Brain Research Reviews 32 Ž2000. 16–28 www.elsevier.comrlocaterbres Short review Electrical synapses, a personal perspective žor history/ Michael V.L. Bennett ) Department of Neuroscience, Albert Einstein College of Medicine, 1300 Morris Park AÕenue, Bronx, NY 10461, USA Abstract Gap junctions are the morphological substrate of one class of electrical synapse. This memoir records the author’s involvement in the development of our knowledge of the physiology and ultrastructure of electrical synapses. The answer to whether neurotransmission is electrical or chemical is either. One lesson is that Occam’s razor sometimes cut too deep; the nervous system does its operations in a number of different ways and a unitarian approach can lead one astray wM.V.L. Bennett, Nicked by Occam’s razor: unitarianism in the investigation of synaptic transmission, Biol. Bull. 168 Ž1985. 159–167x. Electrical synapses can do many things that chemical synapses can do, and do them just as slowly. The new molecular, cellular and physiological techniques will clarify where gap junctions and electrical coupling do and do not occur and permit experimental manipulation with high specificity. q 2000 Elsevier Science B.V. All rights reserved. Keywords: Gap junction; Electrical synapse; Connexin; Coupling Contents 1. Introduction . ....................................................................... 2. An aside on ‘‘ephapses’’ and other nomenclatural niceties . 16 .............................................. 17 ...................................................... 18 ....................................................... 20 ................................................................... 22 6. Electrical versus chemical ................................................................ 24 7. An aside about evolution . ................................................................ 24 .................................................................. 25 ..................................................................... 27 .......................................................................... 27 3. The first connexin based electrical synapses! 4. What next? Electric organ control systems 5. Fast motor systems . 8. Romance in academia Acknowledgements . References 1. Introduction It seems like yesterday that I started to work on electrical communication between neurons. Well, actually not, it ) seems quite a long time ago, and the field, and I, have matured significantly over the years. The operation of neurons at the level of electrical signaling, at least at the level of questions that most of us were asking more than Tel.: q1-718-430-2536; fax: q1-718-430-8944; e-mail: [email protected] 0165-0173r00r$ - see front matter q 2000 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 5 - 0 1 7 3 Ž 9 9 . 0 0 0 6 5 - X M.V.L. Bennettr Brain Research ReÕiews 32 (2000) 16–28 40 years ago, is pretty well understood, not completely of course, but there are no very large black boxes. It seems much more like yesterday, and is, that I last wrote a review on the development of knowledge of electrical transmission w73x. What has happened in the last couple of years? Aside from substantial but incremental progress, there has been the discovery of neuron specific connexins w26,63x and of connexin diseases that involve the nervous system w15,22,38x. Although no diseases of electrical synapses have been discovered yet, there are sure to be mutations of neuron specific connexins, whether or not the mutations lead to an observable phenotype. Interestingly, mutations of Cx26 and Cx32, which are expressed by some neurons in the CNS, have yet to show any phenotypic changes in those neurons. Gene targeting is progressing through the connexin gene family. And there is the discovery that gap junctions of the Ecdysozoa Žthe nematode, arthropod line. are encoded by a gene family completely unrelated to the connexins, although there are many convergent Ž not conserved. properties between the junctions encoded by that gene family and by the connexins. Since the new connexins and connexin diseases are extensively discussed in this volume by others more directly involved in the research, and I just participated in writing a review of connexin diseases w15x, I will make this presentation more an oral history than an integrative summary and analysis. One does not have to have been at Oxford to define a synapse as a specialized site of functional interaction between neurons, although Sherrington and Eccles Žand I. were. By this definition gap junctions form one class of electrical synapse w13x. Another kind of electrical synapse mediates short latency inhibition of the Mauthner cell of teleost fishes and possibly mammalian cerebellar Purkinje cells; this form of electrical transmission is not mediated by gap junctions, and involves different junctional specializations Žsee below. w28x. In addition, there probably are electrical interactions that occur between closely apposed cells without obvious gap junctions or specializations other than the absence of interposed glia w28,36,67x. Whether these sites are to be considered synapses, i.e., specialized, or ephapses, i.e., incidental or accidental sites of interaction, may become clear with greater knowledge of the developmental mechanisms. Without deciding on a name one can still describe the electrical interaction, which does indeed appear to be uniquely associated with the close appositions. 2. An aside on ‘‘ephapses’’ and other nomenclatural niceties Angelique Arvanitaki w1x coined the term ephapse from the Greek to mean an apposition that is not quite so close as a synapse Žaccording to my Greek–American colleague, George Dimitrios Pappas.. She used it to denote what she thought of as artificial synapses, which she made by laying 17 one axon along side another where the ‘‘action currents’’ generated during an impulse in one axon altered the excitability of the other axon. This terminology suggests that she thought synaptic transmission was electrical, at least that is my recollection, and, in the spirit of oral history, I will not go to the library to try to confirm that view. Ephapse then came to be used for incidental contacts in the nervous system, particularly where activity in one or more axons excited other axons. These days one might wonder if actual gap junction electrical synapses were formed between axons in injured tissue. My mentor at that time, Harry Grundfest, had embraced the idea that chemical transmission was mediated at electrically inexcitable membrane, i.e., in explicit, modern terms that the conductance of the neurotransmitter receptors was independent of membrane potential and only a function of transmitter concentration. To keep transmission at synapses chemically pure, he decided to use ephapse to denote morphological specializations between neurons where transmission was electrical. Although my Oxford education led me to disagree with this practice, to keep peace in the laboratory I used ‘‘electrotonic junctions’’ for one type of what I now freely term electrical synapses. At that time I felt that ephapse connoted artificiality or incidentality, which downgraded the importance of my work. Electron microscopy now makes it clear that gap junctions are closer appositions than occur at chemical synapses, and, if one were starting over, one would call chemical synapses ephapses and gap junctions between neurons synapses. In my view both he and Jack Žlater Sir John. Eccles made a mistake in thinking that only one mode of transmission could be synaptic, but Eccles was better at changing his mind w14x. I recall when David Potter presented his and Edward Furshpan’s work w31x on the crayfish giant motor synapse at the Monday night electrobiology seminars at the Marine Biological Laboratory ŽMBL, Woods Hole.. This work and the independent studies of Akira Watanabe w70x on the cardiac ganglion of the mantid shrimp were the first unequivocal demonstrations of electrical transmission between neurons. The electrobiology sessions were organized by Harry, who welcomed airing of all views, and were known as the Monday Night Fights because of the sometimes heated discussion and by analogy with the Friday Night Fights, a popular program at the time showing professional boxing matches. Harry suggested in the discussion period, or possibly before, since interruptions were not uncommon, that they should call their rectifying synapse