INTEG. AND COMP.BIOL., 42:705±715 (2002)

Modulation of the Cray®sh Escape Re¯exÐPhysiology and Neuroethology1

FRANKLIN B. KRASNE2 AND DONALD H. EDWARDS Department of Psychology, UCLA, Los Angeles, California 90095-1563 and Department of Biology, Georgia State University, Atlanta, Georgia 30302-4010

SYNOPSIS. We review here factors that control the excitability of the giant neuron-mediated tail-¯ip escape behavior in cray®sh, focusing especially on recent ®ndings concerning serotonergic modulation. Serotonin can either facilitate or inhibit escape depending on concentration and pattern of application. Low concen- trations facilitate while high ones inhibit; however, if high concentrations arise gradually they facilitate instead of inhibiting. The effects of serotonin can also be altered by social experience, with application regimens that cause facilitation in social isolates coming to produce inhibition after an extended period of living as a subordinate. Attempts to understand both the possible physiological basis of some of these com- plexities and their possible function are discussed. Neuroethological investigations indicate that giant neuron- mediated escape is inhibited during the initial ®ghts that establish social relationships and is facilitated in their immediate aftermath. Once the relationship of a pair is well-established, the presence of the dominant tends to suppress giant neuron-mediated escape (but not tail-¯ip escape mediated by non-giant circuitry) in

the subordinate, but the presence of the subordinate has relatively little effect on the dominant. These Downloaded from patterns of modulation can be seen as consistent with the known variations in serotonin's effect as a function of concentration and social experience and may provide a biological reason for these variations.

For most neurobiologists it is an article of faith that unique opportunities: (1) The connection between cel- the behavior which emerges from nervous systems is lular phenomena and behavior is often much clearer in http://icb.oxfordjournals.org/ the product of a neural machine. But the machine is invertebrates due to the relative simplicity of some of one in which a given neural circuit does not always their behavior-producing neural circuitry; thus, we can work the same way. Operational properties of a circuit discover the natural uses made of instances of plastic- can change due to learning and due to modulation by ity and modulation. (2) The very diversity of inverte- other circuits, imparting to the behavior of a given brates inevitably exposes us to a wider array of phe- individual the great variety and irregularity that makes nomena than we would see from studying mammals the behavior of and ourselves interesting and alone; thus it helps us distinguish what is general from a challenge to our understanding. To a great degree, what is not. by guest on February 13, 2012 though not entirely, changes in the properties of neural We review here work on plasticity seen in the neural circuits are due to changes in the functional properties circuitry which mediates escape behavior in cray®sh, of their synapses. with a focus on recent surprising ®ndings on seroto- The last several decades have seen remarkable pro- nergic modulation, and their possible functional sig- gress in uncovering various forms of synaptic plastic- ni®cance. ity induced either by activity or by chemical modula- tors. Some of the ®rst forms of synaptic plasticity to THE NEURAL CIRCUITRY UNDERLYING be described and related to behavioral plasticity were ESCAPE BEHAVIOR in invertebratesÐspeci®cally in Aplysia (see articles Escape can be mediated in two rather different ways by Sutton and Carew, 2002, and Sherff and Carew, as indicated separately on the left and right of Figure 2002 [this issue]; Kandel, 1976) and also in cray®sh 1A (for a review see Edwards et al., 1999). The cir- (Krasne, 1969; Zucker, 1972; Zucker et al., 1971). In cuitry on the left produces either of two distinct types parallel with these studies on invertebrates were dis- of response, depending on locus of stimulation. Each coveries beginning at roughly the same time on mam- response type is associated with a giant command neu- malian hippocampal LTP (Bliss and Collingridge, ron. The medial giants (MGs) sum input from anterior 1993), which is now widely regarded as a possible sensory channels and make output connections with mechanism of associative learning. This latter line of giant ¯exor motor neurons (motor ``giants''ÐMoGs in work has led to a veritable frenzy of activity directed both at working out cellular and molecular mecha- Fig. 1) that cause a dart backwards when the excitation nisms of long-term potentiation (LTP) and at trying to produces even one spike in the MGs. The lateral gi- fathom what might be its actual roles visa vi behavior. ants (LGs) sum input from posterior channels and Research on LTP has somewhat eclipsed invertebrate cause an upward rotation that distances the hind end research, but invertebrate systems continue to provide of the from the disturbing . MG and LG responses are often referred to as ``re¯ex'' respons- es or as ``giant ®ber (GF)'' responses for the giant 1 From the Symposium Recent Advances in Neurobiology pre- sented at the Annual Meeting of the Society for Integrative and axons of the MGs and LGs, which run the length of Comparative Biology, 2±6 January 2002, at Anaheim, California. the nerve cord. GF responses are very prompt (muscle 2 E-mail: [email protected] potentials can begin within 3 msec) and are good at

705 706 F. B. KRASNE AND D. H. EDWARDS Downloaded from http://icb.oxfordjournals.org/ by guest on February 13, 2012

