Developmental alterations in the functional properties of excitatory neocortical synapses

Abstract

In the neocortex, most excitatory, glutamatergic synapses are established during the first 4–5 weeks after birth. During this period profound changes in the properties of synaptic transmission occur. Excitatory postsynaptic potentials (EPSPs) at immature synaptic connections are profoundly and progressively reduced in response to moderate to high frequency (5–100 Hz) stimulation. With maturation, this frequency-dependent depression becomes progressively weaker and may eventually transform into a weak to moderate EPSP facilitation. In parallel to changes in the short-term plasticity, a reduction in the synaptic reliability occurs at most glutamatergic neocortical synapses: immature synapses show a high probability of neurotransmitter release as indicated by their low failure rate and small EPSP amplitude variation. This high reliability is reduced in mature synapses, which show considerably higher failure rates and more variable EPSP amplitudes. During early neocortical development synaptic vesicle pools are not yet fully differentiated and their replenishment may be slow, thus resulting in EPSP amplitude depression. The decrease in the probability of neurotransmitter release may be the result of an altered Ca²⁺ control in the presynaptic terminal with a reduced Ca²⁺ influx and/or a higher Ca²⁺ buffering capacity. This may lead to a lower synaptic reliability and a weaker short-term synaptic depression with maturation.

Synaptic connections are highly dynamic with respect to their efficacy (i.e. the average unitary postsynaptic potential (PSP)) and reliability (i.e. the failure rate and variability of the PSP). The response of a postsynaptic neuron either increases or decreases in response to a presynaptic neuron firing action potentials at moderate rates to high rates (5–100 Hz). The strength of a synaptic connection is therefore not constant but shows a dynamic, firing rate-dependent ‘gain’ control of the synaptic response (Abbott et al. 1997). This so-called short-term synaptic plasticity develops within a time scale of milliseconds and recovers in less than a second. It has been attributed to presynaptic mechanisms such as the probability of Ca²⁺-dependent neurotransmitter release (P_r) and the neurotransmitter vesicle depletion in the axon terminals of the presynaptic neuron (Katz & Miledi, 1968; for reviews see Thomson, 2000; Zucker & Regehr, 2002). In studies on mouse and rat, short-term synaptic plasticity has been demonstrated to be target neuron specific, and thus dependent on the type of synaptic connection (Thomson, 1997; Markram et al. 1998; Reyes et al. 1998; Scanziani et al. 1998; Kozloski et al. 2001; Koester & Johnston, 2005; Watanabe et al. 2005; Helmstaedter et al. 2008). Furthermore, the history of previous activity is also a determinant of the short-term plasticity of a given synaptic connection (Markram & Tsodyks, 1996; Abbott et al. 1997; Tsodyks & Markram, 1997; Finnerty et al. 1999; Hardingham et al. 2007). In addition, changes in the intracortical concentration of neuromodulators such as acetylcholine, dopamine, endocannabinoids and adenosine, which are released during transitions from one behavioural state to another (suppression of REM sleep, wakefulness, arousal, high metabolic activity, etc.) can affect the efficacy and dynamics of synaptic connections (Gao et al. 2001; Seamans et al. 2001; Fontanez & Porter, 2006; Levy et al. 2006; Sjöström et al. 2007; Levy et al. 2008).

In summary, the functional properties of synapses are not fixed but are finely tuned in a context-dependent manner and may represent a flexible mechanism for temporal information processing in higher cortical integration.

Information about the structural and functional properties of many cortical and extracortical connections has been obtained mainly using dual or multiple intracellular recordings of synaptically connected neurons (for reviews see Silberberg et al. 2005; Watts & Thomson, 2005; Lübke & Feldmeyer, 2007). Many of these studies were performed on late postnatal animals (between 2 and 3 weeks of age in rodents). However, this period is still characterised by a substantial synaptogenesis and maturation as studies on rat and mouse neocortex have shown (Micheva & Beaulieu, 1996; De Felipe et al. 1997), and the formation and refinement of cortical connectivity ensues to a considerable degree even after this early postnatal stage (Bender et al. 2003; Bureau et al. 2004; for reviews see e.g. Cohen-Cory, 2002; Garner et al. 2006). In order to understand the flow of activity in columnar circuits in young and more mature cortex, it is therefore essential to study the properties of specific connections during the ongoing refinement of cortical circuitry. In the current review we will focus on the profound alterations in synaptic efficacy and short-term dynamics mainly of identified neocortical synaptic connections during early and late postnatal development, i.e. between the 2nd and 4th–5th postnatal week. As most studies have been performed on rodents and in particular on rats, only the use of other species will be identified in the text. The subspecies, brain regions and cell types used in the cited studies as well as their postnatal ages are shown in Supplemental Table 1.