an ephapse, because it was electrical although it was electrically inexcitable. David pointed out to him with evident pleasure that the junctional conductance was a function of voltage and thus was electrically excitable. Harry did not have a good retort, which was unusual for him. Of course gap junctions between segments of the septate axon are electrically linear over a wide range w37,71x, while connexin based gap junctions all show some 18 M.V.L. Bennettr Brain Research ReÕiews 32 (2000) 16–28 degree of dependence on transjunctional voltage Že.g., Ref. w34x.. In a collection of gap junction papers, it should not be necessary to obsess about the relative merits of chemical and electrical synapses. We workers in the field for the most part think highly of what we do. Still, in this contribution it may be worthwhile to do a little complaining, qvetching or whining, depending on one’s ethnic origin. For example, consider the ‘‘medical subject headings’’ ŽMeSH. for the PubMed data base. Under ‘‘Synapses’’ is: Specialized junctions at which a neuron communicates with a target cell. At classical synapses, a neuron’s presynaptic terminal releases a chemical transmitter stored in synaptic vesicles which diffuses across a narrow synaptic cleft and activates receptors on the postsynaptic membrane of the target cell. The target may be a dendrite, cell body, or axon of another neuron, or a specialized region of a muscle or secretory cell. Neurons may also communicate through direct electrical connections which are sometimes called electrical synapses; these are not included here but rather in GAP JUNCTIONS. And then under ‘‘Synaptic Transmission’’: The communication from a neuron to a target Žneuron, muscle, or secretory cell. across a synapse. In chemical synaptic transmission, the presynaptic neuron releases a neurotransmitter that diffuses across the synaptic cleft and binds to specific synaptic receptors. These activated receptors modulate ion channels andror second-messenger systems to influence the postsynaptic cell. Electrical transmission is less common in the nervous system, and, as in other tissues, is mediated by gap junctions. Thus, to look for the latest in electrical transmission or electrical synapses, one has to take a somewhat devious route and examine all those other citations that come along with gap junctions and nervous system. Looking for earlier papers is more complicated because Gap Junctions as a MeSH term was not introduced until 1994, and you cannot search for the phrase ‘‘electrical synapse’’, although ‘‘electrically synaptic transmission’’ is in the Compound Word Dictionary and yields 12 citations. Another bit of whining for the in-group: even some workers in the gap junction field have trouble using electrical with respect to PSPs, no doubt influenced by Harry. Korn and Faber write about ‘‘coupling potentials’’ at electrical synapses rather than PSPs. J.G.R. Jefferys w36x in a Physiological Review considers ‘‘four classes of non-synaptic interaction, mainly in the mammalian brain’’ of which the first is ‘‘Electrotonic Žand chemical. coupling through gap junctions’’. Yet he also writes of ‘‘gap junctions, which commonly serve as electrical synapses in invertebrates but appear to be used less often for electrical signaling in vertebrates’’. The reader may feel that there is too much discussion of terminology here. There probably was too much quarrelling over nomenclature, but some of the controversy represented real differences in concepts rather than egodriven preference. We should all be familiar with Feldberg’s Dictum, which is that a scientist would rather use another scientist’s toothbrush than his terminology. I believe I heard this from ŽSir. Bernard Katz, who cited it in one of his lectures. It is a delightfully apt phrase in that words have a flavor of their origin andror meaning. When I was a child, a not uncommon punishment for use of foul language was washing the offender’s mouth out with soap; thus, the mystical view seems to be that dirty words physically soil the speaking apparatus. In my own case, speaking of alpha and beta connexins make me want to brush my teeth with my toothbrush. Jean Paul Changeaux once chided me for calling gap junctions between neurons electrical synapses, when the same structures were called gap junctions when they occurred between non-neuronal cells. The venue was a sidewalk cafe in Paris, and the statement should not be taken very seriously. Moreover, as chemical interactions between non-neuronal cells have become more widely described and as the molecules responsible for exocytosis and endocytosis at synapses have proved to have homologs in non-neuronal cells, the same criticism might be lodged about the terminology for chemical synapses. 3. The first connexin based electrical synapses! The supramedullary neurons of the puffer fish, Spheroides maculatus, were the subject of my first experiments at the MBL in Woods Hole. These large neurons Ž0.2–0.3 mm in diameter. sit on the dorsal surface of the medulla and can be seen with the naked eye, at least with my eyes at that time ŽFig. 1.. I no longer remember where Harry Grundfest found out about them, possibly from Shigehiro Nakajima and Susumu Hagiwara, who later studied their action potential generation, but the cells were known to early comparative anatomists including Sigmund Freud. Large neurons were of interest, because they were relatively easy to study with sharp intracellular microelectrodes and patch electrodes were far in the future. The function of the neurons was unknown, and we were able to show that they were effector cells sending their axons out the dorsal roots to the skin. But nothing obvious happened in the skin when they were stimulated. Recent data demonstrate that they contain gastrinrcholecystokinin and innervate mucous glands w29x, and a secretomotor function should be more carefully investigated. But that is comparative physiology, which is primarily of interest to NSF. What is more relevant to general physiology and this M.V.L. Bennettr Brain Research ReÕiews 32 (2000) 16–28 19 Fig. 1. Anterior spinal cord of the puffer viewed from the dorsal side. The posterior limit of the cerebellum is to the left. The supramedullary neurons are the large round cells, about 250 mm in diameter, that are located on the surface of the cord. Several of the cells that had been penetrated for intracellular recording are dark because of increased staining by toluidine blue applied to the surface wfrom Ref. w19xx. discussion is the observation that they fire synchronously in response to cutaneous inputs. The initial observation of synchronous firing was the accidental result of putting two electrodes in adjacent cells when trying to get two into one cell for separate current application and voltage recording. What was one to think in 1957 when one saw synchronous firing? I thought that a higher level synchronizing center was exciting the supramedullary neurons, and that the large depolarization that initiated the overshooting spike was a PSP Žimplicitly chemically mediated. generated by inputs from that center. One afternoon Eccles was visiting the laboratory while I was recording. He looked at the oscilloscope screen, saw the two component spike and said of my synchronizing input ‘‘That’s an initial segment spike’’. He advised me to advance an electrode beneath the cluster of cell bodies to record from the axons directly. I tried it, and, of course, he was right. It was quite easy to find two component spikes that characterized an axonal recording and then to hyperpolarize one by one the overlying somata until the soma of origin was identified. But in addition to finding axon spikes and clarifying the nature of the two components of the spike recorded in the soma, I also found coupling. The cells are coupled by gap junctions between their axons w20x. Thus, when one hyperpolarized an overlying cell body that did not belong to an axon being recorded from, the hyperpolarization due to coupling was larger than in the cell body giving rise to that axon and occasionally big enough that even the unprepared mind could not miss it. We were rapidly convinced that the coupling was responsible for the synchronization. What had not been obvious is that mutual excitation between the cells was electrical. An action potential directly evoked in one cell could spread to the rest of the cells in the cluster and this spread showed paired