FIG. 1. Escape tail-¯ip circuitry of the cray®sh. A. Schematic circuit. Primary mechanosensory afferents of the abdomen (black neurons in top row) excite LG neurons directly and via an intervening layer of (black neurons in second row). The LGs in turn excite giant motor neurons (MoGs) to muscles (marked by solid black squares) whose contraction causes an upward, forward trajectory of movement due to bending at joints indicated by solid black circles over the cray®sh depicted at the bottom of the ®gure. The overall movement produced is indicated by the black silhouette at the lower right of the ®gure. Stippled neurons and marks show a corresponding arrangement for the production of backward-directed tail-¯ips commanded by MGs in response by to stimulation of anterior sensory neurons, and the resultant response is indicated by the stippled drawing at the lower left of the ®gure. Non-giant tail¯ip circuitry activates an independent set of ¯exor muscles (FFs) that innervate the same ¯exor muscles. GFs also recruit FFs via the segmental giant neurons (SGs), which excite motor and pre-motor neurons within the non-G circuitry (modi®ed from Edwards et al. [1999] which should be consulted for further explanation). The ef®cacy of synapses at levels I and II are particularly subject to alteration by past activity and by neuromodulators, as discussed in this review and summarized in Figure 2. B. Form of EPSP evoked in LGs by a volley of sensory root activity. Explanation in text.

getting the cray®sh moving away from the source of sophisticated responses of the non-G circuitry come at stimulation rapidly. a price: They are far from prompt (latencies are about The circuitry on the right has no giant neurons (non- 100 msec). G circuitry), is much more complex, and is far from Before beginning to discuss forms of synaptic mod- fully charted. Whereas the giant-containing circuitry ulation that have been studied in the GF circuitry it produces only two very stereotyped forms of response should be noted that only the synapses between pri- (``back'' and ``upward rotation'') and always single mary afferents and the sensory interneurons of the GF ¯exions, the responses generated by the non-giant cir- circuitry are conventional chemical synapses. The re- cuitry have a seemingly in®nite variety of possible mainder are voltage-gated electrical synapses, which forms and can occur in repetitive strings (``swims''). pass current effectively only when the presynaptic side Using this circuitry cray®sh can move directly away of the synapse is made positive relative to the post- from an oblique stimulus, avoid obstacles, and move synaptic side by the arrival of a presynaptic spike (Ed- toward speci®c locations. Unlike GF responses, which wards et al., 1991; Furshpan and Potter, 1959; Giaume ordinarily occur only in response to abrupt and fairly et al., 1987; Jaslove and Brink, 1986). Qualitatively, vigorous stimulation, non-G responses are often these voltage-gated electrical synapses have many of prompted by gradually developing threats. The more the same properties as chemical synapses. These in- MODULATION OF THE CRAYFISH ESCAPE 707

thresholds when, for example an animal has found a food source and is avidly feeding (Krasne and Lee, 1988) or when an animal is ®rmly clutched by a ®sh- erman and re¯ex tail¯ips would be useless (Krasne and Wine, 1975). Recently, work done by former students from each of our laboratories has begun to suggest that transmis- sion at the voltage-gated electrical synapses on the GFs is also subject to intrinsic, activity-dependent change. In both their experiments, as in other experi- ments to be discussed here, brief electrical test shocks to sensory nerves were delivered every few minutes and responses recorded in the LGs with intracellular microelectrodes. The shocks produce compound excit- atory postsynaptic potentials (EPSPs) (Fig. 1B), which have a ®rst elevation (the ␣ component) resulting from monosynaptic input from the primary afferents and a second (the ␤ component), which is due to input ar- Downloaded from riving via the sensory interneurons. Shi-Rung Yeh (personal communication) has found that several sec- ond long trains of 4 Hz stimuli to the afferents cause both the monosynaptic and disynaptic components of FIG. 2. Factors regulating the excitability of GF escape. The levels (I and II) at which synaptic transmission is affected are those indi- LG EPSPs to grow and stay high for many hours (Fig. cated in Figure 1. At the right are listed known forms of behavioral 3, top; Fig. 2, line 3). The augmentation is speci®c to http://icb.oxfordjournals.org/ modulation. Arrows indicate possible causality of forms of behav- input pathways that were given the 4 Hz stimulation ioral modulation with dashed lines marked by ??? indicating entirely and seems to be entirely prevented if the calcium ion speculative causality. Details in text. chelators BAPTA or EGTA were previously injected into the LGs. Thus, this phenomenon is similar to LTP clude polarized transmission, temperature sensitive at glutamatergic synapses in that its induction is de- synaptic delay, and modulation of EPSP size by post- pendent on a transient elevation of calcium ions in the synaptic membrane potential level. Because of these postsynaptic neuron. Since these are voltage-depen-

