Abstract
Potassium channels have been implicated in central roles in activity-dependent neural plasticity. The giant fiber escape pathway ofDrosophila has been established as a model for analyzing habituation and its modification by memory mutations in an identified circuit. Several genes in Drosophila encoding K+ channel subunits have been characterized, permitting examination of the contributions of specific channel subunits to simple conditioning in an identified circuit that is amenable to genetic analysis. Our results show that mutations altering each of four K+ channel subunits (Sh,slo, eag, and Hk) have distinct effects on habituation at least as strong as those ofdunce and rutabaga, memory mutants with defective cAMP metabolism (Engel and Wu, 1996). Habituation, spontaneous recovery, and dishabituation of the electrically stimulated long-latency giant fiber pathway response were shown in each mutant type. Mutations of Sh (voltage-gated) andslo (Ca2+-gated) subunits enhanced and slowed habituation, respectively. However, mutations ofeag and Hk subunits, which confer K+-current modulation, had even more extreme phenotypes, again enhancing and slowing habituation, respectively. In double mutants, Sh mutations moderated the strong phenotypes of eag and Hk, suggesting that their modulatory functions are best expressed in the presence of intactSh subunits. Nonactivity-dependent responses (refractory period and latency) at two stages of the circuit were altered only in some mutants and do not account for modifications of habituation. Furthermore, failures of the long-latency response during habituation, which normally occur in labile connections in the brain, could be induced in the thoracic circuit stage in Hk mutants. Our work indicates that different K+ channel subunits play distinct roles in activity-dependent neural plasticity and thus can be incorporated along with second messenger “memory” loci to enrich the genetic analysis of learning and memory.
- habituation
- learning and memory
- giant fiber escape response circuit
- Drosophila
- insects
- invertebrates
- K+
- potassium channels
- α subunit
- β subunit
- Shaker
- Sh
- slowpoke
- slo
- ether à go-go
- eag
- Hyperkinetic
- Hk
Experience-dependent neural plasticity implies the modification of synaptic strength or neural firing properties by activity. K+ channels are likely to be crucial, both as regulators of neural activity and as targets of modulation leading to lasting changes in synaptic efficacy (Rudy, 1988; Alkon, 1990; Hille, 1992; Klein, 1995; Byrne and Kandel, 1996). However, roles of different channel subunits are difficult to dissect using conventional physiological and pharmacological approaches, in which effects are rarely limited to one channel type. Therefore there is value to a genetic approach that allows identified channels to be mutated in intact animals, linking behavioral alterations to cellular or molecular defects.
Several genes identified by hyperexcitable phenotypes (i.e., leg shaking) have been shown to encode K+ channel subunits (Kaplan and Trout, 1969; Hall, 1982; Wu and Ganetzky, 1992) of different families that contribute to a variety of K+ channels in Drosophila and other animals including vertebrates (Jan and Jan, 1990; Rehm and Tempel, 1991; Salkoff et al., 1992; Chouinard et al., 1995; Ganetzky et al., 1995; Baro et al., 1997). Furthermore, mutations that modify ion channels can affect behavioral plasticity, e.g., in courtship (Sh, eag) (Cowan and Siegel, 1984; Griffith et al., 1994) and classical olfactory conditioning (Sh) (Cowan and Siegel, 1986). However, the functional contributions of these subunits to channels in neural circuits are not well understood.
Three such genes encode α (pore-forming) subunits of multimeric K+ channels with different properties.Shaker (Sh) channels are voltage-gated (Salkoff and Wyman, 1981; Wu and Haugland, 1985; Iverson et al., 1988; Timpe et al., 1988), whereas slowpoke (slo) channels are activated by cytoplasmic Ca2+ (Komatsu et al., 1990;Atkinson et al., 1991). Both could regulate neuronal excitability and synaptic transmission (Jan et al., 1977; Tanouye et al., 1981; Ganetzky and Wu, 1982; Elkins et al., 1986; Gho and Ganetzky, 1992). Theether à go-go (eag) subunit seems to coassemble into channels with other subunit types, and eagmutations affect several K+ currents, including those mediated by Sh and slo channels (Zhong and Wu, 1991b, 1993b; Chen et al., 1996). Sequence analysis and physiological results suggest that the eag subunit is a target for channel modulation by phosphorylation and cyclic nucleotide binding (Warmke et al., 1991; Brüggemann et al., 1993; Zhong and Wu, 1993b; Griffith et al., 1994). Hyperkinetic(Hk) encodes a β (auxiliary) subunit (Chouinard et al., 1995) that associates with heterologously expressed Shchannels to confer modulation (Rettig et al., 1994; Chouinard et al., 1995; Rhodes et al., 1995), and Hk mutations alter the amplitude and kinetics of Sh-type currents, especially in near-threshold voltages, in Drosophila muscle (Wang and Wu, 1996) and cultured neurons (Yao and Wu, 1995).