Development of short term plasticity

The gain of neocortical synaptic connections, i.e. the amplitude of the postsynaptic signal, is adjusted according to its recent history of activation, but this gain control differs between immature and adult synapses. In the early postnatal neocortex, the transmission of trains of action potentials is rather ineffective. Firing activity as low as 10 Hz in a presynaptic neuron leads to substantial depression of unitary EPSPs in the postsynaptic neuron, and thus only the first action potential in the train may lead to a superthreshold response and is likely to be propagated. A strong paired-pulse depression is a feature of many synaptic connections in immature neocortex as has been demonstrated for several excitatory connections, including intralaminar connections between excitatory neurons in layers 2/3, 4, 5A and 5B (Reyes & Sakmann, 1999; Zhang, 2004; Frick et al. 2007; Oswald & Reyes, 2008; Radnikow, Lübke & Feldmeyer, in preparation; see Fig. 1A), translaminar connections between layer 2/3 and 5B pyramidal neurons (Reyes & Sakmann, 1999) and callosal–layer 5 pyramidal cell connections (Kumar & Huguenard, 2001). However, strong paired-pulse depression is not only characteristic for synaptic connections between excitatory neurons in immature neocortex but has also been reported for layer 5B pyramidal cell–‘fast-spiking’ interneuron connections (Angulo et al. 1999) and for cortico-striatal connections (Choi & Lovinger, 1997). Furthermore, synaptic connections in other brain regions outside the neocortex such as those between cerebellar stellate interneurons and Purkinje cells (Pouzat & Hestrin, 1997) and the calyx of Held in the brainstem (Taschenberger & von Gersdorff, 2000; Iwasaki & Takahashi, 2001) also show strongly depressing synaptic signals in the immature brain.

Figure 1. Developmental alterations in the short-term plasticity and reliability of the intracortical synaptic connections between layer 5A pyramidal neurons.

Aa, recordings from synaptically coupled layer 5A pyramidal neurons at postnatal day (P) 14 and P29 showing trains of three presynaptic action potentials at 10 Hz (top traces) and six consecutive traces of associated EPSPs (grey middle traces). The bottom traces show the average EPSPs for each connection. In this example, the immature (P14) connection showed strong short-term depression whilst the more mature connection (P29) displayed paired pulse facilitation, an observation made in ∼40% of connections of at this age. Ab, summary graph showing the change the paired pulse behaviour with age. Three age groups (P14–P15, P18–P20, P25–P29) are shown; circles are paired pulse ratios of individual connections, are averages paired pulse ratios of the three age groups. The red line indicates the average paired pulse ratio for immature synaptic connections, the blue line that for more mature connections. Note that whilst almost all connections at P14–P15 are depressing, those at older ages (P25–P29) vary between weakly depressing and facilitating. Ba, recordings from synaptically coupled layer 5A pyramidal neurons at P15 and P28. Seven consecutive postsynaptic responses (grey traces, middle row) to action potentials evoked in the presynaptic neuron (top row) are shown. While the response amplitude is stable for the P15 connection, the amplitude is more variable in the mature connections and occasionally failures can be observed in P28 connections. The bottom row gives the average EPSP for each connection, which was generally smaller for more mature connections. Bb, summary graph showing the developmental alteration of the coefficient of EPSP amplitude variation (CV). Three age groups (P14–P16, P18–P20, P25–P29) are shown; circles are CVs of individual connections, diamond average CVs of the three age groups. The red line indicates the average CV for immature synaptic connections, the blue line that for more mature connections. With maturation, the CV increases, which suggests a reduction in the probability of transmitter release at this synapse. Figure modified from Frick et al. 2007.

