body surface [8]. The phase sensitive neurons in the ELL respond to D=jA jB, where jA and jB are phase at body area A and B, respectively. Dj may switch its sign for different orientations of neighbor’s electric field. The switching of the sign of Dj is equivalent to the switch of the sign of Df and thus neurons may err in determination of the sign of Df showing orientation ambiguity [6]. In the present study, the effect of the orientation of the electric field of the neighbor on the selectivity to the sign of Df was also examined to determine the degree of ‘orientation ambiguity’ in neurons in the torus.
2. Material and methods Fifty-five Gymnarchus niloticus (12–16 cm) were used. They were obtained from a local dealer and maintained under a light–dark condition of L:D=12:12. Fish were initially anesthetized with MS-222 (1:10 000) and then immobilized with curare (Flaxedil, 0.03%, 3 to 8 ml IM). The preparation was submerged in water in the tank identical to the one used in a previous behavioral study [6]. The gills were perfused with oxygenated water during surgery and physiological experiments. Water temperature and conductivity were maintained at 25–27 C and 100–200 mS/cm for maintenance tanks, the experimental tank, and perfusion water. After local application of Xylocaine (2%), a small hole was drilled above the torus. The medial edge of the optic tectum was gently pressed outwards to visualize the anterior tip of the torus. The MD and LA subdivisions of the torus were visually localized [1]. The use of animals in all experiments were approved by the Animal Care and Use Committee of University of Virginia (protocol #1904). Electrosensory stimuli were given with the ‘free-field’ configuration used in Kawasaki [6]. A sinusoidal signal, S1, which mimicked the fish’s own silenced EOD, was given between an electrode in the mouth and an electrode placed near the tail. The frequency of S1, f1, was constant and was within 20 Hz of EOD frequencies before curarization of the fish. A sinusoidal signal, S2, which mimicked the EOD of a neighbor, was given through one of four pairs of electrodes as shown in Fig. 1. The frequency of S2, f2, was selectable so that the frequency difference, Df=f2 f1, was 4, 2, 1, +1, +2 or +4 Hz. Amplitude of S1 was set at 1–2 mV/cm near the gill cover. Amplitude of S2 was set at 20% of S1. Stimulus signals were created by custom made
software and delivered through a DA converter (DA3-4, Tucker Davis Technology) and FET based stimulus isolators. Neurons in the torus were recorded by the in vivo whole-cell recording technique, which allowed longterm (typically more than 30 min) stable recording [9,10]. Intracellular potentials were recorded in current clamp mode (Axoclamp 2B, Axon Instruments) and fed to an AD converter (AD-3, Tucker Davis Technology) for on-line and off-line analyses. Neurons were sought using a stimulus condition Df= 2 and +2 Hz. When extracellular spikes that were modulated with each cycle of df were detected, an attempt was made to establish a giga-Ohm seal by applying gentle suction to the electrode. A seal with more than 1 G was typically established. The cell membrane was opened by applying a brief (500 ms) positive current (1–2 nA). Access resistance was maintained less than 100 M during the entire recording session. After establishing whole-cell recording, all four orientations of S2 were first quickly tested to determine the orientation at which the modulation of spiking was most strongly heard from the audio monitor. In this orientation, histograms were made for Df= 4, 2, 1, +1, +2, and +4 Hz to determine which |Df| produces the highest spike rates. With this best |Df|, spike histograms were made for Df < 0 and Df > 0 for all four orientations of S2. In each orientation, two parameters indicating differential responses to the different signs of Df were computed. SS was computed as SS=(spkDf >0 spkDf <0)/ (spkDf > 0+spkDf <0) where spkDf < 0 and spkDf > 0 are mean spike rates for Df < 0 and Df > 0, respectively. p was computed as the significance value of the unpaired t-test between spike counts for Df < 0 and Df > 0. Whereas SS expresses the response ratio between the difference and the mean spiking rate, p expresses the
Fig. 1. Four orientations of S2. S2 were delivered through one of four pairs of electrodes located in a circular (15 cm in diameter) fashion. S1 was given between two electrodes in the mouth and at the tail (not shown).
M. Kawasaki, Y.-X. Guo / Journal of Physiology - Paris 96 (2002) 531–537
significance of the difference. Neurons were determined to be ‘sign-selective’, or to have ‘preferred’ a sign of Df in a given orientation of S2 whenever p < 0.05 regardless of the magnitude of SS.
