Synapses: Multitasking Global Players in the Brain

Historical background

For centuries, the belief that the structure of the brain and its elements, namely neurons, their dendrites, axons and synapses, and non-neuronal astro- and oligodendrocytes, reflect its function had been a ‘driving force’ for investigations and the source of major discoveries in neuroscience. One of the most important ones was, beside the definition of the neuron, the introduction of the term ‘Synapse’ originally termed more than 100 years ago by Charles Sherrington, a well-known electrophysiologist at that time. Ramon y Cajal, one of the most influential neuroanatomists and the founder of the neuronal doctrine, later adopted this term. The word ‘Synapse’ comes from the Greek synapsis (συνάψις), meaning ‘conjunction’, and from συνάπτεὶν (συν ‘together’) and ἅπτειν (‘to fasten’). This early fundamental discovery and description is even more intriguing because both Sherrington and Ramon y Cajal suggested a direct connection between neurons via these structures, but without ever seeing them.

The introduction of EM and its further development leading nowadays to high-end, fine-scale transmission electron microscopy (TEM), focused ion beam scanning electron microscopy (FIB-SEM), CRYO-EM and EM tomography combined, for example with high-pressure freezing (see Studer et al. 2014; Imig et al. 2014) extended the structural investigations from the light microscopic visible neuron to the subcellular and even the molecular level of the synapse, the elementary building block of any neural networks in the brain. It has to be mentioned thought, that CRYO-EM is more typically suited for the analysis of the structure of proteins at high resolution, rather than subcellular structures.

In addition, modern light microscopic techniques, like Stimulated Emission Depletion (STED) and direct stochastic optical reconstruction microscopy (dSTORM) aimed to visualize synaptic structures at the nanoscopic level, however focused more on the abundance of various synaptic proteins at active zones. Furthermore, the introduction of CRYO-correlative light- and electron microscopy (CRYO-CLEM) allows both fluorescence microscopy as well as three dimensional (3D) CRYO-EM tomography to reveal the ultrastructure of significant target molecules with specific cellular functions at high temporal and spatial resolution (Plitzko et al. 2009). Since then, a wealth of information was obtained that deepened our understanding of synaptic structures, in the developmental and adult brain in health and disease based on studies undertaken in various animal species, including rodents, higher mammals, non-human primates and even humans.

Although synapses had been looked at from different viewpoints, summarized in meanwhile thousands of original publications, reviews and numerous textbooks, a detailed, comprehensive and quantitative knowledge about their morphology is still limited to a relative small number of CNS synapses in different brain regions (Calyx of Held: Rowland et al. 2000; Sätzler et al. 2002; Wimmer et al. 2006; Cochlear bushy cell synapses: Nicol and Walmsley 2002; Climbing fiber synapses: Xu-Friedman et al. 2001; Cerebellar mossy fiber: Xu-Friedman and Regehr 2003; Hippocampal Mossy Fiber Bouton: Chicurel and Harris 1992; Rollenhagen et al. 2007; Synapses in the dentate gyrus: Marrone et al. 2005; Area CA1 synapses: Sorra and Harris 1993; Harris and Sultan 1995; Spacek and Harris 1998; Schikorski and Stevens 1997, 2001; Ribbon synapses in the retina and cochlear: Sikora et al. 2005; Moser et al. 2006; Michanski et al. 2019; Olfactory cortical synapses: Schikorski and Stevens 1999). Such detailed descriptions, however, are required to understand and link structural and functional components of the signal cascades underlying synaptic transmission and plasticity.

An important first step towards an improved understanding of synaptic function were simultaneous patch-clamp recordings from a glutamatergic giant synapse, the so-called Calyx of Held terminating on the principal neurons in the medial nucleus of the trapezoid body in the auditory brainstem by Sakmann, Neher and co-workers (for example see Borst and Sakmann 1996, 1998, 1999; Takahashi et al. 1996; Schneggenburger et al. 1999, Schneggenburger and Neher 2000; for review see also Schneggenburger et al. 2002). However, it turned out that the Calyx of Held is rather the exception than the rule with respect to its synaptic properties perfectly adapted to audition. Hence the investigation of the Calyx of Held synapse strongly suggested that synapses are ‘unique’ entities, in both structural and functional terms.