Differences in the action and modulation of Sh,eag, slo, and Hk subunits suggest that they could play distinct functional roles in neural plasticity. However, differential expression and alternative RNA splicing of these genes in different excitable tissues (Schwarz et al., 1988, 1990;Stocker et al., 1990; Tseng-Crank et al., 1991; Becker et al., 1995;Mottes and Iverson, 1995), with the potential for heteromeric subunit combinations (Haugland and Wu, 1990; Isacoff et al., 1990; McCormick et al., 1990; Ruppersberg et al., 1990; Wu and Chen, 1995; Chen et al., 1996), raise the question of how a diversity of K+channel types regulates different circuit components. Such functional complexity could be dissected via the use of mutations to analyze plasticity in a well defined circuit.
The study of mechanisms of neural plasticity has been facilitated by simple conditioning paradigms such as habituation (e.g., Thompson and Spencer, 1966; Klein et al., 1980; Fitzgerald et al., 1990; Krasne and Teshiba, 1995). However, only recently has the genetic analysis of habituation been applied to an identified neural circuit in the fly in which molecular and behavioral phenotypes can be linked to physiology (Engel and Wu, 1996). The giant fiber escape pathway mediates a visually evoked jump-and-flight escape response (Levine and Tracey, 1973; Tanouye and Wyman, 1980; Wyman et al., 1984). Different intensities of electrical stimulation, passed between electrodes in the eyes, can be used to activate afferent elements in the brain (long-latency response) or to bypass them and trigger the thoracic stage (short-latency response) (Fig. 1), enabling the analysis of distinct circuit components with different response properties (Elkins and Ganetzky, 1990; Trimarchi and Schneiderman, 1993; Engel and Wu, 1996; Lin and Nash, 1996). We have demonstrated that the long-latency response shows characteristic parameters of habituation (Thompson and Spencer, 1966) and have used the memory mutations rut and dnc (Dudai, 1988;Tully, 1991; Davis, 1996) to show that perturbing cAMP metabolism affects habituation in this circuit (Engel and Wu, 1996), consistent with their effects on behavioral plasticity (e.g., Corfas and Dudai, 1989). We report here that mutations of identified K+ channel subunits induce specific and distinct effects on habituation in the giant fiber circuit of the fly, to an extent as extreme as rut and dnc.
MATERIALS AND METHODS
Controls were Canton-Special (CS) flies (n = 41). Mutant stocks were as follows: for Shaker,ShKS133 (n = 15),y cho f ShrKO120 (n= 11), and Sh5 (n = 6) (Wu and Haugland, 1985); for slowpoke, slo st(n = 17) (Elkins et al., 1986) andslo98 (n = 2) (Komatsu et al., 1990); for ether à go-go,eag1 (n = 11) andeag4pm (n = 5) (Ganetzky and Wu, 1983); and for Hyperkinetic,Hk1 (n = 5) and the viable deletion HkIE18 (n = 6) (Wang and Wu, 1996). Combinations wereeag1ShKS133 (n = 10) andeag1ShrKO120 (n = 9) (Ganetzky and Wu, 1983) and Hk1 f ShrKO120 (n = 5),Hk1 f Sh5 (n = 5), andHk1 g eag1ShKS133 (n = 5) (Yao and Wu, 1995). Values of n refer to the primary data set displayed in the Results (see Figs. 2, 3, 6, 7, Table 1); results for mutant alleles of each gene did not differ and were pooled, except thatShKS133 is contrasted withShrKO120 andSh5 (see Figs. 3,6A, Table 1). Viable genetic markers of eye color [chocolate (cho), garnet(g), and scarlet (st)], body color [yellow (y)], and bristle morphology [forked (f)] are described by Lindsley and Zimm (1992). Ether-induced shaking behavior was present in these stocks as reported previously, and in eag Sh double mutants, the characteristic occurrence of flies with wings held downward and an indented notum was also noted (Engel and Wu, 1992).
Physiological methods were described in Engel and Wu (1996). High- and low-voltage pulses from electrodes in the eyes were used to trigger the cervical giant fiber (CGF) or afferent inputs to the CGF in the brain (Fig. 1). For the short-latency response, stimuli were ∼2 V above threshold; for the long-latency response, stimuli were set 0.2–0.4 V below the threshold for shorter latency responses (Engel and Wu, 1996), except that stimulus voltage was reduced further for some trials as noted (see Figs. 5, 10). Responses were recorded from a tergotrochanteral muscle (TTM) (leg extensor) and a fiber of the contralateral dorsal longitudinal muscle (DLM) (wing depressor) (Fig.1). DLM fiber a was most commonly recorded; previous work indicates that ipsilateral DLM fibers respond and fail together in habituation of the long-latency response (Engel and Wu, 1996).