3. Results Whole-cell recordings were made from 165 neurons in 81 fish. Fig. 2 shows two examples of intracellular potentials from the whole-cell recording. The subthreshold postsynaptic potential was generally 5–15 mV in amplitude but it often exceeded 25 mV as in the case shown in the figure. Spikes commonly reached nearly 0 mV but in some cases spike amplitude was smaller. Histograms were made using a spike threshold set at approximately the half height of the spikes. In 77 neurons, spike histograms were successfully made for Df < 0 and Df > 0 for all four orientations of S2, and response
Fig. 2. Whole-cell intracellular potential in response to joint modulation of amplitude (solid line) and phase (broken line). Neuron in A preferred Df<0 to Df>0. Neuron in B responded equally to Df<0 and Df>0.
533
properties were confirmed to be unchanged after repeated presentation of stimulus sets. Other neurons that showed altered responses during the course of recording were discarded from further analysis. All 77 neurons except for one showed sign-selective responses in at least one orientation. Four of them consistently preferred one sign of Df for all four orientations (Fig. 3A and B). Other neurons showed a variety of response patterns for the sign of Df depending on the orientation. An example shown in Fig. 3C showed strong selectivity to Df < 0 in orientations 1 and 2, and only weak non-selective responses in orientations 3 and 4, whereas the example illustrated in Fig. 3D preferred Df > 0 in orientation 3 but the response magnitude in this orientation was much smaller than those in nonselective responses in the other three orientations. To evaluate the importance of the sign-selective responses over non-selective responses in different orientations, the relative strength of sign-selective responses compared with the non-selective responses was examined in all 77 neurons. First, the means of spike counts for Df < 0 and Df > 0, msel, were computed for each orientation where sign-selectivity existed. The means of spike counts for Df > 0 and Df < 0, mnonsel, were also computed for each of all other orientations where no sign selectivity was shown. R was computed as max(msel)/ max(mnonsel) where max(msel) and max(mnonsel) were the largest means. The distribution of R is shown in Fig. 4. Neurons were divided into Type-I (n=60) and Type-II (n=17) by an arbitrary value of R=1. Type-I neurons (R > 1) were analyzed further since they showed stronger responses in a sign-selective orientation than in nonselective orientations and are likely to participate in the control of the JAR. Type-I neurons showed sign-selectivity in one (n=1), two (n=15), three (n=21), or all four orientations (n=23). When multiple orientations showed sign selectivity, some neurons showed consistent preference to the sign of Df (Fig. 5A and B); other neurons switched preference to the sign of Df according to the orientation (Fig. 5C–F). When the preference switched with the change of orientation, the magnitude of responses was different for different orientations in most cases (e.g. Fig. 5F). The response patterns of Type-I neurons are summarized in Table 1. The orientation with the strongest differential responses to the two signs of df is referred to as ‘best orientation’ and it is defined as the orientation in which the difference between spike counts for Df < 0 and Df > 0 was the largest. All orientations were equally likely to be the best orientation in Type-I neurons. The preferred signs of Df at the best orientation across neurons showed no statistically significant bias towards one sign (Table 2). Signselective neurons showed a range of SS. |SS| at the best orientation was 0.36 0.19 (Table 2). Examples of Type-II neurons are shown in Fig. 6. These neurons were primarily driven by either
534
M. Kawasaki, Y.-X. Guo / Journal of Physiology - Paris 96 (2002) 531–537
amplitude or phase and showed little preference in the orientation in which response magnitude was large.
4. Discussion and conclusions The first finding of this study is that neuronal sensitivity to the temporal patterns of inputs that are specific to the sign of Df appears in the torus of Gymnarchus. The neurons in the ELL that project to the torus respond to either amplitude modulation (solid sinusoidal lines in Fig. 2) or phase modulation (dotted lines in Fig. 2) which are identical for Df < 0 and Df > 0 except for their mutual temporal relation. The neurons in the ELL, and thus the inputs to the torus, do not differentiate the sign of Df [9]. The difference in the temporal relation in the input, however, was detected by neurons in the torus. Although the magnitude of the differences
Fig. 4. Distribution of R=max(msel)/max(mnonsel) across neurons. max(msel)=0 was assigned for a neuron that did not show sign-selectivity in any orientations. R was infinite (Inf) when all orientations showed selectivity and thus max(mnonsel) was zero.
Fig. 3. Neurons showing sign-selective responses with consistent preference to the sign of df for all orientations (A, B). Neurons with sign-selectivity in limited orientations (C, D). In each of A–D, rows are for four orientations of S2, left and right columns are for Df<0 and Df>0, respectively. In this and all following histogram figures, the ordinate shows spike rate (spikes/s) and the abscissa shows time in ms. Sinusoidal curves in the histograms show the time course of amplitude (solid) and phase (broken) modulation as in Fig. 2. ‘=’ marks between the histograms indicate that histograms for Df <0 and Df>0 did not show a statistically significant difference in spiking rate. ‘ <’ and ‘> ’ marks indicate that the difference in spike rates was statistically significant. p is the p-value by the t-test.
