Transient cortical circuits match spontaneous and sensory driven activity during development

Abstract

At earliest developmental stages, spontaneous activity synchronizes local and large-scale cortical networks. These networks form the functional template for the establishment of global thalamocortical networks and cortical architecture. The earliest connections are established autonomously. However, activity from the sensory periphery reshapes these circuits as soon as afferents reach the cortex. The early generated, largely transient neurons of the subplate play a key role in integrating spontaneous and sensory driven activity. Early pathological conditions, such as hypoxia, inflammation, or exposure to pharmacological compounds, alter spontaneous activity patterns, which subsequently induce disturbances in cortical network activity. This cortical dysfunction may lead to local and global miswiring and at later stages can be associated with neurological and psychiatric conditions.

Graphical Abstract

Early spontaneous synchronized neuronal activity sculpts cortical architecture.

A. Schematic outlines of brain development from embryonic to adult. B and C: Prenatal cortical circuits are dominated by early generated, largely transient neurons in the subplate (SP) and marginal zone (MZ) before maturation of cortical plate (CP) neurons. Transformation of early subplate driven circuits to the adult-like six-layered cortex requires spontaneous synchronized burst activity (D) that also controls programmed cell death (apoptosis), arrangement of neurites and axons, and formation and awakening of synapses. Most subplate neurons disappear with development; a few survive in rodents as layer (L) 6b neurons or in primates as interstitial white matter cells (G). During pre- and early postnatal stages, pathophysiological conditions such as hypoxia-ischemia, drugs, infection or inflammation may alter spontaneous activity (E,F). These altered activity patterns may disturb subsequent developmental programs, including apoptosis (H). Surviving subplate neurons that persist in white matter/L6b may support altered circuits that could cause neurological or psychiatric disorders.

One Sentence Summary:

During early development, transient neuronal populations integrate spontaneous and externally-generated activity patterns to form mature cortical networks.

Background:

During early mammalian brain development, transient neurons and circuits form the scaffold for the development of neuronal networks. In the immature cerebral cortex, subplate neurons in the lower cortical layer and Cajal-Retzius cells in the marginal zone lay the foundations for cortical organization in horizontal layers and translaminar radial circuits (“cortical columns”). Patterns of spontaneous activity during early development synchronize local and large-scale cortical networks, which form the functional template for generation of cortical architecture and guide establishment of global thalamocortical and intracortical networks. These networks become established in an autonomous fashion before the arrival of signals from the sensory periphery and before the maturation of cortical circuits. The subplate which is a transient structure in located below the developing cortical plate orchestrates alignment of these autonomously established pathways by integrating spontaneous and sensory driven activity patterns during critical stages of early development.

Advances:

The subplate contains heterogeneous neuronal populations with distinct characteristics, such as origin, birthdate, neurotransmitters, receptor expression, morphology, projections, firing properties and their participation in unique intra- and extracortical connectivity. The transformation of this early subplate driven circuit to the adult-like cortex requires patterned spontaneous activity and depends on the awakening of silent synapses in the cortical plate when thalamic inputs are progressively integrated. Moreover, a subpopulation of the glutamatergic and GABAergic subplate neurons has widespread axonal projections that establish early large-scale networks. The early circuits are remodeled when Cajal-Retzius and subplate neurons largely disappear by programmed cell death. Both the programmed cell death and the remodeling of circuits may be also controlled by the transition from spontaneous synchronized burst to sensory driven activity.

Outlook:

Functional impairments of these transient circuits (that include both transient and more permanent cell types) have great clinical relevance. Genetic abnormalities or early pathological conditions such as in utero infection, inflammation, exposure to pharmacological compounds, or hypoxia-ischemia, induce functional disturbances in early microcircuits, which may lead to cortical miswiring at later stages and subsequent neurological and psychiatric conditions. A better understanding of the transition from early transient to permanent neuronal circuits will clarify mechanisms driving abnormal distribution and persistence of subplate neurons as interstitial white matter cells in pathophysiological conditions. Exploring the transition from transient to permanent circuits helps us to understand causal foundations of certain pharmaco-resistant epilepsies and psychiatric conditions and to consider novel therapeutic strategies to treat such disorders.