For habituation trials, stimulation was given at 5 Hz, found previously to induce habituation within tens to hundreds of stimuli in a variety of genotypes (Engel and Wu, 1996), except for some trials as noted (see Figs. 5, 10). Habituation was indicated by a criterion of five consecutive failures of the DLM flight muscle response (Engel and Wu, 1996), and trials were terminated after five consecutive failures to test recovery or after 1000 stimuli (200 sec) if criterion was not obtained first. Air puffs (Engel and Wu, 1996) were given immediately after five consecutive failures to test dishabituation. For statistical analyses, data for stimuli-to-failure criterion (see Fig. 3) and refractory periods (see Fig. 8) were log-transformed to improve normality (Engel and Wu, 1996).
RESULTS
Habituation in single mutants
We demonstrated previously six of the parameters of habituation ofThompson and Spencer (1966) for the long-latency giant fiber response, including frequency dependence, spontaneous recovery, faster rehabituation, habituation beyond zero response, dishabituation by air puff or light flash, and habituation of dishabituation (Engel and Wu, 1996). A seventh parameter, stronger habituation with weaker stimulation, is described below. To characterize habituation in several mutants of four gene loci, this work focused on the most essential parameter, loss of the response, at a standard stimulus strength and frequency (see Materials and Methods). Recovery and dishabituation were also examined to demonstrate reversibility. Long-latency giant fiber responses to electrical stimulation and habituation of the response were obtained in flies of every genotype.
Mutations affecting different channel subunits led to profound differences in the rate of habituation (Fig.2, Table 1) and the time required to reach a criterion level of habituation (i.e., five consecutive failures; Fig. 3). Different mutant alleles of each gene produced consistent results, except that two alleles of Sh that modify channel function differed from an allele that eliminates function (see below). Therefore, alleles have been combined in figures and tables unless indicated otherwise. Similarities between mutant alleles of a gene locus indicate the phenotypes are attributable to those loci rather than unidentified variability in other parts of the genome.
Sh and slo encode subunits with homologous structure, with the major distinction that Sh subunits are voltage-gated whereas slo subunits require cytosolic Ca2+ for activation (Wu and Ganetzky, 1992). Twoslo mutations, slo1 andslo98 , markedly reduced the rate of habituation (Fig. 2, Table 1) and the time to reach five-failure criterion (Fig. 3), so that many slo mutant flies did not achieve five consecutive failures within the trial length limit of 1000 stimuli. slo1 andslo98 may be amorphic (null), because both alleles eliminate a Ca2+-activated K+ current in muscle (Elkins et al., 1986; Komatsu et al., 1990). In contrast, Sh mutants habituated more rapidly than did controls (Fig. 2, Table 1), but there were interesting differences between alleles. Sh5 and ShrKO120 , which express functional but altered subunits, led to earlier habituation (Fig. 3, Table 1). The Sh5 mutation, in an exon common to all Sh transcripts (Gautam and Tanouye, 1990;Lichtinghagen et al., 1990), alters channel-gating kinetics and voltage sensitivity (Salkoff and Wyman, 1981; Wu and Haugland, 1985); theShrKO120 lesion site is not known but may affect specific splicing variants because its phenotype is far more extreme in neurons than in muscle (Wu and Haugland, 1985; Wu and Ganetzky, 1992). ShKS133 , a missense mutation that affects the pore-forming region of all transcripts (Lichtinghagen et al., 1990), is an antimorph (Haugland and Wu, 1990) that eliminates a transient K+ current,IA , in muscle (Salkoff and Wyman, 1981; Wu and Haugland, 1985). Surprisingly, its effects on habituation were less extreme than were those of Sh5 andShrKO120 (Fig. 3, Table 1).
Mutations of the two subunits implicated in channel modulation,eag and Hk, had even stronger effects on habituation than did Sh and slo, respectively.eag1 andeag4pm increased the rate of habituation (Fig. 2, Table 1) and reduced the number of stimuli to attain five consecutive failures (Fig. 3).Hk1 and the amorphic deletionHkIE18 both dramatically reduced the rate of habituation (Fig. 2, Table 1) and delayed attainment of five consecutive failures (Fig. 3), so that the majority ofHk mutants did not attain this criterion within 1000 stimuli. It should be noted that habituation was assessed according to DLM muscle responses (Engel and Wu, 1996), but in many Hkflies, the TTM muscle continued to respond after the DLM pathway failed (an otherwise rare phenomenon, described below). In this respect, the phenotype of Hk mutants could be even more extreme than is indicated by these plots (Figs. 2, 3).
