The direction to chemical specificity of memory
If remembering of specific information is somehow reflected in origination of chemical specificity in the brain, these chemical traces must be extremely tiny and subtle in order to contain the quantity of information that we acquire during life. Such traces are difficult to find by direct experimental measurement. Nevertheless, some indirect evidences exist. One such piece of evidence is transmission of acquired behavior through metamorphosis [842, 1276]. Let us evaluate, now, if neuronal memory may be transmitted through heredity.
Some forms of relatively complex behavior are genetically predetermined. Moreover, inborn actions may become aware. For example, breathing and heart palpitation can be controlled by consciousness, although such control is impeded. Genetically predetermined actions may be not only automatic. Newborn primates snatch things by palm without any visible learning and newborn rats avoid lit places. Some forms of behavior look innate and faintly depend (if depend at all) on learning. Inborn kittens, blind and scarcely dried off, demonstrate an aggressive reaction, and hiss and grin under danger [1273]. Adult cats keep secret the location of the nest and bury their faeces. Astonishingly, cats living in domestic conditions after defecation also scratch the floor and try to bury faeces, although this is impossible to accomplish in the firm floor. When a cat fulfills the innate command "to bury" its faeces, it is calm and does not note faeces further, although they are openly present on the floor. Interestingly, this instinct does not decay and cats during the years continue these senseless actions. These observations do not prove that same habits may be absolutely innate and are displayed without any learning. So, salmon find their mother’s river from the ocean, but homing is not innate behavior; salmon remember the smell of the river they left in their youth [1095]. In the human, launching, entraining and expulsion events are interpreted causally by young infants, but there is no good evidence that these representations are innate [1088]. However these facts demonstrate an inclination to some types of learning. For instance, imprinting arises rapidly, is very stable and does not decay. A chicken interprets as the mother the first moving thing that it sees and never changes this habit. Almost any signal appearing in the sensitive period of infancy acquires traits of the "key" signal for imprinting. If learning is rapid (one accidental pairing, as during imprinting), it may slip away from attention in an experiment. The principal importance of the existence or absence of inborn actions is determined by the possibility to transfer acquired information to posterity. Rates of learning in young animals and in the adult are different. They are much higher in the young, and at a venerable age the rate of acquiring new information continues to decrease [1379]. An instrumental conditioned reflex in 2-week old rats is acquired after one or, rarely, two pairings of correct movement with food reinforcement, compared with 5 pairing needed in adults [1252]. Young animals have a predisposition to rapid and robust learning. Their learning is similar to imprinting; it is rapidly automated and resistant to decay after abolition the reinforcement. Therefore, early acquired habits are easy to identify as innate behavior. Nevertheless information can probably pass through a genetic barrier. When, in infancy, an animal promptly remembers and firmly keeps in mind association between the signal and the vital event, this cannot be any unspecific signal. Only the signals that are similar to the natural signal may play a role as a key signal. For example, when 1-3-days chicks take the first moving object for their mother, this object must be by size, color, direction and speed of relocation similar to a fowl [1040]. Then they remember this object and follow it. Although this is the acquired habit, there are restrictions for a key signal, which have to correspond to some determined features.
This means that some general information is inherent and may be transmitted with the genetic material. Another remarkable example of the dependence of behavior on genetic chemical properties is seen in Drosophila flies. Males and females fight with distinctly different styles, and males but not females establish dominance relationships. As well, males (but not females) perform an elaborate and innate courtship ritual directed toward females (but not males). Male courtship requires products of the fruitless gene, which is spliced differently in males and females. When the male form of the protein is expressed in females, the females will mount and direct the courtship toward other females or toward males that have been engineered to produce a characteristic female odor [308]. Hence, peculiarities of behavior are determined by the genes.
