- Email: [email protected]
Flies, genes, learning, and memory
Neuron, Vol. 15, 747-750, October, 1995, Copyright© 1995 by Cell Press
Flies, Genes, Learning, and Memory
Ralph J. Greenspan Department of Biology Center for Neural Science New York University New York, New York 10003
Drosophila was not always the darling of developmental biology that it is today. Mutants such as Antennapedia and wingless were once considered to be intriguing oddities but were relegated to a backwater of embryology because they bore no relevance to the way "real" animals developed. The turnaround followed a convergence of gene cloning, sequencing, and germline transformation with the ability to isolate and analyze mutations affecting pattern. Suddenly, a mutant could be cloned, its identity proven, and its time, place, and mode of action charted and manipulated. The finding that vertebrates (even humans!) had the same genes doing many of the same things didn't hurt either. The rest is history. A similar history is now being played out in the study of learning and memory. From a rocky beginning in which the very existence of fly learning was contested, the field has staggered through reckless youth and troubled adolescence and is now emerging into the bloom of earlyadulthood. Not the early adulthood of Generation X--trapped in a going-nowhere "McDiscipline"--nor the early adulthood of the culture of science welfare-trapped in an endless cycle of handouts given on the unfulfilled promise of great things to come. Instead, fly learning has earned its place as a mature and serious discipline, through hard work and adherence to traditional values. The Paradigm and "Forward" Genetics What were the ingredients of success in fly development and what are their counterparts in fly learning? The first was the establishment of a simple, reliable, and informative assay for function and mutant isolation. The segmented pattern of the larval cuticle, from which myriad mutants with tantalizing defects were amassed, did it for development. An olfactory avoidance paradigm is playing this role for fly learning (Tully and Quinn, 1985), establishing it as bona fide Pavlovian conditioning and laying to rest criticisms of its lack of rigor that came from traditional behavioral quarters (Hirsch, 1986). In the paradigm, flies are presented with two different fragrances. During exposure to one of them, they are also given an electric shock. Under these conditions, flies will learn to avoid the odor that had been accompanied by shock and wilt remember it for a day or so. Their memory can be extended to a week if they are given ten training sessions with suitable intervals in between (Tully et al., 1994). (The intervals are crucial since ten sessions in rapid succession are no better than one.) For flies in the wild, one week is long enough to arrive at midlife crisis, so this qualifies as long-term memory. One may be tempted to ask, at this point, what relevance
Minireview
learning has to the life-style of a fruit fly in the wild? A fly's ability to be conditioned during courtship appears to affect its mating success. In a laboratory version of this phenomenon, a male fly learns not to court when in the presence of unreceptive females, that is, females who have recently mated (Siegel and Hall, 1979). The larval segmentation assay served as the basis for isolating a wide array of developmental mutants identifying gap, pair-rule, and segment polarity genes as well as anterior-posterior, dorso-ventral, and terminal patterning genes (Bate and Martinez Arias, 1993). The ability to obtain more than one mutation of each gene gave a clearer picture of the range of their effects. Even before their molecular identities were known, these mutant categories stood out as obvious and distinct alterations of pattern. With their cloning, they have been organized into pathways of signal transduction and transcription. For learning studies, the olfactory avoidance paradigm has yielded mutants that all show essentially the same overt phenotype: stupidity. This leaves molecular identity as the primary basis for organizing them into understandable categories. Some affect components of the cAMP second messenger system, already well established from studies in Aplysia: a phosphodiesterase gene (dunce), an adenyl cyclase gene (rutabaga) (Davis, 1993), and a putative neuropeptide gene (amnesiac; Feany and Quinn, 1995). Others are either uncloned and unknown, like latheo (Boynton and Tully, 1992) and radish (Folkers et al., 1993), or cloned and unfamiliar, like linotte (Bolwig et al., 1995 [this issue of Neuron]). The rate of return is significantly slower for learning mutants than for those affecting development, because the learning assay uses adults and takes more time. This has also hampered the ability to obtain multiple lesions of each gene and assess its range of effects. The courtship paradigm has not been used for isolating new mutants but has served to generalize the roles of dunce and rutabaga in learning (Siegel and Hall, 1979; Gailey et al., 1984). (The etymology of the mutants' names is as follows: dunce, after John Duns Scotus, 13th century opponent to the revival of classical learning; rutagaba, because it's as dumb as a vegetable; radish, see rutabaga; amnesiac, as the name suggests; latheo, a Greek word meaning forgetful, related to the name of one of the rivers of Hades in Greek mythology, Lethe, whose waters caused all who drank from them to forget all of their former life; linotte, from the French expression t~te de linotte having the same idiomatic meaning as bird-brain.) The virtue of such a "forward genetic" approach is that it makes the fewest assumptions about which genes will be involved and lets the luck of the draw pull out those capable of producing the right kind of mutant. Part of the power of the olfactory avoidance paradigm is its ability to weed out spurious mutants that are unable to move, smell, or react to shock in task-relevant controls. Without the ability to make such distinctions, linotte, with a low initial
Neuron 748
learning score but a normal rate of memory decay, could not have been recognized as affecting initial learning. On the other side, an olfactory mutation in a ubiquitous sodium channel, smellblind (Lilly et al., 1994), would otherwise have been mistaken for a learning mutant.