Dishabituation, spontaneous recovery, and failure patterns
It is important to consider whether the conditioning induced in these mutants is habituation as defined previously in this system (Engel and Wu, 1996). As defined previously, habituation can be localized to a portion of the circuit in the head (Fig. 1) because DLM and TTM muscle responses failed in synchrony (an exception is described below), and the pathway in the thorax can respond at frequencies well above the 5 Hz used to induce habituation (Engel and Wu, 1992; see Fig.8B for comparison). Response latencies were only slightly altered in mutants (see below), and there were no abrupt changes in latency during habituation that could have indicated a novel activity-dependent circuit response.
Moreover, two critical characteristics of habituation, spontaneous recovery and dishabituation induced by a novel stimulus (air puff), were observed in flies of all genotypes. Figure4 shows spontaneous recovery after 30 sec of rest after habituation by 5 Hz stimuli. eag,Sh, and slo mutants recovered at least as fully as did controls. Hk flies are not represented in Figure 4because they did not readily reach the five-failure habituation criterion at 5 Hz, but Hk animals did recover from habituation when tested with increased stimulus frequencies in order to attain habituation (see below).
Figure 5 shows examples of dishabituation in each of the mutant types. To demonstrate dishabituation in slowly habituating Hk and slo mutants, we could use a higher stimulation frequency or a lower stimulus voltage to induce habituation (Fig. 5, see Fig. 10). The latter approach takes advantage of a negative relationship between stimulus intensity and the strength of habituation (Thompson and Spencer, 1966), a parameter of habituation that was not reported previously in this system (Engel and Wu, 1996).
The mutant genotypes may be distinguished further by examining the clustering of response failures. Again, eag andHk produced the most pronounced deviations from controls. Figure 6A plots the cumulative frequency of switches between responses and failures. Theslope of a curve indicates both the frequency of failures and the degree of clustering (longer strings of successes or failures lead to fewer switches). The curve for eag was similar to that for ShrKO120 andSh5 during the first 100 stimuli but was steeper than that forShKS133 , reflecting a greater number of failure strings. After approximately stimulus 90, theeag curve flattened abruptly because of the termination of faster-habituating trials at five-failure criterion (Fig.6A, inset).
The pattern of failures also differed between slo andHk mutants (Fig. 6A), even though both showed diminished habituation (Figs. 2, 3). Failure onset was delayed in slo mutants; the earliest failure (in 19 trials) occurred at the 16th stimulus (Fig. 6A). After failure onset, response-and-failure switching was as frequent as in controls until approximately stimulus 175, when the rate diminished indicating that the remaining subset of slo animals (i.e., trials not yet terminated; Fig. 6A, inset) switched less frequently. In contrast, failure onset was much later in Hktrials, no earlier than stimulus 62 (in 11 trials), and the frequency of switching (slope in Fig. 6A) never attained the level seen in controls.
Figure 6B shows the frequency of strings of consecutive failures in the first 100 stimuli, grouped by string length. The first bar in each histogram in Figure6B [no failures (nf)] indicates the proportion of trials without a single failure in 100 stimuli. The last bar (5+) indicates the proportion of trials in which a string of five or more failures occurred (equal to or more than five because trials were terminated after five consecutive failures). The most notable result in Figure 6B is that the rankings for those two categories are consistent with the rates of habituation indicated in Figures 2 and 3; for nf, Hk >slo > controls > Sh >eag, and for 5+, Hk < slo< controls ≤ Sh < eag. Distribution patterns of one to four failure strings were less markedly different between genotypes. Slower habituation in slo is attributable to shorter failure strings, because the average number of failures in 100 stimuli was similar to that in controls and Sh (Fig.6B, failures/trial). In contrast,Hk failed much less than did slo because of late failure onset (compare Fig. 6A). eagmutants showed the greatest failure frequency, even though the true frequency is underestimated because most eag animals habituated in under 100 stimuli (Fig. 6A,inset).
Habituation in double mutants
Of the channel subunit genes studied, eag mutants showed the most extreme increase in strength of habituation, whereasHk mutants showed the most extreme decrease (Figs. 2, 3, 6, Table 1). There is strong physiological evidence (reviewed in the introductory remarks) that both eag and Hksubunits can interact with Sh channels. Therefore, double- and triple-mutant combinations were examined to determine how these interactions contribute to habituation. eag andSh mutations generally interact to produce phenotypes more extreme than that produced by either mutation alone for characters including motoneuron excitability, terminal branching and synaptic transmission at neuromuscular junctions, leg shaking, novel wing position, and defective flight (Ganetzky and Wu, 1983; Budnik et al., 1990; Engel and Wu, 1992). This is consistent with the finding that in muscle, eag contributes to several K+currents besides the Sh IA (Zhong and Wu, 1993b; Wu and Chen, 1995). Therefore, it was surprising that rates of habituation in eag Sh double mutants were no more extreme than the rates in either Sh or eagalone (Figs. 2, 3, Table 1). Failures began early in eag Shdouble mutants (Fig. 7A), as in eag alone, but isolated failures (string length = 1) were common (Fig. 7B). Spontaneous recovery from habituation in eag Sh double mutants is shown in Figure 4.