similarities it can be very dif®cult to distinguish volt- dent electrical synapses and not glutamatergic ones, by guest on February 13, 2012 age-gated electrical synapses from chemical ones, and this is an interesting parallelism. Sun Hee Lee has as we shall see, EPSPs produced by these electrical found that similar stimulation (5 Hz) given for the synapses are also subject to modulation of kinds that much longer time of 5 min causes a depression that one normally associates with chemical synapses. also lasts a long time and is blocked by BAPTA in the LGs (Lee, 1996; Lee and Park, 1997). This long-term MODULATION AND PLASTICITY IN GF CIRCUIT depression (LTD)-Iike phenomenon seems to occur The ®rst form of plasticity discovered in GF cir- mainly in the disynaptic (␤) EPSP but presumably still cuitry circuit was intrinsic depression of transmitter involves depression at synapses directly on the LGs, release as the result of repetitive presynaptic activity since it is blocked by preventing calcium ion elevation at the cholinergic synapses made onto sensory inter- in the LGs. The functional consequences of LTP are neurons that innervate the LGs (Fig. 2, line 1; Krasne, quite unknown, but it seems possible that the LTD may 1976; Miller et al., 1992; Zucker, 1972). Like probably play some role in habituation (Fig. 2, line 4). all escape behavior, GF responses habituate to repeated stimulation, and this presynaptic activity-dependent Aminergic neuromodulation of transmission to the depression is at least in part responsible. Its discovery LGs in the late 60s was exciting, because it was one of the The story that is the major focus of this review be- ®rst times that a simple kind of learning had been gan about 20 yr ago when Ed Kavitz and his students traced to a speci®c neural mechanism. (Livingstone et al., 1980; Kravitz, 1988) discovered It was soon after found that a second in¯uence af- that injection of serotonin into lobster or cray®sh al- fecting whether cray®sh will escape to threats is a tered synaptic transmission at certain neuromuscular GABA-ergic inhibitory input directly to the GF den- synapses and caused animals to adopt a posture that drites (Fig. 2, line 2; Krasne and Wine, 1975; Vu and was seen as similar to that adopted by socially domi- Krasne, 1993; Vu et al., 1993). This inhibition, which nant animals, whereas octopamine, the arthropod an- is controlled by higher centers, is turned on under a alog of adrenaline, caused a subordinate-like posture. variety of circumstances and greatly elevates the stim- It seemed likely that since these agents affected pos- ulus threshold for producing GF escape. This ``tonic tures associated with ®ght and ¯ight, they might also inhibition'' contributes to habituation (Krasne and affect escape. In particular, it was conjectured that oc- Teshiba, 1995) and also serves to elevate GF escape topamine might facilitate GF escape while serotonin 708 F. B. KRASNE AND D. H. EDWARDS Downloaded from http://icb.oxfordjournals.org/

FIG. 3. LTP and LTD-like changes at synapses on the LGs. Left: Changes in amplitude of sensory root shock-evoked test EPSPs in LGs resulting from short (above) and long (below) trains of 4±5 Hz stimulation via the electrodes used to produce test EPSPs (but set higher than test voltages to recruit more afferents ®bers during trains of LTP and LTD inducing stimulation). Right: Comparison of the effects of trains under normal conditions and when BAPTA had been infused into LGs prior to the experiment. LTP data from Shi-Rung Yeh, personal communication, with permission. LTD data from Sun Lee (Lee, 1996; Lee and Park, 1997, and personal communication, with permission).

might suppress it. This conjecture was soon veri®ed. mission at the ®rst synapse that leads to behavioral by guest on February 13, 2012 Octopamine enhances transmission at the ®rst synapse sensitization of GF escape (Krasne and Glanzman, (Fig. 2, line 5; Bustamante and Krasne, 1991; Glanz- 1986); the possibility that this is mediated by octopa- man and Krasne, 1983), and serotonin was initially mine is obvious, but it has not been established. Lim- found to inhibit transmission (but see further below), ited evidence also suggests that serotonin might under at least in part by an action at the level of the LGs some circumstances share with GABA a role in the (Fig. 2, line 6; Glanzman and Krasne, 1983; Vu and tonic inhibition mentioned above (Glanzman and Kras- Krasne, 1993). It was found soon thereafter that trau- ne, 1986, but see Vu and Krasne, 1993). matic stimulation also causes a facilitation of trans- Serotonergic modulation Though the discovery of serotonergic inhibition was consistent with the conjectures that had prompted the ®rst test of serotonin's effects, it has recently become clear that these effects are much more complex than originally believed. Experiments done in the Edwards lab by Shi-Rung Yeh, over a decade after the ®rst ex- periments showing an inhibitory effect, consistently found serotonin to have a facilitatory effect on trans- mission to the LGs (Yeh et al., 1996, 1997). After a period of some confusion it eventually became clear that the difference lay in the regimen of serotonin ap- plication (Teshiba et al., 2001). When serotonin is in- troduced as rapidly as possible (FAST in Fig. 4) and left in place for only 10±15 min (SHORT in Fig. 4), FIG. 4. Mean time course of ␤ component EPSP amplitude during as was done in the original experiments, inhibition de- and following 5-HT exposure. Beta component amplitudes were nor- malized by their values on the last stimulation before the start of 5- velops over 5±10 min and washes out at the same rate HT exposure. The EPSP amplitude on the last trial in 5-HT is taken (Fig. 4, solid triangles). However, when serotonin lev- as the 0 time point for the washout graphs (Teshiba et al., 2001). els are allowed to increase only gradually (SLOW in MODULATION OF THE CRAYFISH ESCAPE REFLEX 709