Gene Identification and Manipulation A second ingredient in the success of fly development was the facility of cloning and identifying genes and of manipulating their expression. These techniques are identical in studies of either development or learning. Cloning has been accelerated by inducing mutations with transposons, mobile DNA elements that disrupt the gene and then provide a tag for retrieving the disrupted sequences. A variant of the transposon approach incorporates a reporter lacZ gene in the mobile element so that new putative mutants can be prescreened for their pattern of 13-galactosidase expression. Inducible expression of cloned genes, following germline transformation, made it possible to manipulate and define critical periods for developmental genes. This has recently been achieved for learning genes as well. The dunce mutant has been partially rescued after induction of a cAMP phosphodiesterase cDNA under the control of a heat-shock promoter (Dauwalder and Davis, 1995). The lack of complete rescue of its learning defect may indicate a developmental requirement for the dunce gene, echoing a nagging complaint about mutations affecting learning and memory: are they actually developmental mutations? Full rescue of the linotte mutation by a heat-inducible cDNA proves that adult expression is all that is needed for its action (Bolwig et al., 1995).
Pinpointing Key Steps A further ingredient in the success of developmental genetics is the ability to identify critical regulatory steps by the reciprocal effects of loss of function and gain of function changes in the gene. Such effects can be produced either by mutations or by induced expression of activating or inhibitory transgenes. Reciprocal phenotypes for developmental genes have generally been defined in two ways: those producing alternative cell fates, such as the neurogenic gene Notch, whose loss of function produces a neural fate and whose gain of function produces an epidermal fate; or those producing alternative patterns, such as the maternal gene Toil, whose loss of function produces a dorsalized embryo and whose gain of function produces a ventralized embryo (Bate and Martinez Arias, 1993). What would be reciprocal phenotypes for learning? Smart versus dumb flies. All of the mutants are poor learners by definition because that is how they were isolated. The creation of a superstrain of flies that remembers more effectively than normal fell out of a study of long-term memory as influenced by the cAMP-dependent transcription factor, CREB. A long history of studies had shown a requirement for protein synthesis in long-term memory of many animals, including Drosophila (Tully et al., 1994). Long-term facilitation in Aplysia had been shown to depend on protein synthesis and transcription of genes containing cAMP response elements (CREs; Dash et al., 1990). In flies, when expression of a negatively acting CRE-binding protein (dCREB2-b) is transiently induced
prior to training, it selectively abolishes long-term memory but leaves short-term memory unscathed (Yin et al., 1994). When a positively acting form (dCREB2-a) is similarly expressed in flies, it enables them to form long-term memory far more efficiently than normal, after only a single training session as opposed to the usual ten (Yin et al., 1995). Opposing actions of CREB produce reciprocal phenotypes: forgetfulness on one hand and photographic memory on the other. This leaves little question as to its importance. Manipulation of CREB was effected by pure reverse genetics. No mutations in the genes have yet been isolated in screens for learning mutants, but its key role was predicted from earlier studies and the selectivity of its effect was shown with the same behavioral controls used in the mutant screens. An analogous approach has established the involvement of the calcium/calmodulin-dependent protein kinase II (CaMKll) in the courtship paradigm. Heat-shockinduced expression of a specific peptide inhibitor of the kinase produces a graded effect on conditioning in this paradigm: a mild reduction in enzyme activity blocks retention of training, whereas a slightly higher level affects acquisition (Griffith et al., 1993, 1994). Genetic and biochemical interactions between CaMKII and the potassium channel subunit eag have indicated the channel as a likely target for the kinase's action in this paradigm (Griffith et al., 1994). For both CREB and CaMKII, the inducible nature of transgene expression removed any doubts about possible developmental contributions to the learning and memory defects.