Introduction

In the adult brain, neuronal communication is mediated primarily through chemical synapses and neurons interact within a short time frame. The developing brain is not just a smaller version of the adult brain, but rather has different types of interactions between immature cells. These interactions are slower, not as well stereotyped and predicted, and rely more on spontaneous activity patterns than interactions in the adult brain. These early spontaneous activity patterns are mediated through transient neuronal networks that continue to exist during the gradual establishment of permanent networks. Transient alterations in activity during crucial developmental periods can lead to persistent changes in functional connectivity and therefore might underlie the manifestation of neurological and psychiatric conditions (1). Thus, fundamental knowledge on early steps of activity-dependent circuit formation has general biological, as well as practical clinical implications.

The development of neural circuits starts early in embryogenesis. The preplate is the first postmitotic cortical neuronal layer (1) (Fig. 1) and is split into marginal zone (MZ) and subplate (SP) by later-arriving cortical plate (CP) neurons. Then, unfolding genetic programs of neurogenesis and neuronal migration interact with various forms of neuronal communication to establish the mature circuitry.

Fig. 1. Cellular interactions in developing brain.

A: Cross-section of 18 gestational weeks human brain (2) (blue: subplate SP; pink: germinal zones (ventricular zone, VZ; subventricular zone, SVZ)). B: Schematic illustration of cellular cortical components present. Dividing radial glial progenitors in VZ are in contact with cerebrospinal fluid and receive endocrine signals, some through blood vessels (C). Immature neurons interact via paracrine (D) and autocrine (E) mechanisms, or couple into local networks via electrical (G) and chemical (H) synapses (F). These various forms of communications co-exist in the developing brain.

Before completion of neurogenesis and migration (Fig. 2A, B), neocortical areas display distinct spontaneous and sensory-driven activity patterns, which can influence production and release of growth factors, maintenance of gap junctions, regulate transmitter release, and guide the precise topography of developing projections (3). Overall-levels of activity can also contribute to cell survival, including of interneurons (4), Cajal-Retzius (CR) (5), and thalamic neurons.

Fig. 2. Corticogenesis and relationship with spontaneous activity patterns.

(A) Approximate time points of major developmental events in human neocortex at postconceptional weeks (pcw) and postnatal months (pnm) and in mouse cortex at embryonic (E) and postnatal (P) days (12). (B) Delta brush EEG activity and average Fast-Fourier transformation (FFT) spectrum evoked by hand movement (vertical bar) (13) in preterm human or following whisker stimulation in newborn rodent (vertical bar). Note similarity in spectrum. (C) Schematic illustration of developmental changes in spontaneous activity. (1) CR and subplate neurons (yellow and red bars, respectively) already discharge faster action potentials and at higher frequency than CP neurons (black bars). (2) Neurons electrically-coupled via gap junctions either generate local synchronized activity or propagating activity waves. (3) Discharges become faster and local networks discharge in synchronized bursts. Transient early-born neurons start to disappear during this phase. (4) Appearance of adult-like sparse desynchronized activity independent of transient neurons and circuits. (D) Subplate neurons are gap-junction-coupled when thalamocortical projections arrive. Thalamic fibers first establish synapses with subplate neurons before innervating L4 neurons. Subplate and L4 connections transiently co-exist to reinforce the more permanent thalamic projections that remain after subplate neurons lose their contact with thalamic projections, and also lose their contact to L4 themselves. Few L6b neurons survive to adulthood (14).

Developing neurons often show transient depolarizations, which can be transmitted to other neurons via gap junctions. Subsequently, neurons become capable of releasing various neurotransmitters, depending on the frequency and intensity of their firing. Subplate and CR neurons can also release trophic factors and in this way directly influence cellular targets (2, 6).

Given the multitude of neuronal communication pathways, we review progress in understanding the role of neuronal electrical activity in the earliest neuronal networks of the cerebral cortex. We focus on a critical developmental stage, when cortical networks containing transient neurons begin to interact with the emerging inputs from the sensory periphery, and on reciprocal connectivity between the thalamus and cortex, as well as on connections within the cerebral cortex.