The second, more central synapse, where paired recording became possible was the mossy fiber bouton-CA3 pyramidal cell synapse in the hippocampus, a synapse involved in learning and memory processes (Geiger and Jonas 2000; Bischofberger and Jonas 2002; Hallermann et al. 2003; Engel and Jonas 2005; Alle and Geiger 2006). The work on this synapse strongly supported and extended the view that synapses are unique in their structural and functional properties. Thus, the dream to create a general ‘model synapse’ for the brain was over.

Nevertheless, the simultaneous recordings from two different CNS synapses and their target structures made it possible for the first time to measure transmitter release under defined internal and external ionic and membrane potential conditions. In addition, the size and time course of action potential evoked Ca²⁺ influx (Borst and Sakmann 1996, 1998; Bischofberger and Jonas 2002), the occupancy of the putative Ca²⁺ sensor driving vesicle fusion (Bollmann et al. 2000; Schneggenburger and Neher 2000), the equilibration of intracellular Ca²⁺ with the endogenous Ca²⁺ buffer, and the eventual Ca²⁺-clearance (Helmchen et al. 1997) can be accurately measured. Furthermore, the latency, size and time course of evoked quantal and multiquantal EPSCs (Borst and Sakmann 1996; Silver et al. 2003; Molnar et al. 2016; Holderith et al. 2016; Seeman et al. 2018; Rollenhagen et al. 2018; Vaden et al. 2019; reviewed by Neher 2015; Chamberland and Toth 2016) can be determined. However, there are still steps in the signal cascades that at present can only be simulated (Yamada and Zucker 1992; Bertram et al. 1999; Meinrenken et al. 2002, 2003; Freche et al. 2011). This includes the site, time- and space-dependent build-up and collapse of Ca²⁺-domains around the pore of Ca²⁺ channels at a synaptic contact and the buffered diffusion and the subsequent interaction of free Ca²⁺ with the Ca²⁺ sensor.

Thus, realistic values of the geometry of synaptic boutons, including the number, size and shape of active zones, and the three functionally defined pools of synaptic vesicles (Rizzoli and Betz 2005), namely the readily releasable (RRP), the recycling (RP) and resting pool are essential for constraining realistic geometrical models of synaptic structures.

On the postsynaptic side, the time course and amplitude of spontaneous and evoked excitatory postsynaptic potentials (EPSCs) were used to infer the characteristics of quantal release (Henze et al. 1997; Silver et al. 2003; Biro et al. 2005; Szabadics et al. 2006; Saviane and Silver 2006; Rollenhagen et al. 2018; Vaden et al. 2019). This interference requires simulations of the transient increase of the glutamate concentration in the synaptic cleft, reversible binding of glutamate to appropriate glutamate receptors and eventual uptake and diffusion of glutamate out of the cleft (Freche et al. 2011). To a large extent, these processes are governed by the geometry of the synaptic cleft and the shape and size of pre- and postsynaptic densities. These parameters can only be estimated from 3D-reconstructions of synaptic structures.

This review will focus on recent findings on the detailed quantitative structural description of the most common type of synaptic bouton in the CNS: excitatory synaptic boutons in different layers of the rodent, non-human primate and human neocortex. Their small size and the great diversity of neurons and synaptic boutons in different layers of a cortical column made this investigation a challenging and difficult task. Such detailed morphological descriptions are useful to directly correlate structure with function of synapses and may therefore explain their different and specific functional performance and computational properties within the network in which they are integrated. On the other hand, such quantitative data provide the basis for numerical and/or MonteCarlo simulations of various synaptic parameters that are still only partially accessible for experiment, at least in the human brain.

Methodological considerations

One possible way to describe synaptic boutons and their target structures in such great detail are either 3D-volume reconstruction based on serial ultrathin sections using TEM (Fig. 1A) or FIB-SEM (Fig. 1B). With the second approach, serial digital EM images were obtained by constant milling a defined area of the sample containing the area of interest by a gallium ion laser beam and subsequent imaging of the block surface (block-face imaging).

