Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
TINS November i 985 478 - Neural and molecular mech isms underlying in rrnation storage in Aplysia: imp.cations for learning and memory John H. Byrne The marine mollusc Aplysia /s one of several experimental preparations that neuroscientists are using to help understand the mechanisms underlying learning and memory. Simple behavioral modifications such as sensitization and classical conditioning can be related to changes in the ability o f previously formed synaptic connections to release neurotransmitter substance. The modifications of transmitter release in turn are regulated by the intracellular second messengers Ca2+ and cAMP. The second messenger systems may act separately in the case o f non-associative learning (sensitization) but their specific interactions may underly forms o f associative learning such as classical conditioning. A fundamental problem in neurobiology is to understand the basic mechanisms of learning and memory. What parts of the nervous system are involved in learning and memory, how is information about a learned event acquired and encoded at the cellular level, how is the information stored, and once stored how is it retrieved or read-out? These are questions for which few answers are available at a mechanistic level. While most neuroscientists believe that changes in individual nerve cells (or at least small groups of cells) are critical for learning and memory, there is considerably less agreement on the underlying mechanisms for such changes. Mechanisms such as growth of new synaptic contacts, alterations in synaptic efficacy at previously existing synapses, and reverberatory activity in neural circnits~ have been proposed, but only recently have they been the subject of direct experimental analyses. Considerable attention is now being focused on the hypothesis that a change in synaptic transmission at previously formed synaptic connections may underlie examples of learning such as habituation, sensitization and classical conditioning. A number of experimental preparations are being examined to test this hypothesis and to study the neural mechanisms of learnings . One that seems particularly useful for a molecular analysis is the marine mollusc Aplysia, the subject of this review. This animal has a relatively simple nervous system with large, identifiable neurons that are accessible for detailed biochemical and biophysical studies. © 1985, Elsevier Science Publishers B.V., Amsterdam The connections between sensory neurons and motor neurons controlling the defensive siphon withdrawal reflex and the defensive tail withdrawal reflex, have been the focus of recent analyses6,7. These connections exhibit a number of plastic properties including synaptic depression, posttetanic potentiation, presynaptic facilitation, activity-dependent neuromodulation and long-term potentiation6-1 a. The synaptic modifications in turn have been related to a variety of behavioral modifications including habituation, sensitization and classical conditioning7,9,N . Neural and molecular mechanisms contrilmting to ram-associative information storage Neural and molecular mechanisms contributing to sensitization have been analysed extensively in the siphon and gill withdrawal reflex of Aplysia, and similar mechanisms appear to contribute to sensitization of the tail withdrawal reflex as well. Sensitization is a simple form of non-associative learning in which a non-specific enhancement of the response to a test stimulus is produced by a second averswe stimulus applied to another part of the animal. For example, a weak mechanical stimulus delivered to the siphon of Aplysia produces a small withdrawal of the siphon and nearby gill. If, however, a noxious stimulus is applied to the animal's head or tail. subsequent test stimuli delivered to the siphon produce much greater responses 7. Both short-term sensitization, lasting tens of minutes, and long-term sensitization, lasting days to weeks, have been analysed in Aplysia but the mechanisms for the short-term effects are best understood and will be the focus of this review. Sensitizing stimuli to one part of the animal lead to the release of a modulatory transmitter that appears to produce presynaptic facilitation of many sensory neurons including those that do not have their receptive fields in the region of the sensitizing stimulus. The effects of presynaptic facili- Sensitizing stimulus to animal Activation of modulatory intemeurons Modulatory,4,..