Spontaneous activity in early development

Spontaneous electrical activity, electrical events occurring without apparent external generation, is a general feature of all developing networks, but the cellular mechanisms underlying the various activity patterns may differ and change profoundly during specific stages of development (Fig. 2A) (7–9). Although the underlying mechanisms and functional role of the various activity patterns have been mostly studied in animals, data from preterm and newborn human babies indicate a similar sequence. At distinct developmental periods, spontaneous activity influences or even controls neurogenesis and neuronal migration, synaptogenesis, apoptosis, and myelination (10, 11).

In general, the sequence of spontaneous activity in developing neuronal networks shows four distinct phases (Fig. 2C):

(1) At early embryonic stages (1 in Fig. 2C), spontaneous activity is sparse and asynchronous, because neurons are not connected yet. Neurons in isolation reveal long calcium transients that appear stochastically and are influenced by ionotropic and metabotropic receptors activated by ambient or paracrine release of GABA and glutamate. Subplate and CR neurons already reveal relatively mature electrophysiological properties, e.g. faster action potentials and higher discharge rates. Thalamic stimulation elicits long responses in rodent SP, which likely represent action potentials followed by prolonged depolarizations (Fig. 3A)(15). Subplate neurons in pcw17-23 fetal human cortex in vitro generate spontaneous depolarizations that depend on gap junctional coupling (16)

Fig. 3. Transient circuit topologies during thalamocortical development.

A. Evoked responses in subplate and L4 after stimulation of the optic radiation in cat (45), of thalamus in mouse thalamocortical slices (15), and in vivo in ferret (39). Responses emerge and latencies are always shortest in SP.

B. Topography is emerging in the subplate. Plotted is the difference in correlation of tuning curves between neighboring recording sites. Early-evoked responses in subplate show differencing responses at larger distances (39).

C: Left: The integration of subplate neurites shows an age-specific pattern (57). Middle: Subplate axons target the septa in S1 barrel cortex (52). Right: Ablating a row of whiskers at birth changed the distribution of the corresponding neurites (57).

D: Various transient connections only present during specific stages of development and not in the adult (34, 40, 59).

(2) In the next stage (2 in Fig. 2C), bursts of correlated neuronal activity are generated by intrinsic mechanisms or synaptic interactions including extrasynaptic receptors activated by ambient GABA or glutamate. Active neurons are coupled in small networks via electrical and/or chemical synapses and generate synchronized burst activity, which may be restricted to a local neuronal network, propagates as a “wave” to neighboring cortical areas or activates remote cortical and subcortical regions via axonal projections (17). Highly synchronized spontaneous activity in the form of calcium waves emerges as early as E14.5 in mouse thalamus and propagates among sensory thalamic relay nuclei, thereby coordinating patterning of the cortical sensory area (18). At this age, the brainstem has not yet innervated the thalamus, indicating that this activity is generated within the thalamus. Thalamocortical systems are already assembled when the first peripheral inputs reach them. Therefore, initial thalamocortical loop formation occurs in an autonomous fashion, and periphery-related activity from sensory organs can subsequently modify these connections (19).

(3) With maturation of intrinsic and synaptic electrophysiological properties, neurons fire in bursts and synchronize local networks (3 in Fig. 2C). Spindle burst or delta brush activity is a physiological hallmark of this distinct period in human and rodent development (Fig. 2B). Spindle bursts are triggered by signals from the sensory periphery (20), endogenous thalamic activity (21), or by activity from other neocortical areas. Short gamma bursts are present in rodent cortex during the same developmental period and also synchronize local radial neuronal networks, functional pre-columns (9, 22, 23). Subplate neurons are required for this local burst activity and subplate ablation causes disarrangement in the cortical architecture (24). At embryonic and neonatal stage, this activity originates in the thalamus and controls the formation of the cortical map (21). The cholinergic system elicits some of these early activity patterns (7, 25, 26).