Fig. 4), reaching full concentration over some 20±30 min, and are allowed to remain in place for 30±45 min (Fig. 4, solid circles), facilitation rather than inhibition is seen, and this facilitation persists for as much as 5 hr or more even during washout (Yeh et al., 1997). The persistence of facilitation during wash requires ex- posures of longer than 10 to 15 min; the longevity of serotonin-induced facilitation in Aplysia is also well known to depend on duration of exposure (see Sutton and Carew, 2002 and Sherff and Carw, 2002 [this is- sue]). One gets persistent facilitation even with fast application if one uses a low dose of 5-HT (i.e., 10Ϫ8± 10Ϫ6 M as opposed to 10Ϫ4 and aboveÐFig. 4, open circles). It is not yet known whether sertotonin alters the properties of the electrical junctions on the LGs. How- ever, one can in part understand these modulations as re¯ections of altered ionic conductances in the LGs. Downloaded from The inhibition is associated with an increased conduc- tance and small depolarization postsynaptically (Vu and Krasne, 1993), which given that the equilibrium FIG. 5. Hypothesis to explain effects of concentration and rate of potential for chloride is about 8 mV above the resting application of serotonin on its effect. Fl, F2, and I are unidenti®ed signaling molecules conjectured to mediate modulatory effects of level, is probably due to increased chloride conduc- serotonin on EPSP amplitude. gK and gCl, respectively, are potassium tance. The facilitation that is caused at low 5-HT con- ion and chloride ion conductances, changes of which may be re- http://icb.oxfordjournals.org/ centrations (Fig. 2, line 7) is associated with a de- sponsible for alterations of EPSP amplitude. Conjectured 5-HT creased conductance and also a depolarization (Yeh et threshold and onset characteristics needed to cause production of signaling molecules are speci®ed at the left. The predictions made al., 1997), which seems most likely to be due to a by the model, explained in the text, are indicated at the bottom of decreased potassium ion conductance. Thus inhibition the ®gure. and facilitation are due to non-mutually exclusive causes that could co-exist. Another way in which the facilitation and inhibition F2, so facilitation then develops. With SLOW-HIGH

are independent is that that their underlying intracel- exposures F2 builds up before 5-HT concentration by guest on February 13, 2012 lular progenitors can apparently co-exist. Thus, the reaches a level that can activate the inhibitory path- precursors of the facilitation appear to develop even at way, and F2 has reached a level where it can suppress high doses that cause inhibition; however the inhibi- the formation of I by the time serotonin is concentrated tion prevents or masks expression of the facilitation. enough to exceed the threshold of the inhibitory path- This can be seen when one washes out a high dose of way. These ideas have been developed into a compu- serotonin that has been in place long enough for per- tational model that correctly predicts (qualitatively) all sistent facilitation to develop. Inhibition is seen for as of the types of modulation observed (Teshiba et al., long as the 5-HT is present, but when it is washed 2001). away, the inhibition gives way to facilitation (Fig. 4, The various delivery regimens used in our experi- open triangles, FAST, LONG, HIGH). ments may correspond to different modes of serotonin Although the signaling molecules that mediate 5- delivery that occur naturally. Serotonin is released HT's effects in the LGs are not yet known, Figure 5 both synaptically within abdominal ganglia neuropile, proposes the logic of an intracellular signaling scheme which could provide natural FAST, HIGH exposures that could account for the complex effects of 5-HT and is also released into the blood as a hormone from exposure regimen on modulatory effect (see Teshiba a variety of sites, presumably providing a SLOW, et al., 2001). A pathway with a low 5-HT threshold LOW form of delivery (Beltz and Kravitz, 1983; Yen and relatively slow onset produces facilitation, while et al., 1997). a pathway with a high 5-HT threshold and faster onset produces inhibition. The buildup of the ®nal stage sig- Social dependence of 5-HT effects naling molecule of each pathway suppresses the for- There is yet another, and rather extraordinary layer mation of buildup of the ®nal stage molecule of the of complexity to this story. Cray®sh have long been other pathway. In this scheme FAST-LOW exposures known to form social hierarchies (Bovbjerg, 1953; cause facilitation because only the facilitatory pathway Lowe, 1956). When two cray®sh are brought together, gets activated. FAST-HIGH exposures cause inhibition one generally becomes dominant and the other sub- because signaling molecule I builds up ®rst and pre- ordinate after a short period of interaction (see below). vents formation of F2; however F1 still builds despite Quite remarkably, social experience alters the effects the manifest inhibition, and when 5-HT is washed out, of serotonin on transmission to the LGs: Whereas in persisting F1 is allowed to promote the formation of social isolates low concentrations of serotonin facili- 710 F. B. KRASNE AND D. H. EDWARDS Downloaded from http://icb.oxfordjournals.org/ by guest on February 13, 2012 FIG. 6. Effects of 5-HT depend on animal's social history. Center: Change in EPSP amplitude (measured at disynaptic ␤ peak of EPSP) produced by 5-HT in social isolates kept for varying periods with a dominant partner. Left and right: Examples of EPSPs before (CNT) and during 5-HT exposure (5-HT) in an isolate (left) and after 12 days of subordinance (right). (Data from Yeh et al., 1997).