M. Kawasaki, Y.-X. Guo / Journal of Physiology - Paris 96 (2002) 531–537
in responses for Df < 0 and Df > 0 varied from neuron to neuron, a majority of neurons in the torus showed sensitivity to the sign of Df (Table 2). How do inputs from amplitude and phase sensitive neurons interact in the sign-selective neurons in the torus to create the selectivity to the sign of Df ? The mode of interaction could be understood in some
535
neurons by inspecting response histograms. For example, the neuron in Fig. 5A responded when amplitude reached its maximum both in Df < 0 and Df > 0 histograms in orientation 2, and this response appeared to be differentially gated by the phase input. Similarly, the neuron in Fig. 5D responded to the same phase value and these responses were differentially gated by ampli-
Fig. 5. Type-I neurons showing a variety of the pattern of sign-selectivity over different orientations. A, consistent selectivity in orientations 1, 2, and 4. B, consistent selectivity in orientations 1 and 4. C, D Mixed preferred signs. E, F, Strong selective responses in one orientation with weaker and opposite selective responses in other orientations.
536
M. Kawasaki, Y.-X. Guo / Journal of Physiology - Paris 96 (2002) 531–537
tude changes. In other neurons response patterns were more complex. Further analyses of synaptic interaction of inputs require independent modulation of amplitude and phase components of the signal that was not possible in the present stimulus configuration. Intracellular recording of postsynaptic potential using the phase chamber in which amplitude and phase can be independently modulated [6] will be the subject of future study. The second finding of the present study is that neurons in the torus generally show highly variable responses with different orientation of S2 as predicted by the computational rules that have been determined by behavioral studies [6]. Response magnitudes, sign selectivity expressed by SS, and preferred sign of Df in these neurons were generally different for different orienta-
tions. Amplitude and phase modulation that affect these neurons occur at particular parts of the body surface with a given orientation of S2 and this spatial pattern of modulated sites shifts when orientation of S2 is altered. The variability of the responses of these neurons may be explained by the shift of modulation sites and a small size of receptive field of these neurons. These neurons cannot encode the sign of Df unambiguously because of the orientation dependent responses. Behavioral experiments have shown that the fish performs the correct JAR with any one orientation of S2 indicating that fish obtains unambiguous information about the sign of Df with any given orientation. The ambiguous neurons in the torus, however, may participate in coding of the sign of Df by means of population coding. For the population coding, the participating neurons showing signselective responses in one orientation of S2 should not interfere with or cancel the sign-selective responses of other neurons that are sign-selective in other orientations. Some neurons found in the present study fulfill this condition having large values of R (Fig. 3C) or showing differential magnitudes of responses when the preference to Df switches with orientation (Fig. 5E and F), whereas other neurons are not suited for encoding of the sign of Df (e.g. Fig. 3D).
Table 1 Type-I neurons (n=60) Number of sign-selective orientations
Consistent sign of preference over orientations
Mixed signs of preference over orientations
1 2 3 4
n=1 n=12 n=7 n=4
n=1 n=14 n=19
Table 2 Comparison of sign selectivity in different best orientations in Type-I neurons Best orientation
Number of neurons total (Df<0/Df>0) Mean of |SS| Standard deviation of |SS| Maximum of |SS| Maximum of |SS|
Total
1
2
3
4
12* (4/8) 0.35 0.24 0.91 0.16
11* (7/4) 0.44 0.15 0.67 0.22
19* (10/9) 0.37 0.16 0.62 0.11
18* (5/13) 0.35 0.20 0.59 0.06
60(26**/34**) 0.36 0.19 0.91 0.06
* Difference across orientations not significant (w2-test, p=0.34). ** Difference between positive and negative selectivity not significant (sign test, p=0.18).
Fig. 6. Examples of Type-II neurons. A, Weak sign-selective responses in orientations 1 and 3 and strong non-selective responses in orientations 2 and 4. B, no selectivity in any orientations.
M. Kawasaki, Y.-X. Guo / Journal of Physiology - Paris 96 (2002) 531–537
Four neurons that showed consistent preference to one sign of Df over all four orientations (Fig. 3A and B) may be higher order neurons that receive inputs from other ambiguous neurons in the torus and represent the unambiguous information about the sign of Df. If the neurons in Fig. 3A and B, respectively make excitatory and inhibitory connections to the pacemaker that drives the EOD and they are mutually inhibited, the correct JARs would occur in any orientation of S2. The data obtained in Gymnarchus in this study give insights into comparative physiology of JAR in Eigenmannia and Gymnarchus [5,11]. In both genera, the first site of interaction between amplitude and phase information was found to be the midbrain structure, the torus semicircularis. The degree of interaction, or abundance of neurons that integrate amplitude and phase information, appears to be similar in these fish.