Cortical early network oscillations (cENOs), developmentally followed by cortical giant depolarizing potentials (cGDPs) represent the two activity patterns in newborn rodents during further development (9, 27). Endogenous activity in the somatosensory system often arises from central pattern generators in motor regions (8). However, because the somatosensory system is tightly linked with the motor system, it is often not possible to clearly separate “true” spontaneous activity from sensory activity evoked by motor action. Twitch-related activity is present in the somatosensory cortex (28) shortly after birth, indicating that circuits from periphery to cortex are functional from early ages (29, 30).

Developing neurons do not oscillate at frequencies of spindle bursts and oscillations are abolished by manipulating thalamocortical circuits and subplate neurons, suggesting that oscillations are generated by specific circuits, rather than individually oscillating neurons (14, 24, 31–34). Death of neurons, changes in circuits or receptor composition, and changes in (intracellular) ion composition could all contribute to the developmental changes in oscillations. Moreover, since early activity patterns seem to be coordinated across the brain, the above considerations point to a key role of the thalamus in relaying, integrating, and sculpting early spontaneous and sensory driven activity patterns.

In sensory systems, the periphery (7, 35) as well as central sources might initially independently generate spontaneous activity (17), but their interaction is unclear. Spontaneous peripheral activity is relayed via the thalamus to the developing cortex (36). Thus, it is not just a transient subplate, but rather a series of transient networks and structures that all contribute to the changing spontaneous activity.

(4) With further maturation of intrinsic membrane and synaptic properties, spontaneous activity becomes sparse and desynchronized (9) (4 in Fig. 2C). This gradual developmental shift from the dominant burst pattern to “adult-like” low-amplitude desynchronized activity is accompanied by a progressively stronger impact of the peripheral sensory input. However, sensory cortices can be activated by external stimuli (light, sound, touch and muscle twitches) at surprisingly early stages, e.g. in preterm human babies or early postnatal periods in altricial animals (13, 37–40), before the sensory organs are fully functional. In ferrets, primary auditory cortex (A1) responds to sounds as early as P21 (39), while in mice, whisker stimulation elicits a cortical responses at E18.5 (21). During the first postnatal week in rodents, L4 neurons exhibit spontaneous activity in a barrel map-like pattern, which is driven by the periphery, but largely independent of self-generated whisker movements (29, 30, 41). This activity may be generated by gap junction coupled dorsal root ganglia or brainstem neurons. In the visual cortex of ferrets, long-range correlated spontaneous activity is generated through short-range interactions in the form of distributed coactive domains (42).

Instructive role of the subplate during development

In all three sensory systems, the thalamocortical projection is subserved by multiple subnuclei projecting to the respective cortical areas. Only primary thalamic nuclei receive direct input from the sensory periphery, while higher order thalamic nuclei receive their input from cortex and relay their output to other cortical areas providing pathways for transthalamic cortico-cortical communications. During development, higher order thalamic nuclei seem to provide most of the early projections to matched cortical areas (43) indicating that primary sensory areas are induced by sensory activity at later ages (44). The sensory systems differ with respect to timing, relative maturity, peripheral receptors and brainstem circuits. Moreover, the early developmental period is characterized by a high degree of multisensory projections to especially higher order thalamic nuclei (43). Thus, early activity in each sensory system can influence the activity across sensory cortical areas and deprivation can lead to cross-modal rewiring.

During development, the thalamocortical and corticothalamic systems undergo a dramatic transformation due to the presence of transient circuits. Thalamic projections accumulate closer to the cortex and corticothalamic projections accumulate outside the thalamus. Both assemble transient circuits with the subplate (14) (Fig. 2D) and the thalamic reticular nucleus (19). The period in which the thalamic connections are accumulating below the cortex (so-called waiting period) is a crucial developmental period characterized by highly dynamic and essential interactions between thalamus and cortex (15, 39, 45, 46). Subplate neurons are among the first cortical neurons to mature and receive thalamic inputs before these innervate their eventual target (L4) (45–49) (Fig. 3A). In the auditory system in vivo, subplate neurons can respond to sensory stimuli before L4 neurons respond to sound (39) (Fig. 3A) and an early topography of sensory responses exists in subplate (Fig. 3B). Thus, given these developmental dynamics, this early stage should be more appropriately designated as the “proto-organizational” period.