In contrast to eag and Sh, Hk andSh mutations generally do not interact to produce more extreme phenotypes than produced by Sh alone for characters including neuromuscular synaptic facilitation and aberrant neuronal spiking (Stern and Ganetzky, 1989; Yao and Wu, 1995; W.-D. Yao and C.-F. Wu, unpublished observations), and in both muscle and neurons,Hk has only been shown to affect the Sh IA current (Yao and Wu, 1995; Wang and Wu, 1996). Yet, habituation in Hk Sh double mutants was markedly retarded, as in Hk mutants, rather than more rapid as inSh (Figs. 2, 3, Table 1). The strong effect of Hkwas apparent even in Hk1eag1ShKS133 triple mutants, which habituated more slowly than did controls, unlike eag,Sh, or eag Sh (Figs. 2, 3). In Hk Shand Hk eag Sh mutants, failure onset was delayed compared with Sh or eag Sh (Fig. 7A), yet rates of response-and-failure switching (Fig. 7A) were more normal than in Hk alone.
Double- and triple-mutant results indicate that the modification ofSh subunits alone has less impact on habituation than does altering Hk or eag subunits in the presence of functional Sh subunits. Unlike the previously studied excitability phenotypes listed above, habituation is an activity-dependent process. Apparently, Hk andeag mutations affect conditioning processes more severely than does Sh, even though Sh affects nonactivity-dependent parameters of the giant fiber response, including the latency and refractory period as described below.
Resting properties: refractory period and latency
Refractory period and response latency are temporal characteristics on a millisecond scale that might influence the process of habituation, a type of activity-dependent conditioning lasting seconds or longer. These processes could be mediated by overlapping mechanisms at overlapping sites in the circuit. In fact, we found that alterations in refractory period and response latency in K+ channel mutants do not account for their effects on habituation. To analyze response properties at two stages of the giant fiber pathway, low-voltage stimuli were used to trigger the long-latency response that begins in labile-afferent pathways in the brain where habituation is mediated (Engel and Wu, 1996), whereas stronger stimuli were given to bypass the afferent stage (see Materials and Methods) and trigger a short-latency response in the thoracic portion of the circuit (Fig. 1).
Refractory periods of the long-latency response tended to be shorter than normal in all K+ channel mutants examined except in eag Sh double mutants (Fig.8A), which were normal. This indicates that habituation rate is not coupled to refractory period in this system, because the two parameters are not altered in the same pattern across genotypes. A similar lack of correlation was shown with cAMP-metabolic mutations of dunce andrutabaga (Engel and Wu, 1996). Short-latency response refractory periods, mediated by the nonlabile thoracic pathway, were affected quite differently in the same mutants (Fig.8B); they were not shortened in any K+ channel mutants but were prolonged in eag Sh (cf. Engel and Wu, 1992), Hk, and Hk Shmutants.
Response latencies were also affected differently in cephalic and thoracic stages of the pathway (Fig. 9). The long-latency response was delayed slightly in Sh andeag Sh mutants. Sh, eag Sh, andslo mutations induced small but statistically significant delays in short-latency TTM responses, whereas Hk andeag mutations led to earlier short-latency DLM responses. The small scale of these differences indicates that the giant fiber pathway is still functional (cf. Baird et al., 1990) and weighs against the recruitment of potential collateral pathways in slo andHk mutants (see below). Although latency increased gradually during habituation (Engel and Wu, 1996), sudden increments were not observed. The most notable feature of these results, however, is that the pattern of nonactivity-dependent properties of nerve conduction and synaptic transmission in the circuit, including marked effects on the long-latency response refractory period, do not correlate with habituation phenotypes. The lack of a simple relationship among habituation, response latency, and refractory period in K+ channel mutants implies that activity-dependent plasticity could involve multiple cellular mechanisms at different sites in the circuit, influenced by K+ currents, that shape efficacy in axonal conduction and synaptic transmission.