--~-b ' ~ ' ~ " ~ - ' - ' ~ - - '~- ' ~ ~ transmitter adenylate cyclase release by ~'~activity in intemeurons ~ Increased levels of cyclic AMP Increased ~ . . . . ~ - - " ~ activity of protein kinase Increas.ed ~ protein phosphorylation S. p i k ~ . . . . . . . . . . _ . ~ ~ a s e d m ~ Ca2+ influx sensory neurons ~ !nto broadened terminal Increased.,4~ EPSP in motor neurons Increased .~_~activation of motor neurons ~ Decreased ,~_ K + ....-- conductance Increased transmitter ~ release from ....--- sensory neurons Enhanced . ~ behaviora! response Fig. 1. Sequenceof events contributing to sensitizationin Aplysia. Modifiedfrom Ref. 23. 0378 5912/85/$02.00 TINS- 479 N o v e m b e r 1985 tation are expressed when subsequent test stimuli are presented. Presynaptic facilitation leads to an enhanced release of neurotransmitter from sensory neurons and thus an enhanced probability of activating the motor neurons and an enhanced behavioral response (Fig. 1). The effects of the natural neuromodulator released by sensitizing stimuli can be mimicked by application of serotonin or SCP13 ('small cardio-acceleratory peptide B') 11-15. These chemicals, like the natural modulator, exert their effects through changes in the level of the intracellular second messenger cAMP 14-21. One consequence of the elevated cAMP levels is reduction of a resting steady-state K + current by the closure of an 'S-channel' (so named because it is serotonin-sensitive) that contributes significantly to the repolarization of the action potential ~5.22. As a result of closure of the S-channel, repolarizing K + currents are decreased and action potentials initiated in sensory neurons by test stimuli are broader. Consequently, the Ca 2+ influx that normally occurs during the action potential is prolonged. This allows for greater transmitter release and the enhanced activiation of the motor neurons, that contributes to sensitization. Since short-term sensitization and changes in synaptic efficacy that contribute to sensitization are relatively persistent (15-30min), an obvious question concerns which of the biochemical steps illustrated in Fig. 1 is critical for storing the short-term memory. To address this issue Castellucci et al. 23 injected an inhibitor of cAMP dependent protein kinase into sensory neurons just after the application of serotonin. As a result of the injections the extent of the spike broadening was decreased dramatically23. Thus, the duration of the memory for the short-term sensitization seems to depend primarily on the persistence of an active protein kinase and not the persistence of a phosphorylated protein. What then determines the time-course of cAMP levels that keep the protein kinase active? Two possibilities must be considered. One is the time-course of cAMP degradation (controlled by phosphodiesterase) and the second is the time-course of cAMP synthesis. To examine the possible role of increased synthesis, Castellucci et al. 24 utilized GDPI~S, an irreversible inhibitor of the cyclase. Injection of GDP[3S just after the spike broadening induced by serotonin caused a rapid return of the spike to its pre-serotonin duration 24,25. These results indicate that the elevation of cAMP levels is dependent upon continued activation of the cyclase and that a molecular locus for the shortterm memory is at the level of the adenylate cyclase complex. Neural model of associative learning While the analysis of short-term sensitization is fairly extensive 7, little was known until recently about the types of storage mechanisms that may underly memories produced by associative learning. One hypothesis is that the mechanism contributing to the associative learning in this animal is simply an elaboration of mechanisms already in place that contribute to sensitization. Before discussing this possibility it is useful to review briefly the features of associative learning and compare them with sensitization. One example of associative learning that has been examined extensively is classical or Pavlovian conditioning. In classical conditioning, a test stimulus (known as the conditioned stimulus, CS) becomes effective in eliciting a response (the conditioned response, CR) when the CS is paired with another stimulus (the reinforcing or unconditioned stimulus, US) that reliably produces a response (the unconditioned response, UR). For example, in salivary conditioning the tone (CS) becomes effective in eliciting salivation (CR) if the tone is paired repeatedly with another stimulus (US, e.g. meat powder) that reliably produces a response (UR). Classical conditioning and sensitization share some common features. In both, the effectiveness of a test stimulus in eliciting a response is modified by a reinforcing stimulus. They