Subplate neurons differ in their origin, birthdate, molecular profile, and morphology as well as local and long-distance connectivity (33, 49–52), but the distinct roles of each subpopulation are unknown. Many subplate neurons provide excitatory input to L4 and other cortical layers including L1 (46, 52, 53) and thus form a relay of thalamic information to future thalamorecipient layers. Because of the excitatory nature of subplate neuron projections, these neurons have a possible instructive role in thalamocortical as well as intracortical connectivity (14, 54). Besides projecting into the developing cortical plate, subplate neurons also pioneer corticothalamic projections (14). Subsets of subplate neurons target higher order thalamic nuclei (51, 52). Subplate neurons also receive inputs from the developing cortical plate including L4 (33), as well as from both glutamatergic and GABAergic subplate neurons (33, 49). While at young ages subplate neurons receive inputs from L4 via NMDA receptor only connections (33, 54), intra-subplate and thalamocortical synapses on subplate neurons are not silent (55). Thus, at young ages there is an NMDA-R mediated feedback from the eventual thalamorecipient L4 to the currently thalamorecipient subplate neurons. The emerging connectivity diagram points to an integrative role of subplate neurons at this key stage of development (54) (Fig. 2D).

Besides subplate neurons, Cajal-Retzius neurons in the MZ/L1 represent another population of early generated and transient neurons (56). Cajal-Retzius neurons receive mainly GABAergic synaptic inputs of mostly unknown origin, but some arise from subplate. The axonal targets of Cajal-Retzius cells and their role in spontaneous activity are unclear.

The early circuits that are dominated by transient neuronal cell types coexist with circuits that will prevail in the adult for a period of time. Subsequently, the increasing influence of the sensory periphery will trigger the dismantling of the transient networks. Arrangement of subplate neurites is regulated by sensory input, since sensory deprivation delays the remodeling of subplate neurites (57, 58) (Fig. 3C).

Besides the largely transient subplate neurons, other transient connections exist within the cortical plate. For example, subsets of GABAergic L5A neurons receive thalamic inputs and transiently project to L4 (Fig. 3D); transient widespread connections exist from subgranular to supragranular layers (40)(Fig. 3D); and GABAergic connections transition from early-depolarizing to hyperpolarizing action (60, 61).

Instructive role of subplate in cortico-thalamo-cortical circuits

The corticothalamic projections start to extend towards the internal capsule immediately after the generation of the first preplate cells (Fig. 4A). Subplate projections cross the pallial-subpallial boundary simultaneously with the thalamic projections and they co-fasciculate, providing vital scaffolds for thalamocortical ingrowth (62). Removal or repositioning of subplate neurons at this age prevents thalamic innervation of cortex (14). Subplate neurons pioneer the outgrowth of corticothalamic projections (14), and corticothalamic projections accumulate outside the thalamus before ingrowth (63).

Fig. 4. Establishment and plasticity of thalamo-cortical-thalamic circuits.

A. Development of thalamocortical connectivity in mouse (E15, P4 and P56). Thalamocortical projections cross the pallial-subpallial boundary simultaneously with corticothalamic projections, co-fasciculate, providing mutual guidance, and accumulate in subplate or outside the thalamus, possibly in TRN. L5 projections give side-branches selectively to higher order thalamic nuclei. L6 projections innervate both nuclei.

B. Selective innervation of first and higher order thalamic nuclei. White circles illustrate dLGN and VB. L5 and L6b projections selectively innervate higher order thalamic nuclei, whereas L6a lack such preference (51, 63, 66). L5 terminals from S1 to PO are larger than L6a or L6b terminals from S1 to VB (66).

C. Reciprocal thalamocortical connectivity during development and in adult.

D. After sensory loss, cortical and thalamic connectivity in primary (1) and secondary (2) cortical area is changed.

At later ages, thalamocortical projections show anatomical rearrangements, e.g. into ocular dominance columns in V1 or barrels in rodent S1. Subplate neurites (57) and axons (52) also show patterned projections (Fig. 3C). Selective subplate lesions have shown that subplate neurons are required for the emergence of patterned thalamocortical projections and for functional maturation and plasticity of thalamocortical and potentially intracortical circuits (24, 64, 65). The role of subplate neurons outside primary sensory areas has been enigmatic.