From the resulting z-stacks of EM images quantitative 3D-models of synaptic boutons and their prospective target structures can then be generated using different commercially or self-made reconstruction software tools running on high-performance computer systems. Both, TEM and FIB-SEM have advantages, but also disadvantages. Serial sectioning and subsequent TEM examination of ultrathin sections within a series is a very labor-intensive and thus time-consuming process with a comparable low throughput of tissue samples. Secondly, in ultrathin sections, the tilting of the electron beam restricts the area of interest, and during the cutting and imaging process, malformations or distortions or the complete loss of the tissue sample can be a limiting factor. However, the major advantage of using serial ultrathin sections and TEM imaging is their very high quality at high resolution that is required for the detailed analysis of important structural subelements such as the number, size and shape of active zones and the organization and size of the three functionally defined pools of synaptic vesicles (Figs. 1A, 2C, D, 3A-E).

Fig. 1:

Comparison of the ultrastructure between TEM and FIB-SEM

A, Low power electron micrograph of the neuropil in the lower part of layer 1 in the human temporal lobe neocortex as visualized with TEM. Note that even at this relatively low EM magnification several synaptic complexes between synaptic boutons and either dendritic shafts or spines, are clearly identifiable.

B, Low power electron micrograph of the neuropil in layer 4 of the human temporal lobe neocortex taken with FIB-SEM. Scale bar in A and B 1 µm.

In both electron micrographs, synaptic boutons are given in transparent yellow and postsynaptic structures in transparent blue. Several thick astrocytic (ast) processes are clearly visible.

Note the differences in the appearance of active zones and synaptic vesicles due to the use of a different EM protocol required for SEM-FIB.

In contrast, FIB-SEM (Fig. 1B), a relatively new, modern EM technology, allows a much higher throughput of tissue samples because the time and labor-intensive step of serial ultrathin sectioning is no longer required. Secondly, a larger area of interest ~50 by 50 µm can be obtained compared to TEM where the area of interest is limited to ~10–20 by 10–20 µm. Finally, since the surface of the block is milled and polished rather no malformations or distortions are expected thus no or minor alignment processing of adjacent images is required. The major disadvantage, however, are limitations in the resolution of active zones and synaptic vesicles which appear structurally different due to the use of a different EM embedding protocol required for FIB-SEM (Fig. 1B; Movie 1).

In the future, the combination of both TEM and FIB-SEM will be the method of choice to address specific questions and further unravel the ‘microcosms’ of the brain, for example in describing the ‘connectomics’ and synaptic organization of various layers, nuclei and brain regions.

Synaptic boutons in the neocortex of rodents and non-human primates

Meanwhile numerous publications described structural and functional aspects of synaptic transmission and plasticity in different layers mainly in the rodent neocortex using paired or multiple recordings and subsequent morphological analysis of synaptically coupled pairs filled with biocytin or fluorescent dyes during recording (reviewed by Lübke and Feldmeyer 2007; Feldmeyer 2012; Feldmeyer et al. 2013; Qi et al. 2015; Radnikow and Feldmeyer 2018). It has been demonstrated that, beside similarities huge differences exist between intralaminar (synaptic connections within a given layer) and translaminar (synaptic connections across layers) excitatory-excitatory, excitatory-inhibitory and inhibitory-inhibitory synaptic connections with respect to synaptic efficacy, strength, release probability, short-term plasticity and contribution of various neurotransmitter receptors and their subunits, for example different glutamate and GABA receptors. However, these studies aimed to correlate structural with functional properties of a given synaptic connection rather than on the structural composition of individual synaptic contacts.