pulse facilitation, although we did not call it that. The period of increased excitability could be as long as 200 ms, which we thought suggested a chemical rather than electrical mechanism, but subsequently it proved to be explained by a long-lasting depolarizing afterpotential. And the rest is history. Let us be frank here. The presence of coupling was put in a footnote in the 1959 puffer papers and discussed at slightly greater length in a few abstracts. Full publication took about 7 years w9,20x. NIH was more forgiving and the race for priority was not so hectic as it is now. Nor did we know that connexins and Ecdysozoan gap junction proteins were different families. Although many of the implications of electrical coupling of supramedullary neurons were not immediately obvious and only became clear as other examples of coupling were discovered, the system has relevance to mammalian systems. First, electrical synapses can serve a synchronizing function, but the degree of synchronization need not be very precise; the spikes in different cells in the cluster can be quite dispersed in time. Propagation of impulses between cells can be slow compared to the delays at chemical synapses and in some species the safety factor for propagation can be less than one in that an impulse in 20 M.V.L. Bennettr Brain Research ReÕiews 32 (2000) 16–28 one cell is not always accompanied by an impulse in all the other cells; the safety factor depends on the presynaptic action potential, strength of coupling and excitability of the postsynaptic cell. These conclusions could also have been drawn from Watanabe’s mantid shrimp data w70x. He observed coupling of bursting neurons controlling heart rate. He thought the cells were connected by cytoplasmic bridges, which is probably wrong, but recognized the synchronizing function of the coupling. Second, the synchronizing function of electrical synapses involves hyperpolarization of more positive cells, as well as depolarization of more negative cells. Coupling is a two way street, and a significant fraction of a cell’s input conductance can be the input conductance into its electrical synapses with other cells. Third, impulses are more likely to spread between cells when they are depolarized by other synaptic inputs, chemical as well as electrical, and there can be both spatial and temporal summation of electrical PSPs. All these features are trivial if not obvious consequences of coupling by gap junctions. Although we were concerned about the morphological basis of electrical coupling, the puffer was not a great preparation in which to look for gap junctions, and associating gap junctions with electrical transmission came later. 4. What next? Electric organ control systems This was not the time when hunting for further electrical synapses crossed my mind, but it proved possible to blunder upon them. Harry Grundfest had been interested before I arrived at P & S Žthe College of Physicians and Surgeons of Columbia University. in how weakly electric fishes generated their electric pulses and how the organ discharge was controlled. These fishes, depending on the species, emit brief pulses with relatively long intervals between them or pulses that are separated by an interval comparable to the pulse duration ŽFig. 2.. The former group, pulse fish, modulated their discharge frequency in response almost any mode of stimulation, whereas the latter group, wave fish, tended to have a very constant frequency. Stimulating the spinal cord of wave fish did not cause acceleration. When examined carefully, there was a slight phase advance, which we now know to be due to depolarization from antidromic activity spreading into the pacemaker nucleus in which the frequency is set. Still, the constancy was impressive in the face of a stimulus that activated sensory inputs and caused a dramatic acceleration in the pulse fish. Akira Watanabe, he who had shown the coupling between cardiac ganglion cells of the mantid shrimp, took a more subtle approach. Asking what a wave fish would do when presented with a stimulus of nearly its own frequency, which would certainly happen in the gregarious species, he discovered the jamming avoidance response. Presented with a sinusoidal stimulus near its own frequency, the fish either accelerates or decelerates its own Fig. 2. Patterns of electric organ discharges in teleosts. ŽA. An electric catfish, Malapterurus electricus, activity recorded head positivity upward. Mechanical stimulation evoked a train of five pulses with a X maximum frequency of ;190rs. ŽA . A single pulse could also be evoked. Recorded at a faster sweep speed. ŽB–D. Discharge of weakly electric gymnotids, South American fishes, recorded head positivity upwards. ŽB. A pulse fish, Gymnotus carapo, emits pulses at a basal frequency of ; 35rs. Touching the side of the fish at the time indicated by the downward step in the lower trace caused an acceleration up to X ;65rs. ŽB . At a faster sweep speed, the single pulses show three phases, initially head negative. ŽC. Sternopygus macrurus, a wave fish, discharges at ; 55rs. The horizontal line indicates the zero potential level. The discharge has little DC component. ŽD. Sternarchus (Apteronotus) albifrons, a high frequency wave fish, emits biphasic pulses at ;800rs. The horizontal line indicates the zero potential level. Calibrations in volts and milliseconds wfrom Ref. w11xx. discharge to increase the frequency difference and thereby reduce interference. The central pathways and physiology of this response and of electroreception in general were extensively and productively explored by Walter Heilegenberg, who was tragically killed in an airplane crash w32x. Walter had many gifted collaborators. With the background of gross stimulation of the electric fish and after more or less exhaustingly reporting the modes of operation of electric organs w8,11x, Emilio Aljure and I looked in the spinal cord and then medulla of mormyrid electric fishes. These species are pulse fishes and generate very brief discharges - 0.5 ms in duration. Since the generating cells, or electrocytes, emit bi- or triphasic pulses, very precise synchrony is required to prevent cancellation of out of phase activity. Although the electromotor neurons were not visualizable in the spinal cord, it was not that hard to penetrate neighboring cells and demonstrate electrotonic coupling directly. In these species, the electromotor neurons showed quite large diameter dendrodendritic connections Žand with uniform staining the cells can appear syncytial or multinucleate, Fig. 3.. Now here was a preparation that one could, without shame, ask one’s anatomical colleagues to examine. Yasuko Nakajima and George Pappas soon showed that there were close membrane appositions between the dendrites ŽFig. 3. w17x. We would now call these structures gap junctions, although the gap was not resolved in the early pictures. We did suggest that ultrastructural examination could prove useful for identifying synapses between dendrites where electrical measurements were hard to obtain. My colleague, Dominick Purpura, at that time a mammalian neurophysiologist, did not approve of this suggestion, but M.V.L. Bennettr Brain Research ReÕiews 32 (2000) 16–28 21 Fig. 3. Neurons appearing syncytial with the light microscope, medullary electromotor relay neurons of the African electric fish, Gnathonemus sp. ŽA. In a silver stained preparation ŽRomanes’ method. a thick process appears to connect the two cell bodies without there being any intervening membrane. ŽB. Electron microscopy reveals membranes across these processes, between the arrows in this example. A capillary is present on the lower left. Axon terminals Ža. form morphologically defined chemical synapses on the cell somata ŽS.. ŽC. At a similar region of apposition between spinal neurons, higher magnification shows large regions, where the extracellular space is greatly diminished, and the membranes appear fused between the arrows. Quasi-periodic dots in the center of the apposition in this osmicated preparation result from superposition of images of stained channels wfrom Ref. w17xx. then he thought he was recording activity of chemical synaptic inputs to the dendrites. Examination of several South American Žgymnotid. electric fishes also showed electrical coupling in the electromotor system, and we put forth the generalization that if an organism wanted to perform a highly synchronous act, such as an electric organ discharge, the controlling neurons should be electrically coupled. In these electromotor sys- tems, the precision of synchronization was greater than could be provided by reciprocal chemically mediated excitation with its attendant synaptic delay. The existence of reciprocal presumably excitatory chemical synapses between less synchronously active neurons, the amacrine cells of the retina, was pointed out to me by Paul Fatt, and reciprocal excitatory chemical synapses also exist in coelenterates w72x. At a meeting of the Neuroscience Research 22 M.V.L. Bennettr Brain Research ReÕiews 32 (2000) 16–28 Program, I floated the name ‘‘69 synapses’’, by analogy with Gray’s type I and II as well as an uncommon sexual practice, but this nomenclature never caught on. It may occur to the perceptive reader that synchronizing a controlling nucleus precisely does not mean that activity of the effector cells will also be precisely synchronized, since the conduction distances from the control center can be quite different. I will return to this point below in considering the issue of synaptic delay. 