Strictly speaking, not only macromolecules of deoxyribonucleic acid (DNA) may serve as a substrate of transmission, but also a cytoplasm of the egg. Hereditary predisposition to specific behavior is not, of course, transmitted in each individual case. This event is (if exists at all) extremely infrequent. However, if this is principally feasible, it would be a decisive argument for the chemical nature of memory. Even a weak possibility for transmission of acquired behavior to posterity is important for comprehension of the memory mechanism, which in this case has a chemical nature. If the particular unit of memory is kept in the salient feature of the neural network, a way for altered brain structures to influence the genome in absent. This way must pass through the gametes. Particularly, sperm motility varied depending on the social status of male mice and becomes enhanced due to environmental factors experienced by male mice during maturation [669]. Besides, this again brings up the question of concerning the genetic acquisition of features obtained during an individual life: the question that was brought up by J.B. Lamarck and was closed down by Ch. Darvin.
Does a possibility for memory transmission through heredity exist? Let us believe that production of counterfeits (forgery) is a normal way of technical progress. When one fabricates some item, say a calculator, in an underground factory there is a possibility unintentionally to improve this item, since sometimes a mistake in the fabrication may result in an upgrading. Further fabrication of this upgraded item may lead to a new useful mistake and so on, until the factory will produce a computer. Formation the brain in evolution by means of mutations one may compare with the fabrication of spoilt counterfeits. If this long way can be considered to be probable, then in order to inherit habits one must seat a monkey at the computer and wait till a new version of Windows is created.
It is difficult to believe that neuronal memory may arise as the result of mutation, but nobody rejects this option determinedly. Chemical transmission of memory from parents to offspring needs two factors: chemical memory must exist and the way for a spread of modified memory through heredity must exist, too. We may only conclude that some primitive forms of behavior are hereditary and now will demonstrate that influence of acquired memory to the genome at least, does not contradict experimental data. One of the possible forms of influence of acquired information to offspring (epigenesis) is transmission of factors affecting activity of preexisting genes.
One of the most elusive questions in the study of memory is the nature of the enduring molecular changes that underlay memory storage. New memory may be connected with a new macromolecule, modification of preexisting macromolecule, change in gene activity or with mixtures of relatively simple substances. The separate question is how chemical alterations may become persistent.
Proteins are the main working horse of cells and neurons exceed all other cells by their intensity of biosynthesis [927]. Biological activity of proteins is determined by their three-dimensional structure, which depends on a succession of amino acids, constituting the protein. On the other hand, a one-to-one correspondence between amino acid sequence and spatial structure is absent, and there are several stable configurations of the same protein. Actual configuration may depend on conditions of assembling, microenvironment, pospho-rylation, etc. Thus proteins structure is a candidate mechanism for memory storage. In particular, prion-like proteins represent auto-replicative structures that may serve as a persistent form of information. Prions (infectious proteins) exemplify a novel mechanism of biological information transfer based on self-propagating changes in protein conformation, rather than inheritance of amino acid sequence. A prion may be intrinsically pathogenic, but may function as a cytoprotective molecule [519]. It has been proposed that prions play a role in memory storage [612].
Another possibility for information storage is spatial travel of preexisting proteins. Trafficking of chemoreceptors and potential-dependent channels plays a critical role during long-term plasticity, as we have described for LTP. However, supposed variants of proteins in memory coding (with the exception of synthesis of macromolecules, when information can be coded in their amino acid sequence) do not explain the variety of memory elements. Although a protein is a very complex molecule, a restricted number of its stable conformations exist. And although protein modifications sometimes may be persistent, the variety of possible memory elements is a decisive factor. As an example, scars also may be persistent, but only at a stretch of the imagination a scar may be considered as a memory, and we will argue further that LTP is something like a scar in the brain. Such chemical processes ensure only spatial modification of the network or morphological modification of the neuron itself. Especially in-tracellular integration of synaptic influences (the binding problem) still waits an explanation. It was supposed also that memory is stored in astrocytes by organizing the activity of ion channels, as a closed, high resistance, state of the gap junctions, and is not associated with a strict physical location [220]. This function may be served by astrocyte gap junctions and suggests that agents that selectively block these gap junctions should disrupt memory. Indeed, general anesthetics disrupt both gap junctions and consciousness and during loss of consciousness, memory does not function. Nevertheless, it is incompletely clear how memory may be recovered after one regains consciousness. Besides, not every impact, that decreases awareness, blocks also gap junctions.
