Long-term potentiation and related forms of synaptic plasticity, sensitization and postsynaptic depression are often used for investigation of learning and memory. The advantage of LTP over conventional forms of learning (habit-uation, classical and instrumental conditioning) is the possibility of inducing long-term and strong alterations in the neural system by means of a short treatment. Persistent synaptic enhancement is easily induced in several structures within the brain by a brief afferent tetanus or by an electric or chemical sensitizing stimulus. When the journal "Behavioral and Neural Biology" altered its title to "Neurobiology of Learning and Memory", it presented at once two reviews devoted to LTP [581, 793] in the first issue. This phenomena actually leaves traces in the neural system and have left a trace in the neurosciences. However, it is not clear whether it can serve as a model of even primitive learning. There is a certain skepticism in evaluating an accuracy of a parallel between LTP and memory, and evidence of LTP involvement in normal brain function is considered to be inconclusive [1255, 273, 632, 817, 785]. Damage also leaves traces, but a trace not necessarily represents the memory. There are many similarities between synaptic plasticity and excitotoxic damage, especially since they both are a consequence of strong excitation. On the other hand, an LTP is related to developmental plasticity. Investigations of LTP have given a basic system of proofs to relative synaptic plasticity as a key mechanism of memory. Somewhat parallel between the stressed memory and LTP does exist. This parallel is determined by their common dependence on hurtful impacts [660]. Let us demonstrate that LTP is a pathologic state of neural tissue and cannot be considered as an example of memory trace.
Is LTP something like an excitotoxicity?
Excessive excitation produces both excitotoxicity and LTP. An attack must generate postsynaptic firing. In other cases, LTP cannot arise. Initiating excitations may be of various kinds. Postsynaptic firing, which must accompany an initiating excitation, may be evoked by current injection, by extracellular stimulation of enough amplitude [1148], or by antidromic stimulation of soma [586, 414]. Besides a classical combination of presynaptic glutamatergic excitation and strong activation of postsynaptic membrane [54, 1148], long-lasting potentiation may be produced by intensification of the cyclic AMP pathway and consequent upregulation of AMPA receptors by phosphorylation [1159], while inhibition of the cyclic AMP pathway blocks long-term facilitation [77]. Enhancement of intracellular Na+ is the first consequence of the excitatory current and postsynaptic Na+ is known to induce cell damage [313] and LTP [894, 374]. Augmentation of inhibition or hindrance of excitation prevents LTP development and counteracts damage, while restriction of inhibition shifts cellular balance to excitation and thus promotes LTP and damage [632, 54, 1148]. For instance, augmentation of synaptic inhibition and membrane hyperpolar-ization exerts strong preventative action on cell LTP [54, 1148]. In the same way, after local blockage of inhibitory receptors in piriform cortex, LTP was increased [601]. Although LTP resembles the consequence of damage, this phenomenon is not similar to death of neurons. The origin of LTP corresponds to activation of compensatory protective processes. Nevertheless, potentia-tion of compensation may turn out to be insufficient and then neurons after LTP induction really die. So, slow-onset of potentiation in the hippocampus is directly associated with cell death in the CA1 region in vivo [785]. Long-term depression is induced by smaller and protracted excitations [746], which usually initiate a compensatory mechanism and produce cell protection. This completely corresponds to conditions of the origin of protective preconditioning by means of a preliminary weak or moderate attack. Prolonged harmful influence allows, probably, for the development of compensatory protective processes, which restrict neuronal excitation by homeostasis and overcom-pensation may lead to inhibition and protection. Long-term depression, like LTP, requires Ca2+ entry through the NMDA receptor and this suggests that long-term depression is a reversal of LTP, and vice versa [94].
Parallelism between damage-protection and LTP
There is almost complete parallelism between characteristics of neuronal damage and LTP. Both LTP and damage are induced by a strong excitation [746] and in both cases an intrinsic excitability is reorganized together with a synaptic efficacy [135]. Both necrotic damage and the LTP are persistent, but reversible phenomena [290]. Long-term potentiation is presented in the hippocampus, cortex and cerebellum [291], areas the most sensitive to damage. After a strong impact LTP is developed slowly [1432, 124], as the necrotic damage is developed, in the cases where compensation counteracts the damage. Growth of dendritic branches and synaptic terminals is also observed in both cases [1195].