Subplate projections to higher order thalamic nuclei

Besides receiving ascending inputs from the sensory periphery, the thalamus is target of cortical inputs. The thalamus comprises of two functionally distinct components; the first and higher order thalamic nuclei (67). Only the first order nuclei (e.g. VB) receive direct input from the sensory periphery, whereas higher order nuclei (e.g. PO) receive most inputs from cortex (Fig. 4A–C). Thalamocortical circuits have been studied most extensively in sensory cortices, but higher order thalamic nuclei have received increasing attention because of their presumed involvement in higher cognitive functions. First-order thalamic projections target L4, and provide inputs to L6a. In contrast, higher-order thalamic inputs target L5 and L1, potentially providing feedback modulation and integration with multisensory processing in L1.

First and higher order thalamic nuclei have very different relationships to TRN, since only first order thalamic nuclei give collaterals to this thin sheet of GABAergic neurons (Fig. 4C). They also have different relationships to the corticothalamic projections from different origins. In adult, the corticothalamic inputs to first order thalamic nuclei arise from L6 neurons, that target both types of thalamic nuclei as well as the TRN (Fig.4B). L5 neurons only send projections to higher order thalamic nuclei and do not innervate TRN (51). The corticothalamic pathway is pioneered by subplate neurons (14), and specific populations of subplate neurons provide input to higher order thalamic nuclei (51, 52). However, L5 projections enter thalamus first (68) and provide strong input to higher order nuclei. Retinal ablation at birth elicited L5 innervation of first order dorsal lateral geniculate nucleus (LGN), suggesting that early peripheral activity can regulate corticothalamic innervation (63)(Fig. 4D).

Subplate neurons are on the nexus between thalamocortical and corticothalamic loops, thus their activity can change the function of these pathways and act like a gate (Fig. 4). Indeed, subplate neurons are targets of multiple neuromodulators (14) and subplate remnant L6b neurons are modulated by neurotensin and orexin, which is wake promoting, consistent with a gating role (69). Subplate neurons share many similarities with claustral neurons, suggesting a functional homology (see Box).

BOX: Subplate and claustrum share similarities: control of large-scale networks?

The claustrum has the strongest connectivity in the adult human brain, links all cortical areas, and is associated with higher cognitive functions (70, 71).

Claustrum shares extensive similarities with the subplate. Their principal neurons are amongst the earliest born and many markers expressed in the subplate/L6b are also expressed in the claustrum (72). Adult L6b neurons also link distant cortical areas (73). We hypothesize that subplate and claustrum perform similar functions as a key nexus between corticocortical, thalamocortical and corticothalamic loops. Subplate neurons might function as developmental “operational” hub cells, similar to those of the developing hippocampus (74). Claustrum may perform comparable roles in the adult.

What does early activity do?

Neural activity regulates functional maturation of cells and circuits, by driving the expression of ion channels and receptors or by changing morphology. Activity can also change cell identity by driving genetic programs, likely by adjusting final differentiation (75).

Higher order thalamus enables cross-modal connections (43), i.e. while manipulating peripheral spontaneous activity can alter ascending circuits in the matched system (7, 76, 77), effects can also be widespread. Cross-modal thalamocortical plasticity occurs when a sensory organ changes its input and another sensory modality takes over these pathways on thalamic level (Fig. 4). In addition, cross-hierarchical corticothalamic plasticity exists, in which first order thalamic nucleus adopts a ‘higher order thalamic phenotype’ by receiving input from L5 (63, 78). These two forms of plasticity might occur in parallel when sensory driven activities are altered. Changes after peripheral manipulations are typically interpreted in the framework of thalamocortical competition, but developmental and cross-hierarchical changes have to be considered as well.

