The dynamics of excitatory synapse formation on dendritic spines

Abstract

Dendritic spines, the postsynaptic compartments of most functional excitatory synapses in the Central Nervous System (CNS), are highly dynamic structures, having the ability to grow, change shape, or retract in response to varying levels of neuronal activity. This dynamic nature of spines allows modifications in brain circuitry and connectivity, thus participating in fundamental processes such as learning, recall, and emotional behaviors. Although many studies have characterized the precise molecular identities and signaling pathways by which spines initially form, little is known about the actual time course over which they mature into functional postsynaptic compartments of excitatory synapses. A recent publication in Neuron addresses this issue by studying dendritic spine growth in response to multiphoton glutamate uncaging, simultaneously monitoring the amplitudes of the resultant postsynaptic currents and intracellular Ca²⁺ transients within individual spines in CA1 pyramidal neurons in organotypic cultures of postnatal hippocampal slices. The authors describe that dendritic spines are able to respond to glutamate shortly after their formation, leading to the conclusion that spine growth and glutamate receptor recruitment are closely coupled temporally. AMPA receptor-mediated currents exhibited similar amplitudes in newly formed spines compared with older, more mature spines when their volume was taken into account. In addition, NMDA receptor-mediated currents also appeared early after spine formation, although the amount of Ca²⁺ entry through these receptors was significantly lower in newly formed spines compared to older, mature spines. Within just a couple of hours, these newly formed spines were contacted by presynaptic terminals, thus acquiring a morphological appearance indistinguishable from already existing mature excitatory synapses.

Dendritic spines were discovered by Santiago Ramón y Cajal in brain tissue stained with the Golgi silver impregnation technique at the end of the 19th century (Cajal, 1891) and have attracted the attention of neuroscientists ever since. In his first descriptions of dendritic spines, Cajal proposed that these structures increased the dendritic surface where synapses – the site of contact between neurons – could be formed (Cajal, 1909). Indeed, as it has turned out, the vast majority of excitatory synapses in the CNS are composed of these contacts between a presynaptic axonal terminal from one neuron and a postsynaptic spine grown from the dendrite of another. Since then, dendritic spines have been shown to contain a plethora of neurotransmitter receptors, intracellular signaling molecules, protein translation machinery, cisterns of endoplasmic reticulum, and a highly active cytoskeleton meshwork composed of actin microfilaments, shortly-lived tubulin microtubule structures, and cytosolic myosins (Tada and Sheng, 2006; Sheng and Hoogenraad, 2007).

Perhaps it’s the dynamic nature of dendritic spines and their sensitivity to neuronal activity levels and behavioral states, however, that caught the attention and imagination of modern neuroscientists. Because the cytoskeleton of dendritic spines is composed of actin filaments, it is possible for spines to develop, grow, change shape, and retract in response to synaptic input and overall neuronal activity. The density of spines can be used as a measure of synaptic strength, as changes in activity can alter both the number and even the shape of spines. In addition to activation of postsynaptic glutamate receptors, other regulators of spine dynamics exist, which include small GTPases, neurotrophins such as brain-derived neurotrophic factor, and neuroactive steroids such as estrogen (Bourne and Harris, 2008; Chapleau and Pozzo-Miller, 2008).

Questions regarding the time course and stepwise progression by which a nascent spine and a presynaptic bouton eventually become involved as an excitatory glutamatergic synapse have been addressed recently. It’s important to keep in mind that dendritic protrusions resembling spines emerge from CA1 pyramidal neuron dendrites within seconds-to-minutes from high frequency afferent stimulation in cultured hippocampal slices (Maletic-Savatic et al. 1999; Engert & Bonhoeffer 1999), and that most existing spines display glutamate responsiveness, as estimated by multiphoton glutamate uncaging (Matsuzaki et al. 2004; Béïque et al. 2006; Busetto et al. 2008). Prior work by Nagerl et al. (2007) utilized theta-burst stimulation of Schaffer collaterals in hippocampal slice cultures to study the time course of synaptogenesis. They reported that, although the new spines that occur after stimulation form physical contact with a presynaptic bouton within a couple of hours, mature synapses are not created until 15 or more hours later. Conversely, studies conducted by Friedman et al. (2000) found that synaptic vesicles appear within 30 minutes of initial axo-dendritic contact; presynaptic proteins such as Bassoon are then recruited to this area. Shortly following, postsynaptic elements such as PSD-95 begin to aggregate and an actual spine forms thereafter. Friedman et al. (2000) report that these events occur within a couple of hours, a vastly shorter time scale than that found by Nagerl et al. (2007). On the other hand, an in vivo time-lapse imaging study of spine gain/loss caused by ‘chessboard’ whisker trimming showed that spines in the barrel cortex form excitatory glutamatergic synapses over a longer time course, probably exceeding 1 day (Knott et al. 2006).

The authors of Zito et al. (2009) set out to determine the amount of time required for a newly formed dendritic spine to become the functional postsynaptic compartment of an excitatory synapse. By transfecting hippocampal pyramidal neurons with eGFP and performing multiphoton excitation microscopy, they were able to identify and subsequently classify spines based upon their appearance during time-lapse imaging: mature, which were present from the outset; new persistent, which appeared by the second time point, 11 hours later; and new spines, those that appeared at the third time point, 12.5 hours after the start of imaging. Simultaneous electrophysiological recordings were then performed from the cell body in order to gauge the postsynaptic response to multiphoton glutamate uncaging at a particular spine head, a tall order when considering the infinitesimal magnitude of these currents (>20pA) along with the large access resistance measurements reported (20–40MΩ for chronic time lapse and 30–95 MΩ; for acute time lapse experiments). Pharmacological antagonists of the different subtypes of ionotropic glutamate receptors were used to isolate glutamate uncaging currents mediated by either AMPA or NMDA receptors.