tate transmission to the LGs (as discussed above), after chloride conductance-increasing type of inhibition pro- a cray®sh has lived for 1±2 wk as a subordinate, se- duced by high 5-HT, but a hyperpolarizing, presum- rotonin comes to inhibit transmission to the LGs (Fig. ably potassium conductance-increasing inhibition. 6; Yeh et al., 1996, 1997). This is not the depolarizing, Thus, whereas in isolates low concentrations of 5-HT decrease potassium ion conductance, in subordinates serotonin increases potassium ion conductanceÐa di- rectly opposite effect (Fig. 2, line 8). Living as a dom- inant causes a more subtle change: Whereas in isolates prolonged exposure to serotonin causes facilitation that persists after washout of serotonin, dominants do not show this persistence of facilitation. Preliminary pharmacological experiments suggest that exposure to a vertebrate 5-HT2 receptor agonist [␣-methyl 5-HT] mimics the faciltatory effects of se- rotonin, while a 5-HT1 agonist [1-(3-chlorophenyl) pi- perazine] has little effect in isolates but has a large inhibitory effect in subordinates (Fig. 7; Yeh et al., 1997). This interesting observation suggests the pos- sibility that the transformation of serotonin's net effect from facilitatory to inhibitory might be due to the in- sertion of a species of 5-HT1 receptor into LG mem-

FIG. 7. Effects of serotonin and agonists on isolated and subordi- brane or to some alteration in such a receptor's down- nate cray®sh (Data from Yeh et al., 1997). stream signaling effects. However, the effects of the MODULATION OF THE CRAYFISH ESCAPE REFLEX 711 agonists in dominants belie this straight-forward inter- provide any obvious reason for the changes in sero- pretation, since 5-HT1 agonists also produce an inhib- tonin's effect as a function of social status. Indeed, if itory effect in dominants, while 5-HT2 agonist effects one assumes that serotonin is released during agonistic are not altered. The cloning of cray®sh 5-HT receptors encounters, one is faced with the conundrum that se- is in progress, and soon receptor antibodies will be rotonergic modulation should make subordinates less used to look for changes in the amount of speci®c likely to escape during social interactions than are receptor types on the LGs as the result of social ex- dominants!! Obviously some information on whether perience (Spitzer et al., 2001). and how escape is in fact modulated during social in- These are fascinating observations. The ®nding that teractions would help. experience can alter the qualitative effect of a neuro- modulator seems both new and remarkable. Moreover, NEUROETHOLOGY OF GF MODULATION DURING when one considers that mental illnesses often seem SOCIAL INTERACTIONS to involve abnormalities in neuromodulation and that When a pair of cray®sh previously unknown to each mental illness seems to have important experiential de- other is placed in the same living space, three phases terminants, these observations provide a possible mod- of interaction ensue (Herberholz et al., 2001; Issa et el for why life experiences might be so important. Es- al., 1999): pecially interesting from this point of view is the ob- I. The animals engage in a ®ght during which one servation that serotonin's effect in a subordinate can animal often grasps the other and engages in ``of- Downloaded from again take on the typical isolate pro®le if the subor- fensive'' tail¯ips that appear to be demonstrations dinate is re-isolated, or can take on the pro®le typical of dominance and strength; in lobsters seemingly of a dominant if the subordinate is paired with an an- similar behavior can result in the target animal's imal over which it becomes dominant. However, once dismemberment (Huber and Kravitz, 1995). After an animal has been dominant long enough for its re- a few minutes phase I ends and phase II is ushered sponse to serotonin to take on the typical dominant http://icb.oxfordjournals.org/ in when the animal that is getting the worst of it pro®le, this pro®le is retained even if the animal later gives up and ¯ips away. becomes subordinate (Yeh et al., 1997). Thus, some II. The new dominant frequently harasses the new experience-caused changes in neuromodulation are subordinate, which backs or ¯ips away. readily reversible, but others are not. III. After some days the relationship becomes ®rmly We now have a rather long list of ways in which established. The subordinate cowers at the margin the GF circuitry is modi®able, and we have some plau- of the living space and appears to try to keep as sible connections to behavior (Fig. 2). It is interesting low a pro®le as possible, resting close as possible

to note that each known form of behavioral control by guest on February 13, 2012 to the substrate, while the dominant moves about may be due to a multiplicity of modulatory mecha- freely, maintains a tall posture, and very occa- nisms. Thus, the roles played by each and the possi- sionally harasses the subordinate, which slinks or bility of their interaction becomes an interesting topic ¯ips away. The transition to phase III is illustrated for future investigation. That there are interactions in Figure 8 (Issa et al., 1999), which shows that seems clear. For example, if GABA is infused along encounters greatly decrease over days and that re- with 5-HT, the persistent effects of 5-HT usually pro- treats replace escape as the most common subor- duced by prolonged application do not develop (Ghiu- dinate evasion response. seli and Krasne, unpublished). And experiments in progress seem to show that in the presence of picro- Edwards' group recently developed a noninvasive toxin, which blocks the chloride channel normally procedure to record tail¯ips from freely behaving opened by GABA, LTD may not be produced (Lee, animals during their initial encounter in a 30 min Shirinyan, and Krasne, unpublished). This raises in- session (Herberholz et al., 2001). They found (Fig. teresting questions for the future. 9) that during phase I there were almost no GF-me- We turn next to the role of serotonergic modulation diated ¯ips; the frequent offensive tail ¯ips are me- in the natural economy of the cray®sh. It seems plau- diated by non-giant circuitry. Since the animals are sible to assume that serotonergic modulation of escape ®ghting vigorously and providing many of the sorts re¯exes plays an important role during social interac- of stimuli that would ordinarily evoke GF escape, tions, since otherwise its change of effect as a function the lack of responses suggests that GF escape is in- of social status makes little sense. This belief is sup- hibited. Subsequently, during phase II, the dominant ported by serotonin's promotion of a dominant stance harasses the subordinate a great deal and the sub- (Livingstone et al., 1980) and by the more recent ®nd- ordinate commonly escapes either with GF or non- ing that injection of serotonin reduces the extent to G responses. It appears that during this period GF which animals will retreat during ®ghts (Huber et al., excitability is elevated in both subordinates and 1997). However, the general notion that serotonergic dominants, since both animals typically make a few modulation of the GFs might be used during social GF escape responses per observation period that do interactions does not help to explain either the reason not appear to be caused by the sorts of sudden stim- for the complex mix of facilitatory and inhibitory ef- uli that are ordinarily needed to trigger GF escape. fects as a function of application regimen nor does it Additional GF responses are made during this period 712 F. B. KRASNE AND D. H. EDWARDS Downloaded from http://icb.oxfordjournals.org/ by guest on February 13, 2012