Various pathological conditions, including hypoxia and ischemia, leading to cell damage also establish the LTP or long-term depression [1369, 1429]. Glutamatergic synaptic transmission in hippocampal CA1 areas is facilitated after traumatic brain injury [205]. The same impact (halothane anesthesia) producing severe neuronal damage can simultaneously induce in the same brain area (CA1 region of hippocampus) a synaptic potentiation and synaptic depression of nearby neurons [1369].
Damage and LTP are provoked by a similar means. Besides the chief role of strong excitation, they can be produced by the same pharmacological treatments. The substances that exacerbate neuronal damage, as a rule, also increase susceptibility to LTP and on the other hand, the treatments that protect neurons, such as glutamate receptor antagonists and GABA^ agonists [1367] retard the onset of LTP [581, 793]. Facilitation, such as damage, is greatly reduced by postsynaptic injection of a rapid Ca2+ chelator or by postsynaptic hyperpolarization during tetanic stimulation [85]. Dopamine-containing neurons are very stable [1380] and accordingly, the same tetanus in the presence of dopamine leads to long-term depression, but not to more LTP in cortex slices. The metabolic pathways responsible for LTP and neu-ronal damage-protection are almost completely coincidental. Long-term po-tentiation, as damage, depends on Ca2+ influx [107, 1316] and on activation of NMDA [581, 793, 107, 751, 1316], metabotropic [555, 1290] and AMPA [731, 966] glutamate receptors. At the same time, application of an NMDA receptor antagonist blocks induction of LTP [632] and prevents cell injury. Moreover, early phases of both LTP [581, 374] and neurotoxicity [993] depend on NMDA, while their late stages depend on AMPA receptors. Intracellular acidosis, being protracted in time, is a harmful influence and leads to an irreversible depression of synaptic responses in the hippocampus, but if acidosis was shorter by 30 minutes, depression would become reversible [1366], as is observed at the early stages of necrotic damage. LTP and damage induce also a similar hormonal response and increase the synthesis of brain-derived neurotrophic factor and nerve growth factor.
Correspondingly, GABA^ activation results in a decrease in LTP [593, 54, 756] and neuronal damage, at the same time as attenuation of inhibitions by GABA^ receptor antagonists facilitates LTP induction [632, 581, 793, 1383] and exacerbates damage. Sometimes tetanic stimulation failed to elicit LTP unless a GABA^ receptor antagonist was applied to the slices [1007, 632]. The maintenance of LTP in the hippocampus is accompanied by impairment of GABA^ receptor function [1177].
Enhancement of levels of second messengers, intracellular Ca2+ and cyclic AMP, is also an important attribute of LTP [606, 713, 290, 1159] and damage. Both postsynaptic intracellular [581, 793, 925, 1367] and presynaptic [894] Ca2+ enhancement is an important condition for LTP and neuronal damage.
Retrograde messengers, the endogenous cannabinoids, the gases NO and CO, as well as arachidonic acid participate in development of both the LTP and neuronal injury [111, 925, 23]. Hippocampal and amygdala neuronal damage and LTP similarly depend on cytokines [659, 948]. Intracellular processes activated within Aplysia sensory neurons by injury, and those activated during long-term behavioral sensitization, also overlap significantly [1195]. Cell death [623] and LTP [39] may depend on protein kinase C translocation towards the membrane.
Systems of retrograde and second messengers are involved in processes of damage-protection and LTP in a similar way. Particularly, as a consequence of repeated stimulation, retrograde messenger NO modulates cellular function that leads to LTP, and NO effects depend on excitation power in the same way that excitotoxic damage depends on excitation power [991]. Similarly, activation of the second messenger cyclic AMP participate in the LTP induction similarly in fine details [811], as is observed during cell injury. High postsynaptic Ca?+ levels produce LTP and moderate levels induce LTD [124] and at the same time, a high level of intracellular Ca2+ is lethal and a moderate level is beneficial and induces tolerance to consequent damage [126]. Damage as well as LTP [581, 811, 374, 1367] depend on protein synthesis. Nevertheless, chemical similarity between cellular processes does not mean, generally speaking, a correspondence between the mechanisms of functioning. If, say, proteins participate in both a muscle contraction and thermoregulation, mechanisms of these forms of activity are not obliged to be similar. To say that something depends on Ca2+, K+, G proteins, cyclic AMP, ATP, and so on means to say almost nothing, because it is difficult to find some process in the living tissue without their participation. Of course, coincidence in multiple points is conceivable and participation of the same specific proteins in damage and LTP in a similar way is essential and this includes, for instance nitric oxide syn-thase, adenylate cyclase, protein kinases A, protein kinase C, phosphatidyl inositol kinase or haem oxygenase-2. A whole complex of biochemical and physiological processes that proceed during injury and during LTP is almost coincidental. Correspondingly, the main basis of proofs for a role of persistent synaptic plasticity in memory is defeated.