differ, however, in that classical conditioning requires a close temporal association between the CS and the reinforcer (US). If the pairing of the tone and meat powder are separated in time, the effectiveness of the tone in eliciting salivation is reduced. Response enhancement that occurs independent of the timing of the CS and US is sensitization. At the mechanistic level how might sensitization and classical conditioning be related? One possibility is that electrical activity in the CS pathway is capable of amplifying the effects of sensitizing or reinforcing stimuli on the CS pathway. Such a notion is consistent with the theoretical work of some neurobiologists and psychologists who have assumed that the formation of associations depends upon the contiguous activation of sensory pathways and modulatory arousal centers~-~L A general model of how sensitization and classical conditioning might be related at the cellular level is A. LEARNING B. MEMORY PAIRED R STIMULUS .-D,,.( E S P 0 G UNPAIRED STIMULUS "~" I -<• O I ~ s E SYNAPSE(OUTPUT) OMA AND DENDRITES(INPUT) PAIREDACTIVITY Fig. 2. General model of activity-dependent neuromodulation. (A) Learning. Stippling indicates temporally contiguous activity. A motivationally potent reinforcing stimulus activates a neural response system and a modulatory system that regulates the efficacy of diffuse afferents to the response system. Increased spike activity in the paired afferent (1) immediately before the modulatory signal amplifies the degree and duration o f the modulatory effects, perhaps through the Ca2+ sensitivity of the modulatory evoked second messenger. The unpaired afferent neuron (2) does not show an amplification of the modulatory effects. (B) Memory. The amplified modulatory effects cause longterm increases in transmitter release and~or excitability of the paired neuron, which in turn strengthens the functional connection between the paired neuron(l) and the response system. Modified from Ref. 8. 77NS November 1985 480 illustrated in Fig. 2. Assume that two sensory pathways, designated as sensory neuron 1 and sensory neuron 2, make weak subthreshold connections to a response system. The two sensory neurons might represent two cells of the same modality with different receptive fields or sensory neurons representing two different modalities. (In general the two input cells could be interneurons as well.) Reinforcing stimuli have two effects. One is to activate directly the response system and produce the unconditioned response (UR). The second is to activate a diffuse modulatory or facilitatory system that non-specifically enhances the connections of all the sensory neurons. Reinforcing stimuli may enhance synaptic transmission by initiating an increase in intracellular cAMP levels, which in tum affect neurotransmitter release through the sequence of events illustrated in Fig. l. The properties of the system at this level are rather non-specific since the modulator influences all of the target sensory neurons and this modulation occurs whether or not there is a CS. How is the temporal specificity characteristic of associative learning obtained from such a seemingly non-specific system? At the cellular level, temporal specificity could be achieved if input from the CS can amplify the biochemical sequence of events initiated by the reinforcing stimulus (or US). One possibility is that spike activity in a sensory neuron which occurs just prior to the modulatory US causes a selective amplification of cAMP levels over and above the increase caused by the modulatory input alone. Thus, the extent of the modulatory effects in a target neuron would be dependent upon whether that neuron had recently been active. This mechanism, termed activity-dependent neuromodulation 8,9, allows both spatial and temporal specificity to be obtained from a diffuse modulatory system. The amplification of the modulatory effects in the paired sensory neuron would lead to enhancement of the ability of that sensory neuron to activate the response system and produce the conditioned response (Fig. 2B). If the activity-dependent neuromodulation hypothesis for classical conditioning is correct, then a classical conditioning paradigm applied to individual sensory neurons in Aplysia should specifically modify the monosynaptic connections between these cells and the motor neurons. This was A. H ETEROSYNAPTIC FAClLITATION ~ ~ PK ~ MODULATORY TRANSMITTER © cAMP k ATP-.