FIG. 8. Decline in frequency of aggressive interactions following formation of a dominance hierarchy in groups of 5 juvenile cray®sh. Top: Average frequencies (and SD) of encounters between any two animals and of each of four behaviors. Measurements were made over 15 min perods during the ®rst hour of group formation and over one hour periods during the following days. Bottom: Relative proportions of each behavior (Issa et al., 1999). There is a gradual decline in the number of encounters and a gradual increase in the extent to which the subordinate reacts to threats by walking retreats rather than tail-¯ip escape reactions.

when the dominant attacks the subordinate or when that could be used to directly test the excitability of the subordinate bumps into the dominant while ex- the LG (Figure 10; Krasne et al., 1997). When the ecuting a non-G . sensory excitability of the animals tested apart and to- Animals that have had considerable experience with gether was compared, it was found that, while the an- one another (phase III) have been studied with chron- imals are together, the sensory threshold of subordi- ically implanted stimulating and recording electrodes nates is about triple its value when they are apart; in MODULATION OF THE CRAYFISH ESCAPE REFLEX 713

FIG. 10. Effect of encounter of a subordinate with its dominant partner on LG threshold. The ability of a 0.25 msec voltage pulse to sensory nerves via chronically implanted electrodes to ®re LGs (and also a large sensory , A) was tested every few min- utes with the level of shock being varied so as to continuously eval- uate the threshold voltage for ®ring (from Krasne et al., 1997). FIG. 9. Ethogram of a pair of juvenile cray®sh at their ®rst meeting. The abscissa is the ordinal number of the observed behavioral ac- Downloaded from tion; this, rather than time, is used so that actions occurring very close together in time are somewhat spread out along the axis. Giant neuron-mediated (GF) and non-G mediated (NG) tail ¯ips were dis- tinguished with the aid of bath recording electrodes. ``Attack'' refers to rapid approaches leading to contact. ``Approach'' refers to slower approach. ``Retreat'' refers to movements away not involving tail http://icb.oxfordjournals.org/ ¯ips. ``Swims'' are strings of tail-¯ips with about a 100 msec inter- ¯ip interval (Herberholz et al., 2001).

contrast, the threshold of the dominant is elevated just slightly while the animals are together. It may seem odd that the subordinates should inhibit