5. Fast motor systems By this time one could generalize that synchronous activity of electric organs required electrical transmission Žor a single cell command nucleus.. However, cats do not have electric organs, and my medical school mammalian colleagues could easily ignore the data, i.e., the work of my colleagues and me. To be sure, Purpura invited me to present the story at a meeting on the thalamus and I argued that electrical synapses might contribute to thalamic oscillations, although the oscillations are explicable in terms of recurrent inhibition and spontaneous spike activity w9x. No one, particularly Eccles, liked the idea. There is still the possibility of electrical synapses in this system, and evidence for electrical coupling in other pools of synchronously active inhibitory neurons accumulates Že.g., Ref. w51x.. Another approach was to look further in fishes, with which I was familiar, and investigate motor behavior that involved fast and synchronous activity. The sonic motoneurons of the toadfish were a candidate, since the sonic muscle, which wraps around the swimbladder, contracts at the fundamental frequency of the sound, 100–200 Hz, a frequency comparable to that of many electric organ discharges. This fish was common in Woods Hole and was being used for its isolated pancreatic islet tissue. Its muscle had also been studied because of the fast rate of contraction and relaxation. After a modest amount of development, it became possible to record from the sonic motoneurons identified by antidromic stimulation. The cells were smaller and less accessible than the electromotor neurons we had previously studied, and I only succeeded in recording from one cell at a time. However, graded antidromic stimulation showed short latency graded depolarizations, which identified electrical coupling, and George Pappas showed gap junctions both between motoneuron dendrites and between presynaptic fibers and motoneurons w57x. The antidromic potentials, electrical PSPs, showed paired pulse facilitation, ascribable to increased degree of antidromic invasion as a result of depolarization remaining from the previous response. During this period I met Hans Lissman, who with Machin had discovered the discharges of the weakly electric fishes and demonstrated their role in electrolocation. Although the early anatomists had found the electrogenic tissue of a number of weakly electric fishes, they were unaware of the discharges. I cannot remember whether I actually applied Galvani’s rheoscopic frog preparation to a weakly electric fish, but it certainly would have detected the discharges. Hans was a zoologist and knew about many wonderful animals. Among them was the hatchetfish, which was thought to fly. It is shaped like a hatchet and has a large breast bone to which the pectoral fin depressor muscles attach ŽFig. 4.. Hans had seen it taxi along the surface, which was much more reasonable in terms of wing loading. Because of high speed and short latency of action, the system appeared a promising preparation in which to prospect for electrical synapses. On one of my occasional trips to the tropical fish dealers looking for different species of electric fishes, I saw hatchetfish and brought a few back to the laboratory for anatomical study. The medulla proved to have obvious large axoaxonic synapses between the Mauthner fibers and other large fibers that it did not take a Cajal to see. The large fibers, which we called giant fibers, made contact with both Mauthner fibers, and the giant fibers coursed into the motor nuclei innervating the pectoral fin depressor muscles. This circuitry suggested that each Mauthner fiber could excite pectoral fin motoneurons on both sides of the body. It also suggested that the Mauthner fiber, giant fiber synapses should be electrical to shorten latency and rectifying to keep one Mauthner fiber from exciting the other via the giant fibers. Somewhat to my disappointment, Al Auerbach unequivocally demonstrated that the Mauthner fiber, giant fiber synapse was chemical w2x. The giant fiber, motoneuron synapse proved to be electrical, which was fine, and it was also rectifying, so that this property was not restricted to the crayfish giant motor synapse w3x. Our teleological explanation of the rectification was that it permitted subsets of the motoneurons to be excited by independent inputs to them in order to generate small movements without exciting the giant fibers and thereby all the other motoneurons. Dave Hall using the modern tracer HRP was able to show that there are gap junctions at the rectifying synapses that the giant fibers make on the motoneurons w33x. We now would predict that there are different connexins in pre- and postsynaptic hemichannels w5,56,69x. Mahlon Kriebel and subsequently Henri Korn and I also investigated oculomotor neurons in puffer and goldfish, because saccadic eye movements are fast. The cell bodies were coupled, to which we assigned the task of synchronizing eye movements either for the fast phase of vestibular nystagmus or for eye withdrawal in response to stimulation of cutaneous nerves w41–43,45x. Inputs from the ipsilateral horizontal canal initiated impulses out in the dendrites, where we presumed the cells were not coupled in order to allow graded but rapid saccadic movements. We also hypothesized that there is electrical coupling in the saccade generator. We proposed a possible solution to the problem of generating a synchronous but graded vol- M.V.L. Bennettr Brain Research ReÕiews 32 (2000) 16–28 23 Fig. 4. The Mauthner fiber, pectoral fin circuitry of the hatchetfish. The lower diagrams show side and front views of the fish with the pectoral fins, brain, spinal cord, and innervation of the pectoral fin depressor muscle ŽM. by anterior ŽaN. and posterior ŽpN. nerves. The upper diagram shows the Mauthner cells ŽM. giving rise to the Mauthner fibers ŽMF. which decussate; it also shows a single giant fiber of the several that occur on each side. The cell of origin ŽG. of the giant fiber ŽGF. lies on the side contralateral to its main axon. The GF crosses dorsal to the near MF and forms a single large chemical synapse Žcs. with it. It then passes ventral to the other MF with which it makes several synapses. It sends a process caudally to the pectoral fin depressor motoneurons with which it forms rectifying electrical synapses. It also sends a process rostrally which subsequent work showed ends on jaw muscle motoneurons, probably to close the jaw during fast movement wfrom Ref. w2xx. ley, which was analogous to an electrophysiological stimulator. One nucleus Žor a transistor or a vacuum tube, which at that time still existed in stimulators. would generate an all-or-none synchronous volley using positive feedback, i.e., in neurons, the spike generating mechanism and electrical coupling of the cells. The synchronous volley would then generate a rapidly rising PSP in the appropriate oculomotor neurons that, superimposed on a pre-existing ‘‘excitatory state’’ due to activation of other graded inputs, would determine what fraction of the neurons were excited by the rapidly rising PSP Žor in the stimulator analogy, how large a current flowed through the output stage.. The observation of coupling in the teleost oculomotor system was particular gratifying, because the many