References [1] A.H. Bass, C.D. Hopkins, Comparative aspects of brain organization of an African ‘‘wave’’ electric fish, Gymnarchus niloticus, J Morph 174 (1982) 313–334. [2] W. Heiligenberg, Jamming avoidance responses: model systems for neuroethology. In: T.H. Bullock, W. Heiligenberg (Eds.), Electroreception, Wiley & Sons, New York, 1986, pp. 613–649.
537
[3] W. Heiligenberg, Neural nets in electric fish, The MIT Press, Cambridge, MA, 1991. [4] W. Heiligenberg, C. Baker, J. Matsubara, The jamming avoidance response in Eigenmannia revisited: the structure of a neuronal democracy, J Comp Physiol 127 (1978) 267–286. [5] W. Heiligenberg, G. Rose, Gating of sensory information: Joint computations of phase and amplitude data in the midbrain of the electric fish Eigenmannia, J Comp Physiol 159 (1986) 311– 324. [6] M. Kawasaki, Independently evolved jamming avoidance responses employ identical computational algorithms: a behavioral study of the African electric fish, Gymnarchus niloticus, J Comp Physiol 173 (1993) 9–22. [7] M. Kawasaki, Comparative analysis of the jamming avoidance response in African and South American wave-type fishes, Biol Bull 191 (1996) 103–108. [8] M. Kawasaki, Y.-X. Guo, Neuronal circuitry for comparison of timing in the electrosensory lateral line lobe of an African wavetype electric fish, Gymnarchus niloticus, J Neurosci 16 (1996) 380– 391. [9] M. Kawasaki, Y.-X. Guo, Parallel projection of amplitude and phase information from the hindbrain to the midbrain of the African electric fish, Gymnarchus niloticus, J Neurosci 18 (1998) 7599–7611. [10] G. Rose, E. Fortune, New techniques for making whole-cell recordings from CNS neurons in vivo, Neurosci Research 26 (1996) 89–94. [11] G.J. Rose, W. Heiligenberg, Neural coding of difference frequencies in the midbrain of the electric fish Eigenmannia: reading the sense of rotation in an amplitude-phase plane, J Comp Physiol 158 (1986) 613–624.
Emergence of temporal-pattern sensitive neurons in the midbrain of weakly electric fish Gymnarchus niloticus Masashi Kawasakia,b,*, Yuan-Xing Guoa a
Department of Biology, Gilmer Hall, University of Virginia, Charlottesville, VA 22903, USA PRESTO, Japan Science and Technology Corporation, Kawaguchi, Saitama 332-0012, Japan
b
Abstract Sensitivity of neurons in the torus semicircularis of a weakly electric fish, Gymnarchus niloticus, to two stimulus parameters that are critical for its behavior the jamming avoidance response was examined. The first parameter is the sign of frequency difference between discharge frequencies of fish’s own electric organ and that of a neighbor’s. The second parameter is the spatial orientation of neighbor’s electric field. Whereas neuronal ambiguity of frequency coding for different orientations of neighbor’s electric field is predicted, unambiguous JAR occurs at the behavioral level. Most neurons in the torus semicircularis showed sensitivity to the sign of frequency difference. Although a small number of neurons showed preference to a consistent sign of the frequency difference, the coding of the sign of frequency differences was found to be ambiguous with a highly variable pattern of responses for different orientations in most of neurons. # 2003 Elsevier Ltd. All rights reserved. Keywords: Electric fish; Combination sensitivity; Jamming avoidance response; Spatio-temporal pattern; In vivo whole-cell recording
1. Introduction In the jamming avoidance response (JAR) of weakly electric fishes, fish shift their frequencies of electric organ discharge (EOD) away from each other to avoid mutual jamming of the electrosensory system [2]. Behavioral studies showed that a fish shifts its own discharge frequency in the direction that increases the frequency difference without a trial-and-error behavior. Thus, the computational task for this behavior is to determine if a neighbor’s EOD frequency is higher or lower than its own. If the frequency difference, Df, between two fish is defined as Df=f2 f1, where f2 is frequency of the EOD of the neighboring fish, and f1 is that of the fish’s own electric organ, the fish’s task is to determine the sign of Df. Computational algorithms and neuronal mechanisms of the jamming avoidance response of weakly electric fishes have been extensively studied in two genera of electric fishes, Eigenmannia and Gymnarchus, which lack common electric fish ancestors [3,7]. Both fish determine the sign of Df by joint evaluation of two stimulus parameters, amplitude and phase modulation in the electrosensory stimulus. Amplitude and phase * Corresponding author. Tel.: +1-434-982-5763. E-mail address: [email protected] (M. Kawasaki). 0928-4257/03/$ - see front matter # 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0928-4257(03)00008-1