In contrast, only a few coherent and quantitative structural studies exist for synaptic boutons in the rodent neocortex (Rollenhagen et al. 2015, 2018; Dufour et al. 2016; Bopp et al. 2017; Hsu et al. 2017; Rodriguez-Moreno et al. 2018) and non-human primate neocortex (Anderson and Martin 2006, 2009; Freese and Amaral 2006; Hsu et al. 2017). These studies have demonstrated, for example, layer, region and gender specific differences in the density of synaptic boutons (Alonso-Nanclares et al. 2008). Most strikingly, synaptic boutons beside layer and area-specific differences (see also Rollenhagen 2015, 2018; Bopp et al. 2017; Hsu et al. 2017), differ substantially not only in their shape and size, but even more importantly in the number, size and shape of active zones and in the organization and size of the three pools of synaptic vesicles summarized in Table 1. Interestingly, some structural parameters such as bouton size, pre- and postsynaptic density surface area, content of mitochondria, and synaptic vesicles pools are in some cortical synapses well correlated but in others, no or only a weak correlation between several structural subelements are found (Rollenhagen 2015, 2018; Dufour et al. 2016; Hsu et al. 2017; Bopp et al. 2017; Rodriguez-Moreno et al. 2018). The most striking difference at cortical synaptic boutons is the total pool, and the three functionally defined pools of synaptic vesicles, namely the RRP, the RP and resting pools. Is has to be noted that a structural correlate for the functionally defined pools is not identifiable at the EM level due to the ‘randomly’ distribution of synaptic vesicles within the synaptic bouton. However, an attempt was made to sort synaptic vesicles with respect to their distance from the presynaptic density by a perimeter analysis. This approach allows the identification of synaptic vesicles belonging to one of the functionally defined pools (for criteria see Rizzoli and Betz 2005) and match nearly perfectly with functional estimations of the RRP (Hallermann et al. 2003) and RP (Rollenhagen et al. 2018). However, it is for example, still controversially discussed whether only so-called ‘docked’ vesicles identified at the ultrastructural level (Figs. 3F, 4C inset) already fused at the presynaptic density represent the RRP or also vesicles very close (10–20 nm from the active zone) belong to the RRP (Rollenhagen et al. 2015, 2018; Yakoubi et al. 2019a, b). The same holds true for the RP, how is it defined, and how many vesicles it contains at which distance from the presynaptic density is largely unknown for most of the CNS synapses (but see Rollenhagen et al. 2018). For the role and importance of the resting pool of synaptic vesicles that is thought not to be recruited under ‘normal’ physiological conditions, rather no information is available. However, it has been shown that vesicles from the resting pools can be transferred to the RP and RRP under the control of mitochondria (Verstreken et al. 2005). As a consequence, further experiments using a combination of labeling synaptic vesicles, for example with SM1–43, and high-resolution STED or two-photon laser microcopy using in vitro acute brain slices or neuronal cell cultures can address such questions (Verstreken et al. 2008) in more detail.

Very special entities: Synaptic boutons in humans?

One of the major question in synaptic neuroscience is whether results obtained in experimental animals can be transferred one-to-one to the human brain. Research on the human brain was for a long-time restricted to postmortem brains. However, for fine-scale, high-resolution EM it turned out that tissue samples from postmortem brains are not suitable because the time window between the removal and fixation of the brain is far too long to guarantee an excellent preservation of the ultrastructure, a pre-requisite for EM investigations at the cellular and subcellular level. To overcome this problem access, non-epileptic tissue from epilepsy- or brain tumor surgery became the method of choice. Here, care was taken that the tissue samples were selected far away from the epileptic focusas monitored by magnetic resonance imaging and electrophysiology and may thus be regarded as non-affected (non-epileptic) as also demonstrated by other studies using the same experimental approach (Alonso-Nanclares et al. 2008; Navarrete et al. 2013; Mohan et al. 2015; Molnar et al. 2016; Seeman et al. 2018). After its removal, tissue sample can be either immediately immersion-fixed or even prepared for acute brain slice preparations. Meanwhile several studies have studied structural (for example: Alonso-Nanclares et al. 2008; Blazquez-Llorca et al. 2013; Morales et al. 2014; Liu and Schumann 2014, Yakoubi et al. 2019a, b) and functional aspects (for example: Holderith et al. 2016; Molnar et al. 2016; Seeman et al. 2018) of synaptic transmission and plasticity in humans. However, coherent and comprehensive studies about the synaptic organization of the human brain, in particular quantitative 3D-models of synaptic boutons in humans are still very rare (but see Yakoubi et al. 2019a, b).

Using non-affected neocortical access tissue taken from epilepsy surgery, we have started to study the layer-specific synaptic organization of the temporal lobe neocortex (TLN), a typical example of a six-layered granular associational neocortex (Fig. 2–4B, C). The growing interest in the TLN is motivated by its importance in high-order brain functions as audition, vision, memory, language processing, and various multimodal associations. Moreover, the TLN is also involved in several neurological diseases most importantly as the area of origin and onset of TL epilepsy (TLE). TLE is the most common form of refractory epilepsy characterized by recurrent, unprovoked focal seizures that may, with progressing disease, also spread to other areas of the brain. Taken together, the TLN represents an important region in the normal and pathologically altered brain in humans.