Shift of long-latency response failure site in K+ channel mutants
Habituation of the long-latency response is characterized by synchronous failures of DLM and TTM responses (Fig. 5) (Engel and Wu, 1996), indicating that failures occur in the brain (Fig. 1). However, in some animals of certain genotypes, the DLM branch of the circuit failed more readily than did the TTM branch. This pattern was seen most often in flies with Hk mutations (Engel, 1995) orHk combinations (Fig. 10), less frequently in slo (Fig. 5) and eag Sh (data not shown) mutants, and rarely in other mutants or controls (even in individual flies that habituated unusually slowly, approaching the level of Hk or slo mutants). These independent DLM failures probably occur after the circuit has bifurcated in the thorax (Fig. 1), and consistent with this, a similar pattern of DLM failures coupled with TTM responses could be observed when short-latency responses were induced by increasing stimulus intensity (Engel, 1995).
This pattern could indicate a shift in the site of failures from normally labile synapses in the head (e.g., site 1 in Fig.1) to electrical or cholinergic synapses of the DLM branch in the thorax (sites 2 or 3). One plausible explanation could be the recruitment of normally silent afferents in the brain leading to stronger transmission of the signal to the giant fiber (i.e., at site 1). Decreasing stimulus intensity or interstimulus intervals typically promotes habituation (Thompson and Spencer, 1966), and to extend the above scenario, extra afferent inputs activated in mutants could fail under these stimulus conditions. Indeed, after the two muscle responses had become decoupled, it was sometimes possible to induce the “typical” pattern of synchronous DLM and TTM failures either by decreasing stimulus intensity to just above the long-latency response threshold (slo1 in Fig. 5,Hk1eag1ShKS133 in Fig. 10) or by increasing the rate of stimulation (Hk1 rKO120 in Fig. 10).
Conversely, drastically reduced TTM failures in Hk, Hk Sh, and Hk eag Sh mutants could indicate recruitment of parallel inputs to the thoracic TTM branch. Although the exact site of modification in the pathway awaits further investigation, the long-latency TTM response is clearly strengthened by Hkmutations even more than is the long-latency DLM response. The typical pattern of synchronized muscle responses could still occur in these mutants (Figs. 5, 10), which indicates that the normal site of habituation in the brain is labile in these flies as well. The apparent shift in the site of failures under certain stimulus conditions inHk mutants reveals that there is considerable plasticity for circuit function that could be regulated by the Hk channel subunit in normal flies.
DISCUSSION
This is the first use of Drosophila mutations to examine the influence of identified subunits of K+channels on habituation. The giant fiber pathway system provides a means to link molecular components to cellular physiology within a defined circuit that shows plasticity (Engel and Wu, 1996). Our results establish that different K+ channels have distinct functional consequences for habituation in the giant fiber response, with effects as extreme as those of the “learning mutations”rutabaga and dunce (Engel and Wu, 1996).
Involvement of four potassium channel subunits in habituation
Potassium currents regulate signal transmission throughout the nervous system, affecting presynaptic neurotransmitter release, postsynaptic integration, and the shape and frequency coding of action potentials (Rudy, 1988; Hille, 1992; Turrigiano et al., 1994). In each of these functions, K+ channels have been implicated in neural plasticity (Alkon, 1990; Klein, 1995; Byrne and Kandel, 1996;Debanne et al., 1997; Hoffman et al., 1997). K+channels could have a dual role, as regulators of activity-dependent modulation of neural processes leading to plasticity and as targets of modulation. Besides a range of mechanisms at the level of single neurons, habituation may involve various forms of plasticity at the circuit level: conduction of excitatory paths may be depressed, or previously latent inhibitory inputs may be facilitated and recruited (cf. Charpier et al., 1995; Krasne and Teshiba, 1995), and there is often a component of sensitization that could itself involve K+ currents (Groves and Thompson, 1970). Therefore, it is essential to study habituation in a defined circuit in the intact organism, to complement work done with model synapses in reduced nervous systems. In this manner, analysis of mutants of identified K+ channel subunits has produced results that would have been difficult to predict based on other experimental approaches.