There are similarities between stressful conditions and mechanisms that occur in long-term potentiation. In both cases, neurons become hyperexcitable and the cyclic AMP system is activated [280], while Na+,K+-ATPase activity decreases [1365]. Synaptic change resembling LTP has been observed in the thalamus-amygdala pathway after naturally occurring fear conditioning [1035]. One-trial inhibitory avoidance produces the same changes in hip-pocampal glutamate receptors as induction of LTP and is associated with the delivery of AMPA receptors to synapses [1335], but unlike normal memory, the changes in AMPA receptors are not detectable for longer than 1 hour after training. Besides, chronic exposure to numerous types of drugs of abuse induces a long-term potentiation-like state in ventral tegmental area dopamine neurons, mediated via increases in AMPA glutamate receptor responsiveness and cyclic AMP response element binding protein [887]. Damage-protection and LTP demonstrate also many other common features. For instance, during induction of LTP, correspondence between neuronal excitability, input resistance and membrane potential is disturbed [374], as is usually observed during cell damage. As in the case of damage, LTP does not always depend on the presence of all its features for normal development and may proceed, for example, without participation of NMDA receptors and postsynaptic Ca2+.
Signaling between astrocytes and neurons also participates in LTP [27] and damage of neural tissue. So, tetanic stimulation of corticothalamic fibers caused a long-lasting reduction in electrical coupling strength. Metabotropic glutamate receptor agonist reduces electrical coupling to a degree comparable with the effect of tetanus. Thus, metabotropic glutamate receptor may play a role in regulating the spatial and temporal coordination of inhibition to the dorsal thalamus [701]. LTP in one cell could decrease spike latency in a nearby cell by pairing low-frequency stimulation with postsynaptic depolarization. The enhancement in the neighboring cell, resulting from action of a diffusible messenger, may be blocked by loading the cell with a Ca2+ chelator and hyperpolarizing the cell during the time when LTP has being induced in the first cell [94]. Properties of sensitization in Aplysia also have symptoms of damage. Sensorimotor synapses in Aplysia, the basic object for investigation of sensitization, are glutamatergic [47]. Both cellular damage and synaptic facilitation in Aplysia depend on release of Ca2+ from postsynaptic intracel-lular stores, postsynaptic exocytosis, and modulation of postsynaptic AMPA receptor efficacy [731]. Growth of processes during sensitization and dependence upon the cyclic AMP system [77, 811] are also indicators of early stage of protection after damage.
However, the role of some biochemical pathways needs clarification. For example, participation of nitric oxide in LTP, in memory formation and in cell damage is not completely clear. In addition, strong hypoxia or ischemia-evoked LTP may be more sensitive to a blockade of NMDA or metabotropic glutamate receptors, than long-term depression, but non-symmetry between long-term potentiation and depression corresponds to a non-symmetry between strong damage and protection, since strong damage lead to inhibition of homeostasis, particularly to inhibition of Na+, K+ -pump activity [1022]. The role of protein phosphorylation also may be different in the induction of long-term potentiation and depression [632].This is determined by a parallel development of damage and compensation. Besides, homeostasis is responsible for general cell survival and is not obliged to support each particular parameter in a cell at the stable level.
Development and LTP
During early development, neurons exhibit intensive growth. Developmental overproduction has been observed in many parts of the central nervous system. Some of the neurons at this stage die through apoptosis. Parts of neurons are apparently more sensitive to damage, especially in the hippocampus. The injured adult central nervous system exhibits features that are also observed during development: the same properties that central neurons display during regeneration following damage to peripheral nervous system [858, 359]. This growth is accompanied by increased excitability. High threshold neurons in the leech respond by sprouting at the lesion site in the nerve root and by sprouting additional processes from the axon hillock region, i.e. from the region of highest excitability.