~ t Ca 2+ B. ACTIVITY-DEPENDENT NEUROMODULATION Ca 2+ \ PK ~ cAMP, AT[ K+ Ca 2+ Fig. 3. Model of possible molecular events contributing to heterosynaptic facilitation and activitydependent neuromodulation. (A) Heterosynaptic facilitation (which may account for sensitization). Modulatory transmitter binding to receptor (R) activates adenylate cyclase (C) via a regulatory subunit ((7). The resulting cAMP activates one or more protein kinases (PK) whose action(s) includes closure of steady-state K + channels. Closing K + channels results (indirectly) in increased Caz+ influx and increased transmitter release. (B) Activity-dependent neuromodulation rwhich may be mechanism for associative learning). An initial influx of Cae+ due to spike activity affects one or more components of the adenylate cyclase complex. Subsequent transmitter activation of the cyclase by the modulatory transmitter results in an ampliJ~cation of cAMP production, ultimately resulting in enhanced transmitter release relative to that produced by the modulatory transmitter alone. Additional and longer term effects of amplified cAMP production (mediated by increased activiation of protein kinases) are also possible. Modified from Ref. 18. shown in two studies that used a classical conditioning procedure with intracellular activation of individual tail sensory cells and siphon sensory cells as a neural analog of the conditioned stimulus (CS) and shock to the skin as the unconditioned stimulus (US) s,9. While the sensory neurons are normally activated by mechanical stimulation of the skin, artificial intracellular activation was used to control which cells were activated as well as the number of action potentials and the exact timing of the spike activity relative to the modulatory input. The US did not activate the sensory neurons being examined, but did activate diffuse neuromodulatory input, producing heterosynaptic facilitation of all the connections of the sensory neuronsS~ 9. Paired presentation of the ~ and US resulted in a significant enhancement of the excitatory postsynaptic potential 481 T I N S - November 1985 compared to the facilitation produced when the CS and US were separated in time or when the US was presented alone s,9. The unpaired procedure where the CS and US were separated in time was particularly important because it illustrated that spike activity and reinforcing stimuli must be temporally close in order for the associative change in synaptic efficacy to an identical procedure applied to the remnants of the pleural ganglia failed to produce any pairing-specific enhancement of c A M P levels 17,18. Similar conclusions using slightly different procedures were reached by Abrams et al. 31. The measurements of c A M P levels were obtained with isolated sensory neuron clusters and therefore the occur. effects that were monitored presumably occurred in the somata and not Molecular mechanism for associative the synaptic terminals. This effect is information storage interesting by itself because of the Since enhanced cyclase activity is possibility that cyclic A M P may inbelieved to be the critical step for the fluence the synthetic capabilities of the enhancement of synaptic efficacy that cell soma (see below). If similar effects underlies short-term sensitization, a were also taking place at the terminal, further prediction of the activity- a more immediate read-out of the dependent neuromodulation model ~ memory might be an enhancement of for classical conditioning is that cAMP transmitter release. A molecular levels in a sensory neuron receiving a model is illustrated in Fig. 3. This paired presentation of spike activity model incorporates the notion that a and the modulatory effect of the US mechanism for the acquisition and should be enhanced relative to a cell short-term retention of the associative receiving unpaired stimuli. information is simply an extension of As a first step in testing this mechanisms already in place for nonhypothesis, sensory neurons involved associative information storage. Rein the tail-withdrawal reflex were inforcing stimuli cause the release of surgically isolated and exposed to a serotonin, or a related neuromodulamodification of the conditioning parator, that activates the adenylate cydigm used in the electrophysiological clase complex in the sensory neurons studies. Rather than attempting to (Figs 1 and 3A). At least one effect of measure c A M P levels in a single the c A M P produced is to close resting neuron, however, Ocorr et al. 17,18 K + channels, presumably by activating applied a modified conditioning pro- one or more protein kinases that in cedure to a homogeneous cluster of turn phosphorylate proteins associated about 100 sensory neurons. A brief with K + channels 22. According to the exposure to high K + artificial