Without sensory input, thalamocortical circuits remain in a status resembling higher order circuits (44). We thus hypothesize that higher order thalamic nuclei are the early endogenous pattern generators in thalamocortical systems. Acquiring characteristics of higher order thalamic nuclei might be the default developmental pathway when there is no sensory peripheral input. This default endogenous thalamocortical activity is suppressed together with the default differentiation path in first order thalamic nuclei with normal spontaneous activity and normal sensory inputs (Fig. 4D). Thus, a specific pattern of spontaneous synchronized activity at a distinct time point may activate or inactivate a developmental program in a subset of immature neurons. We started to appreciate the variety and complexity of spontaneous activity patterns, and the precise roles of each pattern in specific neurons at specific developmental stages are unknown.

Regulation of subplate neuronal death during development

Spontaneous activity also plays a role in controlling the number of surviving versus dying neurons. GABAergic neurons do not only control the pattern of synchronized spontaneous activity and the emergence of functional network topography in developing cortex, but also control apoptosis of interneurons (79, 80). Blockade of electrical activity for 6 hours in vitro doubles the number of apoptotic neurons (81). In contrast, synchronized network activity resembling physiological spindle bursts and gamma oscillations promotes neuronal survival and reduces apoptosis. This pro-survival effect is mediated by brain-derived neurotrophic factor (BDNF), which is released in response to synchronized burst discharges at 20 to 50 Hz (82). Thus, cell survival in the developing cortex is controlled by distinct patterns and not by the level of activity. Synchronized spindle bursts and gamma oscillations may be the adequate activity patterns to control the fate of cortical neurons at a developmental period, when apoptosis coincides with the expression of these patterns (Fig. 2A–C). Beside pattern specificity, activity-dependent apoptosis may be cell dependent. Cajal-Retzius and subplate neurons both highly express neurotrophin receptor p75 (p75NTR), but this activates a pathway that induces death in Cajal-Retzius neurons (83), but survival in subplate neurons (84), suggesting that the same activity pattern may activate different intracellular pathways.

The normal integration, remodeling, and eventual death of subplate and also MZ neurons is controlled by spontaneous activity and possibly early sensory activity (39, 57). Alterations in cortical activity patterns therefore could alter the distribution and number of surviving neurons. Understanding the dendritic maturation and programmed cell death of these transient neurons is key to the understanding of the subtle anatomical changes observed in the number and distribution of interstitial white matter and MZ cells in some neurological and psychiatric disorders (85). Damage of early generated subplate neurons has been implicated in hypoxia-ischemia, autism, and epilepsy (86–88). The link between subplate gene expression patterns and cell numbers and distribution alterations in disorders is supported by several studies (50, 89).

Disturbances in spontaneous activity patterns during distinct stages of early development will interfere with apoptosis programs. Alcohol and general anesthetics modify spindle bursts and gamma oscillations in newborn rats and can induce widespread cortical cell death (90).

Clinical relevance

The spread of spontaneous activity in the thalamocortical network and cerebral cortex is broader in the human preterm as compared to the term infant (Fig. 5A). Wide-spread cortical activation in the preterm is also evident from resting state EEG connectivity in the 8 to 15 Hz frequency band (spindle burst) (Fig. 5B). Premature babies show the full repertoire of resting state dynamics emerging during the period of rapid neural growth before term (~40 pcw) (91). Sensory, motor, default mode, frontoparietal, and executive control networks develop at different rates, suggesting that they are formed before the sensory periphery is fully functional and before acquisition of cognitive competencies takes place in later childhood. At the early stages, subplate is fully integrated into cortical circuits and may influence resting state networks before term birth. Indeed, resting state network activities in extremely immature human cerebral cortex are mostly restricted to lower cortical layers (91). We postulate that early transient circuits form the basis for activity patterns in the preterm and that the pathophysiological persistence of these circuits is involved in the manifestation of neurological and psychiatric disorders (Fig. 5C). All these changes could occur according to a default timetable that is adjusted by sensory activity, general embryonic maturation, and various environmental factors, e.g. nutrition, inflammation, maternal stress etc.

Fig. 5. Spontaneous activity in preterm and adult schizophrenics.

A, resting state fMRI (92). B, resting state EEG connectivity matrix at 8–15 Hz, showing stronger connectivity (red and yellow) in preterm (93). C, resting state fMRI data showing significant differences in thalamic connectivity between healthy (CON) and individuals with schizophrenia (SCZ) (94).