Damage induces neuronal alterations, which are very similar to the properties of sensitized neurons and in the artificially sensitized neurons, LTP is easier to induce. For example, it is difficult to induce LTP in chronic preparations of the neocortex. However, LTP can be reliably induced in neocortical slices or in acute preparations [1007]. The damage is magnified under in vitro conditions and this may be the reason for slice susceptibility to LTP. As well, the treatments which induce abnormally high excitability in mammals and invertebrates evoke growth of axon branches and synaptic connections. Strong neuronal activation leads to synthesis of vesicular and other synaptic components. Neural tissue displays also developmental-like properties during LTP [1255]. Both developmental plasticity and LTP require the participation of postsynaptic targets [476]. It is directly established that LTP of excitatory synaptic transmission can regulate the rate of neurogenesis in the adult rat dentate gyrus in vivo [171]. Therefore, LTP and sensitization may be the consequences of developmental plasticity induced by transient or persistent excitotoxic damage.
Long-term potentiation and depression are easier to induce in young animals and embryonic cultures [1432, 1159], during developmental instability of the neuronal structure. Crair and Malenka [273] have examined LTP in the thalamocortical synapses that form whisker barrels in the somatosensory cortex. The period during which LTP can be induced closely matches the critical period during which the barrels can be modified by sensory perturbations. Susceptibility to LTP in the visual cortex also coincides with the critical period of naturally occurring experience-dependent synaptic modification and, like the critical period, susceptibility to LTP can be prolonged by rearing animals in the dark [632]. Therefore, properties of LTP correspond to a normal mechanism of experience-dependent synaptic modification in the developing mammalian brain. Such a critical period of neuronal growth and death depends on the environment and this therefore reminds the memory phenomenon, but does not coincide with it, at least with the aware, declarative memory which is frequently identified with the LTP [20, 845, 86]. Yet, these findings support the hypothesis that LTP reflects a mechanism of experience-dependent synap-tic modification in the developing or injured neural tissue, but there is too large a distance between the developmental plasticity and normal learning.
Temporal scopes of damage, LTP and learning
Long-term memory is remotely similar to LTP and the important likeness is that both phenomena are protracted and may exceed 24 hours [660]. Longevity of long-term memory may be almost infinite, although this is known only from behavioral experiments. Nevertheless, learning does not need a large time in order to remember new knowledge. Information is rapidly recorded in a memory and is stored for a long time. Declarative memory can be recollected immediately after acquiring new information. Procedural memory needs repetition of learning procedures, but this is necessary only in order to reveal regularities in the environment. Brain does not spend much time in remembering; this will be evident if one takes into account probing movements during an instrumental conditioning, when an animal examines a working hypothesis relative to the method of the achievement of a useful result. Long-term memory requires hours for consolidation, but this phenomenon is related to preservation of memory and its resistance to deleterious influences. Rapidity of remembering for the motor memory is also rather high. Of course, memorizing cannot be instantaneous, but the time needed does not exceed seconds. As we already have indicated, development of necrotic damage takes minutes or even tens of minutes. Compensatory processes can extend this time for hours and days, until healing or death will occur. Certainly, the first excitation in a response to injury may arise rapidly, but it is still a reaction of healthy tissue. So, anoxia-induced depolarization started at about 100 s, but the presence of a little O2 (5%) lead to significant delays in depolarization responses [688]. Synaptic activity regulates the surface distribution of neurotransmitter receptors for glycine, glutamate and GABA^ and these processes are rather extended (minutes days) [633]. Advance of morphological symptoms of necrotic damage is even more protracted. During a later phase of LTP, the newly synthesized proteins have to be selectively transported to activated synapses and this takes many hours or days and evokes growth of cells [1316]. Such a time-consuming route cannot quantitatively account for fast memory formation, but it is in accord with the development of cellular protection and slow change in states of neural system, such as motivational states, cycle of sleep-awareness, etc. As for apoptotic damage, it develops too slowly for considering its role in the current behavior.