Mouse Knock.Out Mutants and Learning The molecular technology of homologous recombination has given mouse geneticists a powerful means of generating mutants affecting learning and memory. The chosen targets have been genes for kinases and receptors previously implicated in long-term potentiation. Mouse learning is assayed in an associative spatial task where the animal is set loose in a pool of murky water and asked to find its way to a submerged platform, a paradigm that bears a striking resemblance to graduate training. The pool has various landmarks around it as visible cues that help the mouse to become more proficient at finding the hidden platform after repeated trials. Mice mutant in the a subunit of CaMKII perform poorly in this test (Silva et al., 1992), as do mice mutant for the fyn receptor tyrosine kinase (Grant et al., 1992), the metabotropic glutamate receptor GluR1 (Aiba et al., 1994), and the ~1 subunit of the N-methyI-D-aspartate receptor (Sakimura et al., 1995). Milder defects accompany mutation of the 7 isoform of protein kinase C (Abeliovich et al., 1993) and of type I adenyl cyclase (Wu et al., 1995). Mutants in mouse CREB have defective tong-term memory in the water maze test and in a foot-shock fear conditioning paradigm (Bourtchuladze et al., 1994). These CREB studies establish a strong phylogenetic link between fly and mammalian mechanisms of learning and memory. Powerful as it is, the knock-out approach must live with the bias of testing known components. Can mouse genetics take affirmative action to identify new learning genes by forward genetics? Techniques of chemical mutagenesis in
Minireview 749
Olfactory Avoidance
100 ] I~ r
STM:short-term MTM:medium4erm ARM:anesthesia-resistant LTM:long-term
801 \Observed 6o STM
~
linotte
Initial Learning
~ rutabaga
C
E O 03 13-
dunce
200
I
1
I
i
I
3
I
5
hours
I
I
7
i,
,i
i
i
2
i
4
Short-term Memory
days
Data are from Tully et al., 1990, 1994.
~
rutabaga CaMKII (severe)
Initial Decrement
~ dunce CaMKII (mild)
Persistent Decrement
amnesiac
Retention time Figure 1. Time Course of Memory Phases
Courtship
Medium-term Memory radish~ ~kdCREB2
the mouse improved enormously in the late seventies and early eighties, yielding a new generation of induced mouse mutants. With sufficient mouse cages and patience for testing offspring, new learning mutants could be isolated in the same way a circadian rhythm mutant was recently found (Vitaterna et al., 1994), by taking advantage of the observation that such mutations often have semidominant effects. The conundrum remains, however, of how to distinguish developmental from adult gene action. Organizing Mutant Genes into Pathways Fly development began to make sense when a rational order could be imposed on the various genes. This came about by studying their ability to regulate each others' expression and to mitigate or exacerbate each others' mutant defects. In fly learning, the imposition of order has come from the correspondence between phases of memory and the action of particular genes (see Figures 1 and 2 and Tully et al., 1990, 1994). Initial learning is defined as the extent of conditioning when tested immediately after training. Memory phases are then defined temporally and operationally (Figure 2). Short-, medium-, and long-term refer to duration as measured by retention of conditioning. A fifth phase of learning is called anesthesia-resistant, as defined by the residual conditioning that persists after the animals have been subjected to a cold shock. None of these phases is apparent from the smooth decay curve of memory in normal flies, but phases are revealed by subtracting the decay curves of various mutants or treatments from the curve for normal flies. All of the mutants bring about some kind of change in the shape of the memory decay curve except linotte, which starts with a lower learning score but has a normal rate of decay, and is thus classified as an acquisition mutant. A comparable, albeit less complete, pathway can be drawn for courtship conditioning (Figure 1). Acquisition is measured directly by the decrement in a male's courtship during exposure to a mated female. Retention is then determined by testing the lackluster persistence of his courtship with a receptive female. In this paradigm, rutabaga and dunce act at different steps, and the site of CaMKII
Anesthesiaresistant Memory
Long-term Memory
Figure 2. Pathways of Learning Genes Data are from Tully et al., 1990, 1994; Siegel and Hall, 1979; Galley et al., 1984; Griffith et al., 1993, 1994.