similarities to mammalian oculomotor systems made it more likely that electrical coupling would be found in them. Henri looked in the cat and failed to find coupling of oculomotor neurons. We presumed that coupling at this level was not necessary in the cat, because there was a sufficiently synchronous volley from a higher center to drive the oculomotor neurons to generate a saccade. We still think it possible that there is coupling of neurons in the higher center in which saccades are generated. Mahlon used to go snorkeling at lunch time to net his own puffers off a nearby beach. Atlantic puffer was becoming fairly rare, since it had been accepted as a food fish and served in restaurants as sea squab. Many of us growing up on the East coast knew how good puffer was, and how easy to remove the meat from the skin. There were some rumors about poison, but I do not recall their being connected to the Japanese reality of fugu and tetrodotoxin. In those days, Americans did not eat ugly fish such as puffer, skate, and monk fish Ža euphemism for Lophius piscatorius, an anglerfish.. There is now a fugu website Žhttp:rrfugu.hgmp.mrc.ac.uk. devoted to sequencing its genome, which has much less intronic DNA than other vertebrates. Fugu are also raised in mariculture, although I understand that cultured fugu lack tetrodotoxin, presumably because a dietary requirement is missing. Do tetrodotoxin-free fugu sell at a premium or a discount in Tokyo? Could our own S. maculatus make tetrodotoxin, given the right diet? While summering in Woods Hole, Henri found electrically transmitting inputs to primary vestibular neurons in toadfish w43x, which was an extension of Furshpan’s studies of inputs to the goldfish Mauthner cell w30x. He and Constantin Sotelo also found electrical coupling and gap junctions at these synapses in the rat w44x. It was reasonable to think that electrical transmission at this synapse would shorten reaction time for postural adjustment. Interestingly, this result did not generalize to cats. Perhaps cats are so big that the small decrease in response latency provided by electrical transmission at this synapse is not significant. Then think of an elephant. If there is no advantage to electrical transmission at this synapse, one still needs to account for there being chemical transmission. Evolutionary history is a possible explanation and the last refuge of teleological rascals. Startle circuitry remains 24 M.V.L. Bennettr Brain Research ReÕiews 32 (2000) 16–28 interesting as a site where electrical transmission might be present to speed response, but the advantage is going to be restricted to small and fast animals. Considering how rapidly bats can emit distinct cries, we perfused a bat or two for electron microscopy, but availability questions, and the possibility of rabies, kept us from getting deeply involved. We thought about hummingbirds, and Susumu Hagiwara and Ted Bullock captured a few and measured wing beat frequency. Their work on this interesting subject to my knowledge never saw the light of peer review. wThere is nothing in PubMed, which goes back to just about the time these studies were being done.x Shosaku Obara investigated the reflex action of the nictitating membrane of the chicken, a remarkably fast blink. The gross anatomy is intriguing. There are two muscles; one is attached to the membrane by a long tendon and the other provides a stirrup through which the tendon makes an 1808 turn. Thus, the distal part of the tendon moves the distance that the first muscle contracts plus twice the distance that the second muscle contracts, and the speed of membrane movement is nominally three times as fast as the muscles shorten. Sho never did succeed in getting good intracellular recordings from the motoneurons, so the question of electrical coupling remains unresolved. This work only appears in an abstract in the Proceedings of the XXIVth International Congress of Physiology. In chicken and most other birds, the eyeballs are almost in contact back to back. This arrangement makes it difficult to record from the muscles and stimulate the relevant oculomotor nerves. Owls look forward so that their eyes are much more favorably placed to get at the eye muscles. We had a cute burrowing owl around for a long time, but its cuteness and the fact that owls tend to shift gaze by moving their heads rather than just their eyes aborted this approach. 6. Electrical versus chemical The bulk of the initial work on electrical transmission and synchronization was published in 1966 and 1967. The basic hatchetfish story came out in 1969. By this time one could argue Žand I did in too many reviews, e.g., Refs. w7,10,12,13x., that chemical and electrical synapses could each do most things required in the nervous system. The point was not so much that electrical synapses were as good as chemical synapses, but that one had to be careful about establishing what the mode of transmission actually was. There also remained a somewhat adversarial atmosphere. In most instances it was easy to show that inhibition was chemically mediated, but the evidence for excitation was often not compelling. Electrical transmission is faster, but the latency difference is not so great in warm blooded animals, and because of postsynaptic input time constant the measured delay can easily be longer at an electrical synapse than at a chemical synapse. When one is dealing with 0.2 ms for a chemical synapse, it is hard to exclude time to excite and conduction delay in the presynaptic fiber. Chemical synapses are fundamentally unidirectional, but then most single synapses do not excite their postsynaptic targets in any case. Electrical synapses are fundamentally bidirectional, unless they are rectifying, and that may prove to be where they are most useful for mammals. At that time there were few indications of plasticity in the electrically coupled systems, although relatively long-lasting actions could be observed mostly under artificial conditions. The lack of plasticity may have been in part a result of working on systems in which plasticity should not occur, and now numerous instances of cellular controls of junctional conductance have been found Že.g., Refs. w18,35,48,58,60x.. There is a contrarian position to be taken about delays and electrical transmission. As noted above, output cells may be at quite different distances from a synchronously firing command nucleus. The general solution to the problem of differing distances is to slow conduction in the pathway to the more proximal regions of the organ by making the fibers going to that region have a smaller diameter Žand perhaps an internodal distance short enough to reduce conduction velocity w53x andror by making the fibers take a more devious route w11x. In the electric eel, for example, the anterior and posterior extremes of the organ may be separated by over 1 m and at least 10 ms conduction time. Spinal electromotor neurons, which innervate the electrocytes are activated by spinomedullary fibers from a relay nucleus in the medulla w53x. Transmission from the descending fibers is electrical, rise times of the EPSPs are short, and the compensatory delays are primarily if not exclusively in conduction times in the preterminal fibers and the peripheral nerves to the electric organ. Thus, electrical transmission along the axon, by essentially the same in mechanism as transmission across a gap junction, provides the delays. Similar mechanisms may operate in the auditory system where very precise arrival time comparisons are made to permit source localization w24x. An interesting problem is how the system wires itself up to achieve the observed precision. It seems likely that there is feedback in the developmental mechanisms to promote synchronization, but the nature remains obscure. Eccles raised this question in 1959, but I admit that at the time I was more interested in the finished product than how it was made. 7. An aside about evolution The widely held opinion that electrical transmission is characteristic of lower forms probably derives from the large cell systems that were studied in the initial period of intracellular recording, which hardly constitute a reasonable sample. There may be a kernel of truth in the idea, M.V.L. Bennettr Brain Research ReÕiews 32 (2000) 16–28 since synaptic delay is shorter at mammalian body temperature and this advantage of electrical transmission is less important. With respect