So far, the synaptic organization of layer 4, the receiving input layer of signals from the sensory periphery thus representing the first station of intracortical information processing and layer 5 the major output layer was quantitatively analyzed in the TLN (Yakoubi et al. 2019a, b). The final goal of our investigations is to describe the synaptic organization of a cortical column, the elementary building block of the neocortex also in humans, exemplified for the TLN.

Synaptic boutons in the human TLN have an average size of ~2.5 to 6 µm² and are, beside similarities, strikingly different in some structural parameters from their counterparts in experimental animals (Table 1). Like in rodents and non-human primates so-called en passant (Fig. 2B; 4A) and endterminal synaptic boutons (Fig. 2C) contact either dendritic shafts (Figs. 2D; 3A, 4B), but the vast majority (~90 %) of excitatory synaptic boutons was established on dendritic spines of different types including stubby (Figs. 2D, 3A, B), mushroom (Figs. 3C, 4A, C), filopodial and elongated spines which is different to various animal species. Secondly, the majority of spines (~90 %) contained a so-called spine apparatus (Figs. 2D, 3A, B), a derivate of the endoplasmic reticulum, responsible for spine motility and stabilization of the synaptic complex during single or repetitive high-frequency stimulation. Thus, it was hypothesized that spines containing a spine apparatus partially contribute in modulating short-term synaptic plasticity (for example Holtmaat et al. 2006; for review see Knott and Holtmaat 2008). Interestingly, so-called dendro-dendritic synapses, regarded as a feature of the developmental brain, occur more frequently in the human TLN when compared to the neocortex in experimental animals. In addition, so-called clathrin-coated pits were frequently observed in synaptic boutons, some of which are located near the active zone (Fig. 2C). Clathrin-coated vesicles selectively sort cargo at the cell membrane, trans-Golgi network, and endosomal compartments for multiple membrane traffic pathways, for example exo- and endocytosis. A subpopulation is used in synaptic vesicle formation at the active zone.

Finally, also astrocytes receive direct synaptic input (Fig. 3D), although infrequently, supporting their involvement in synaptic transmission and plasticity (Min and Nevian 2012). Astrocytes have long been thought to act as nutrition suppliers and providing a stabilizing corset for neurons in the brain. However, it is now well established that astrocytes also play an important role in synaptic function, acting not only as physical barriers to glutamate diffusion, but also mediate transmitter uptake by glutamate transporters (Min and Nevian 2012; for review see Allen 2014; Dallerac et al. 2018). A striking common feature in both the human and the animal neocortex is the tight ensheathment of synaptic complexes with astrocytic processes forming the ‘tripartite’ synaptic complex (Fig. 3A, C), in contrast to MFBs and calyx of Held synapses, where astrocytic processes were never located close to individual active zones (Rollenhagen et al. 2007; Müller et al. 2009). This may explain the occurrence of glutamate spillover, synaptic cross talk and the switch from asynchronous to synchronous release upon repetitive stimulation as shown for the MFB (Hallermann et al. 2003) and Calyx of Held synapses (reviewed by von Gersdorff and Borst 2002). Astrocytes can actively take-up excessive or ‘spilled’ neurotransmitter when close to the synaptic cleft; hence they modulate the temporal and spatial neurotransmitter concentration thus controlling the induction, maintenance and termination of synaptic transmission but also modulate short-term synaptic plasticity in the neocortex.

Fig. 2:

Synaptic organization in different layers of the human TLN

A, Low power TEM micrograph of the neuropil in layer 2/3. For better visualization, some structures are highlighted in different colors. Transparent magenta: a pyramidal cell (pyr) and an astrocyte (ast) close to another adjacent pyramidal cell. The nuclei are given in transparent blue. Numerous dendritic profiles of different shape and size in the neuropil are highlighted in transparent yellow and synaptic boutons in green. Mitochondria in all structures are given in transparent blue. Note the axon initial segment (ais) originating at the base of one pyramidal cell soma. The ais is innervated by several synaptic boutons (marked by asterisks). Scale bar 5 µm.