From their data, the authors determined that new spines, indeed, contain AMPA-type glutamate receptors. Although new spines responded to multiphoton glutamate uncaging with current amplitudes significantly lower than those recorded from persistent and new persistent spines, normalizing these amplitudes to spine head volume revealed no significant differences among all three spine ages. Additionally, as new spines increased in age, so too did their volume; subsequently, AMPA receptor-mediated currents increased in direct proportion to spine volume.

Likewise, nascent spines were probed for the presence of NMDA receptor-mediated currents. Multiphoton excitation time-lapse imaging was once again utilized in order to identify spines, which were then classified into the three age groups. Immediately following, the imaged neuron was filled with a Ca²⁺ sensitive fluorescent dye through the whole-cell electrode and NMDA receptor-mediated currents arising from the uncaging of glutamate were recorded from the cell body; these experiments were performed in low Mg²⁺ extracellular solutions to facilitate the activation of NMDA receptors in voltage-clamped neurons at their resting membrane potential (the SERCA pump was inhibited by thapsigargin, and Ca²⁺-induced Ca²⁺ release from intracellular stores was blocked by ryanodine). When normalized to the average current amplitude from persistent and new persistent spines in the same dendrite, there were no significant differences in NMDA receptor-mediated current amplitudes among the various spine-age classifications. However, newly formed spines showed significantly smaller NMDA receptor-mediated spine Ca²⁺ transients compared to persistent spines. Because diffusion of the highly mobile indicator-bound Ca²⁺ from the spine head to the parent dendrite likely overestimates spine Ca²⁺ decay times, diffusional coupling through the spine neck was estimated by multiphoton photoactivation of GFP (Bloodgood and Sabatini, 2005). Indeed, the relaxation phase of paGFP within individual spines revealed faster exponential decay times (τ) in newly formed spines, indicating that they have a better diffusional coupling with their parent dendrite, potentially contributing to the smaller Ca²⁺ transient amplitudes (Sobczyk et al. 2005). The role of spine geometry – especially that of their necks – in the temporal profile of evoked Ca²⁺ signals within spine heads is, however, still controversial because these measurements are highly sensitive to unavoidable alterations of the endogenous Ca²⁺ buffer capacity of spines by Ca²⁺ indicators, as well as to the temperature at which the imaging/recordings are performed (Majewska et al. 2000; Sabatini et al. 2002; Noguchi et al. 2005; Cornelisse et al. 2007; Higley and Sabatini, 2008).

Zito et al. (2009) further delved into the issue of whether and when newly formed spines were contacted by a presynaptic bouton. After time-lapse multiphoton imaging and electrophysiological recordings, slices were fixed and processed for retrospective serial section electron microscopy and 3D reconstruction (Knott et al. 2006; Nagerl et al., 2007). New spines were found to be directly apposed to a presynaptic bouton, which led to the conclusion that new spines become part of a synapse within a couple of hours from their first appearance as a dendritic appendage. The authors report that, at the electron microscopy level, the newly formed spine synapses were indistinguishable from already existing synapses on other spines, and have all the well-described features of Gray type-I asymmetric synapses on dendritic spines (Gray, 1959; Westrum and Blackstad, 1962; Harris and Stevens, 1989).

Despite the tour de force these experiments represent, they did not yield a quantitative time-course over which new spines form synapses, nor did they take into account whether the newly formed synapses were ‘mature’ in terms of their morphological appearance. According to the authors, their data demonstrate that the temporal relationship of new spine formation and AMPA receptor recruitment occur ‘within a few tens of minutes of each other’. It is not readily obvious how they arrived at this range of time. Most studies regarding synaptogenesis have notoriously yielded different results depending upon the experimental model that was utilized. While studies based on single snapshots of fixed tissue have concluded longer time periods required for spine synapse formation, time-lapse studies of live neurons have yielded much shorter lengths of time. Questions regarding the state of the postsynaptic AMPA and NMDA receptors still linger in the Zito et al. study. Perhaps in the future, immunolabelling could be performed after time-lapse multiphoton imaging and glutamate uncaging on individual spines in order to yield a more quantitative view of postsynaptic receptor expression and density at newly formed spines. Additionally, immunolabelling of known dendritic spine molecules will shed light on the maturation of the intracellular signaling machinery of functional spine synapses. Lastly, quantitative biophysical analyses of postsynaptic current amplitudes and fluctuations may also lend evidence as to whether there are an increasing number of receptors or just a change in conductance of the ones already there.

Figure 1. The structure of dendritic spines of hippocampal pyramidal neurons.

Using particle-mediated gene transfer (a.k.a. gene gun), organotypic slice cultures were transfected with cDNA coding for eYFP. Top panels: Laser-scanning confocal microscopy images of a pyramidal neuron in area CA1 are shown at different magnifications to illustrate the complexity of their dendritic arbor and the abundance of dendritic spines in secondary and tertiary branches. Bottom left panel: A maximum-intensity projection of z-stacks shows a dendritic segment studded with the most common spine morphologies, i.e. stubby, mushroom and thin. The cartoon illustrates the geometrical dimensions measured in individual spines to categorize them. Bottom right panel: A mushroom dendritic spine (outlined in green) forms an asymmetric synapse with a single presynaptic terminal (outlined in red) in stratum radiatum of area CA1 in organotypic slice culture. We thank Dr. Christopher Chapleau for these images.