Any long-term trace of memory must be embodied in some steadfast change in the brain, whether we are speaking about restructuring of brain construction or chemical reorganization. Nevertheless, long-term memory cannot be considered as utterably stable and, once being formed, kept intact for an entire life. A short (few minutes) disruption of the blood circulation evokes deep amnesia, more serious than mechanical brain damage. Morphology of synaptic connections usually does not change after short-term impacts, and neurons also survive an accident overload. Life is more resistant to failure of oxidative metabolism than memory. The neural network is too well-built in order to be the source of memory. On the other hand, a memory trace cannot be as firm as, for example, the DNA structure, which is not destroyed after a short arrest of blood circulation. Memory, evidently, needs energy and has a dynamic nature. Nevertheless, complete blockage of brain electrical processes during hibernation, or, contrarily, complete distortion of electrical activity during an epileptic seizure disrupts consciousness, but leaves memory intact. This means that a dynamic process connected with memory has a chemical rather than an electrical character. These chemical traces concern, evidently, memory retrieval, since amnesia may sometimes be reversible even after a severe crash, as after the vegetative states of consciousness. What constitutes long-term memory itself is still unclear at the present time.
Neurons, like other cells, are controlled by the portions of DNA in the cell nucleus that are expressed. Proteins governing the structure and function of a neuron depend upon which RNA chains are produced by DNA in the nucleus – which, in turn, depends upon the regulatory proteins bound to the DNA. RNA and DNA codes consist of nucleotide sequences (bases, which are slightly different for DNA and RNA, but we will not go deeper into details). One amino acid in a protein is coded by a sequence of the three bases.
An expression of the protein in the cell is very stable. So, one minute exposure to nerve growth factor in early developmental converts a cell into a neuron and to acquisition of membrane excitability through a signaling pathway requiring immediate-early genes for induction of sodium channels. With few exceptions, protein have a relatively short half-life (hours or few days) compared with the duration of memory (days, weeks or even years).
RNA is even less stable than proteins. DNA is a stable molecule. Although DNA is persistently attacked by nucleases and mutagens (in each human cell, about 500 bases suffer oxidative damage per day [1132], DNA structure is constantly repaired. The normal direction for informational flow in a cell is: DNA RNA protein substrate. The process of making RNA based on DNA information is called transcription, while the process by which RNA produces proteins is called translation. The reverse route (DNA RNA protein substrate) is fraught with difficulties that are insurmountable, since the reasonable way for influence of the substrate to the consequence of amino acids in the protein is unknown, doubtful and may be absent.
Nontemplate RNA synthesis?
It is known that basic RNA synthesis is accomplished by copying DNA with DNA- dependent RNA polymerases participating, and triphosphates of ri-bonucleosides are used for the synthesis. In addition to DNA-dependent RNA synthesis, there are a number of pathways of nontemplate RNA synthesis that are catalyzed by various enzymes [990, 265]. In this case ribonucleo-side triphosphates are used in the process of synthesis, and the enzyme that polymerizes ribonucleoside diphosphates is known (polynucleotide phosphory-lase). There is almost nothing we can say about the biological function of most of these enzymes. Nevertheless, if learning induced nontemplate synthesis or modification of RNA in neurons is possible, in spite of a rapid destruction of these RNAs, acquired information may be stored in DNA, since reverse transcription of RNA DNA is a carefully established fact. Reverse transcription has in fact been shown to be activated after learning [1070]. If RNAs participate in the process of memorization (as a template for synthesis of specific proteins or peptides of memory or in some other fashion), their synthesis, modification or degradation must take place during active learning.