to primitiveness, I have argued that unicellular organisms evolved the basic machinery of chemical transmission for release of and response to chemicals, but have no functional equivalent of electrical synapses. Thus, gap junctions are the more advanced form of transmission. This view is strengthened by the recent findings that Ecdysozoan and vertebrate gap junctions are formed by different gene families. It is a good bet that invertebrates of the Deuterostome line ŽEchinoderms, Ascidians, Chordates. will also have connexin based gap junctions, and ascidian blastomeres have voltage dependent gap junctions very similar to those of amphibians Žpublished only in abstract form w40x.. The term ‘‘innexins’’, coined as a contraction of invertebrate connexins, has been proposed to denote the Ecdysozoan family of gap junction proteins. This term will prove something of a misnomer, in the likely event that connexins are found in invertebrate Deuterostomes or even Lophotrochozoa Žannelids and molluscs among others.. Since homologs of mammalian glutamate receptors are found in Drosophila, it is clear that glutamatergic transmission is more primitive, or at least phylogenetically older, than gap junction mediated communication in either or both the vertebrates and arthropods. Nice looking gap junctions are seen in coelenterates diverging before either bilaterian group; it will be of interest to determine if coelenterate gap junctions are encoded by either of the two described gap junction gene families. As a further indication of the advanced nature of gap junctional communication, many of the proteins recently implicated in neurotransmitter release have homologs involved in secretion in yeast w23x, but homologs of connexins have not been reported outside of vertebrates w54x. Septate junctions are found in the Protostome line ŽEcdysozoa and Lophotrochozoa. and also in the Deuterostome line and apparently serve a similar function to tight junctions. Although they are prominent in Echinodermata, only a few ‘‘septate-like’’ junctions have been described in vertebrates, at the initial segment of Mauthner cells and of cerebellar Purkinje cells where there is electrical inhibition and also in the testis between Sertoli cells w27x. Tight junctions have largely replaced septate junctions in epithelia. There is homology between caspr, a molecule at the axon, Schwann cell junction in the perinodal region, and a neurexin in Drosophila that participates in formation of septate junctions w6x. In the vertebrate electrical inhibitory synapses the septate-like junction may be serve a barrier function to increase access resistance from initial segment to surrounding tissues. In myelinated nerve and Sertoli cells it may also be acting as a barrier. It is entertaining that the most important electrical synapse molecules in vertebrates, the connexins, are not homologous with the gap junction proteins of the Ecdysozoan line, whereas these other molecules, which may be found at the rela- 25 tively rare electrical inhibitory synapses, have Ecdysozoan homologs. 8. Romance in academia Gunther Stent has said that a research field has romantic and academic phases. In a romantic phase, there are new discoveries and whole new vistas open up. His prime example was the discovery of DNA and the genetic code. Intracellular recording opened a romantic phase in neurophysiology, now a little used term, and I remember meetings of the American Physiological Society at which one could go to every paper involving intracellular recording Žor of the Society for Neuroscience where one could see all the posters involving patch clamping.. As the number of impulses and number of neurons increased, the field entered an academic phase, where knowledge was being filled in. The endeavor was valid and important for further work, but great new dishabituating insights were not immediately forthcoming. Then cloning and patch clamping opened a new romantic phase of burgeoning excitement. We are now well into a phase of brute force science, in which it is possible to sequence entire genomes and with gene chips determine all the changes in gene expression and protein levels associated with a given physiological or pathological response. It is hard to consider this direction romantic or even hypothesis driven, other than the belief that something interesting must be happening and will be found. Going on a fishing expedition is rarely fundable at the NIH, even when it is clear that there are fish in these here waters. Major funding is going to genome sequencing; draining the lake may not be a clever approach to finding fish, but it certainly is effective. I felt romantically involved with electrical transmission in the 1960s, but in my view the field was in an academic phase through the seventies. There had been a few instances of electrical transmission demonstrated in mammals, e.g., in the mesencephalic trigeminal nerve nucleus and the inferior olive w4,47x. There were a few morphological reports of gap junctions between mammalian neurons. Particularly gratifying to me were the dendrodendritic gap junctions in the sensory cortex of the monkey, a primate no less w62x, as well as in the olfactory bulb and hippocampus w61x. Still, the data w74x came slowly in that fixation of the CNS for electron microscopy was difficult, Lucifer Yellow, neurocytin, and other tracers were not yet available, and neurons could not be visualized in brain slices. In the early 1980s there were exciting new findings of gating at gap junctions both by application of transjunctional voltage and by cytoplasmic acidification w34,64x. The recording of single channels by Neyton and Trautman w55x through use of patch clamping of small high resistance cells was inspirational if confusing because of slow opening and closing of single channels that often occurred interspersed with faster transitions like those we had come 26 M.V.L. Bennettr Brain Research ReÕiews 32 (2000) 16–28 to expect from single channels. Slow transitions have now been seen in many cell and junction types, and are ascribable to transitions through substates differing little in conductance, e.g., Ref. w65x. When single channel recording was combined with exogenous expression of cloned connexins, wildtype and modified by site directed mutagenesis, it was a romantic moment, which of course is leading into the academic enterprise of understanding the structure and function of connexins. Over my scientific lifetime there have been many changes. I grew up with NIH, so have no personal understanding of how it was before. After the Žsecond World. War the US Congress was persuaded to support the health sciences broadly speaking, and, although I am in a position of conflict of interest, I believe that it was good for the country and humanity, as well as me. A memorable quote from Charles Wilson, Secretary of Defense under Eisenhower, when testifying before Congress: he said that ‘‘Basic research is where you don’t know what you are doing’’. He was of course right in one sense, because you would not have to do the research if you knew the outcome. Hypothesis driven research means you have an idea about the outcome and is all very well. I do have a genuine distaste for inventing history in the way some investigators formulate clever hypotheses after they have made fortuitous observations. It is scientific fraud in terms of method, even if the actual data are real. I believe Wilson would have supported the human genome project. But in spite of him there was a time that I could write in a grant application that the results had no direct relevance to human health, but that they were important for general understanding. And get the grant. I still believe that that is the relevance of most of my work, but I would not state it so baldly in seeking NIH funding. Over the years, in response to budgetary exigencies we have become more skilled in giving a potential clinical significance to our work. For example, acidosis following cardiac arrest would decrease junctional conductance between cardiocytes, although pH dependence of junctional conductance was demonstrated in blastomeres. Electrical coupling of neurons via gap junctions may be important in seizure generation, a nice idea with little direct evidence. Loss of coupling may lead to reduced growth control and promote Žnot initiate. carcinogenesis. These few examples are typical of the not very strongly based arguments for the