With maturation the short-term dynamics in all of these strongly depressing synaptic connections transform into either weakly depressing (Taschenberger & von Gersdorff, 2000; Iwasaki & Takahashi, 2001; Oswald & Reyes, 2008; Radnikow, Lübke & Feldmeyer, in preparation) or even facilitating ones (Pouzat & Hestrin, 1997; Angulo et al. 1999; Reyes & Sakmann, 1999; Kumar & Huguenard, 2001; Zhang, 2004; Frick et al. 2007) (Fig. 1A). In the neocortex this developmental regulation of the short-term synaptic plasticity (Fig. 1Ab; indicated by the blue and red dashed lines) occurs between the 2nd and 4th–5th postnatal week. In other brain regions similar changes take place but sometimes within different time windows. Some of the more mature neocortical connections show a more variable short-term plasticity, ranging from depression to strong facilitation (Angulo et al. 1999; Frick et al. 2007), which might permit a more dynamic fine-tuning of the PSP amplitudes to the preceding activity of the neuronal network.

In the hippocampus, the synaptic connection between CA3 and CA1 pyramidal neurons, which is weakly facilitating in the immature brain, displays a similar tendency and develops into a stronger and more robust facilitation in older animals (Bolshakov & Siegelbaum, 1995; Wasling et al. 2004). These changes have generally been interpreted as resulting – at least in part – from a decrease in the probability of neurotransmitter release from early to late development (Bolshakov & Siegelbaum, 1995; Choi & Lovinger, 1997; Taschenberger & von Gersdorff, 2000; Iwasaki & Takahashi, 2001; Frick et al. 2007).

Changes in short-term plasticity from strong to weak depression or even facilitation with ongoing maturation may not be a feature of all synaptic connections in the neocortex. Developmental changes from facilitation in the immature to depression in the mature neocortex have been reported for excitatory inputs (activated by white matter stimulation) onto neurons in rat and ferret visual cortex (Ramoa & Sur, 1996) and for rat cortical neurons targeted by geniculo-cortical fibres (Jia et al. 2006). The presynaptic axons of these connections were of extracortical origin; this may account for the difference in the developmental change in the paired-pulse behaviour. However, at least in one case (Ramoa & Sur, 1996) the change of the short-term synaptic dynamics has also been correlated with the development of intracortical inhibitory neuronal networks, and it has been proposed to reflect inhibition in the postsynaptic neuron as well as effects at the presynaptic site. Consistent with this hypothesis, antagonists of the inhibitory neurotransmitter GABA altered the paired-pulse behaviour of this connection.

In contrast to the connections described above, the mossy fibre–CA3 pyramidal cell synapse in murine hippocampus (Mori-Kawakami et al. 2003) is a very low P_r synapse (<0.03) that displays a developmental reduction from strong to weak facilitation between the 3rd and the 9th postnatal week. However, this reduction was not due to a change in P_r since both the failure rate and the coefficient of variation of the EPSP amplitude were unaltered at both ages. Rather, the residual Ca²⁺ concentration decreases with development because of an increased Ca²⁺ buffering as a result of an increased expression of the fast Ca²⁺ binding protein calbindin-D_28K (Celio, 1990). This would lead to an earlier Ca²⁺ buffer saturation in 3-week-old animals and hence to a larger paired pulse ratio. This study indicates that not only short-term plasticity itself but also the direction of its developmental alteration may be pathway dependent.

Developmental alteration of synaptic efficacy and reliability

Apart from changes in the short-term dynamics, synaptic connections display also developmental alterations with respect to their efficacy and reliability. However, these changes have only been investigated in a few identified synaptic connections. A decrease in the unitary EPSP amplitude with age has been reported for neocortical pyramidal neuron connections within layers 2/3 and 5 and between these two cortical layers both in somatosensory and auditory cortex (Reyes & Sakmann, 1999; Frick et al. 2007; Oswald & Reyes, 2008); however, the decrease in synaptic efficacy is not always monotonic but may be preceded by an initial increase and a subsequent reduction (Frick et al. 2007; see Fig. 1Bb). Changes in synaptic efficacy are generally accompanied by an increase in both failure rate (i.e. the percentage of failures to respond to a presynaptic action potential) and the coefficient of EPSP amplitude variation (Reyes & Sakmann, 1999; Frick et al. 2007; Oswald & Reyes, 2008), i.e. an altered reliability of transmission. For the synaptic connection between layer 5A pyramidal neurons such a developmental regulation can be seen in Fig. 1Bb, where the CV changes from 0.38 ± 0.20 at P14–P16 via 0.30 ± 0.16 at P17–P20 to 0.54 ± 0.20 at P24–P29. The most straightforward explanation for these data is that the presynaptic probability of neurotransmitter release at the existing synaptic contacts is reduced during neocortical development. Such developmental changes have also been described for the hippocampal CA3–CA1 connection (Bolshakov & Siegelbaum, 1995), where P_r was 0.9 in 4- to 8-day-old rats and decreased to less than 0.5 at an age of 2–3 weeks, at least in a subpopulation of CA3–CA1 synapses (Wasling et al. 2004). Similar reductions of P_r have also been suggested for the cortico-striatal (Choi & Lovinger, 1997), cerebellar (Pouzat & Hestrin, 1997) and pyramidal–interneuron connections in the neocortex (Angulo et al. 1999).