Origin of long-term potentiation and depression need a short and strong clash. However, exhibition of potentiation requires minutes or tens of minutes, before potentiation or depression reaches a steady level [586, 15, 414, 1432, 124, 374, 1367]. These time scopes are closer to organization of compensation during advance of damage than they are to formation of memory. So, ripening of LTP in pyramidal cells of hippocampus takes more than 10 and sometimes 20-30 minutes after induction, in order to be displayed [414, 374]. After potentiation, neurons in the cerebellar deep nuclei develop LTP gradually, during 10-20 minutes [15]. Some forms of potentiation develop even more slowly. Pairing low-frequency orthodromic stimulation with high-frequency antidromic conditioning of pyramidal cells in area CAI of the rat hippocampus resulted in long-lasting potentiation, but the amplitude of such potentiation took up to 60 minutes to reach its peak, much longer than standard synaptic LTP [586]. Moreover, after stimulating hippocampal CA1 pyramidal neurons with synap-tic inputs correlating with postsynaptic neuronal spikes, evoked by current, the threshold decreases to minimum during 100-120 minutes [1367]. Induction of long-term depression is usually slower than induction of LTP. Thus, temporal scopes for establishment of memory and LTP are different, but LTP and response to injury are developed approximately with equal rate. As for the time required for information storage in memory, there are distinctions from LTP. The early phase of LTP is not long-term. Phosphorylation-dependent modification of synaptic potentiation is capable of supporting LTP only for 13 h. Further LTP is expired spontaneously. This extinction cannot be identified with a short-term memory, which is not terminated spontaneously, but is only susceptible to outer influences. Therefore, development in time memory and LTP contradict the hypothesis about their tight relationship, while temporal similarity between formation of LTP and development of damage and protection evidences their common mechanism.
Depotentiation and protection
We have already mentioned that post-tetanic change in synaptic efficacy and excitability may be reversed at the early phase of potentiation and depression. For example, the intracellular application of an inhibitor of the inositol pathway eliminated the LTP in pyramidal cells 30 minutes after paired pre-and postsynaptic activation [100]. The forces ensuring de-potentiation and de-depression do not create a new point of equilibrium depending on their mutual power: properties of neurons are returned to an initial state. In particular, a depression can erase potentiation [1022]. Inhibitors of the enzyme haem oxygenase-2, which catalyses the production of carbon monoxide, prevent the induction of LTP in CA1 pyramidal cells. Furthermore, they can erase LTP that is already established and the percentage decrease of response size closely equaled the percentage increase produced earlier by tetanus [1181]. This corresponds to a homeostatic return to the norm.
Potentiation is energy and substance consuming process and it does not have a quit status, suitable for the parsimonious storage of experience. Thus, blockage of CO production can erase LTP and for LTP exhibition, it is necessary continuously to support production of CO [1181]. Likewise, rat sensory neurons are capable of undergoing a long-term sensitization that does not down-regulate, but requires the continual presence of a sensitizing agent, specifically, cyclic AMP and therefore the ability of sensory neurons to be sensitized has to be maintained [147]. Augmentation of the cyclic AMP system means intensification of the compensational capability of cells. At the same time, the treatments that generate LTP or long-term depression may be short and their presence is not necessary for maintenance of alterations, which already have arisen. This concerns activity of NMDA and metabotropic glutamate receptors. Posttetanic long-term alterations of responses are insensitive to blockade of glutamate receptors [1022]. On the contrary, activation of NMDA and metabotropic glutamate receptors has been shown to be required for depotentiation and these general features of depotentiation were found in animals of all ages [1432]. This supports the notion that LTP is the trial of the neural system to intensify protection and recover detrimental function, particularly by means of the cyclic AMP system. Contrarily, it is impossible to conceive of the memory system needed for energetic replenishment.
Preconditioning of LTP and compensation of damage
We have demonstrated that weak injurious preconditioning ameliorates consequences of later more severe injury, but severe preconditioning augments damage. Evidently, a weak preliminary attack augments a protection, while a strong preliminary attack leaves mainly traces of damage.
Correspondingly, it would be unsurprising if harmful preconditioning would affect the subsequent LTP. If we consider an LTP, as a protective reaction of neural tissue, it would be natural to expect that weak preconditioning would augment LTP. Really, a brief acute swim stress experience enhances LTP [660]. As well, a combination of short episodes (2 min) of hypoxia with tetanus produces LTP significantly larger than potentiation alone and, consequently, potentiation induced by a weak hypoxia is additive with tetanus-LTP [771]. Obviously, here the compensations are added and not the damages themselves. Experiments with the stronger preconditioning (with high frequency stimulation that induced LTP) have demonstrated a reduction of the next LTP [474]. Correspondingly, low acute doses of cocaine augmented the induction (10 min before LTP induction) of LTP at the excitatory synapses of the hippocampus, but once LTP had been established, cocaine had no effect on the potentiated response. At the same time, high doses of cocaine block LTP induction [1153].