modulation are created by summation of EODs from fish’s own electric organ and that from a neighboring fish. The information about the sign of Df cannot be determined by evaluation of amplitude or phase modulation alone since both of them occur at the absolute frequency difference, |Df|. The temporal relation between the time course of amplitude and phase modulation, however, is specific to the sign of Df. Behavioral experiments demonstrated that joint evaluation of amplitude and phase modulation is necessary for JAR [4,6]. In Gymnarchus, neurons in the first electrosensory brain station, the electrosensory lateral line lobe (ELL), process amplitude and phase information independently. These neurons are sensitive to either amplitude modulation or phase modulation and commonly project to the medial dorsal (MD) and lateral anterior (LA) subdivisions of the midbrain structure, the torus semicircularis (hereafter simply ‘torus’) [1,9]. In the present study, neuronal sensitivity to the temporal pattern of amplitude and phase modulation was explored in the common projection areas in the torus. In the JAR, a fish raises or lowers its discharge frequency according to the frequency of a neighbor. The ’correct’ JAR, i.e. the frequency shift of EOD resulting in increase of the frequency difference between two fish,
532
M. Kawasaki, Y.-X. Guo / Journal of Physiology - Paris 96 (2002) 531–537
occurs whatever the spatial orientation of a neighbor with respect to the fish. The phase computation mentioned above, however, is inherently sensitive to the orientation of the electric field created by a neighbor since phase information is detected by phase sensitive neurons in the ELL as the phase difference between two areas on the body surface [8]. The phase sensitive neurons in the ELL respond to D=jA jB, where jA and jB are phase at body area A and B, respectively. Dj may switch its sign for different orientations of neighbor’s electric field. The switching of the sign of Dj is equivalent to the switch of the sign of Df and thus neurons may err in determination of the sign of Df showing orientation ambiguity [6]. In the present study, the effect of the orientation of the electric field of the neighbor on the selectivity to the sign of Df was also examined to determine the degree of ‘orientation ambiguity’ in neurons in the torus.
2. Material and methods Fifty-five Gymnarchus niloticus (12–16 cm) were used. They were obtained from a local dealer and maintained under a light–dark condition of L:D=12:12. Fish were initially anesthetized with MS-222 (1:10 000) and then immobilized with curare (Flaxedil, 0.03%, 3 to 8 ml IM). The preparation was submerged in water in the tank identical to the one used in a previous behavioral study [6]. The gills were perfused with oxygenated water during surgery and physiological experiments. Water temperature and conductivity were maintained at 25–27 C and 100–200 mS/cm for maintenance tanks, the experimental tank, and perfusion water. After local application of Xylocaine (2%), a small hole was drilled above the torus. The medial edge of the optic tectum was gently pressed outwards to visualize the anterior tip of the torus. The MD and LA subdivisions of the torus were visually localized [1]. The use of animals in all experiments were approved by the Animal Care and Use Committee of University of Virginia (protocol #1904). Electrosensory stimuli were given with the ‘free-field’ configuration used in Kawasaki [6]. A sinusoidal signal, S1, which mimicked the fish’s own silenced EOD, was given between an electrode in the mouth and an electrode placed near the tail. The frequency of S1, f1, was constant and was within 20 Hz of EOD frequencies before curarization of the fish. A sinusoidal signal, S2, which mimicked the EOD of a neighbor, was given through one of four pairs of electrodes as shown in Fig. 1. The frequency of S2, f2, was selectable so that the frequency difference, Df=f2 f1, was 4, 2, 1, +1, +2 or +4 Hz. Amplitude of S1 was set at 1–2 mV/cm near the gill cover. Amplitude of S2 was set at 20% of S1. Stimulus signals were created by custom made
software and delivered through a DA converter (DA3-4, Tucker Davis Technology) and FET based stimulus isolators. Neurons in the torus were recorded by the in vivo whole-cell recording technique, which allowed longterm (typically more than 30 min) stable recording [9,10]. Intracellular potentials were recorded in current clamp mode (Axoclamp 2B, Axon Instruments) and fed to an AD converter (AD-3, Tucker Davis Technology) for on-line and off-line analyses. Neurons were sought using a stimulus condition Df= 2 and +2 Hz. When extracellular spikes that were modulated with each cycle of df were detected, an attempt was made to establish a giga-Ohm seal by applying gentle suction to the electrode. A seal with more than 1 G was typically established. The cell membrane was opened by applying a brief (500 ms) positive current (1–2 nA). Access resistance was maintained less than 100 M during the entire recording session. After establishing whole-cell recording, all four orientations of S2 were first quickly tested to determine the orientation at which the modulation of spiking was most strongly heard from the audio monitor. In this orientation, histograms were made for Df= 4, 2, 1, +1, +2, and +4 Hz to determine which |Df| produces the highest spike rates. With this best |Df|, spike histograms were made for Df < 0 and Df > 0 for all four orientations of S2. In each orientation, two parameters indicating differential responses to the different signs of Df were computed. SS was computed as SS=(spkDf >0 spkDf <0)/ (spkDf > 0+spkDf <0) where spkDf < 0 and spkDf > 0 are mean spike rates for Df < 0 and Df > 0, respectively. p was computed as the significance value of the unpaired t-test between spike counts for Df < 0 and Df > 0. Whereas SS expresses the response ratio between the difference and the mean spiking rate, p expresses the