We argue that spontaneous activity patterns are mediated through higher and first order thalamocortical systems and that the two pathways have to interact to produce the normal thalamo-cortico-thalamic circuits for the emergence of cognitive functions (Fig. 4). The two systems have different maturation timelines, and process different aspects of sensory information and brain states. These spatiotemporal features have to be linked and we speculate that unlinking or failure to link may cause cognitive disorders. Thus, a key developmental event is to keep the first order and higher order thalamic loops linked and aligned with the intracortical connections. Since subplate neurons connect to each other over long distances (33) and also project to the thalamus, they might form a network linking primary and higher order areas and with the dissolution of subplate this link is abolished.

The transient circuits are vulnerable and the sensitive periods for disorders overlap with times when circuits process largely spontaneous activity patterns (Fig. 2). These circuits are vulnerable to hypoxia-ischemia (87, 95) and pharmacological manipulations, e.g. prenatal valproate exposure (86) and disrupting them will have implications on the development of the thalamo-cortico-thalamic networks. Large numbers of WM and deep interstitial neurons without disturbances of the normal cortical layering are pathological characteristics seen in temporal lobe epilepsy (TLE). Neocortical surgical specimens of patients suffering from pharmaco-resistant TLE demonstrated that these deep neurons expressed transcription factors that are highly expressed in subplate (96–98). Subplate-like neurons have also been identified in pediatric epilepsy patients (99), further supporting the hypothesis that subplate neurons may survive and may be causal to TLE.

A better understanding of the development of transient circuits, their roles in brain wiring, higher brain functions and the consequences of retained white matter networks in cognitive conditions is needed to design new diagnostic and therapeutic approaches in neurodevelopmental disorders. However, investigations of changes in various disorders at adult stages can only give us limited clues on abnormal developmental processes. The interstitial white matter cells in these pathologies and miswiring are only the remnants of or consequence of malfunction of a transient neuronal population. Specific biomarkers are needed to detect and correct malfunction of transient populations at early ages

Open questions / perspectives

1. Do all spontaneous activity patterns fulfill a particular functional role in development or are some of them just epiphenomena?

2. What activity pattern is normal and what is abnormal? The presence or absence of synchrony and coordinated development between first and higher order circuits might not manifest clinically during development when temporal processing is imprecise and lead to, on the surface, normal locally stable brain function. At later ages, if the brain is challenged by more rapid changes, e.g. environmental insults, puberty etc., such deficits could be unmasked.

3. How do different spontaneous activity patterns interact (e.g. local burst activity vs propagating wave; periphery-driven vs cortical)?

   Many neural structures are capable of producing spontaneous activity. Non-linear interactions might mediate the integration and transition between activity patterns due to the presence of silent synapses in subplate.

4. Are higher order thalamic nuclei the early endogenous pattern generators in the thalamocortical system? We speculate that first order nuclei suppress the default endogenous thalamocortical activity based on emerging peripheral inputs and thereby induce a different transcriptional profile in targeted cortical neurons.

5. Enlarged subplate and higher order thalamic circuits are a key feature of the primate brain.

   The human subplate can be five times larger than the cortical plate, indicating a primate specialization. Since key elements that determine projection neuron identity are shared between rodent and primates, are there subplate neurons that are only present in humans and did subplate and higher-order nuclei co-evolve?

6. Does a selective vulnerability of subgroups of subplate neurons contribute to distinct pathologies?

   We speculate that subtypes of subplate neurons mediate lemniscal, paralemniscal, and non-lemniscal development and that subplate circuits differ between primary and higher order sensory cortical areas. Would alterations in different subplate circuits be the tipping point to enter distinct pathological developmental trajectories? Are some diseases higher order thalamic diseases and others due to network asynchrony?

7. The role of persistent subplate/Cajal-Retzius neurons in cognitive disorders.

   Subplate neurons can synapse on migrating cortical plate neurons (100) and might alter neuronal differentiation. Therefore, early changes in subplate function could influence the delivery, positioning and differentiation of cortical plate neurons, causing their accumulation in the subplate contributing to abnormal connections.