Preconditioning of LTP with any detrimental preconditioning has a definite time window, requiring over 2 minutes to develop, being very effective at 1020 minutes post preconditioning, and then ceasing to be operative at 45 minutes after the preconditioning stimulation [474]; however, precise parameters of time depend on conditions of experiments. This delay approximately corresponds to the time window for beneficial effects of weak harmful preconditioning.
The protective results of strong preconditioning sometimes were ambiguous. Severe preconditioning of hippocampal slices by NMDA (chemical insult) improves recovery following acute insults and increased cell survival. but has deleterious effects on LTP [1384]. An effect of strong preconditioning may exert a non-linear effect to LTP. Severe attack may induce stronger protection and this will exhaust the cell possibility for compensation. As a result, tetanization cannot increase this compensation additionally. In this case, cells will survive better, but LTP will not correspond to the real value of compensation. In other cases, severe preconditioning may hurt tissue, impair the protective mechanism itself and, hence, decrease LTP. Lastly, tissue may fall into a paradoxical phase, when additional inhibitory protection may animate and excite a neuron. For example, protracted intermittent hypoxia (during 3 days) decreased the LTP in rat hippocampal slices. The diminishment was biphasic, i.e. more pronounced after a 3-day intermittent hypoxia than a 7-day intermittent hypoxia. Influence of cocaine withdrawal to the magnitude of LTP also may be dependent on longevity of the withdrawal period. After 3-day cocaine withdrawal the magnitude of LTP was increased. A positive correlation exists among the 3-day cocaine withdrawal group between the amounts of cocaine ingested vs. the magnitude of the LTP observed in slices obtained from these rats. After the 30-day cocaine withdrawal. LTP magnitudes were similar to LTP magnitudes after saline administration and after the 100-day cocaine withdrawal. LTP decreased [1226]. Memory traces do not allow such an easy manipulation with their intensity. Typically, a trace of memory is either absent or present forever.
Specificity of LTP
Is distinctiveness of damage specific in respect to a specific insult? Naturally, this question has sense only for an early stage of damage, when neurons still stay alive. Apoptotic damage, certainly, cannot be specific, since it is under central control from the nucleus. Grow of necrotic tissue certainly can be local and specificity may be determined by the place of damage or by the manner of hurtful action, particularly by the variables that have been distorted. Pathways to death also may be different after different destructive impacts. Thus, anoxia and glucose depletion induce a notable acidosis in rat dorsal vagal neurons while metabolic arrest does not affect intracellular pH [1015]. It is unclear if such specificity is essential and has an informational component. However, the most fundamental problem is existence (or absence) of specificity of compensation or homeostasis and, consequently, protection. Compensation is, evidently, directed to recovery of loose function and its contribution to a current neuronal behavior is almost completely predetermined. However, if compensation could interact with the reasons of damage, that is, to be forestalled in respect to proposed damage, its efficacy would be strongly augmented.
The example of specific damage-protection events provides LTP or long-term depression. The induction of LTP in the hippocampus and visual cortex may concern only those synapses that were activated during conditional stimulation [1235, 632, 923, 414, 290], although such a specificity is not ideal and modifications in one input may concern changes in other inputs. It is possible to receive different LTPs of two independent Schaffer collateral pathways converging to the same pyramidal cell in the hippocampus [923]. Activity-induced synaptic modifications in neurons of hippocampus are often input-specific, i.e, only the synapses experiencing repetitive synaptic activation are modified. However, LTP induced at one Schaffer collateral input to the pyramidal cell by high-frequency stimulation could spread to other synapses on the same pyramidal cell when the unstimulated inputs were within 70 ^m from the site of LTP induction. To be exact, specificity is limited by the distance and potentiation/depression is spread to adjacent synapses [290]. Specific changes in potentiation and depression only for part of synaptic input has, usually a simple reason: this is determined by a spatial remoteness of the groups of synapses. So, different synaptic groups may be located at different dendritic branches of hippocampal neurons [414]. It is supposed that there are larger changes in dendritic channels of excitable membrane than in somatic channels [1367]. Long-term depression also can be specific. So, conjunctive irritation of Purkinje cells by pulses of metabotropic receptor agonist and direct cell depolarization induces LTD and input specificity is retained in this reduced system [744]. It is possible even to evoke LTP in specific synapses of one neuron and long-term depression in other specific synapses of the same neuron [1235, 632]. The depotentiation also can be input-specific or may not be [1432].