The order in which nucleotides are joined in a polynucleotide chain in systems in vitro depends on the specificity of the given enzyme and on the concentration of substrates; the order of joining of nucleotides in vivo is not known for most enzymes. The order of nucleotides might depend on properties of the substrate, which absorbs nucleotides and which is altered as the result of learning. In accord with the principle of cross-stereocomplementarity [825] there is a correspondence between spatial structures of a protein and coding its DNA. At least in some cases this principle works well. Correspondingly, if the cellular antigen, as, for instance, modified protein, which was developed during learning, will absorb ribonucleoside diphosphates and polynucleotide phosphorylase will combine these bases into a chain, the protein created by this chain will be complementary to the antigen and thus will serve as the learning-related trace. We do not know if this principle holds true. It is necessary also to consider that thus far we do not know any template-dependent RNA polymerase that utilizes nucleoside disphosphates and it is difficult to express a specific conception as to how nontemplated synthesis can participate in the formation of RNA of a concrete structure.
As is known, the signal reaching the neuron leads to chemical or conformation changes in the membrane. With a change in the biological significance of the stimulus as a result of learning, the chemical process that this stimulus gives rise to in the neuron changes (see previous paragraph), specifically, the state of the excitable membrane. Considering that learning is accompanied by modification of membrane proteins [356], we may hypothesize that specific determinants formed during learning selectively adsorb nucleotides that develop a chain localized near the nontemplate polymerases. Information on the nucleotide sequence carries a specific determinant that develops during learning and serves as a kind of template, such as, for example, the enzyme complex in bacteria that catalyzes nonribosomal (independent of the RNA template) synthesis of peptide antibiotics [490, 1187]. The ability of nontem-plate polymerase to carry out the reaction that is the reverse of synthesis pyrophosphorylation of RNA with the formation of nucleoside diphosphates [990] ensures a rapid destruction of these RNAs after their translation and preparation of substrates of nontemplated RNA synthesis for receiving new bits of information. The existence of such a reverse reaction could explain why learning is made easier not only due to RNA of a brain, but also due to RNA of the liver or of yeasts [899]. Probably, the effect is linked to the enrichment of the supply of nucleotides in the brain. We examined the influence of ri-bonucleoside diphosphates to the process of the simple learning, habituation [1263].
Experiments were carried out on basically identified neurons LPa3 and PPa3 of the snail connected with the reactions of breathing and avoidance. Adaptation to repeated tactile stimulation of the mantle at intervals of 5-7 sec was developed. Specificity of adaptation was controlled by presentation of a rare stimulus (turning off the light) 1-2 min before and 5-7 sec after training. Activity of the neurons was recorded by intracellular glass micro-electrodes filled with a solution of potassium citrate in experiments with the pure control (learning without introducing the substances) and with solutions of corresponding nucleoside diphosphates in the other cases. We tried to determine if learning would be disturbed if the concentration of one of the substrates of template (triphospates) or nontemplate (diphospates ) RNA synthesis were changed. For this purpose, we introduced cytidine triphospate (CTP), cytidine diphospate (CDP), or uridine diphospate (UDP) into the neuron. In subsequent learning, this could lead to synthesis of the RNAs with an excessive content of cytidine or uridine nucleotides. As a control, we introduced cytidine monophosphate (CMP) into the neuron, which is not capable of polymerization but can participate in other metabolic processes proper to cytidine nucleotides. Another control was a mixture of approximately equimolar concentrations of CDP, UDP, adenine diphosphate (ADP), and guanine diphosphate (GDP). If the proportion of concentrations of ribonu-cleoside diphosphates in the neuron was approximately the same, its uniform change would not substantially distort the structure of the RNA synthesized during learning.
After the substances were introduced, the character of spontaneous activity did not change. Habituation training in the pure control was accompanied by a decrease in AP in response to repeated stimuli, and an increase in the threshold of generation and latency of the first synaptic AP elicited by this stimulus. Habituation training did not result in an increased AP threshold elicited by the rare stimulus. The membrane potential remained almost unchanged regardless of training.