action depends on the severity of its defect (Siegel and Hall, 1979; Gailey et al., 1984; Griffith et al., 1993, 1994). The time course of a male's persistent decrement corresponds to short-term memory in the olfactory avoidance paradigm. The pathway analogy has enormous heuristic value, even if it is only an approximation of reality. It often represents the most important functional relationships between the various genes. The analogy falls short in its simplification of these interactions to the extent that it singles out only the strongest genes and shows only their formal relationships. Studies of Ras signal transduction in mammals and flies have shed light on this sort of analysis and have provided an instructive comparison: genetics defined a pathway and biochemistry defined a network (Pawson, 1995). The value of the genetic pathway analogy comes from the conceptual scaffolding it provides and the future directions it indicates. Mutants serve as the shock troops into unknown territory. Ultimate answers require both biochemistry and further definition of the cellular circuitry in which each gene exerts its effect (Greenspan and Tully, 1994). Is a Fly Just a Fly or Is It a Human with Wings? The force of fly developmental genetics has been as much a function of its universality as of its insights, The fly Bithorax and Antennapedia complexes led to the homologous and analogous vertebrate Hox genes specifying anteriorposterior identity. The sevenless and torso receptor tyrosine kinases and Ras and raf genes specifying cell fate provided a normal biological context (along with similar genes in Caenorhabditis elegans) for understanding human oncogenes. The Wnt and Sonic hedgehog families
Neuron 75O
of signaling g e n e s b e g a n as s e g m e n t polarity mutants in the fly e m b r y o and h a v e n o w g r a d u a t e d to be vertebrate brain and limb patterning genes. Most recently, the Notch and Delta genes, which d e t e r m i n e neuronal cell fate in the fly, have been shown to have X e n o p u s h o m o l o g s that perform r e m a r k a b l y similar roles. C o n s e r v a t i o n of biological function gives their s e q u e n c e h o m o l o g i e s even more significance. T h e CREB connection s u g g e s t s that the s a m e m a y s o o n be said for g e n e s affecting learning and memory. August 25, 1995 Selected Reading Abeliovich, A., Paylor, R., Chen, C., Kim, J. J., Wehner, J. M., and Tonegawa, S. (1993). Cell 75, 1263-1271. Aiba,. A., Kano, M., Chen, C., Stanton, M. E., Fox, G. D., Herrup, K., Zwingman, T. A., and Tonegawa, S. (1994). Cell 79, 377-388. Bate, M., and Martinez Arias, A., eds. (1993). The Development of Drosophila melanogaster (Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press). Bolwig, G. M., Del Vecchio, M., Harmon, G., and Tully, T. (1995). Neuron, this issue. Bourtchuladze, R., Frenguelli, B., Blendy, J., Cioffi, D., Schutz, G., and Silva, A. J. (1994). Cell 79, 59-68. Boynton, S., and Tully, T. (1992). Genetics 131,655-672. Dash, P. K., Hochner, B., and Kandel, E. R. (1990). Nature 345, 718721. Dauwalder, B., and Davis, R. L. (1995). J. Neurosci. 15, 3490-3499. Davis, R. L. (1993). Neuron 11, 1-14. Feany, M. B., and Quinn, W. G. (1995). Science 268, 869-873. Folkers, E., Drain, P., and Quinn, W. G. (1993). Proc. Natl. Acad. Sci. USA 90, 8123-8127. Gailey, D. A., Jackson, F. R., and Siegel, R. W. (1984). Genetics 106, 613-623. Grant, S. G. N., O'Dell, T. J., Karl, K. A., Stein, P. L., Soriano, P., and Kandel, E, R. (1992). Science 258, 1903-1910. Greenspan, R. J., and Tully, T. (1994). In Flexibility and Constraint in Behavioral Systems, R. J. Greenspan and C. P. Kyriacou, eds. (New York: John Wiley and Sons), pp. 65-80. Griffith, L. C., Verselis, L. M., Aitken, K. M., Kyriacou, C. P., Danho, W., and Greenspan, R. J. (1993). Neuron 10, 501-509. Griffith, L. C., Wang, J., Zhong, Y., Wu, C.-F., and Greenspan, R. J. (1994). Proc. Natl. Acad. Sci. USA 91, 10044-10048. Hirsch, J. (1986). In Perspectives in Behavior Genetics, J. L. Fuller and E. C. Simmel, eds. (Hillsdale, New Jersey: Lawrence Elbaum Assoc.), pp. 155-200. Lilly, M., Kreber, R., Ganetzky, B., and Carlson, J. R, (1994). Genetics 136, 1087-1096. Pawson, T. (1995), Nature 373, 573-580. Sakimura, K., Kutsuwada, T., Ito, I., Manabe, T., Takayama, C., Kushiya, E., Yagi, T., Aizawa, S., Inoue, Y., Sugiyama, H., and Mishina, M. (1995). Nature 373, 151-155. Siegel, R. W., and Hall, J. C. (1979). Proc. Natl. Acad. Sci. USA 76, 3430-3434. Silva, A. J., Paylor, K., Wehner, J. M., and Tonegawa, S. (1992). Science 257, 206-211. Tully, T., and Quinn, W. G. (1985). J. Comp. Physiol. 157, 263-277. Tully, T., Boynton, S., Brandes, C., Dura, J. M., Mihalek, R., Preat, T., and Villela, A. (1990). Cold Spring Harbor Symp. Quant. Biol, 55, 203-211. Tully, T., Preat, T., Boynton, S. C., and Del Vecchio, M. (1994). Cell 79, 35-47.
Vitaterna, M. H., King, D. P., Chang, A. M., Kornhauser, J, M., Lowrey, P. L., McDonald, J. D., Dove, W. F., Pinto, L. H., Turek, F. W., and Takahashi, J. S. (1994). Science 264, 719-725. Wu, Z. L., Thomas, S. A., Villacres, E. C., Xia, Z., Simmons, M. L., Chavkin, C., Palmiter, R. D., and Storm, D, R. (1995). Proc. Natl. Acad. Sci. USA 92, 220-224. Yin, J. C. P., Del Vecchio, M., Zhou, H., and Tully, T. (1995). Cell 81, 107-115. Yin, J. C. P., Wallach, J. S., Del Vecchio, M., Wilder, E. L., Zhou, H., Quinn, W. G., and Tully, T. (1994)~ Cell 79, 49-58.