In all these examples, what is principally important is an availability of specificity, even if not in all cases. Absence of specificity is a trivial result. After all, even learning is unspecific at the beginning of training (phase of generalization). The generation of long-term potentiation and depression depends upon associative interactions between synapses that converge on individual dendrites. The distance over which these associative interactions occur and consequently, the degree of specificity is limited by synaptic GABAergic inhibition [1235]. GABA^ receptors establish a shunt between inner and outer environments and decrease the interaction of remote synapses and their correlation. Hyperpolarization, similarly to synaptic inhibition, also prevents LTP induction [1148]. When during LTP, the probability of firing of a postsynap-tic neuron to a given EPSP is enhanced and, hence, excitability increases, this effect is partially determined by GABA^ receptors (60% of this effect). 40% of the effect does not depend on GABA^, but depends on NMDA and Ca?+ and this component is input specific; that is, modification of excitability could be restricted to a local dendritic area. Similarly, synaptic depression is NMDA and Ca2+ dependent, partially (40%) GABA^ independent and this part is input specific [291, 414]. These effects are evidently determined by the protective action of GABA^ receptors and the detrimental actions of NMDA receptors and Ca2+ increase. This rather reminds one of local damage of potentiated dendrites. Specificity of LTP is also well displayed in the example of change in intrinsic excitability, which is induced by relatively weaker inputs than potentiation of synaptic efficacy [54, 414, 1367]. Such a change in excitability during LTP may be selective, as was found in the hippocampus, cerebellum and neocortex [586, 291, 414, 374]. Particularly, such specificity can be high and voltage threshold of spikes arising from the potentiated EPSP can be lowered up to 2 mV, compared with the voltage threshold of spontaneous APs, which remained stable [374]. Evidently, during LTP, specificity of effects is relatively low and confined only by a spatial localization of the effects. Such methods of coding already have been rejected when we considered possible mechanisms of memory in topic 1. If the specificity of homeostatic protection is also restricted by a spatial location of protective events, homeostasis cannot anticipate salient features of damage and serve as an instrument of forestalling protection.
Selective changes of excitability during learning, which we have considered earlier, perhaps also conformed to a specific shift in the scale of damage-protection. For example, irresponsiveness of a neuron to habitual stimulus is a specific breakage of excitability to the given stimulus and this breakage proceeds without vague effects of excitotoxicity. A specific change in excitability is, probably, observed during damage, too (particularly, during LTP), although during learning the degree of specificity is higher. In both cases, during damage and during learning, an excitable membrane acquires paradoxical properties, when change in a membrane potential does not correspond to a change in excitability and this indirectly indicates a possible likeness of phenomena. Our discussion reveals that LTP and neuronal damage-protection have many features in common. LTP is, probably, the consequence of a development-related plasticity induced by the excitotoxic damage of neurons. The temporal scope of both damage and LTP coinside, but differ from the temporal scope of memory. Both phenomena are easily induced by a strong excitation, are mediated by glutamate receptors and depend on intracellular Ca2+ accumulation, retrograde messengers, cyclic AMP and protein synthesis. Both LTP and neuronal damage become weaker following treatments with GABA^ receptor agonists and NMDA receptor antagonists. During development, neuronal damage and death are enhanced, as also is susceptibility to LTP. Susceptibility to LTP increases during tissue damage (for example, after preparing slices), but preconditioning with weak injury lessens both damage and LTP. Neuronal growth is observed during damage and development and during LTP. The hippocampus is especially susceptible to LTP and is the most vulnerable to damage. Similar influences protect against cellular damage and evoke depotentiation of LTP. The properties of LTP remind one of the properties of behavioral learning only if one considers high-stressed avoidance behavior, which induces transient damage of specific brain neurons. Maintenance of LTP requires the continous availability of sensitizing agent, while memory store does not need a support. Therefore, LTP and neuronal damage have similar reasons and similar mechanisms. Nevertheless, LTP is not a dying. This is, rather, intensification of compensation or nervous tension directed against damage. Long-term depression is, probably, not always opposite to LTP. Sometimes it may be overcompensation of weak damage and a shift to inhibitory protection (as during weak acidosis), but sometimes long-term depression may be converted into dying (as during strong acidosis) [1366].
