Re-thinking the brain

new insights into early experience and brain development

The brain is a self-organizing system that adapts to its specific environment throughout pre- and postnatal life (Braun and Bock, 2011). Self-organization refers to the spontaneous formation of patterns and pattern change in open non-equilibrium systems. Edelman’s theory of neuronal group selection (Edelman, 1989) highlights this process. Groups of neurons are ‘selected’ or organized into groups or networks that are dynamically organized through epigenetic factors and experience. Developmental selection occurs largely before birth. Processes such as cell division, differentiation and programmed cell death and the mechanisms of neuronal migration are regulated by epigenetic factors. While genetics provides a general blueprint for neural development, the developmental processes are not precisely pre-specified by genes, and produce unique patterns of neurons and neuronal groups in every brain. The result is a diverse pattern of connectivity forming primary repertoires of different neuronal groups. Structural diversity occurs through selective mechanical and chemical events regulated by cell and substrate adhesion molecules. A second process called experiential selection occurs postnatally through behavioural experience, resulting in modifications in the strength of synaptic connections, and creating diverse secondary repertoires. Finally, re-entrant signalling leads to the development of dynamic ‘maps’, an interconnected series of neuronal groups that independently receive inputs from the real world and create coherent perceptual constructs.

Early development

After the initial proliferation of precursor cells and differentiation of neurons, the cells migrate from the site of origin to their final location where they extend neurites and grow in size. Synapses begin to form around this time (refer to Shepherd, 1994, for review). Maturation of the neurons into their final form and function is a process that may take several decades (de Graaf-Peters and Hadders-Algra, 2006). The development of neural connections in many regions of the central nervous system is characterized by an initial overproduction of neurons, axons and synapses, followed at certain periods by the selective elimination or pruning of excess numbers of these structures (see Purves and Lichtman, 1985, for review). A wave of programmed neuronal cell death occurs very early in the prenatal period during which up to 50% of neurons initially formed will die (Oppenheim, 1981). This process is a means of regulating cell numbers to match the capacity of the target structures, and corrects for some errors in positioning. A second wave of cell death occurs in postmitotic neurons and plays a critical role in shaping the nervous system. Cells that survive appear to be successful in competing for growth factors (see Lossi and Merighi, 2003, for review).

Constant trophic input is essential for appropriate nervous system function. Trophic support is provided by the target cells (other neurons, muscle) as well as from local glial cells. However there is also evidence that blood-borne hormonal-like growth factors may also be important. These include insulin-like growth factor-I (IGF-I), fibroblast growth factor-2 (FGF-2) or the neurotrophins (Torres-Aleman, 2000). The implications of these effects will be discussed in a later section.

There has been a prevailing view that neuronal connectivity is more diffuse during early stages of development and that the final shaping of the adult-like pattern of connectivity is achieved by activity-dependent processes that prune non-relevant connections during critical periods (Katz and Shatz, 1996). This view relied heavily on research describing the postnatal development of the ocular dominance columns in visual cortex (Le Vay et al., 1978, reviewed below). In contrast, there is considerable accuracy of connections early during development, as soon as axons first reach their target structures. This selective synapse formation is controlled by specific molecular cues, for example, the patterning of thalamocortical projections by ephrin–A5 (Vanderhaeghen and Polleux, 2004), the control of the laminar origin of the corticospinal tract by Fezf2 (Han et al., 2011) and differential patterns of Lmo4 in sensory neurons (Chen et al., 2005), that influence innervation of the homonymous spinal motoneurons by Ia afferents in the absence of activity (Frank and Jackson, 1986).

It has long been thought that activity-dependent mechanisms during critical periods in the postnatal phase were key factors in the development of motor and sensory systems. A classic example is the development of the ocular dominance system in the visual cortex. The visual system has been a frequently investigated model of sensory systems. Neurons in lamina IV of the primary visual cortex (V1) are segregated into columns dominated alternately by the right or left eye (Hubel et al., 1977). At birth there is considerable overlap of the afferents from the lateral geniculate nucleus (LGN) in V1 from each eye (Hubel et al., 1977), so that many of the neurons in lamina IV are driven by inputs from both eyes. The segregation of these afferents into alternating bands occurs gradually by rearrangement of synapses and is dependent on visual experience. Occlusion of vision of one eye for the first three months (the critical period) leads to a change in the effectiveness of the occluded eye to drive neurons in V1, with a corresponding reduction in the size of the column for the occluded eye, and an expansion of the column for the non-occluded eye.