Flies, Genes, Learning, and Memory
Ralph J. Greenspan Department of Biology Center for Neural Science New York University New York, New York 10003
Drosophila was not always the darling of developmental biology that it is today. Mutants such as Antennapedia and wingless were once considered to be intriguing oddities but were relegated to a backwater of embryology because they bore no relevance to the way "real" animals developed. The turnaround followed a convergence of gene cloning, sequencing, and germline transformation with the ability to isolate and analyze mutations affecting pattern. Suddenly, a mutant could be cloned, its identity proven, and its time, place, and mode of action charted and manipulated. The finding that vertebrates (even humans!) had the same genes doing many of the same things didn't hurt either. The rest is history. A similar history is now being played out in the study of learning and memory. From a rocky beginning in which the very existence of fly learning was contested, the field has staggered through reckless youth and troubled adolescence and is now emerging into the bloom of earlyadulthood. Not the early adulthood of Generation X--trapped in a going-nowhere "McDiscipline"--nor the early adulthood of the culture of science welfare-trapped in an endless cycle of handouts given on the unfulfilled promise of great things to come. Instead, fly learning has earned its place as a mature and serious discipline, through hard work and adherence to traditional values. The Paradigm and "Forward" Genetics What were the ingredients of success in fly development and what are their counterparts in fly learning? The first was the establishment of a simple, reliable, and informative assay for function and mutant isolation. The segmented pattern of the larval cuticle, from which myriad mutants with tantalizing defects were amassed, did it for development. An olfactory avoidance paradigm is playing this role for fly learning (Tully and Quinn, 1985), establishing it as bona fide Pavlovian conditioning and laying to rest criticisms of its lack of rigor that came from traditional behavioral quarters (Hirsch, 1986). In the paradigm, flies are presented with two different fragrances. During exposure to one of them, they are also given an electric shock. Under these conditions, flies will learn to avoid the odor that had been accompanied by shock and wilt remember it for a day or so. Their memory can be extended to a week if they are given ten training sessions with suitable intervals in between (Tully et al., 1994). (The intervals are crucial since ten sessions in rapid succession are no better than one.) For flies in the wild, one week is long enough to arrive at midlife crisis, so this qualifies as long-term memory. One may be tempted to ask, at this point, what relevance
Minireview
learning has to the life-style of a fruit fly in the wild? A fly's ability to be conditioned during courtship appears to affect its mating success. In a laboratory version of this phenomenon, a male fly learns not to court when in the presence of unreceptive females, that is, females who have recently mated (Siegel and Hall, 1979). The larval segmentation assay served as the basis for isolating a wide array of developmental mutants identifying gap, pair-rule, and segment polarity genes as well as anterior-posterior, dorso-ventral, and terminal patterning genes (Bate and Martinez Arias, 1993). The ability to obtain more than one mutation of each gene gave a clearer picture of the range of their effects. Even before their molecular identities were known, these mutant categories stood out as obvious and distinct alterations of pattern. With their cloning, they have been organized into pathways of signal transduction and transcription. For learning studies, the olfactory avoidance paradigm has yielded mutants that all show essentially the same overt phenotype: stupidity. This leaves molecular identity as the primary basis for organizing them into understandable categories. Some affect components of the cAMP second messenger system, already well established from studies in Aplysia: a phosphodiesterase gene (dunce), an adenyl cyclase gene (rutabaga) (Davis, 1993), and a putative neuropeptide gene (amnesiac; Feany and Quinn, 1995). Others are either uncloned and unknown, like latheo (Boynton and Tully, 1992) and radish (Folkers et al., 1993), or cloned and unfamiliar, like linotte (Bolwig et al., 1995 [this issue of Neuron]). The rate of return is significantly slower for learning mutants than for those affecting development, because the learning assay uses adults and takes more time. This has also hampered the ability to obtain multiple lesions of each gene and assess its range of effects. The courtship paradigm has not been used for isolating new mutants but has served to generalize the roles of dunce and rutabaga in learning (Siegel and Hall, 1979; Gailey et al., 1984). (The etymology of the mutants' names is as follows: dunce, after John Duns Scotus, 13th century opponent to the revival of classical learning; rutagaba, because it's as dumb as a vegetable; radish, see rutabaga; amnesiac, as the name suggests; latheo, a Greek word meaning forgetful, related to the name of one of the rivers of Hades in Greek mythology, Lethe, whose waters caused all who drank from them to forget all of their former life; linotte, from the French expression t~te de linotte having the same idiomatic meaning as bird-brain.) The virtue of such a "forward genetic" approach is that it makes the fewest assumptions about which genes will be involved and lets the luck of the draw pull out those capable of producing the right kind of mutant. Part of the power of the olfactory avoidance paradigm is its ability to weed out spurious mutants that are unable to move, smell, or react to shock in task-relevant controls. Without the ability to make such distinctions, linotte, with a low initial
Neuron 748
learning score but a normal rate of memory decay, could not have been recognized as affecting initial learning. On the other side, an olfactory mutation in a ubiquitous sodium channel, smellblind (Lilly et al., 1994), would otherwise have been mistaken for a learning mutant.
