Progression over time from simpler to more complex structure and function

Biological development (des- + voloper, old French for 'to wrap') is manifested in growth, remodelling, and specialization of cells and tissues. This is accompanied by a change in the functional capabilities of the organism, which, in the case of the nervous system, could include modification of learning capabilities (Marcus et al. 1988; Hartshorn et al. 1998; Stanton 2000). The relationship of development and growth to learning raises some of the most fundamental issues in the neurosciences, abutting molecular biology on the one hand and philosophy on the other. Among these: how do brain circuits achieve their complexity and specificity? How much of learning is merely unravelling by experience of information that is already encoded in the genes? Are the cellular mechanisms of learning an extension of, or even identical to, the mechanisms that operate in development and growth? 'Growth and learning', wrote Holt (1931), 'are one continuous process, to the earlier phases of which we give the one name, and to the later... phases we give the other.' He was not the first to suggest that. The notion that when we learn our neural tissue grows was explicit in the scientific and philosophical writings of the late nineteenth century. Furthermore, some of the proposals referred specifically to the primary site where growth should occur—the "synapse, which at that time was not even yet known by that name: 'For every act of memory... there is a specific grouping or coordination of sensations and movements, by virtue of specific growths in the cell junctions' (Bain 1872, cited in Finger 1994). More elaborate experience-dependent growth theories emerged only later (e.g. Kappers 1917; Hebb 1949), paving the way first to the proposal (Monne 1949), and later to the discovery (Flexner et al. 1963; Agranoff and Klinger 1964; Barondes and Cohen 1966), that de novo "protein synthesis, hence modulation of gene expression, is required for the encoding of long-term memory (LTM).

The function LTM -/(Growth), where 'growth' is synaptic remodelling, is hence a "paradigmatic tenet of the current neurobiological "zeitgeist. But it is more than that. It also provides a guideline and framework for experiments on the biological bases of lasting memory. For if growth is concerned, then the study of the cellular mechanisms of LTM can borrow not only concepts but also "methods and data from the study of development. And as developmental processes follow rules shared by different tissues, organisms and phyla (Wolpert et al. 1998; Fraser and Harland 2000; Scott 2000), the demarcation line between the mechanisms specific to LTM and those that occur in other developing tissues that are irrelevant to learning, may be blurred. The implications of this will be further noted below. In the meantime, a few generalizations are noteworthy. Similarly to all other tissues (ibid.), the development of the nervous system involves cell division, the emergence of pattern, change in form, cell differentiation, and growth. In the process, neurons migrate over distances from their place of birth to their place of work, start to express specific gene products, including enzymes, "receptors, and "ion channels, form connections with their target areas, and then establish specific synaptic contacts. Even a brief account of each of these families of processes and mechanisms far exceeds the scope of this discussion (for guides to neurodevelopment, see Jacobson 1991; Goodman and Shatz 1993; Hatten 1999; for comments on the history of the discipline, Cowan 1998). Suffice it to note that in principle, two types of mechanisms work in concert in the formation of functional circuits in brains: activity independent, and activity dependent. The former are guided solely or mostly by genetic instructions, and usually occur even before the neurons become functionally active. The latter depend on extracellular signals, such as hormones, growth factors, and ions. The interaction between genetics and experience is especially important during certain 'critical periods' in development (e.g. Katz 1999; "birdsong, "imprinting, "nutrients).

It is useful to consider the relationship between experience-dependent modifications that take place in development and those that occur in learning, in terms of "levels of organization and analysis.

1. The level of overall computational strategy. "Stimuli that result in experience-dependent modifications may either 'instruct' a system to change, or 'select' one or a few among multiple endogenous states. The 'selectionist' strategy has 'hardware' and 'software' versions. The 'hardware' version proposes that experience selects and stabilizes certain morphological configurations of the system. The 'software' version only assumes that experience selects and stabilizes certain 'pre-representations', i.e. functional states of the system, which may or may not be subserved by specific morphological configurations (see "a priori, "internal representations). 'Instructionist' and 'selectionist' strategies, which refer to multiple spatial and temporal "dimensions in the function of the tissue, were proposed for development and for learning alike (Hebb 1949; Pringle 1951; Changeux and Danchin 1976; Young 1979; Lo and Poo 1991; Edelman 1993; Marler 1997; Quartz and Sejnowski 1997).

2. The level of the *algorithms that implement the strategy. Similar synaptic algorithms are postulated to operate in experience-dependent modification in development and in adult learning. Foremost in current theories and "modelling are Hebbian algorithms, or their conceptual progenies, including some that restrict plasticity only to 'critical periods' (e.g. Bienenstock et al. 1982; "metaplasticity).

3. The level of the biological mechanisms that implement the algorithms. Here again, similar molecular and cellular mechanisms are discovered in development and learning. They involve modulation of gene expression by extracellular stimuli, culminating in tissue remodelling (e.g. Corfas and Dudai 1991; Weiler et al. 1995; Davis et al. 1996; Martin and Kandel 1996; McAllister et al. 1999; Pham et al. 1999; "CREB, "immediate early genes, "intracellular signal transduction cascade, "late response genes). Whether the mechanisms are only similar or indeed identical is another story, the conclusion of which is not yet known. For example, the similarity between experience-dependent synapto-genesis in the developing brain and "long-term potenti-ation in the mature brain may be a bit overstated (Constantine-Paton and Cline 1998). Similarly, the role of neurotransmitters in neural development may require a second thought (Verhage et al. 2000).