The effect of nucleotides on learning was studied soon (in 3-12 min) after their introduction because after a certain time metabolic processes of the neuron would have had a substantial effect on the concentration of the substances introduced. Experiments indicated that the dynamics of decreasing the AP during habituation did not differ after introducing CMP, CDP and CTP (Fig. 1.30).
We must consider, however, that the nucleotides were introduced into a single neuron, but habituation was developed by the whole nervous system of the mollusk. The AP number in the response is heavily dependent on the power of the stimulation reaching the neuron, which is superthreshold in AP generation, and reflects primarily the passive participation of the neuron in brain activity. For this reason, we had to isolate the individual neuron’s contribution to the learning process, which is determined by a local change in excitability of the recorded neuron.
Nucleotides mostly elicited an initial increase in the AP threshold in response to the first stimulus of the habituation series. A comparison of all curves in Fig. 1.31, however, shows that the initial increase in the threshold cannot be the reason for the blocking of its further increase with administration of CDP and UDP. Thus, the initial threshold was approximately identical after the mixture of nucleotides and CDP was introduced, but habituation was accompanied by an increase of the threshold only in the first case.
Changing the threshold of AP generation accounts for the individual contribution of the recorded neuron to the process of habituation since the threshold is the property of the excitable membrane and, under normal physiological conditions; its value is practically independent of stimulus. After administration of CMP, habituation was accompanied by a selective increase in the AP threshold elicited by the repeated stimulus (Fig. 1.31,A 2). Such changes in excitability during development of habituation were also observed when no substances were introduced (Fig. 1.31,A in the frame). CDP and UDP completely prevented an increase in the threshold in response to repeated stimuli (Fig. 1.31,A 1 and B 4). Habituation had become nonspecific. A change in thresholds after administration of CTP had a qualitatively normal character although it was somewhat lower (Fig. 1.31,A 3). This slight drop may be linked to the fact that the CTP preparation contained up to 2-3% CDP formed in the CTP after purification and during storage and use. The CDP + UDP + ADP + GDP mixture did not prevent an increase in the threshold during training, although the increase was retarded (Fig. 1.31,B 5).
Fig. 1.30. Decrease in AP number in neuronal response according to degree of habituation after introducing CMP, CDP or CTP. Abscissa : Number of presentations of the tactile stimulus. Ordinate: AP number in the response. 1) Development of habituation with CDP; 2) with CMP; 3) with CTP. Standard errors are indicated. Methods. The substances were introduced into the neuron by intracellular microionophoreses (current, 0.5-1.0 nA; time, 30-120 sec). One of the substances (or the mixture) was introduced into each recorded neuron in low or high concentration. After the first dose, a short training series (seven stimuli) was administered. After the subsequent second dose, a complete habituation series was administered (30 stimuli). There was a 10-12 min interval between the short and long training series. The amount of substance introduced into the neuron depended on the charge passing through the microelectrode. The first dose was 5•10"9 C in all cases; the second doses: CMP – 2.5 • 10"8; CDP – 3 • 10"8; CTP – 4 • 10"8; UDP – 2.5 • 10"8 C; mixture of nucleotides, 10"7 C. The difference in amount of electricity passing through the electrode for the cytidine nucleotide can be explained by the fact that the molecular charge differs: CMP < CDP < CTP. The mixture of four nucleotides was introduced in doses four times greater than in experiments with a separate administration of nucleoside diphosphates. The approximate amount of substances introduced into the neuron was 10"13 mole; concentration of the substances in the neuron was brought to approximately 0.5 • 10"5 M. Nucleoside phosphates pass through the membrane poorly. For this reason the chemical action was limited by the recorded neurons.