B, En passant axon (ax) giving rise to a synaptic bouton (sb, transparent yellow) innervating a dendritic spine (sp, transparent blue) in layer 2/3. Active zones are outlined in red. Scale bar 2 µm.

C, Two endterminal boutons (sb1, sb2) establishing synaptic contacts with two neighboring spines (sp1, sp2) in layer 4. Note the different shape and size of the boutons and the active zones (red contours) in both synaptic complexes that nearly covers the entire pre- and postsynaptic apposition zone. A so-called coated pit (asterisk) is located close to the active zone. Scale bar 0.2 µm.

D, Representative example of a large stubby spine (sp) emerging from a dendrite (de) in layer 1 innervated by a large synaptic bouton (sb) with two active zones (marked by arrowheads) full of synaptic vesicles. Note also the prominent spine apparatus (framed area). Scale bar 0.5 µm.

E, Large mushroom spine (sp) with a long spine neck (spn) emerging from a small dendrite (de) in layer 1. Note the disk-like shape of the prominent spine apparatus (red contours) that covers nearly the entire volume of the spine head. The spine head is innervated by three synaptic boutons (sb) and a fourth establishes a synaptic contact at the base of the spine neck. Scale bar 0.5 µm.

The most striking difference between synaptic boutons in the human, non-human primate and rodent neocortex is, however, the shape and size of the active zones and that of the three functionally defined pools of synaptic vesicles, namely the RRP, RP and resting pool. Although small in size synaptic excitatory synaptic boutons in layer 4 and layer 5 of the TLN contain active zones that were on average 2-fold larger in size (~0.2 – 0.25 µm² in surface area) when compared with their counterparts of comparable size in other brain regions in rodents or non-human primates (Table 1), or even much larger CNS synapses such as the cerebellar and hippocampal mossy fiber bouton and the Calyx of Held endterminal. In numerous synaptic boutons in the human TLN, the active zones covered most of the pre- and postsynaptic apposition zone (Figs. 2 C, 3A, B) hence enlarging the presynaptic ‘docking’ zone for synaptic vesicles. In addition, more synaptic boutons contain not only a single but up to three active zones (Figs. 2B, 3A-C). Numerous of the active zones were perforated at the pre-, post- or both synaptic densities. Even more striking these boutons containing a total pool of synaptic vesicles (average 1500–1800 synaptic vesicles) that was 2–3-fold larger to that reported in the rodent and non-human primate neocortex suggesting also comparable large RRPs (Fig. 3F), RPs and resting pools. Indeed, the RRPs are by 3–5-fold, the RPs by 2-fold and the resting pools by 2-fold larger than in rodent and non-human primate neocortex. It has been recently shown that the size of the RRP dynamically regulates multivesicular release in mice (Vaden et al. 2019). Thus these large pools suggest reliable synaptic transmission even at high-frequency stimulation; hence a rapid depletion of the RRP and RP could be prevented by replenishment of synaptic vesicles from a large resting pool. It has to be noted though that like in non-human primates and rodents, a huge variability exists in the structural composition between individual synaptic boutons in humans, in particular the size of RRP, RP and resting pool that may partially contribute in modulating synaptic plasticity.

Fig. 3:

Innervation pattern of synaptic boutons in different layers of the human TLN

A, Two opposite synaptic complexes (sb1-sp) and (sb2-de) in layer 1 one of which (sb1) establishes a glutamatergic synapse with a stubby spine (sp) identified by the shape and appearance of the active zone (arrowheads) and the size and more roundish shape of the synaptic vesicles. Sb2 is also glutamatergic as identified by the appearance of the active zone (arrowheads) but directly terminating on the dendritic shaft. Note the large astrocytic finger (ast) close to the dendrite and synaptic boutons and the spine apparatus in the stubby spine. Scale Bar 0.5 µm

B, A large stubby spine (sp) in layer 2/3 innervated by a synaptic bouton (sb) with two separated active zones (arrowheads). The spine apparatus is marked by an asterisk. Scale bar 0.5 µm

C, Typical example of a large mushroom spine in layer 2/3 with a thick spine head (sph) and a smaller but thick spine neck (spn) receiving input by a large synaptic bouton (sb). Note the two active zones (arrowheads) one directly on top of the spine head and the other located aside. The relatively large active zone (arrowheads) covers the entire pre- and postsynaptic apposition zone. de: dendrite; ast: astrocytic profile. Scale bar 0.5 µm.