Fig. 1. Four orientations of S2. S2 were delivered through one of four pairs of electrodes located in a circular (15 cm in diameter) fashion. S1 was given between two electrodes in the mouth and at the tail (not shown).
M. Kawasaki, Y.-X. Guo / Journal of Physiology - Paris 96 (2002) 531–537
significance of the difference. Neurons were determined to be ‘sign-selective’, or to have ‘preferred’ a sign of Df in a given orientation of S2 whenever p < 0.05 regardless of the magnitude of SS.
3. Results Whole-cell recordings were made from 165 neurons in 81 fish. Fig. 2 shows two examples of intracellular potentials from the whole-cell recording. The subthreshold postsynaptic potential was generally 5–15 mV in amplitude but it often exceeded 25 mV as in the case shown in the figure. Spikes commonly reached nearly 0 mV but in some cases spike amplitude was smaller. Histograms were made using a spike threshold set at approximately the half height of the spikes. In 77 neurons, spike histograms were successfully made for Df < 0 and Df > 0 for all four orientations of S2, and response
Fig. 2. Whole-cell intracellular potential in response to joint modulation of amplitude (solid line) and phase (broken line). Neuron in A preferred Df<0 to Df>0. Neuron in B responded equally to Df<0 and Df>0.
533
properties were confirmed to be unchanged after repeated presentation of stimulus sets. Other neurons that showed altered responses during the course of recording were discarded from further analysis. All 77 neurons except for one showed sign-selective responses in at least one orientation. Four of them consistently preferred one sign of Df for all four orientations (Fig. 3A and B). Other neurons showed a variety of response patterns for the sign of Df depending on the orientation. An example shown in Fig. 3C showed strong selectivity to Df < 0 in orientations 1 and 2, and only weak non-selective responses in orientations 3 and 4, whereas the example illustrated in Fig. 3D preferred Df > 0 in orientation 3 but the response magnitude in this orientation was much smaller than those in nonselective responses in the other three orientations. To evaluate the importance of the sign-selective responses over non-selective responses in different orientations, the relative strength of sign-selective responses compared with the non-selective responses was examined in all 77 neurons. First, the means of spike counts for Df < 0 and Df > 0, msel, were computed for each orientation where sign-selectivity existed. The means of spike counts for Df > 0 and Df < 0, mnonsel, were also computed for each of all other orientations where no sign selectivity was shown. R was computed as max(msel)/ max(mnonsel) where max(msel) and max(mnonsel) were the largest means. The distribution of R is shown in Fig. 4. Neurons were divided into Type-I (n=60) and Type-II (n=17) by an arbitrary value of R=1. Type-I neurons (R > 1) were analyzed further since they showed stronger responses in a sign-selective orientation than in nonselective orientations and are likely to participate in the control of the JAR. Type-I neurons showed sign-selectivity in one (n=1), two (n=15), three (n=21), or all four orientations (n=23). When multiple orientations showed sign selectivity, some neurons showed consistent preference to the sign of Df (Fig. 5A and B); other neurons switched preference to the sign of Df according to the orientation (Fig. 5C–F). When the preference switched with the change of orientation, the magnitude of responses was different for different orientations in most cases (e.g. Fig. 5F). The response patterns of Type-I neurons are summarized in Table 1. The orientation with the strongest differential responses to the two signs of df is referred to as ‘best orientation’ and it is defined as the orientation in which the difference between spike counts for Df < 0 and Df > 0 was the largest. All orientations were equally likely to be the best orientation in Type-I neurons. The preferred signs of Df at the best orientation across neurons showed no statistically significant bias towards one sign (Table 2). Signselective neurons showed a range of SS. |SS| at the best orientation was 0.36 0.19 (Table 2). Examples of Type-II neurons are shown in Fig. 6. These neurons were primarily driven by either
534
M. Kawasaki, Y.-X. Guo / Journal of Physiology - Paris 96 (2002) 531–537
amplitude or phase and showed little preference in the orientation in which response magnitude was large.