Assessment of threshold is a labor-consuming operation, while evaluation of AP latency does not have any of the subjective factors that appear when it is necessary to evaluate the threshold in a response to a stimulus connected with learning. For this reason, we studied the effect of nucleotides on the dynamics of change in latency of the first synaptic AP elicited by a repeated stimulus. After CMP, CTP and the nucleotide mixture were administered (Fig. 1.32), habituation was accompanied by the usual increase in AP latency (control: straight line in the frame, obtained in experiments without administration of the substances). After introducing CDP and UDP, however, the coefficient of regression dropped sharply although not to zero (Fig. 1.32). Incomplete blocking of increase in AP latency during the process of habituation after administration of CDP and UDP is evidently coupled with the absence of suppression of excitation in the presynaptic pathways that are not affected by administration of the substances.
Fig. 1.31. Change in neuron excitability during habituation after administration of nucleotides. Ordinate: threshold of AP generation, mV. The threshold was measured from the level of membrane potential to the point of greatest curvature on the leading edge of the first synaptic AP in response to the stimulus (inset, on the right). Fine arrow, administration of a small amount of the substance: 1a-5a: short habituation series after this administration. Thick arrow, introduction of larger amounts of the substances; curves 1-5: long habituation series. A) In the frame: curve of change in excitability during habituation without administration of substances, in the same coordinates. B) 4a and 4: Habituation after administration of UDP; 5a and 5: after administration of the mixture (CDP + UDP + ADP + GDP). Other notations as in Fig. 1.30.
Latency of the AP depends on the threshold of its generation and on the characteristics of the stimulation that reaches the neuron. These results mean that biochemical processes of memory are reflected in the changes of thresholds and latency of action potentials in responses of neurons to a stimulus, which biological significance of novelty changes as a result of habituation. This regulation is based on chemical processes the nature of which is not yet known. Our data indicate that local chemical action on the recorded neuron can lead to a change in the learning process only in that neuron and change in excitability is its contribution in a whole response alteration. Actually the procedures we used did not affect the dynamics of decrease in AP value elicited by the repeated stimulus. Evidently this is linked to the fact that the response of the neuron to superthreshold effects is only slightly dependent on its own excitability. This means that in studying the chemical mechanisms of plasticity by means of action on a single neuron it is scarcely possible to use such parameters of electrical activity as the AP value in the response of the neuron or the frequency of spontaneous activity. The most sensitive parameter is the threshold of AP generation, and its latency is also acceptable. It must be noted that apart from complete objectivity of measuring latency, this parameter can be used not only in intracellular, but also in extracellular recording of neuron activity. Intracellular administration is weakly reflected in the frequency characteristics of peak activity, since the process of learning was hindered in only one neuron.
Fig. 1.32. Change in AP latency during habituation to tactile stimulus presentations after administrati on of ribonucleoside phosphates. Abscissa: number of tactile stimulus (logarithmic coordinates); ordinate: latency, ms. Notation as in Fig. 1.30. Lines of regression constructed. In the frame, change in the latency during habitu-ation without administration of substances in the same coordinates.
The evidence gathered in this study indirectly shows that learning is linked to synthesis of RNA from nucleoside diphosphates and is apparently not dependent on a template. Actually, a significant increase in concentration of one of the substrates of nontemplated RNA synthesis completely blocked the learning process. Blocking was rapid and practically complete, but for template synthesis we might expect an incomplete inhibition because of the great similarity of the complementary template nucleotides. The substrate of template synthesis (CTP) affected learning much less than CDP and UDP. CDP and UDP are not among the agents that have generally toxic effects. The similar effect of CDP and UDP is also evidence of this. The weak effect on learning of the mixture of four nucleoside diphosphates indicates that the participation of all four ribonucleotides is necessary for normal memorization. The only known reaction of this type is the synthesis of recognized RNA sequences. These data are in agreement with the hypothesis that the learning process is linked to nontemplate synthesis, but does not prove this assumption definitively. Besides, we have investigated only short-term effects. If long-term effects of ribonucleosides do exist (we know nothing about it), the same mechanism may work with the participation of reverse transcription.