Gene Identification and Manipulation A second ingredient in the success of fly development was the facility of cloning and identifying genes and of manipulating their expression. These techniques are identical in studies of either development or learning. Cloning has been accelerated by inducing mutations with transposons, mobile DNA elements that disrupt the gene and then provide a tag for retrieving the disrupted sequences. A variant of the transposon approach incorporates a reporter lacZ gene in the mobile element so that new putative mutants can be prescreened for their pattern of 13-galactosidase expression. Inducible expression of cloned genes, following germline transformation, made it possible to manipulate and define critical periods for developmental genes. This has recently been achieved for learning genes as well. The dunce mutant has been partially rescued after induction of a cAMP phosphodiesterase cDNA under the control of a heat-shock promoter (Dauwalder and Davis, 1995). The lack of complete rescue of its learning defect may indicate a developmental requirement for the dunce gene, echoing a nagging complaint about mutations affecting learning and memory: are they actually developmental mutations? Full rescue of the linotte mutation by a heat-inducible cDNA proves that adult expression is all that is needed for its action (Bolwig et al., 1995).
Pinpointing Key Steps A further ingredient in the success of developmental genetics is the ability to identify critical regulatory steps by the reciprocal effects of loss of function and gain of function changes in the gene. Such effects can be produced either by mutations or by induced expression of activating or inhibitory transgenes. Reciprocal phenotypes for developmental genes have generally been defined in two ways: those producing alternative cell fates, such as the neurogenic gene Notch, whose loss of function produces a neural fate and whose gain of function produces an epidermal fate; or those producing alternative patterns, such as the maternal gene Toil, whose loss of function produces a dorsalized embryo and whose gain of function produces a ventralized embryo (Bate and Martinez Arias, 1993). What would be reciprocal phenotypes for learning? Smart versus dumb flies. All of the mutants are poor learners by definition because that is how they were isolated. The creation of a superstrain of flies that remembers more effectively than normal fell out of a study of long-term memory as influenced by the cAMP-dependent transcription factor, CREB. A long history of studies had shown a requirement for protein synthesis in long-term memory of many animals, including Drosophila (Tully et al., 1994). Long-term facilitation in Aplysia had been shown to depend on protein synthesis and transcription of genes containing cAMP response elements (CREs; Dash et al., 1990). In flies, when expression of a negatively acting CRE-binding protein (dCREB2-b) is transiently induced
prior to training, it selectively abolishes long-term memory but leaves short-term memory unscathed (Yin et al., 1994). When a positively acting form (dCREB2-a) is similarly expressed in flies, it enables them to form long-term memory far more efficiently than normal, after only a single training session as opposed to the usual ten (Yin et al., 1995). Opposing actions of CREB produce reciprocal phenotypes: forgetfulness on one hand and photographic memory on the other. This leaves little question as to its importance. Manipulation of CREB was effected by pure reverse genetics. No mutations in the genes have yet been isolated in screens for learning mutants, but its key role was predicted from earlier studies and the selectivity of its effect was shown with the same behavioral controls used in the mutant screens. An analogous approach has established the involvement of the calcium/calmodulin-dependent protein kinase II (CaMKll) in the courtship paradigm. Heat-shockinduced expression of a specific peptide inhibitor of the kinase produces a graded effect on conditioning in this paradigm: a mild reduction in enzyme activity blocks retention of training, whereas a slightly higher level affects acquisition (Griffith et al., 1993, 1994). Genetic and biochemical interactions between CaMKII and the potassium channel subunit eag have indicated the channel as a likely target for the kinase's action in this paradigm (Griffith et al., 1994). For both CREB and CaMKII, the inducible nature of transgene expression removed any doubts about possible developmental contributions to the learning and memory defects.