Newer investigative methods have led to a clarification about the development of ocular dominance columns. It is now realized that development involves two separate phases: an establishment phase, and an activity-dependent remodelling phase (Crowley and Katz, 2002). It appears that the initial establishment of the columns occurs considerably earlier than the critical period. In macaque monkeys, it has been shown that the projections carrying input from both eyes become segregated during the second half of gestation and become partially separated in the cortex about three weeks before birth, i.e., before visual experience (Horton and Hocking, 1996; Rakic, 1976). Development of the ocular dominance columns is relatively rapid and precise, occurring before the cortex responds to visual stimulation, i.e., before the onset of what has been described as the critical period (Crowley and Katz, 2000). Thus Edelman’s primary repertoires are clearly complex and detailed patterns of connectivity. The ocular dominance columns do not develop independently, but in relation to other aspects of visual cortex organization, such as retinotopic maps that arise in part from connections within the LGN, and which are related to the pattern of ocular dominance columns (Le Vay et al., 1985). It is highly probable, therefore, that molecular guidance mechanisms, which play an important role in the establishment of topographical connections within the brain, also affect the development of the ocular dominance columns. This does not negate the effect of visual experience, but rather highlights the fact that a considerable degree of specificity occurs well before the postnatal period, enabled through internally generated spontaneous activity (Katz and Shatz, 1996).

The development of the somatosensory system follows a similar template. The animal model often used is the development of the barrel fields in rodents. Barrels are the largest and probably the most behaviourally important sensory areas in the rodent neocortex. Information from the large facial whiskers arrayed on the snout is transferred via the trigeminal nerve to the ventrobasal nucleus of the thalamus. Thalamic afferents then project to layer IV of the somatosensory cortex to form the distinct barrel pattern observed in the barrel map (Inan and Crair, 2007). Gradients of specific molecules, including ephrin-A5 and EphA4, regulate the guidance of thalamocortical axons into their appropriate cortical areas (Dufour et al., 2003). Neuronal activity is required for topographical specification and refinement of these projections (Jensen and Killackey, 1987).

Thalamocortical projections

The thalamus provides the passage of entry for all sensory information to the cerebral cortex, as well as for all correlated and processed information from the spinal cord, cerebellum, and basal ganglia, including the substantia nigra. The thalamus provides the cerebral cortex with the key information needed to represent both body space and extrapersonal space, i.e., the contextual information for all complex behaviour. A previously widely accepted model of the organization of the thalamocortical projections to the sensorimotor cortex was that each nucleus with its own unique subthalamic input in turn projects to a particular region of cortex. This model implied that the thalamus subserves a largely passive relay function. However, there is considerable territorial convergence of input from two or more thalamic nuclei to each localized zone of cortex. Each thalamic nucleus, specified by its cytoarchitecture and subthalamic input, projects to a quite extensive area of sensorimotor cortex which overlaps with the projections of adjacent thalamic nuclei (Darian-Smith et al., 1996).

There is evidence that early topographic maps between and within individual cortical areas are generated prenatally through the interaction of the guidance cues ephrin-A5 and EphA receptors. Ephrin-A5 in the cortex acts as a graded repulsive cue for thalamocortical axons expressing graded levels of EphA receptors (including EphA4) to generate a topographic somatosensory map (Dufour et al., 2003; Vanderhaeghen and Polleux, 2004).

Subplate neurons play an important role in the development of the thalamocortical projections. The subplate represents a transient layer in the developing cerebral cortex. It comprises a heterogeneous set of neurons located directly under the cortical plate. Subplate neurons have extensive dendritic arborization and widespread axonal projections, and they have substantial glutamatergic or GABAergic synaptic inputs from thalamic, intra-subplate and cortical plate sources. They synchronize neuronal activity by receiving incoming extrinsic and intrinsic signals and distributing them throughout the developing cortical plate (Kanold and Luhmann, 2010). They play a critical role in regulating the maturation of cortical inhibition, as well as in the organization of the functional columns that are the hallmark of all cortical architecture, especially in sensory areas (Kanold and Luhmann, 2010). Subplate neurons are prone to hypoxic injury in the perinatal period (McQuillen and Ferriero, 2005), and their loss will affect the maturation of thalamocortical and other projections. For example, damage of the subplate in visual cortex prevents the maturation of thalamocortical synapses, the maturation of inhibition in layer 4, the development of orientation selective responses and the formation of ocular dominance columns. Subplate removal also alters ocular dominance plasticity during the critical period. Damage to the subplate in other cortical areas is likely to have similar effects, though this has not been investigated fully. The consequences of such damage have been highlighted by modern imaging methods. For example, diffusion tensor imaging has identified a reduction in the posterior thalamocortical tracts associated with periventricular leukomalacia in children with spastic quadriplegia who had normal corticospinal tracts (Hoon et al., 2002).

Corticospinal projections

Corticospinal projections are the only direct link between the sensorimotor cortex and the spinal cord. They comprise parallel, somatotopically organized projections to each level of the spinal cord with unique, though overlapping, patterns of termination. In addition to a dense projection from the motor cortex, corticospinal fibres arise from the premotor cortex, the postcentral cortex, especially the posterior parietal areas, the second somatosensory area and the caudal part of the insula. On the medial surface, there are extensive projections to the spinal cord from the supplementary motor area and the cortex within the cingulate sulcus (Dum and Strick, 1991; Galea and Darian-Smith, 1994). Corticospinal neuron populations thus transmit a complex orchestrated output from a number of different regions of the cerebral cortex to the neuron populations of every segment of the spinal cord.