4. Discussion and conclusions The first finding of this study is that neuronal sensitivity to the temporal patterns of inputs that are specific to the sign of Df appears in the torus of Gymnarchus. The neurons in the ELL that project to the torus respond to either amplitude modulation (solid sinusoidal lines in Fig. 2) or phase modulation (dotted lines in Fig. 2) which are identical for Df < 0 and Df > 0 except for their mutual temporal relation. The neurons in the ELL, and thus the inputs to the torus, do not differentiate the sign of Df [9]. The difference in the temporal relation in the input, however, was detected by neurons in the torus. Although the magnitude of the differences
Fig. 4. Distribution of R=max(msel)/max(mnonsel) across neurons. max(msel)=0 was assigned for a neuron that did not show sign-selectivity in any orientations. R was infinite (Inf) when all orientations showed selectivity and thus max(mnonsel) was zero.
Fig. 3. Neurons showing sign-selective responses with consistent preference to the sign of df for all orientations (A, B). Neurons with sign-selectivity in limited orientations (C, D). In each of A–D, rows are for four orientations of S2, left and right columns are for Df<0 and Df>0, respectively. In this and all following histogram figures, the ordinate shows spike rate (spikes/s) and the abscissa shows time in ms. Sinusoidal curves in the histograms show the time course of amplitude (solid) and phase (broken) modulation as in Fig. 2. ‘=’ marks between the histograms indicate that histograms for Df <0 and Df>0 did not show a statistically significant difference in spiking rate. ‘ <’ and ‘> ’ marks indicate that the difference in spike rates was statistically significant. p is the p-value by the t-test.
M. Kawasaki, Y.-X. Guo / Journal of Physiology - Paris 96 (2002) 531–537
in responses for Df < 0 and Df > 0 varied from neuron to neuron, a majority of neurons in the torus showed sensitivity to the sign of Df (Table 2). How do inputs from amplitude and phase sensitive neurons interact in the sign-selective neurons in the torus to create the selectivity to the sign of Df ? The mode of interaction could be understood in some
535
neurons by inspecting response histograms. For example, the neuron in Fig. 5A responded when amplitude reached its maximum both in Df < 0 and Df > 0 histograms in orientation 2, and this response appeared to be differentially gated by the phase input. Similarly, the neuron in Fig. 5D responded to the same phase value and these responses were differentially gated by ampli-
Fig. 5. Type-I neurons showing a variety of the pattern of sign-selectivity over different orientations. A, consistent selectivity in orientations 1, 2, and 4. B, consistent selectivity in orientations 1 and 4. C, D Mixed preferred signs. E, F, Strong selective responses in one orientation with weaker and opposite selective responses in other orientations.
536
M. Kawasaki, Y.-X. Guo / Journal of Physiology - Paris 96 (2002) 531–537
tude changes. In other neurons response patterns were more complex. Further analyses of synaptic interaction of inputs require independent modulation of amplitude and phase components of the signal that was not possible in the present stimulus configuration. Intracellular recording of postsynaptic potential using the phase chamber in which amplitude and phase can be independently modulated [6] will be the subject of future study. The second finding of the present study is that neurons in the torus generally show highly variable responses with different orientation of S2 as predicted by the computational rules that have been determined by behavioral studies [6]. Response magnitudes, sign selectivity expressed by SS, and preferred sign of Df in these neurons were generally different for different orienta-
tions. Amplitude and phase modulation that affect these neurons occur at particular parts of the body surface with a given orientation of S2 and this spatial pattern of modulated sites shifts when orientation of S2 is altered. The variability of the responses of these neurons may be explained by the shift of modulation sites and a small size of receptive field of these neurons. These neurons cannot encode the sign of Df unambiguously because of the orientation dependent responses. Behavioral experiments have shown that the fish performs the correct JAR with any one orientation of S2 indicating that fish obtains unambiguous information about the sign of Df with any given orientation. The ambiguous neurons in the torus, however, may participate in coding of the sign of Df by means of population coding. For the population coding, the participating neurons showing signselective responses in one orientation of S2 should not interfere with or cancel the sign-selective responses of other neurons that are sign-selective in other orientations. Some neurons found in the present study fulfill this condition having large values of R (Fig. 3C) or showing differential magnitudes of responses when the preference to Df switches with orientation (Fig. 5E and F), whereas other neurons are not suited for encoding of the sign of Df (e.g. Fig. 3D).