D, Astrocytic process (ast) identified by its opaque appearance and the vesicles containing gliotransmitter contacted by a synaptic bouton (sb) in layer 5. This type of contact is rarely found. Scale bar 0.5 µm.

E, Dendro-dendritic synapse (de1, de2) where de1 serves as the presynaptic element identified by the cluster of synaptic vesicles at the active zone (arrowheads) in layer 6. In addition, de2 receives a synaptic bouton (sb) with a non-perforated active zone marked by arrowheads. Scale bar 0.25 µm.

F, High-magnification of a large non-perforated active zone in layer 6 showing three ‘docked’ vesicles (highlighted in transparent green) among the population of synaptic vesicles close to the presynaptic density contacting a dendritic spine at the postsynaptic (post) element. Scale bar 0.25 µm.

Fig. 4:

3D-volume reconstructions of synaptic boutons and their target structures and electron microscopic tomography in the rodent and human temporal lobe neocortex

A, En passant axon (ax, transparent gold) followed over long-distance in consecutive electron micrographs establishing a synaptic contact on a dendritic spine (sp) of a postsynaptic dendritic segment (de, blue) in rodent layer 5. The active zone is marked by arrowheads. Note the association of mitochondria (white) with the pool of synaptic vesicles (green dots). Note the presence of synaptic (green dots) vesicles and dense-core vesicles (magenta dots) within the en passant axon (framed area). Scale bar 1 µm.

B, Two synaptic boutons terminating opposite to each other onto a small caliber dendrite (de) in layer 5. In one bouton, the envelope of the terminal is omitted to better visualize the distribution of synaptic (green dots) and dense-core (magenta) vesicles. The mitochondrion is given in white and the two active zones are marked by arrowheads. Scale bar 1 µm.

C, Synaptic bouton (sb) terminating on a dendritic spine (sp) as visualized with electron microscopic tomography. A coated pit fused with the bouton membrane is shown in the framed area. Inset: High-power magnification of the active zone (arrowheads) with two ‘docked’ vesicles highlighted in transparent green. Pre: presynaptic; post: postsynaptic. Scale bar 0.25 µm.

Taken together, the structural composition of both the presynaptic terminal and the spine as the main target structure suggests high synaptic efficacy and reliability of synaptic transmission but also in the induction, regulation and termination of short-term plasticity at synaptic boutons in the human brain.

This structural heterogeneity was confirmed by a recent structural/functional investigation about synaptic connections in the human TLN using paired recordings (Seeman et al. 2018). This study demonstrated that synaptic connections in the TLN are indeed highly reliable and strong as indicated by large excitatory postsynaptic potential (EPSP) amplitudes when compared to mouse neocortex, but also show layer-specific differences and in modulating short-term plasticity.

Perspectives

In summary, excitatory synaptic boutons in the brain represent ‘unique entities’ in rodents, non-human primates and even more in humans. Their individual composition with marked structural differences strongly suggest that they are perfectly adapted to the ‘job’ they have to fulfill in different neural networks of the brain in which they are embedded. However, there are still a lot of questions remaining that have to be addressed in the future. For example, and most importantly how and when synapses are generated during the development of the human brain and do they undergo the same selective layer-specific pruning or elimination as shown in experimental animals? How and when do they undergo severe structural and functional changes during neurological disorders like schizophrenia, autism and neurodegenerative diseases like Morbus Alzheimer and Morbus Parkinson? How age-, strain-, sex, left/right-, and subregion-specific differences would influence their structural composition? How the three functionally defined pools of synaptic vesicles become differentially recruited during high-frequency brain activity, during different biological rhythms or behavior still remains rather unclear. At the molecular level, how are the numerous pre- and postsynaptic proteins, neurotransmitters and their subunits involved in the induction, maintenance and termination of synaptic transmission and plasticity arranged at the active zone? What is about their density and possible co-localization at individual synaptic complexes? To date we are still far away in our understanding of these fascinating structures.