The relationship of development to learning raises a number of real or apparent conceptual problems, which transcend the discussion of the specific neuronal mechanisms.

1. The specialization paradox. Definitions 1 and 2 above leave open the possibility that development may restrict the potential and "capacity of a system. (To mention an extreme, apoptosis, programmed cell death, is also a developmental programme; Kuan et al. 2000.) For example, the 'hardware' version of'selectionist' models assume selective elimination of synapses. Yet learning by definition adds "internal representations to our mental repertoire. If learning mimics developmental processes, and developmental processes eliminate degrees of freedoms in the system, is there a paradox here? Only an apparent one. First, elimination of alternatives is itself added information. Second, elimination of synapses does not necessarily undermine the potential representational complexity and capacity of a circuit, because this is also expected to be determined by functional synaptic capabilities that could be added in development (definition 3). Third, 'hardware' versions of selectionist processes might occur in development but only in some learning systems, e.g. 'prepared learning' ("birdsong, "imprinting), and not in others. And fourth, synaptic remodelling and growth during adult learning may provide the brain with a continuous supply of substrates for further, ongoing development. Moreover, in some brain regions new neurons are added in adulthood (e.g. Gould etal. 1999a,b; Shors etal. 2001; "birdsong, "hippocampus; but see some doubts concerning neurogenesis in the adult mammalian brain, in Rakic 2002). Nobody knows what these new neurons are doing there, but one possibility is that they compensate for reduced capacity induced by specialization.

2. The relevance issue. When we do detect growth in circuits that learn, how do we know that this is at all related to memory ("criterion)? Growth and tissue remodelling that are triggered by a training situation may fulfil functions other than learning, such as "homeostasis. At the time of writing, there is not even a single case in which learning-related morphological changes in circuits and synapses have been proven without doubt to fulfil functional, not to mention causal role in the use-dependent representational change in the circuit. With powerful model systems such as cultured "Aplysia circuits, that could enable selective lesioning of individual synapses and neurites, the evidence may soon be provided.

3. The persistence problem. If indeed LTM ~ f (Growth), and given that circuits undergo substantial changes with time (e.g. Purves et al. 1986; Segal et al. 2000): How come memory endures in the circuit spite of the turnover of its components? The solution to this problem is discussed under "persistence.

4. What makes memory memory? If cellular mechanisms of development and growth subserve LTM, why not concentrate on development in "simple systems, even in non-neuronal tissues and cell cultures that are much easier to handle than brains, in order to understand the formation of memory? This is what some capable investigators try to do. But a caveat is appropriate. The simple system approach casts some light on "plasticity but not necessarily on "memory. The latter, we should remember, is a functional property of circuits. Therefore, we should not expect to decipher the computations and codes that are used, say, in a cortical circuit, by analysing development in cell culture or a neuromuscular junction. Following the

Fig. 25 Ongoing development and growth in the adult brain: The density of dendritic spines in the rat *hippocampus is affected by experience. One group of rats (trained) was subjected to a complex, stimulating environment that enhanced their ability to learn their way in a water *maze. Another group was composed of individual rats caged in isolation in a quiet environment (isolated), and yet another group of rats caged in pairs in a similar boring environment (paired).The picture on the left depicts extremes of variability in the number of dendritic spines in hippocampal area CA1 in trained (a) and isolated (b) rats. The graph on the right shows the overall increase in spine density in the trained vs. isolated + paired rats. Although the overall difference is small, it is significant and may reflect an important effect of experience on the localized growth of synaptic contacts. Courtesy of Per Andersen; see also Moser et al (1997) and Andersen and Soleng (1998).

Fig. 25 Ongoing development and growth in the adult brain: The density of dendritic spines in the rat *hippocampus is affected by experience. One group of rats (trained) was subjected to a complex, stimulating environment that enhanced their ability to learn their way in a water *maze. Another group was composed of individual rats caged in isolation in a quiet environment (isolated), and yet another group of rats caged in pairs in a similar boring environment (paired).The picture on the left depicts extremes of variability in the number of dendritic spines in hippocampal area CA1 in trained (a) and isolated (b) rats. The graph on the right shows the overall increase in spine density in the trained vs. isolated + paired rats. Although the overall difference is small, it is significant and may reflect an important effect of experience on the localized growth of synaptic contacts. Courtesy of Per Andersen; see also Moser et al (1997) and Andersen and Soleng (1998).

same line of argumentation, even if we understand how neuronal circuits develop, we cannot hope to grasp their contribution to memory and behaviour unless we decipher the representations and computations performed by these circuits in the subsecond range. This is clearly a time-scale ("dimension) very different from that addressed in the study of development.

Selected associations: Consolidation, Immediate early genes, Late response genes, Persistence

0 0

Post a comment