sea model for the neuronal modifications water was used to depolarize the entire that may underly associative learning, cluster of sensory neurons and thus the spike activity (see below) assomimic the electrophysiological spike ciated with the CS enhances the activity used previously as the CS. The synthesis of cAMP in response to a US was a brief exposure to serotonin neuromodulator released by the US. that was used to mimic the modulatory Therefore, even greater levels of input normally produced by reinforc- c A M P would be produced, yielding ing stimuli. Pairing of the stimuli was greater closure of K + channels and accomplished by exposing isolated greater transmitter release compared clusters of sensory neurons to the CS to the enhancement of transmitter immediately followed by exposure to release produced by the modulator the US. Because there are two identi- alone. Thus, increased cAMP levels cal clusters of sensory neurons in each seem to provide a biochemical mecanimal, contralateral clusters of sen- hanism for encoding information sory neurons were used as controls and about the temporal association of received a specifically unpaired pre- separate inputs to these cells 17,18,31 sentation of the CS and US. Ocorr et and therefore may be a primary biochemical step in the memory not al. found that the cAMP levels of sensory neuron clusters receiving the only for sensitization but also for paired presentation of the CS and US classical conditioning. were significantly elevated relative to A particularly intriguing question is the levels seen in the contralateral what aspect of spike activity is capable control clusters receiving the unpaired of enhancing c A M P synthesis? There paradigm is. The effect was produced are a number of possibilities but the one that seems most attractive is Ca 2+ by a single pairing and appeared to be restricted to the sensory neurons since influx. Ca 2+ influx certainly occurs during the spike 32,33 and there are numerous examples of interactions between the Ca/+ and c A M P second messenger systems 3a,35. One possible molecular locus for the induction of the associative change would be the proximate and sequential interaction of Ca 2+ and serotonin (or related neuromodulators) with the adenylate cyclase complex. Evidence supporting such a Ca 2+ interaction with cAMP is provided by studies of vertebrate brain tissue 36 and preliminary evidence indicates the presence of a Ca2+-sensitive cyclase in Aplysia as well 25. More direct support of a critical role for cyclase activity in learning has been provided by studies of Drosophila where it has been shown that a mutant deficient in associative learning also exhibits a loss of Ca2+/calmodulin sensitivity of the particulate adenylate cyclase. Interestingly, the mutants in Drosophila that affect conditioning also affect sensitization 37,38. This provides further support for the hypothesis that these two examples of learning may share common mechanisms. Recently, long-term changes in synaptic transmission and in the excitability of Aplysia sensory neurons have been found in response to simultaneous presentations of spike activity in sensory neurons and modulatory stimuli 9A°. In view of these persistent effects it is interesting to speculate that the increased c A M P content observed in sensory neuron somata is not simply a reflection of events occurring at the synaptic terminals. The fluctuations in the somatic cAMP content might play a role in triggering the long-term changes by acting on cellular elements localized to the soma. Such long-term changes could involve alterations in cell metabolism or in genomic regulation and protein synthesis brought about via cAMP-mediated phosphorylations 34. Altered somatic cAMP levels may not only provide a biochemical link between short- and long-term changes in synaptic efficacy but also between short- and long-term memory. The generality of activity-dependent neuromodulation as a basic mechanism for associative learning and cellular plasticity remains to be established. Indeed, other mechanisms are likely39'4°. It is interesting, however, that activity-dependent neuromodulation has been observed recently at the crayfish neuromuscular junction al and in hippocampal pyramidal cells - systems shown to be involved in long- 482 t e r m c h a n g e s in synaptic efficacy. It m a y also c o n t r i b u t e to the classical conditioning in Drosophila m e n t i o n e d above. In addition, there is g r o w i n g evidence f r o m a n u m b e r of v e r t e b r a t e and i n v e r t e b r a t e animals consistent with synergistic Ca 2+ a n d cyclic