The distinct effects of mutations of four K+ channel subunits indicate differences in their physiological characteristics and localization within neurons of the pathway. Habituation was enhanced in Sh and eag mutants and reduced inslo and Hk mutants. Although Sh andslo subunits are essential for kinetically similar K+ currents in muscle (Wei and Salkoff, 1986; Singh and Wu, 1990; Wu and Ganetzky, 1992), their mutations had opposite effects on habituation, perhaps reflecting differences in their activation mechanisms. Sh-type voltage-activated channels are most likely to regulate the shape and frequency coding of sodium action potentials in axons and dendrites (Tanouye et al., 1981; Debanne et al., 1997; Hoffman et al., 1997). In contrast, slo-type Ca2+-activated K+ channels, which can colocalize with presynaptic Ca2+ channels, would be well suited for regulating the activity-dependent accumulation of presynaptic Ca2+ (Augustine et al., 1988; Gho and Ganetzky, 1992; Robitaille and Charlton, 1992; Mallart, 1993;Warbington et al., 1996). Although Sh and slosubunits are workhorses in the nervous system, each essential for a specific repolarizing current in muscle and neurons, the most extreme positive and negative changes in habituation rates were induced by mutations of eag and Hk subunits that alter K+ current modulation rather than eliminate currents (Zhong and Wu, 1991b, 1993b; Yao and Wu, 1995; Wang and Wu, 1996). Because habituation represents a change over time, it is reasonable that the most important subunits should be those that confer lasting modulation. The opposing direction of eag and Hkmutant phenotypes indicates that plasticity may be regulated by counterbalancing mechanisms (Groves and Thompson, 1970). No channel mutations eliminated habituation, which is likely to be shaped by the interaction of multiple channel types, not by a particular “habituation channel.” Similarly, mutations that eliminate enzymes for cAMP metabolism alter habituation but do not prevent it (Engel and Wu, 1996).
Habituation was enhanced by Sh alleles (ShrKO120 andSh5 ) that alter current density, voltage dependence, or kinetics but was not significantly affected by the antimorphic ShKS133 that (in muscle) eliminates Sh current (Wu and Haugland, 1985). This unexpected result implies that Sh subunits influence habituation but not as a direct result of their modulation because habituation seemed closer to normal in the absence of functionalSh channels than in the presence of the defective but functional subunits in ShrKO120 andSh5 mutants.
The eag channel subunit seems to coassemble into at least four K+ channel types in fly tissues (Zhong and Wu, 1991b, 1993b) and can interact with Sh subunits to form channels when expressed in Xenopus oocytes (Chen et al., 1996) and the two eag alleles studied alter the modulation of several K+ currents in muscle (Zhong and Wu, 1993b; Wu and Chen, 1995). Along with evidence of its modification by kinases and cyclic nucleotides (Warmke et al., 1991; Brüggemann et al., 1993; Zhong and Wu, 1993b; Griffith et al., 1994), this indicates that eag would be an ideal link for the regulation of a broad group of K+ channels. If habituation involves the modulation of channels via eag subunits, then a change in the efficacy of modulation could account for accelerated habituation in eag mutants.
The Hk β subunit is known to associate with and modulateSh channels (Rettig et al., 1994; Chouinard et al., 1995;Rhodes et al., 1995; Yao and Wu, 1995; Wang and Wu, 1996). The importance of this auxiliary subunit in habituation is indicated by the fact that Hk mutations produced the most extreme phenotype, sometimes modifying the site of failures as well as slowing the rate of habituation. Habituation may serve to regulate sensitivity to external or internal stimuli (Fischer and Carew, 1993; Bässler and Nothof, 1994), and it is notable that the visually induced jump response (mediated by the giant fiber pathway) is hypersensitive inHk mutants (Kaplan and Trout, 1969; Levine, 1974). This hypersensitivity is less extreme when Hk is combined withSh (Kaplan and Trout, 1974), consistent with our results in multiple mutants (Figs. 2, 3). Likewise, abnormal spontaneous rhythmic firing seen in cultured Hk neurons becomes less extreme inHk Sh double mutant neurons (Yao and Wu, 1995).
Mutations of Sh affected habituation less strongly than did either eag or Hk. If eag andHk subunits modify Sh currents, as has been proposed, then it seems that plasticity is disrupted by the maladjustment of modulatory mechanisms (in eag orHk mutants) far more than by the alteration or elimination of their target (Sh) channels. Hk subunits have so far only been shown to associate with Sh channels; yet even though ShKS133 eliminates functional Sh channels (Chouinard et al., 1995; Wang and Wu, 1996), Hk1eag1ShKS133 differed markedly fromeag1ShKS133 , and Hk Sh differed from Sh (Figs. 2, 3, 7, Table 1). The Hk subunit may therefore prove to modulate other K+ channels besides Sh, potentially including products of the genessei, shal, shab, and shaw(Butler et al., 1989; Tsunoda and Salkoff, 1995; Titus et al., 1996;Wang et al., 1996) that were not examined here.
Characteristics of habituation for a given behavior can depend critically on the interstimulus interval (ISI) (e.g., Groves and Thompson, 1970; Davis, 1984; Boulis and Sahley, 1988), and it is possible that K+ channels play different roles in habituation at different ISIs. A broad range of ISIs have been used in the study of behavioral habituation in diverse phyla: less than a second (May and Hoy, 1991), seconds (Thompson and Spencer, 1966; Davis, 1984; Wittekind and Spatz, 1988; Corfas and Dudai, 1989), tens of seconds to minutes (Pinsker et al., 1970; Long et al., 1989; Rankin and Broster, 1992; Krasne and Teshiba, 1995; Weil and Weeks, 1996), or hours (Brown et al., 1996). The visually induced jump response inDrosophila habituates at ISIs of 1–10 sec (Engel and Wu, 1996). In the present experiments, an ISI of 0.2 sec (5 Hz) was used; the kinetics of habituation are not expected to be identical because visual stimulation may recruit only a subset of the cervical giant fiber afferents that are activated by the electrical stimulation used here (Fig. 1).