importance of gap junctions. Thus, it gave great pleasure when the first connexin disease was reported w22x, although the relatively mild nature of X-linked Charcot-Marie-Tooth disease and restriction to peripheral nerve had to disappoint some investigators with large investments in Cx32. Consider that Cx32 had not even been described in nerve prior that report. More recently, mutations in Cx26 have been shown to cause deafness w15x. It is virtually certain that other connexin diseases will appear, because it is unthinkable that the other connexins are not important and unlikely that all mutations in other connexins will be embryonic lethal. It has been difficult to predict what a gene knockout will do; one would not have predicted CMT or deafness as the presenting manifestations of Cx32 and Cx26 knockouts. An interesting problem remains, not satisfactorily explained to my knowledge. How is expression selected for in tissues that show no obvious effect of loss of a gene product that they strongly express and when a major decrease in fitness is associated with loss of expression of this gene in another tissue? The first targeted disruption of connexins was a romantic moment. The effort becomes more academic as more genes are targeted. Conditional knockouts are romantic and promise to avoid the developmental and strain differences that plague mouse knockouts. Knockout, knockin is romantic and promises to bring understanding to the functional differences between connexins and between their regulated expression when the proteins can substitute for one another. It is likely to become academic as more and more mutations are evaluated. Knockin of mutant connexins will allow validation of the causes of genetic connexin diseases, but would become academic if used to evaluate the more that 180 known mutations in Cx32. To me the new findings of electrical transmission in mammals are romantic, like an encounter with a former lover for whom one has carried a torch for many years. The cloning of connexins now permits in situ hybridization and antibody labeling, and expression of connexins is common in central neurons. A laborious combination of confocal light microscopy and freeze fracture electron microscopy has revealed an unexpectedly Žand for me delightfully. high incidence of synapses with both gap junctions and morphological characteristics of chemical synapses in rat spinal cord w59x. There is a caveat about morphological approaches. We all know that RNA may not make a protein and a protein may not function, but we have been comfortable with anatomically described chemical synapses and gap junctions as indicative of function. However, presynaptic vesicles, which define an active zone, are also found at axosomatic and axodendritic synapses where electrophysiological findings indicate that transmission is purely electrical w21,44,57x. In fact, it appears that all axodendritic and axosomatic synapses with gap junctions also have active zones, i.e., they are morphologically mixed synapses. Dual electrical and chemical transmission is not that common w25,46,52x. Now come silent synapses that may not express receptors or release transmitter Že.g., Refs. w49,50x.. The presynaptic vesicles and densities at functionally electrical synapses may be involved in membrane recycling of surface proteins or take up of extracellular factors. Conversely, correlation of the number of gap junction channels at club endings on the Mauthner cell with junctional conductance suggests that most of the channels are closed and that gap junctional area is not a good measure of junctional conductance w66x. Feliksas Bukauskas and numerous collaborators in and outside of our group in still unpublished work find that M.V.L. Bennettr Brain Research ReÕiews 32 (2000) 16–28 Cx43 labeled with a green fluorescent protein must form quite sizable plaques before even a single channel starts to open. Also with respect to predictive value, single channel conductance of junctions formed by cloned connexins can vary by an order of magnitude w68x. My summary conclusion as of now is that gap junctions constitute a small but respectable minority of synapses in the mammalian brain, as well as in the brains of ‘‘lower’’ forms. This is a cold calculation rather than a romantic fantasy, and we will progress inevitably through new romantic and academic periods. I will savor each romance, work hard as the relationship matures and eagerly but patiently await the next one. w13x w14x w15x w17x w18x w19x Acknowledgements This work depended on numerous colleagues most of whose names can be found in the reference list; I am deeply indebted to them. Funding came from many NIH and NSF grants over the years, initially to Harry Grundfest. A major support for 26 years has been NS-07512. I am the Sylvia and Robert S. Olnick Professor of Neuroscience. w20x w21x w22x w23x References w1x A. Arvanitaki, Effects evoked in an axon by the activity of a contiguous one, J. Neurophysiol. 5 Ž1942. 89–108. w2x A.A. Auerbach, M.V. Bennett, Chemically mediated transmission at a giant fiber synapse in the central nervous system of a vertebrate, J. Gen. Physiol. 53 Ž1969. 183–210. w3x A.A. Auerbach, M.V. Bennett, A rectifying electrotonic synapse in the central nervous system of a vertebrate, J. Gen. Physiol. 53 Ž1969. 211–237. w4x R. Baker, R. Llinas, Electrotonic coupling between neurones in the rat mesencephalic nucleus, J. Physiol. ŽLondon. 212 Ž1971. 45–63. w5x L.C. Barrio, T. Suchyna, T. Bargiello, L.X. Xu, R.S. Roginski, M.V. Bennett, B.J. Nicholson, Gap junctions formed by connexins 26 and 32 alone and in combination are differently affected by applied voltage, Proc. Natl. Acad. Sci. U. S. A. 88 Ž1991. 8410–8414. w6x H.J. Bellen, Y. Lu, R. Beckstead, M.A. Bhat, Neurexin IV, caspr and paranodin-novel members of the neurexin family: encounters of axons and glia, Trends Neurosci. 21 Ž1998. 444–449. w7x M.V. Bennett, Electrical versus chemical neurotransmission, Res. Publ. — Assoc. Res. Nerv. Ment. Dis. 50 Ž1972. 58–90. w8x M.V.L. Bennett, Modes of operation of electric organs, Ann. N. Y. Acad. Sci. 94 Ž1961. 458–509. w9x M.V.L. Bennett, A Comparative Study of Neuronal Synchronization, in: D.P. Purpura, M.D. Yahr ŽEds.., The Thalamus, Columbia Univ. Press, New York, 1966, pp. 173–181. w10x M.V.L. Bennett, Similarities Between Chemically and Electrically Mediated Transmission, in: F.D. Carlson ŽEd.., Physiological and Biochemical Aspects of Nervous Integration, Prentice-Hall, New York, 1968, pp. 73–128. w11x M.V.L. Bennett, Electric Organs, in: W.S. Hoar, D.J. Randall ŽEds.., Fish Physiology, Vol. 5, Academic Press, New York, 1971, pp. 347–491. w12x M.V.L. Bennett, A Comparison of Electrically and Chemically Mediated Transmission, in: G.D. Pappas, D.P. Purpura ŽEds.., Struc- w24x w25x w26x w27x w28x w29x w30x w31x w32x w33x w34x 27 ture and Function of Synapses, Raven Press, New York, 1972, pp. 221–256. M.V.L. Bennett, Electrical transmission: a functional analysis and comparison to chemical transmission, in: E. Kandel ŽEd.., The Handbook of Physiology, American Physiological Society, Washington, 1977, pp. 357–416. M.V.L. Bennett, Nicked by Occam’s razor: unitarianism in the investigation of synaptic transmission, Biol. Bull. 168 Ž1985. 159– 167, suppl. C.K. Abrams, M.V.L. Bennett, Hereditary Human Diseases Caused by Connexin Mutations, Spectrum Publishers, York, 1999. M.V.L. Bennett, E. Aljure, Y. Nakajima, G.D. Pappas, Electrotonic junctions between teleost spinal neurons: electrophysiology and ultrastructure, Science 141 Ž1963. 262–263. M.V. Bennett, L.C. Barrio, T.A. Bargiello, D.C. Spray, E. Hertzberg, J.C. Saez, Gap junctions: new tools, new answers, new questions, Neuron 6 Ž1991. 305–320. M.V.L. Bennett, S.M. Crain, H. Grundfest, Electrophysiology of supramedullary neurons in Spheroides maculatus: I. Orthodromic and antidromic responses, J. Gen. Physiol. 43 Ž1959. 159–188. M.V. Bennett, Y. Nakajima, G.D. Pappas, Physiology and ultrastructure of electrotonic junctions: I. Supramedullary neurons, J. Neurophysiol. 30 Ž1967. 161–179. M.V. Bennett, Y. Nakajima, G.D. Pappas, Physiology and ultrastructure of electrotonic junctions: 3. Giant electromotor neurons of Malapterurus electricus, J. Neurophysiol. 30 Ž1967. 209–235. J. Bergoffen, S.S. Scherer, S. Wang, M.O. Scott, L.J. Bone, D.L. Paul, K. Chen, M.W. Lensch, P.F. Chance, K.H. Fischbeck, Connexin mutations in X-linked Charcot-Marie-Tooth disease, Science 262 Ž1993. 