Mouse Knock.Out Mutants and Learning The molecular technology of homologous recombination has given mouse geneticists a powerful means of generating mutants affecting learning and memory. The chosen targets have been genes for kinases and receptors previously implicated in long-term potentiation. Mouse learning is assayed in an associative spatial task where the animal is set loose in a pool of murky water and asked to find its way to a submerged platform, a paradigm that bears a striking resemblance to graduate training. The pool has various landmarks around it as visible cues that help the mouse to become more proficient at finding the hidden platform after repeated trials. Mice mutant in the a subunit of CaMKII perform poorly in this test (Silva et al., 1992), as do mice mutant for the fyn receptor tyrosine kinase (Grant et al., 1992), the metabotropic glutamate receptor GluR1 (Aiba et al., 1994), and the ~1 subunit of the N-methyI-D-aspartate receptor (Sakimura et al., 1995). Milder defects accompany mutation of the 7 isoform of protein kinase C (Abeliovich et al., 1993) and of type I adenyl cyclase (Wu et al., 1995). Mutants in mouse CREB have defective tong-term memory in the water maze test and in a foot-shock fear conditioning paradigm (Bourtchuladze et al., 1994). These CREB studies establish a strong phylogenetic link between fly and mammalian mechanisms of learning and memory. Powerful as it is, the knock-out approach must live with the bias of testing known components. Can mouse genetics take affirmative action to identify new learning genes by forward genetics? Techniques of chemical mutagenesis in
Minireview 749
Olfactory Avoidance
100 ] I~ r
STM:short-term MTM:medium4erm ARM:anesthesia-resistant LTM:long-term
801 \Observed 6o STM
~
linotte
Initial Learning
~ rutabaga
C
E O 03 13-
dunce
200
I
1
I
i
I
3
I
5
hours
I
I
7
i,
,i
i
i
2
i
4
Short-term Memory
days
Data are from Tully et al., 1990, 1994.
~
rutabaga CaMKII (severe)
Initial Decrement
~ dunce CaMKII (mild)
Persistent Decrement
amnesiac
Retention time Figure 1. Time Course of Memory Phases
Courtship
Medium-term Memory radish~ ~kdCREB2
the mouse improved enormously in the late seventies and early eighties, yielding a new generation of induced mouse mutants. With sufficient mouse cages and patience for testing offspring, new learning mutants could be isolated in the same way a circadian rhythm mutant was recently found (Vitaterna et al., 1994), by taking advantage of the observation that such mutations often have semidominant effects. The conundrum remains, however, of how to distinguish developmental from adult gene action. Organizing Mutant Genes into Pathways Fly development began to make sense when a rational order could be imposed on the various genes. This came about by studying their ability to regulate each others' expression and to mitigate or exacerbate each others' mutant defects. In fly learning, the imposition of order has come from the correspondence between phases of memory and the action of particular genes (see Figures 1 and 2 and Tully et al., 1990, 1994). Initial learning is defined as the extent of conditioning when tested immediately after training. Memory phases are then defined temporally and operationally (Figure 2). Short-, medium-, and long-term refer to duration as measured by retention of conditioning. A fifth phase of learning is called anesthesia-resistant, as defined by the residual conditioning that persists after the animals have been subjected to a cold shock. None of these phases is apparent from the smooth decay curve of memory in normal flies, but phases are revealed by subtracting the decay curves of various mutants or treatments from the curve for normal flies. All of the mutants bring about some kind of change in the shape of the memory decay curve except linotte, which starts with a lower learning score but has a normal rate of decay, and is thus classified as an acquisition mutant. A comparable, albeit less complete, pathway can be drawn for courtship conditioning (Figure 1). Acquisition is measured directly by the decrement in a male's courtship during exposure to a mated female. Retention is then determined by testing the lackluster persistence of his courtship with a receptive female. In this paradigm, rutabaga and dunce act at different steps, and the site of CaMKII
Anesthesiaresistant Memory
Long-term Memory
Figure 2. Pathways of Learning Genes Data are from Tully et al., 1990, 1994; Siegel and Hall, 1979; Galley et al., 1984; Griffith et al., 1993, 1994.