Table 1 Type-I neurons (n=60) Number of sign-selective orientations
Consistent sign of preference over orientations
Mixed signs of preference over orientations
1 2 3 4
n=1 n=12 n=7 n=4
n=1 n=14 n=19
Table 2 Comparison of sign selectivity in different best orientations in Type-I neurons Best orientation
Number of neurons total (Df<0/Df>0) Mean of |SS| Standard deviation of |SS| Maximum of |SS| Maximum of |SS|
Total
1
2
3
4
12* (4/8) 0.35 0.24 0.91 0.16
11* (7/4) 0.44 0.15 0.67 0.22
19* (10/9) 0.37 0.16 0.62 0.11
18* (5/13) 0.35 0.20 0.59 0.06
60(26**/34**) 0.36 0.19 0.91 0.06
* Difference across orientations not significant (w2-test, p=0.34). ** Difference between positive and negative selectivity not significant (sign test, p=0.18).
Fig. 6. Examples of Type-II neurons. A, Weak sign-selective responses in orientations 1 and 3 and strong non-selective responses in orientations 2 and 4. B, no selectivity in any orientations.
M. Kawasaki, Y.-X. Guo / Journal of Physiology - Paris 96 (2002) 531–537
Four neurons that showed consistent preference to one sign of Df over all four orientations (Fig. 3A and B) may be higher order neurons that receive inputs from other ambiguous neurons in the torus and represent the unambiguous information about the sign of Df. If the neurons in Fig. 3A and B, respectively make excitatory and inhibitory connections to the pacemaker that drives the EOD and they are mutually inhibited, the correct JARs would occur in any orientation of S2. The data obtained in Gymnarchus in this study give insights into comparative physiology of JAR in Eigenmannia and Gymnarchus [5,11]. In both genera, the first site of interaction between amplitude and phase information was found to be the midbrain structure, the torus semicircularis. The degree of interaction, or abundance of neurons that integrate amplitude and phase information, appears to be similar in these fish.
References [1] A.H. Bass, C.D. Hopkins, Comparative aspects of brain organization of an African ‘‘wave’’ electric fish, Gymnarchus niloticus, J Morph 174 (1982) 313–334. [2] W. Heiligenberg, Jamming avoidance responses: model systems for neuroethology. In: T.H. Bullock, W. Heiligenberg (Eds.), Electroreception, Wiley & Sons, New York, 1986, pp. 613–649.
537
[3] W. Heiligenberg, Neural nets in electric fish, The MIT Press, Cambridge, MA, 1991. [4] W. Heiligenberg, C. Baker, J. Matsubara, The jamming avoidance response in Eigenmannia revisited: the structure of a neuronal democracy, J Comp Physiol 127 (1978) 267–286. [5] W. Heiligenberg, G. Rose, Gating of sensory information: Joint computations of phase and amplitude data in the midbrain of the electric fish Eigenmannia, J Comp Physiol 159 (1986) 311– 324. [6] M. Kawasaki, Independently evolved jamming avoidance responses employ identical computational algorithms: a behavioral study of the African electric fish, Gymnarchus niloticus, J Comp Physiol 173 (1993) 9–22. [7] M. Kawasaki, Comparative analysis of the jamming avoidance response in African and South American wave-type fishes, Biol Bull 191 (1996) 103–108. [8] M. Kawasaki, Y.-X. Guo, Neuronal circuitry for comparison of timing in the electrosensory lateral line lobe of an African wavetype electric fish, Gymnarchus niloticus, J Neurosci 16 (1996) 380– 391. [9] M. Kawasaki, Y.-X. Guo, Parallel projection of amplitude and phase information from the hindbrain to the midbrain of the African electric fish, Gymnarchus niloticus, J Neurosci 18 (1998) 7599–7611. [10] G. Rose, E. Fortune, New techniques for making whole-cell recordings from CNS neurons in vivo, Neurosci Research 26 (1996) 89–94. [11] G.J. Rose, W. Heiligenberg, Neural coding of difference frequencies in the midbrain of the electric fish Eigenmannia: reading the sense of rotation in an amplitude-phase plane, J Comp Physiol 158 (1986) 613–624.