Potassium channels and the mutational analysis of learning and memory
The classical approach to the genetic analysis of learning and memory sought mutations or transformants without obvious confounding behavioral or developmental defects. This led to genes involved with various second messenger pathways, including rutabaga,dunce, and amnesiac (cAMP cascade),ala (CaM-kinase), and KCI and turnip (protein kinase C) (Dudai, 1988; Choi et al., 1991; Tully, 1991; Griffith et al., 1993; Feany and Quinn, 1995; Davis, 1996; Kane et al., 1997), but tended to exclude ion channel mutations because these caused easily detectable behavioral defects such as ether-induced leg shaking (Dudai, 1988). It seems unlikely that the changes in habituation performance shown here in channel mutants are a spurious consequence of pleiotropic defects elsewhere in the circuit, because the response attenuation in mutants was localized to the afferent stage of the circuit where habituation ordinarily occurs (see Results) and showed recovery and habituation that appeared grossly normal.
In fact, concerns about pleiotropy also apply to classical learning mutations such as dnc (Dudai, 1988; Qui and Davis, 1993). The distinction between “pure” learning mutations and mutations with secondary physiological, developmental, or behavioral effects is now blurred. It has become clear that mutations of ion channels and second messenger systems have overlapping effects (Dudai, 1988; Davis, 1996; Wu, 1996). Mutations affecting K+ channel subunits alter activity-dependent conditioning of neural excitability and firing (Ikeda and Kaplan, 1970; Tanouye et al., 1981; Ganetzky and Wu, 1982; Saito and Wu, 1991, 1992) and synaptic efficacy (Jan et al., 1977; Ganetzky and Wu, 1983; Mallart, 1993; Delgado et al., 1994;Warbington et al., 1996). However, second messengers modulate ion channels, and mutations of second messenger pathways cause physiological effects resembling channel mutations. rut anddnc mutations that alter cAMP metabolism have been shown to alter muscle K+ currents (Zhong and Wu, 1993a), excitability in cultured neurons (Zhao and Wu, 1997), and facilitation at neuromuscular junctions (Zhong and Wu, 1991a). Branching of motor neuron terminals is increased by eag, Sh, andHk mutations that increase neural activity (Budnik et al., 1990), but cAMP metabolic mutations have identical effects. Indeed, the most extreme terminal branching occurs in dnc eag anddnc Sh double mutants and is suppressed by rut in triple mutants (Zhong et al., 1992). On the other hand, channel mutations can alter behavioral plasticity. Mutations of Shand eag reduce conditioning in courtship (Cowan and Siegel, 1984; Griffith et al., 1994) and alter habituation of a jump response to olfactory stimulation (McKenna et al., 1989; Tully and Koss, 1992) (C.-F. Wu and T. Tully, unpublished observations), and Shmutations alter classical olfactory conditioning (Cowan and Siegel, 1986).
We have shown here that mutations of four different K+ channel subunits alter habituation of the giant fiber response in gene-specific ways, with modulatory eagand Hk subunits appearing particularly influential. In the future, expression of normal or mutated genes may be concentrated in particular regions of the nervous system using gynandromorph mosaics or enhancer trap lines or temporally controlled with conditional promoters (Burg et al., 1993; Ferveur et al., 1995; Zhao et al., 1995; Han et al., 1996; Cambridge et al., 1997). The effects of such novel spatial or temporal patterns of mutant gene expression can be contrasted with the phenotypes of mutations expression patterns closer to normal, such as those used in this study, to give indications of the localization of habituation and the colocalization and potential interaction of different molecular players. More broadly, these results open the possibility to compare the effects of channel mutations and second messenger mutations at cellular, circuit, and behavioral levels within a single system.
Footnotes
- Received June 30, 1997.
- Revision received January 5, 1998.
- Accepted January 6, 1998.
This work was supported by National Institutes of Health Grants NS18500 and NS26528 to C.-F.W. We thank Xianjin Xie for technical assistance and Drs. D. Baro and R. Harris-Warrick for comments on this manuscript.
Correspondence should be addressed to Dr. Jeff E. Engel, Section of Neurobiology and Behavior, Cornell University, Seeley Mudd Hall, Ithaca, NY 14853.
- Copyright © 1998 Society for Neuroscience