action depends on the severity of its defect (Siegel and Hall, 1979; Gailey et al., 1984; Griffith et al., 1993, 1994). The time course of a male's persistent decrement corresponds to short-term memory in the olfactory avoidance paradigm. The pathway analogy has enormous heuristic value, even if it is only an approximation of reality. It often represents the most important functional relationships between the various genes. The analogy falls short in its simplification of these interactions to the extent that it singles out only the strongest genes and shows only their formal relationships. Studies of Ras signal transduction in mammals and flies have shed light on this sort of analysis and have provided an instructive comparison: genetics defined a pathway and biochemistry defined a network (Pawson, 1995). The value of the genetic pathway analogy comes from the conceptual scaffolding it provides and the future directions it indicates. Mutants serve as the shock troops into unknown territory. Ultimate answers require both biochemistry and further definition of the cellular circuitry in which each gene exerts its effect (Greenspan and Tully, 1994). Is a Fly Just a Fly or Is It a Human with Wings? The force of fly developmental genetics has been as much a function of its universality as of its insights, The fly Bithorax and Antennapedia complexes led to the homologous and analogous vertebrate Hox genes specifying anteriorposterior identity. The sevenless and torso receptor tyrosine kinases and Ras and raf genes specifying cell fate provided a normal biological context (along with similar genes in Caenorhabditis elegans) for understanding human oncogenes. The Wnt and Sonic hedgehog families
Neuron 75O
of signaling g e n e s b e g a n as s e g m e n t polarity mutants in the fly e m b r y o and h a v e n o w g r a d u a t e d to be vertebrate brain and limb patterning genes. Most recently, the Notch and Delta genes, which d e t e r m i n e neuronal cell fate in the fly, have been shown to have X e n o p u s h o m o l o g s that perform r e m a r k a b l y similar roles. C o n s e r v a t i o n of biological function gives their s e q u e n c e h o m o l o g i e s even more significance. T h e CREB connection s u g g e s t s that the s a m e m a y s o o n be said for g e n e s affecting learning and memory. August 25, 1995 Selected Reading Abeliovich, A., Paylor, R., Chen, C., Kim, J. J., Wehner, J. M., and Tonegawa, S. (1993). Cell 75, 1263-1271. Aiba,. A., Kano, M., Chen, C., Stanton, M. E., Fox, G. D., Herrup, K., Zwingman, T. A., and Tonegawa, S. (1994). Cell 79, 377-388. Bate, M., and Martinez Arias, A., eds. (1993). The Development of Drosophila melanogaster (Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press). Bolwig, G. M., Del Vecchio, M., Harmon, G., and Tully, T. (1995). Neuron, this issue. Bourtchuladze, R., Frenguelli, B., Blendy, J., Cioffi, D., Schutz, G., and Silva, A. J. (1994). Cell 79, 59-68. Boynton, S., and Tully, T. (1992). Genetics 131,655-672. Dash, P. K., Hochner, B., and Kandel, E. R. (1990). Nature 345, 718721. Dauwalder, B., and Davis, R. L. (1995). J. Neurosci. 15, 3490-3499. Davis, R. L. (1993). Neuron 11, 1-14. Feany, M. B., and Quinn, W. G. (1995). Science 268, 869-873. Folkers, E., Drain, P., and Quinn, W. G. (1993). Proc. Natl. Acad. Sci. USA 90, 8123-8127. Gailey, D. A., Jackson, F. R., and Siegel, R. W. (1984). Genetics 106, 613-623. Grant, S. G. N., O'Dell, T. J., Karl, K. A., Stein, P. L., Soriano, P., and Kandel, E, R. (1992). Science 258, 1903-1910. Greenspan, R. J., and Tully, T. (1994). In Flexibility and Constraint in Behavioral Systems, R. J. Greenspan and C. P. Kyriacou, eds. (New York: John Wiley and Sons), pp. 65-80. Griffith, L. C., Verselis, L. M., Aitken, K. M., Kyriacou, C. P., Danho, W., and Greenspan, R. J. (1993). Neuron 10, 501-509. Griffith, L. C., Wang, J., Zhong, Y., Wu, C.-F., and Greenspan, R. J. (1994). Proc. Natl. Acad. Sci. USA 91, 10044-10048. Hirsch, J. (1986). In Perspectives in Behavior Genetics, J. L. Fuller and E. C. Simmel, eds. (Hillsdale, New Jersey: Lawrence Elbaum Assoc.), pp. 155-200. Lilly, M., Kreber, R., Ganetzky, B., and Carlson, J. R, (1994). Genetics 136, 1087-1096. Pawson, T. (1995), Nature 373, 573-580. Sakimura, K., Kutsuwada, T., Ito, I., Manabe, T., Takayama, C., Kushiya, E., Yagi, T., Aizawa, S., Inoue, Y., Sugiyama, H., and Mishina, M. (1995). Nature 373, 151-155. Siegel, R. W., and Hall, J. C. (1979). Proc. Natl. Acad. Sci. USA 76, 3430-3434. Silva, A. J., Paylor, K., Wehner, J. M., and Tonegawa, S. (1992). Science 257, 206-211. Tully, T., and Quinn, W. G. (1985). J. Comp. Physiol. 157, 263-277. Tully, T., Boynton, S., Brandes, C., Dura, J. M., Mihalek, R., Preat, T., and Villela, A. (1990). Cold Spring Harbor Symp. Quant. Biol, 55, 203-211. Tully, T., Preat, T., Boynton, S. C., and Del Vecchio, M. (1994). Cell 79, 35-47.
Vitaterna, M. H., King, D. P., Chang, A. M., Kornhauser, J, M., Lowrey, P. L., McDonald, J. D., Dove, W. F., Pinto, L. H., Turek, F. W., and Takahashi, J. S. (1994). Science 264, 719-725. Wu, Z. L., Thomas, S. A., Villacres, E. C., Xia, Z., Simmons, M. L., Chavkin, C., Palmiter, R. D., and Storm, D, R. (1995). Proc. Natl. Acad. Sci. USA 92, 220-224. Yin, J. C. P., Del Vecchio, M., Zhou, H., and Tully, T. (1995). Cell 81, 107-115. Yin, J. C. P., Wallach, J. S., Del Vecchio, M., Wilder, E. L., Zhou, H., Quinn, W. G., and Tully, T. (1994)~ Cell 79, 49-58.