However, it has been demonstrated that existing synaptic contacts are eliminated and new ones are formed in an activity-dependent manner, in particular during adolescence but also in the mature animal, (Trachtenberg et al. 2002; Zuo et al. 2005; De Paola et al. 2006; Holtmaat et al. 2006). This may result in the emergence of new synaptic connections with different properties, for example lower quantal EPSP amplitude and/or a reduction in the number of active zones (i.e. functional release sites) per contact or the number of functional contacts per synaptic connection.

In immature neocortical neurons, the input resistance is high resulting in a slow membrane time constant and hence a slow EPSP decay (Frick et al. 2007; Oswald & Reyes, 2008). This allows EPSP summation even at low firing frequencies (10–30 Hz), but prevents a high temporal signal resolution (Oswald & Reyes, 2008). Concurrent with changes in synaptic efficacy and reliability, an acceleration of the EPSP time course with ongoing maturation takes place. This is likely to result from the reduced membrane time constant with development (Frick et al. 2007; Oswald & Reyes, 2008; for a review see Spruston et al. 1994) as well as changes in the postsynaptic glutamate receptor complement. The most notable change is an alteration in the decay time course of the NMDA receptor mediated EPSP, which is slow (∼300 ms) in the immature and faster (∼100 ms) in more mature animals, and which is mediated by a shift in the NMDA receptor subunit expression from NR2B in the immature neocortex to NR2A in the mature neocortex (Carmignoto & Vicini, 1992; Sheng et al. 1994; Flint et al. 1997; Stocca & Vicini, 1998).

Other changes in the glutamate receptor properties affect the paired pulse behaviour. A developmental switch from GluR2 subunit-lacking to GluR2 subunit-containing AMPA receptors has been reported for some neocortical synapses. In the absence of GluR2, AMPA receptors are Ca²⁺ permeable, have a rectifying current–voltage relationship and exhibit paired pulse facilitation via a voltage- and use-dependent relief of an intracellular polyamine block (Rozov & Burnashev, 1999). This polyamine-dependent facilitation disappears with maturation (Kumar et al. 2002; Shin et al. 2005; Brill & Huguenard, 2008).

Finally, changes in the distribution of modulatory presynaptic receptors (e.g. the expression of presynaptic NMDA receptors; Corlew et al. 2007) have been demonstrated to contribute to the developmental maturation of neocortical synapses.

Development and functional maturation of vesicle pools

The studies discussed so far have all been performed at late stages of neocortical development (P10 and older) and therefore represent developmental stages during which synaptic vesicle pools have already matured to a certain degree, though probably not completely. Unfortunately, little information is available concerning the functional properties of early synaptic connections in the neocortex. Classical electron-microscopic studies of synaptogenesis in the visual cortex (Blue & Parnavelas, 1983a,b), occipital cortex (Dyson & Jones, 1980) and spinal cord (Vaughn, 1989) describe an increase in synaptic vesicle number and organisation as one of the most conspicuous signs of synaptic differentiation. In newly forming synaptic contacts (early postnatal in visual cortex and embryonic in the spinal cord) only a few synaptic vesicles can be found in the presynaptic terminal, mostly in the vicinity of the presynaptic terminal. In the neocortex, such immature synapses undergo very rapid fatigue (or even permanent run-down) when stimulated at high rates as has previously been described for transient synaptic contacts onto Cajal–Retzius neurons (Radnikow et al. 2002), subplate neurons (Hanganu et al. 2002) and early postnatal connections between immature excitatory layer 4 spiny neurons (Radnikow, Lübke & Feldmeyer, in preparation).