During development, the first pioneering axons to advance down the spinal cord are those that will innervate the lumbar segments, and these then are followed by a bulk of later arriving fasciculating corticospinal fibres projecting to upper cord segments (Stanfield, 1992). As is the case with other major neural structures, the early development of the corticospinal tract is regulated by transcription factors that play important roles on the specification of corticospinal neurons (Chen et al., 2005), and molecules such as EphA4, which are important for guidance of corticospinal axons to their targets in the spinal cord (Coonan et al., 2001; Dottori et al., 1998). Other molecules are the neural cell adhesion molecule L1, a deficit of which causes failure of decussation at the medulla (Cohen et al., 1998), and netrin-1, which mediates the path finding of corticospinal axons from the cortex to the internal capsule (Richards et al., 1997).

Studies in rodents (Stanfield and O’Leary, 1985), cats (Martin, 2005) and monkeys (Armand et al., 1997; Galea and Darian-Smith, 1995) have highlighted that, although the descent of corticospinal axons to the most caudal parts of the spinal cord occurs before birth, maturation of the corticospinal tract occurs largely postnatally. In monkeys, functional connections with motoneurons which are critical for individuated finger movements are not established until manipulative skills emerge several months after birth (Olivier et al., 1997). Certainly, very few corticospinal terminals have extended into the dorsal and ventral horns at birth, and only by the eighth postnatal month is the terminal arborization comparable with that observed in the mature animal (Galea and Darian-Smith, 1995; Kuypers, 1962). Concurrently there is a threefold reduction in the number of neurons projecting to the spinal cord (Galea and Darian-Smith, 1994, 1995) and a comparable regression of thalamocortical projections to the sensorimotor cortex (Darian-Smith et al., 1990a, 1990b). The spatial and temporal correspondence in the postnatal regression occurring in both the corticospinal neuron populations and in their subcortical inputs from the thalamus highlights the probable interactions between the two populations via the subplate.

Unilateral lesion of the corticospinal system during early postnatal life in cats and rats can result in aberrant corticospinal organization, with maintenance of exuberant projections from the non-damaged side (Castro, 1985; Hicks and D’Amato, 1970; Leonard and Goldberger, 1987b; Leong and Lund, 1973). Martin and colleagues have conducted a series of experiments investigating the role of activity in shaping the developing corticospinal projections in the kitten. Blocking neural activity in the motor cortex by continuous infusion of a GABA_A agonist, muscimol, in postnatal weeks 3–7 resulted in sparse corticospinal terminations to the contralateral spinal cord from the silenced area of cortex, while those from the active cortex maintained an immature bilateral pattern (Martin et al., 2009). The animals showed significant errors in reaching and grasping using the limb contralateral to the infusion, and these behavioural deficits persisted despite further training and practice (Martin et al., 2000). Such infusions have also been found to affect limb locomotor activity and postural reflexes (Leonard and Goldberger, 1987a).

In subsequent experiments, limb use was blocked in kittens by injection of botulinum toxin A into several forearm muscles. As with inactivation of the motor cortex, preventing limb use prevents the maturation of the corticospinal axon terminals and results in the same aberrant pattern of contralateral projections (Martin et al., 2004). Moreover, once the effects of botulinum toxin wore off, the animals showed persistent impairments of grasping (Martin et al., 2004). Thus, activity-dependent refinement of corticospinal axon terminations during the early postnatal period is the mechanism whereby early motor experiences shape the structural and functional organization of the corticospinal system.

Experience and the development of target structures

Activity is not only required for neural development but is also important for the development of target structures. Neuronal activity can influence the maturation of the neuromuscular junction, the development of motoneurons within the spinal cord, and the pattern of outgrowth of motor axons.

Neuromuscular junction

During normal development, muscle fibres are often innervated by more than one motor axon. As postsynaptic development progresses, this polyinnervation is reduced to single innervation during the postnatal period, a process that is activity-dependent (Buffelli et al., 2003; Purves and Lichtman, 1985).

The formation of nerve–muscle contacts, intramuscular nerve branching, and neuronal survival require reciprocal signals from nerve and muscle to regulate the formation of synapses. Following the production of muscle fibres, clusters of acetylcholine receptors (AChRs) are concentrated in the central regions of the muscle fibres, pre-patterned independently of neuronal signals in developing muscle fibres. ACh released by branching motor nerves causes AChR-induced postsynaptic potentials and regulates the localization and stabilization of developing synaptic contacts. These ‘active’ contact sites may prevent AChRs clustering in non-contacted regions and prevent the establishment of additional contacts. A further neuronal factor, agrin, stabilizes the accumulation of AChR at synaptic sites (Witzemann, 2006).

Activity is important for the maintenance of the neuromuscular junction. Disuse, as in spinal cord injury (Burns et al., 2007) or ageing (Valdez et al., 2010), can result in structural changes in the neuromuscular synapse, including synaptic detachment. Children with cerebral palsy aged between 7 and 15 years have been reported to show an abnormal spread of AChRs beyond the confines of the neuromuscular junction, indicating a deficit in development of the neuromuscular synapse (Theroux et al., 2002). They also demonstrate abnormal structural changes, including non-apposition of the synaptic components, which were more numerous in more highly affected patients (Theroux et al., 2005).