nucleotide interactions 34,35. Since Ca ~* a n d cyclic nucleotide control s y s t e m s are so u b i q u i t o u s , it is attractive to think that their specific interactions are involved in g e n e r a l m e c h a n i s m s of synaptic plasticity and learning. Acknowledgements I thank V. CasteUucci, L. Cleary, T. Crow, K. Ocorr and E. T. Waiters for reviewing an earlier draft of this manuscript. Supported by National Institutes of Health grant NS 19895 and Air Force Office of Scientific Research grant AFOSR 84-213. Selected references ! Eccles, J. C. (1964) The Physiology of Synapses, Springer, Berlin 2 Hebb, D. O. (1949) The Organization of Behavior, Wiley, New York 3 Lorente De N6, R. (1939) J. Neurophysiol. 2, 402--464 4 Ramon Y Cajal, S. (1909) Histologie du systeme nerveux de rhomme et des vertebres Vol. 2, Maloine, Paris 5 Hawkins, R. D. (1983) in The Physiological Basis of Memory (Deutseh, J. A., ed.), pp. 71-120, Academic Press, New York 6 Byrne, J. H., Ocorr, K. A., Walsh, J. P. and Waiters, E.T. in Neural Mechanisms of Conditioning (Alkon, D.L. and Woody, C. D., eds), Plenum, New York (in press) 7 Kandel, E. R. and Schwartz, J. H. (1982) Science 218, 433-443 8 Waiters, E. T. and Byrne, J.H. (1983) Science 219, 405--408 9 Hawkins, R. D., Abrams, T. W., Carew, [ I N S -. N o v e m b e r 1985 T. J. and Kandel, E. R. (1983) Science 219, 27 Kety, S. S. (1970) in Fhe Neurosciences. 400-405 Second Study Program (Schmitt, F. O.. ed.). 10 Waiters, E. T. and Byme, J. H. (1985) J. pp. 324-336. Rockefeller University Press. Neurosci. 5,662-672 New York 11 Waiters, E. T., Byrne, J. H., Carew, T. J. 28 Konorski, J. (1%7) Integrative Activity of the and Kandel, E. R. (1983) L Neurophysiol. Brain, Univ. of Chicago Press, Chicago 50, 1543-1559 29 Mackintosh, N. J. (1974) The Psychology o] 12 Brunelli, M., Castellucci, V. and Kandel, Animal Learning, Academic Press, New E. R. (1976) Science 194, 1178-1181 York 13 Abrams, T. W., Castellucci, V. F., Camar30 Kandel, E. R. and Tauc, L. (1%5)J. Physiol do, J. S., Kandel, E. R. and Lloyd, P. E. (London) 181, 28~7 (1984) Proc. Natl Acad. Sci. USA 81, 795631 Abrams, T. W., Bernier, L., Hawkins, 796O R.D. and Kandel, E. R (1984) Soc. 14 Ocorr, K. A. and Byrne, J. H. (1985) Neurosci. Abstr. 10, 269 Neurosci. Lett. 55,113-118 32 Boyle, M. B., Klein, M., Smith, S. J. and 15 Klein, M. and Kandel, E. R. (1980) Proc. Kandel, E. R. (1984) Proc. Natl Acad. Sci. Natl Acad. Sci. USA 77, 6912-6916 USA 81, 7642-7646 16 Bernier, L., Castellucci, V.F., Kandel, 33 Waiters, E. T. and Byrne, J. H. (1983) Brain E. R. and Schwartz, J. H. (1982) J. NeuroRes. 280, 165-168 sci. 2, 1682-1691 34 Nestler, E. J. and Greengard, P. (1984) 17 Ocorr, K. A., Waiters, E.T. and Byrne, Protein Phosphorylation in the Nervov.~ J. H. (1983) Soc. Neurosci. Abstr. 9, 169 System, John Wiley & Sons, New York 18 Ocorr, K. A., Waiters, E. T. and Byme, 35 Rasmussen, H. (1981) Calcium and cAMP as J. H. (1985) Proc. Natl Acad. Sci. USA 82, Synarchic Messengers, John Wiley & Sons, 2548-2552 New York 36 Malnoe, A., Stein, E. A. and Cox, J. A. 19 Pollock, J. D., Camardo, J. S., Bernier, L., (1982) Neurochem. Int. 5, 65-72 Schwartz, J. H. and Kandel, E. R. (1982) 37 Dudai, Y. (1985) TrendsNeuroSci. 79, 18-21 Soc. Neurosci. Abstr. 8, 523 20 Walsh, J. P. and Byrne, J. H. (1984) 38 Aceves-Pina, E. O., Booker, R., Duerr, J.S., Livingston, M.S., Quinn, W.G., Neurosci. Lett. 52, 7-11 Smith, R. F., Sziber, P. P., Tempel, B. L. 21 Belardetti, F., Biondi, C., Brunelli, M., and Tully, T. P. (1983) Cold Spring Harbor Fabri, M. and Trevisani, A. (1984) Brain Symp. Quant. Biol. 48, 831-840 Res. 288, 95-104 39 Deutsch, J. A. (ed.) (1983) The Physio22 Siegelbaum, S. A., Camaxdo, J. S. and logical Basis of Memory, Academic Press, Kandel, E. R. (1982) Nature (London) 299, 413-417 New York 23 Castellucci, V. F., Nairn, A., Greengard, P., --40 Alkon, D. L. and Woody, C. D. (eds) Neural Mechanisms of Conditioning, Plenum Schwartz, J. H. and Kandel, E. R. (1982) J. Neurosci. 2, 1673-1681 Press, New York (in press) 24 Castellucci, V. F., Bernier, L., Schwartz, 41 Breen, C. A. and Atwood, H. L. (1983) Nature (London) 303, 716-7t8 J. H. and Kandel, E. R. (1983) Soc. Neuro42 Hopkins, W. F. and Johnston, D. (1984) sci. Abstr. 9, 169 Science 226-352 25 Schwartz, J. H., Bernier, L., Castellucci, V. F., Polazzolo, M., Saitoh, T., Stapleton, A. and Kandel, E. R. (1983) Cold Spring John H. Byrne is at the Department of: Harbor Syrup. Quant. BioL 48, 811--819 Physiology and Cell Biology, University of Texas 26 Crow, T. J. (1968) Nature (London) 219, Medical School, P.O. Box 20708, Houston, TX 736-737 77225, USA.