With ongoing development the number of synaptic vesicles and their packing density within the terminal and in the proximity of the active zone increases (Dyson & Jones, 1980; Blue & Parnavelas, 1983a,b; Vaughn, 1989). Furthermore, at more mature synapses vesicles appear to be clustered, which is not the case in immature synapses. In the matured synapse vesicles reside in three different pools, the readily releasable, the recycling and the reserve pool (for reviews see Rizzoli & Betz, 2005; Rollenhagen & Lübke, 2006). Experiments on cultured hippocampal neurons have led to the suggestion that these pools undergo three stages of development (Mozhayeva et al. 2002). Initially, synaptic vesicles accumulate to form the recycling pool, which allows synaptic release at a moderate rate. In a second, transitory state the readily releasable pool (with synaptic vesicles docked to the active zone of the presynaptic terminal and therefore allowing fast release) gradually emerges in addition to the recycling pool, and finally the so-called reserve pool is formed. In rodents, the morphological maturation of excitatory neocortical synapses appears to take place over a time period of 4–5 weeks. The strong depression at most immature cortical synapses (as well as at the immature calyx of Held) may be due to a rapid exhaustion of the readily releasable vesicle pool that cannot be replenished at a sufficiently high rate. Matured synapses have a larger readily releasable pool and possibly faster refilling mechanisms and are therefore able to sustain synaptic release at higher rates. Such a mechanism has also been proposed to act at the calyx of Held (Taschenberger & von Gersdorff, 2000; Iwasaki & Takahashi, 2001).

Concurrent with the differentiation of vesicle pools, the molecular components of the release machinery, i.e. the molecules that act as Ca²⁺ sensors of neurotransmitter release and mediate the membrane fusion of synaptic vesicles, are also undergoing refinement (Shimohama et al. 1998; Südhof, 2004; a detailed description is outside the scope of this review). For example, the expression of the putative Ca²⁺ sensors, the synaptotagmins, is developmentally regulated (Berton et al. 1997), which may alter the dynamics of neurotransmitter release. The reduction in P_r could also be the result of developmental alterations in the control of the Ca²⁺ concentration in the presynaptic terminal. Here, the density and subtypes of presynaptic Ca²⁺ channels and hence the Ca²⁺ influx, the expression of intraterminal Ca²⁺ binding proteins, and the Ca²⁺ sensitivity of the molecular release machinery may change (Rozov et al. 2001; Bark et al. 2004; Felmy & Schneggenburger, 2004; Schulz et al. 2004; Koester & Johnston, 2005; Ali & Nelson, 2006; Gonchar et al. 2007; Xie et al. 2007) amongst other factors. This may account for the decrease in P_r observed at many neocortical and extracortical synapses; however, for each individual synaptic connection the exact mechanism responsible for this reduction in P_r remains to be determined.

Implications

Two presynaptic mechanisms seem to contribute to the functional maturation of glutamatergic neurotransmission in the neocortex: a differentiation of the synaptic vesicle pools and an alteration in the intraterminal Ca²⁺ metabolism including the intraterminal Ca²⁺-binding sites. These mechanisms are likely to be responsible for the decrease in the short-term depression, reliability and efficacy of synaptic connections.

In the neocortex, these developmental changes appear to represent a general trend and may contribute to improved temporal signal processing. The strong synaptic depression and the long EPSP decays will reduce the ability of immature neurons to follow rapid AP firing whilst in more mature microcircuits the reduced depression and fast synaptic decays will enhance the temporal fidelity and the ability to follow repetitive AP firing.

On the other hand, the higher reliability and – in some cases – also the higher efficacy of synaptic transmission together with the longer-lasting EPSPs of immature cortical synapses may be a prerequisite for the formation and maintenance of specific synaptic connections. Thus, the synaptic connections in the immature brain appear to be tuned to provide larger and longer synaptic signals in order to stabilise immature synapses. One may speculate that the pre- and postsynaptic properties (e.g. high reliability, slow EPSP decay, different postsynaptic glutamate receptors) contribute to a prolonged Ca²⁺ influx at immature synapses that may induce Ca²⁺-dependent gene expression involved in the regulation of synapse development, maturation, and refinement (Greer & Greenberg, 2008).

Supplemental material

Online supplemental material for this paper can be accessed at:

http://jp.physoc.org/cgi/content/full/jphysiol.2009.169458/DC1