2039–2042. N. Calakos, R.H. Scheller, Synaptic vesicle biogenesis, docking, and fusion: a molecular description, Physiol. Rev. 76 Ž1996. 1–29. C.E. Carr, R.E. Boudreau, Organization of the nucleus magnocellularis and the nucleus laminaris in the barn owl: encoding and measuring interaural time differences, J. Comp. Neurol. 334 Ž1993. 337–355. B.N. Christensen, Distribution of electrotonic synapses on identified lamprey neurons: a comparison of a model prediction with an electron microscopic analysis, J. Neurophysiol. 49 Ž1983. 705–716. D.F. Condorelli, R. Parenti, F. Spinella, A. Trovato Salinaro, N. Belluardo, V. Cardile, F. Cicirata, Cloning of a new gap junction gene ŽCx36. highly expressed in mammalian brain neurons, Eur. J. Neurosci. 10 Ž1998. 1202–1208. C.J. Connell, A freeze-fracture and lanthanum tracer study of the complex junction between Sertoli cells of the canine testis, J. Cell Biol. 76 Ž1978. 57–75. D.S. Faber, H. Korn, Electrical field effects: their relevance in central neural networks, Physiol. Rev. 69 Ž1989. 821–863. K. Funakoshi, T. Kadota, Y. Atobe, M. Nakano, R.C. Goris, R. Kishida, GastrinrCCK-ergic innervation of cutaneous mucous gland by the supramedullary cells of the puffer fish Takifugu niphobles, Neurosci. Lett. 258 Ž1998. 171–174. E.J. Furshpan, ‘‘Electrical transmission’’ at an excitatory synapse in a vertebrate brain, Science 144 Ž1964. 878–880. E.J. Furshpan, D.D. Potter, Transmission at the giant motor synapses of the crayfish, J. Physiol. ŽLondon. 145 Ž1959. 289–325. M. Hagedorn, H.A. Vischer, W. Heiligenberg, Development of the jamming avoidance response and its morphological correlates in the gymnotiform electric fish, Eigenmannia, J. Neurobiol. 23 Ž1992. 1446–1466. D.H. Hall, E. Gilat, M.V.L. Bennett, Ultrastructure of the rectifying electrototonic synapses between giant fibers and pectoral fin adductor motoneurons in the hatchetfish, J. Neurocytol. 14 Ž1985. 825– 834. A.L. Harris, D.C. Spray, M.V. Bennett, Kinetic properties of a voltage-dependent junctional conductance, J. Gen. Physiol. 77 Ž1981. 95–117. 28 M.V.L. Bennettr Brain Research ReÕiews 32 (2000) 16–28 w35x G.I. Hatton, Q.Z. Yang, Incidence of neuronal coupling in supraoptic nuclei of virgin and lactating rats: estimation by neurobiotin and lucifer yellow, Brain Res. 650 Ž1994. 63–69. w36x J.G. Jefferys, Nonsynaptic modulation of neuronal activity in the brain: electric currents and extracellular ions, Physiol. Rev. 75 Ž1995. 689–723, Review. w37x M.F. Johnston, F. Ramon, Voltage-independence of an electrotonic synapse, Biophys. J. 39 Ž1982. 115–117. w38x D.P. Kelsell, J. Dunlop, H.P. Stevens, N.J. Lench, J.N. Liang, G. Parry, R.F. Mueller, I.M. Leigh, Connexin 26 mutations in hereditary non-syndromic sensorineural deafness, Nature 387 Ž1997. 80– 83. w40x J. Knier, V.K. Verselis, D.C. Spray, Gap junctions between tunicate blastomeres: gating similarities and differences compared to amphibia, Biophys. J. 49 Ž1986. 203a. w41x H. Korn, M.V.L. Bennett, Dendritic and somatic impulse initiation in fish oculomotor neurons during vestibular nystagmus, Brain Res. 27 Ž1971. 169–175. w42x H. Korn, M.V.L. Bennett, Electrotonic coupling between teleost oculomotor neurons: restriction to somatic regions and function of somatic and dendritic sites of impulse initiation, Brain Res. 38 Ž1972. 433–439. w43x H. Korn, M.V.L. Bennett, Vestibular nystagmus and teleost oculomotor neurons: functions of electrotonic coupling and dentritic impulse initiation, J. Neurophysiol. 38 Ž1975. 430–451. w44x H. Korn, C. Sotelo, F. Crepel, Electronic coupling between neurons in the rat lateral vestibular nucleus, Exp. Brain Res. 16 Ž1973. 255–275. w45x M.E. Kriebel, M.V.L. Bennett, S.G. Waxman, G.D. Pappas, Oculomotor neurons in fish: electrotonic coupling and multiple sites of impulse initiation, Science 166 Ž1969. 520–524. w46x J.W. Lin, D.S. Faber, Synaptic transmission mediated by single club endings on the goldfish Mauthner cell: I. Characteristics of electrotonic and chemical postsynaptic potentials, J. Neurosci. 8 Ž1988. 1302–1312. w47x R. Llinas, R. Baker, C. Sotelo, Electrotonic coupling between neurons in cat inferior olive, J. Neurophysiol. 37 Ž1974. 560–571. w48x C. Lu, D.G. McMahon, Modulation of hybrid bass retinal gap junctional channel gating by nitric oxide, J. Physiol. ŽLondon. 499 Ž1997. 689–699. w49x R.C. Malenka, R.A. Nicoll, Silent synapses speak up, Neuron 19 Ž1997. 473–476. w50x A. Malgaroli, A silent synapses: I can’t hear you! Could you please speak aloud?, Nat. Neurosci. 2 Ž1999. 3–5. w51x P. Mann-Metzer, Y. Yarom, Electrotonic coupling interacts with intrinsic properties to generate synchronized activity in cerebellar networks of inhibitory interneurons, J. Neurosci. 19 Ž1999. 3298– 3306. w52x A.R. Martin, G. Pilar, Dual mode of synaptic transmission in the avian ciliary ganglion, J. Physiol. ŽLondon. 168 Ž1963. . w53x R.M. Meszler, G.D. Pappas, V.L. Bennett, Morphology of the electromotor system in the spinal cord of the electric eel, Electrophorus electricus, J. Neurocytol. 3 Ž1974. 251–261. w54x A.R. Mushegian, E.V. Koonin, The proposed plant connexin is a protein kinase-like protein, Plant Cell 5 Ž1993. 998–999, letter. w55x J. Neyton, A. Trautmann, Single-channel currents of an intercellular junction, Nature 317 Ž1985. 331–335. w56x S. Oh, J.B. Rubin, M.V.L. Bennett, V.K. Verselis, T.A. Bargiello, Molecular determinants of electrical rectification of single channel conductance in gap junctions formed by connexins 26 and 32, J. Gen. Physiol. Ž1999. 339–364. w57x G.D. Pappas, M.V. Bennett, Specialized junctions involved in electrical transmission between neurons, Ann. N. Y. Acad. Sci. 137 Ž1966. 495–508. w58x A.E. Pereda, D.S. Faber, Activity-dependent short-term enhancement of intercellular coupling, J. Neurosci. 16 Ž1996. 983–992. w59x J.E. Rash, R.K. Dillman, B.L. Bilhartz, H.S. Duffy, L.R. Whalen, T. Yasumura, Mixed synapses discovered and mapped throughout mammalian spinal cord, Proc. Natl. Acad. Sci. U.S. A. 93 Ž1996. 4235–4239. w60x B. Rorig, G. Klausa, B. Sutor, Dye coupling between pyramidal neurons in developing rat prefrontal and frontal cortex is reduced by protein kinase A activation and dopamine, J. Neurosci. 15 Ž1995. 7386–7400. w61x H. Schmalbruch, H. Jahnsen, Gap junctions on CA3 pyramidal cells of guinea pig hippocampus shown by freeze-fracture, Brain Res. 217 Ž1981. 175–178. w62x J.J. Sloper, Gap junctions between dendrites in the primate neocortex, Brain Res. 44 Ž1972. 641–646. w63x G. Sohl, J. Degen, B. Teubner, K. Willecke, The murine gap junction gene connexin36 is highly expressed in mouse retina and regulated during brain development, FEBS Lett. 428 Ž1998. 27–31. w64x D.C. Spray, A.L. Harris, M.V. Bennett, Gap junctional conductance is a simple and sensitive function of intracellular pH, Science 211 Ž1981. 712–715. w65x E.B. Trexler, M.V. Bennett, T.A. Bargiello, V.K. Verselis, Voltage gating and permeation in a gap junction hemichannel, Proc. Natl. Acad. Sci. U.S. A. 93 Ž1996. 5836–5841. w66x R. Tuttle, S. Masuko, Y. Nakajima, Freeze-fracture study of the large myelinated club ending synapse on the goldfish Mauthner cell: special reference to the quantitative analysis of gapgap junctions, J. Comp. Neurol. 246 Ž1986. 202–211. w67x T.A. Valiante, J.L. Perez Velazquez, S.S. Jahromi, P.L. Carlen, Coupling potentials in CA1 neurons during calcium-free-induced field burst activity, J. Neurosci. 15 Ž1995. 6946–6956. w68x R.D. Veenstra, H.Z. Wang, D.A. Beblo, M.G. Chilton, A.L. Harris, E.C. Beyer, P.R. Brink, Selectivity of connexin-specific gap junctions does not correlate with channel conductance, Circ. Res. 77 Ž1995. 1156–1165. w69x V.K. Verselis, C.S. Ginter, T.A. Bargiello, Opposite voltage gating polarities of two closely related connexins, Nature 368 Ž1994. 348–351. w70x A. Watanabe, The interaction of electrical activity among neurons of lobster cardiac ganglion, Jpn. J. Physiol. 8 Ž1958. 305–318. w71x A. Watanabe, H. Grundfest, Impulse propagation at the septal and commissural junctions of crayfish lateral giant axons, J. Gen. Physiol. 45 Ž1961. 267–308. w72x J.A. Westfall, J.C. Kinnamon, Perioral synaptic connections and their possible role in the feeding behavior of Hydra, Tissue Cell 16 Ž1984. 355–365. w73x M.V.L. Bennett, Gap junctions as electrical synapses, J. Neurocytol. 26 Ž1997. 349–366. w74x A.J. Pinching, T.P. Powell, The neuropil of the glomeruli of the olfactory bulb, J. Cellsci. 9 Ž1971. 347–377.