GF escape in the presence of their dominant partner. by guest on February 13, 2012 However, inhibition of GF escape does not mean in- hibition of all escape. Subordinates do in fact execute escape responses when harassed by the dominant; however, these are virtually always non-G rather than GF responses (Krasne et al., 1997). Our interpretation of the suppression of GF responses by subordinates depends on the differing capacities of the giant and non-giant circuitry. GF responses are useful for a sur- FIG. 11. Summary of GF excitability changes during social en- prise attack when an animal is taken unawares, but the counters. Top row: Estimates of GF re¯ex excitability during each more sophisticated non-G responses are more adaptive phase of a developing social relationship are indicated by circles. if an animal has a chance to see its attacker coming Estimates of excitability when the conspeci®cs are separated (the and time to prepare. GF and non-G escape are in-effect ``apart'' level) are indicated as a dashed line, and the direction of change from this level is emphasized by the arrows. For phase I the incompatible strategies. An animal that is watching a measure of excitability used is the number of GF response per abrupt possible attacker approach and is preparing to execute stimulus, which is virtually zero. Although we do not know how an optimal non-G response at an optimal moment common GF responses would be to the same physical stimulus if should avoid the production of a stereotyped GF tail- animals were apart, we believe from experience with cray®sh that they would be fairly common. We have therefore placed a dashed ¯ip, and cray®sh do generally inhibit GF escape cir- line with question marks to indicate the ``apart'' level of responses; cuitry when preparing to make a non-G responses the placement of this line at 1 is arbitrary, but its placement above (Krasne and Wine, 1975, 1984). We suggest that sub- zero is consistent with the known sensory sensitivity of GF respons- ordinates are in a continual state of vigilance and that es. For phase II we have graphed the number of GF responses per the dominant probably never takes them by surprise. animal that occurred during the approximately 25 min observation period without any abrupt stimulus of the kind that would normally If so, it is probably adaptive for subordinates to sup- be needed to trigger GF escape (based on study of the high speed press GF-mediated responding. It might also make video data of Herberholz et al., 2001). Normally, in animals outside sense for excitability of non-G responses to increase a social situation, GF responses would not occur without clear abrupt in subordinates when animals are together, but this has stimuli; therefore, we have placed the dashed ``apart'' level line at zero. For phase III, we have used the increase of the directly mea- not been tested. sured sensory threshold for LG escape, when animals that were sep- These neuroethological observations are summa- arated were brought together, as a measure of the excitability in- rized at the top of Figure 11. GF escape appears to be crease of GF escape. 714 F. B. KRASNE AND D. H. EDWARDS inhibited during phase I, facilitated during phase II, Furshpan, E. J. and D. D. Potter. 1959. Transmission at the giant and inhibited in subordinates during phase III. motor synapses of the cray®sh. J. Physiol. (London) 145:289± 325. It has not yet been possible to measure serotonin Giaume, C., R. T. Kado, and H. Korn. 1987. Voltage-clamp analysis levels in freely behaving animals. However, if we think of a cray®sh rectifying synapse. J. Physiol. 386:91±112. of serotonin as being released when animals are en- Glanzman, D. L. and F. B. Krasne. 1983. Serotonin and octopamine gaged in an agonistic encounter, we might conjecture have opposite modulatory effects on the cray®sh's lateral giant escape reaction. J. Neurosci. 3:2263±9. that serotonin levels would be high in both individuals Glanzman, D. L. and F. B. Krasne. 1986. 5,7-Dihydroxytryptamine during the contest phase and would be reduced sub- lesions of cray®sh serotonin-containing neurons: Effect on the stantially but not to zero during the early post-reso- lateral giant escape reaction. J. Neurosci. 6:1560±9. lution phase. Once the relationship is ®rmly estab- Herberholz, J., F. A. Issa, and D. H. Edwards. 2001. Patterns of lished, dominants seem to go about their business neural circuit activation and behavior during dominance hier- archy formation in freely behaving cray®sh. J. Neurosci. 21: while largely ignoring the subordinate; thus we might 2759±67. conjecture that they are not then releasing serotonin. Huber, R. and E. A. Kravitz. 1995. A quantitative analysis of ago- In contrast, the subordinates, which must continually nistic behavior in juvenile American lobsters (Homarus amer- be on guard against harassment by the dominant, icanus L.). Behav. Evol. 46:72±83. Huber, R., K. Smith, A. Delago, K. Isaksson, and E. A. Kravitz. might be expected to have a steady low level of re- 1997. Serotonin and aggressive motivation in crustaceans: Al- lease. This conjectured pattern is portrayed at the bot- tering the decision to retreat. Proc. Nat. Acad. Sci. U.S.A. 94: tom of Figure. 11. The middle row of Figure 11 shows 5939±42. Downloaded from the modulatory effects these conjectured levels of 5- Issa, F. A., D. J. Adamson, and D. H. Edwards. 1999. Dominance HT might produce given what we know about the ef- hierarchy formation in juvenile cray®sh procambarus clarkii. J. Exp. Biol. 202 Pt 24:3497±506. fects of 5-HT on EPSPs in the LGs. During the contest Jaslove, S. W. and P. R. Brink. 1986. The mechanism of recti®cation phase when serotonin is high, we might expect chlo- at the electrotonic motor giant synapse of the cray®sh. Nature ride conductance-increase inhibition to operate. During 323:63±5. the post-resolution phase, when 5-HT levels are non- Kandel, E. R. 1976. The cellular basis of behavior: An introduction http://icb.oxfordjournals.org/ to behavioral neurobiology. Freeman, San Francisco. zero but low and the animals response to serotonin has Krasne, F. B. 1969. Excitation and habituation of the cray®sh escape not yet been transformed from its isolated state, we re¯ex: The depolarizing response in lateral giant ®bres of the might expect GF escape to be hyperexcitable in both isolated abdomen. J. Exp. Biol. 50:29±46. members of a pair. During Phase III we would expect Krasne, F. B. 1976. Invertebrate systems as a means of gaining in- little modulation in the dominant under the assumption sight into the nature of learning and memory. In M. R. Rosen- zweig and E. L. Bennett (ed.), Neural mechanisms of learning that serotonin release is negligible, whereas the con- and memory, pp. 401±29. MIT Press, Cambridge. Massachu- jectured low-level of 5-HT release in the subordinate setts. Krasne, F. B. and D. L. Glanzman. 1986. Sensitization of the cray®sh should cause potassium conductance-increase based by guest on February 13, 2012 inhibition, since the response to 5-HT has had time to lateral giant escape reaction. J. Neurosci. 6:1013±20. Krasne, F. B. and S. C. Lee. 1988. Response-dedicated trigger neu- become transformed. Thus, the pattern of modulation rons as control points for behavioral actions: Selective inhibi- of GF escape that rather plausibly might be conjec- tion of lateral giant command neurons during feeding in cray- tured based on what is known of serotonin's effects on ®sh. J. Neurosci. 8:3703±12. the LGs matches fairly well the observed pattern of Krasne, F. B., A. Shamsian, and R. Kulkarni. 1997. Altered excit- GF excitability. Of course it seems likely that modu- ability of the cray®sh lateral giant escape re¯ex during agonistic encounters. J. Neurosci. 17:709±16. lation by GABA-ergic tonic inhibition also contributes Krasne, F. B. and T. M. Teshiba. 1995. Habituation of an invertebrate to the observed modulations of excitability. Neverthe- escape re¯ex due to modulation by higher centers rather than less we can now see possible functional correlates of local events. Proc. Natl. Acad. Sci. U.S.A. 92:3362±6. the complex serotonergic modulatory phenomena that Krasne, F. B. and J. J. Wine. 1975. Extrinsic modulation of cray®sh physiological studies have uncovered. escape behaviour. J. Exp. Biol. 63:433±50. Krasne, F. B. and J. J. Wine. 1984. The production of cray®sh tail®p escape responses. In R. C. Eaton (ed.), Neural mechanisms of REFERENCES startle behavior, pp. 179±211. Plenum, New York. Beltz, B. S. and E. A. Kravitz. 1983. Mapping of serotonin-like Kravitz, E. A. 1988. Hormonal control of behavior: Amines and the immunoreactivity in the lobster nervous system. J. Neurosci. 3: biasing of behavioral output in lobsters. Science 241:1775± 585±602. 1781. Bliss, T. V. and G. L. Collingridge. 1993. A synaptic model of mem- Lee, S. C. 1996. Long-term depression of cray®sh lateral giant re- ory: Long-term potentiation in the hippocampus. Nature 361: sponse induced by a train of low frequency stimulation of pri- 31±9. mary afferents. Soc. Neurosci. Abstr. 26:1405. Bovbjerg, R. V. 1953. Dominance order in the cray®sh Orconectes Lee, S. C. and B. J. Park. l997. Long term depression of the root- virilis (Hagen). Physiol. Zool. 26:173±178. evoked PSP in the cray®sh lateral giants. Neurosci. Abstr. 23: Bustamante, J. and F. B. Krasne. 1991. Effects of octopamine on 228. transmission at the ®rst synapse of the cray®sh lateral giant Livingstone, M., R. Harris-Warrick, and E. Kravitz. 1980. Serotonin escape reaction pathway. J. Comp. Physiol. A 169:369±377. and octopamine produce opposite postures in lobsters. Science Edwards, D. H., W. J. Heitler, and F. B. Krasne. 1999. Fifty years 208:76±79. of a command neuron: The neurobiology of escape behavior in Lowe, M. E. 1956. Dominance-subordinance relationships in the the cray®sh. Trends Neurosci. 22:153±61. cray®sh Cambarellus shufeldtii. Tulane Stud. Zool. 4:l39±l70. Edwards, D. H., W. J. Heitler, E. M. Leise, and R. A. Fricke. 1991. Miller, M. W., E. T. Vu, and F. B. Krasne. 1992. Cholinergic trans- Postsynaptic modulation of rectifying electrical synaptic inputs mission at the ®rst synapse of the circuit mediating the cray®sh to the LG escape command neuron in cray®sh. J. Neurosci. 11: lateral giant escape reaction. J. Neurophysiol. 68:2174±84. 2117±29. Sherff, C. M. and T. J. Carew. 2002. Behavioral, cellular, and mo- MODULATION OF THE CRAYFISH ESCAPE REFLEX 715

lecular analysis of memory in Aplysia II: Long-term facilitation. Vu, E. T., S. C. Lee, and F. B. Krasne. 1993. The mechanism of Integr. Comp. Biol. 42:736±742. tonic inhibition of cray®sh escape behavior: Distal inhibition Spitzer, N., B. L. Antonsen, D. J. Baro, and D. H. Edwards. 2001. and its functional signi®cance. J. Neurosci. 13:4379±93. Partial localization of a putative serotonin receptor in the cray- Yeh, S. R., R. A. Fricke, and D. H. Edwards. 1996. The effect of ®sh nervous system. Soc. Neurosci. Abstr. 27: Program # social experience on serotonergic modulation of the escape cir- 942.11. cuit of cray®sh [see comments]. Science 271:366±9. Sutton, M. A. and T. J. Carew. 2002. Behavioral, cellular, and mo- Yeh, S. R., B. E. Musolf, and D. H. Edwards. 1997. Neuronal ad- lecular analysis of memory in Aplysia I: Intermediate-term aptations to changes in the social dominance status of cray®sh. memory. Integr. Comp. Biol. 42:725±735. Teshiba, T., A. Shamsian, B. Yashar, S. R. Yeh, D. H. Edwards, and J. Neurosci. 17:697±708. F. B. Krasne. 2001. Dual and opposing modulatory effects of Zucker, R. S. 1972. Cray®sh escape behavior and central synapses. serotonin on cray®sh lateral giant escape command neurons. J. II. Physiological mechanisms underlying behavioral habitua- Neurosci. 21:4523±9. tion. J. Neurophysiol. 35:621±37. Vu, E. T. and F. B. Krasne. 1993. Cray®sh tonic inhibition: Pro- Zucker, R. S., D. Kennedy, and A. I. Selverston, 1971. Neuronal longed modulation of behavioral excitability by classical GA- circuit mediating escape responses in cray®sh. Science 173: BAergic inhibition. J. Neurosci. 13:4394±402. 645±50. Downloaded from http://icb.oxfordjournals.org/ by guest on February 13, 2012