Introduction

Chemical synapses are specialized junctions that pass signals among neurons or from neuron to non-neuronal cells such as those in muscles or glands. In the CNS, synapses have two structural and functional different regions: (i) the active zone (AZ), a specialized area of the plasma membrane located in the presynaptic neuron, which contains the molecular machinery that regulates neurotransmitter exocytosis from synaptic vesicles; and (ii) the postsynaptic density (PSD), a membrane region of the postsynaptic neuron which has neurotransmitter receptors and signaling apparatus. The strength of synapse communication at the CNS depends on pre- and postsynaptic factors. At the presynapses, synaptic efficacy depends on the frequency and number of vesicles released after stimulation. Synaptic vesicles at the presynaptic terminal are organized, transit between three different pools, and undergo a cycle of exo- and endocytosis. The readily releasable pool (RRP) is located closest to the AZ plasma membrane, followed by the recycling pool and resting pool [1]. AZ proteins maintain the structural and functional integrity of these pools, and also recruit voltage-gated Ca2+ channels (VGCC) to the presynaptic membrane to allow fast and synchronous coupling between excitation and release, thereby modulating synaptic strength and presynaptic plasticity.

The conservation of a collection of proteins found in invertebrate and vertebrates AZs suggest that they form a minimal platform necessary for neurotransmitter release. These proteins include Rab3-interacting molecules (RIMs), RIM-binding proteins (RIM-BPs), ELKS, liprin-α, and Munc13. Other vertebrate AZ proteins such as Piccolo and Bassoon are less conserved among species, and they might play a more specialized role in higher organisms. Hence, AZ proteins form a dense accumulation called the cytomatrix at the active zone (CAZ), which is a macromolecular complex that regulates synaptic vesicle trafficking cycle. The synaptic vesicle cycle involves several steps: SV are recruited to the AZ where they are tethered and docked. Then they are primed so that they can fuse rapidly in response to calcium entry triggered by an action potential [2, 3]. To complete the cycle, synaptic vesicles undergo endocytosis, recycle, and refill with neurotransmitters for a new round of exocytosis.

Synaptogenesis, the formation of synapses between neurons in the nervous system, is a multistep process that requires cell adhesion, transport of synaptic components along neurites, trapping, stabilization of synaptic proteins at sites of newly forming synapses, and maturation. This process and the establishment of protein interactions must be timely and spatially regulated to avoid the development of neurological diseases [4]. There is a multivesicular-mediated mechanism for the transport of AZ and synaptic vesicle proteins from the cell soma into newly forming synapses at distal axonal sites [5,6,7,8]. During trafficking, cytoskeletal and motor proteins participate actively [9], and some AZ proteins play a role as adapters between motor proteins and vesicles [10]. Table 1 summarizes the main described functions of these proteins. In the last 20 years, accumulating evidence has shown the involvement of the actin cytoskeleton and regulatory signaling mechanisms participating in the formation and stabilization of new synapses in which the Wnt signaling pathway plays an important role.

Table 1 Main functions of vertebrate active zone proteins
Full size table

Active Zone Proteins: Role in Presynaptic Organization and Function

The overall function of the AZ is to translate an action potential into a chemical signal releasing neurotransmitters into the synaptic cleft. Therefore, AZ proteins have to interact in a coordinated manner for normal synaptic function to be accomplished. During synapse formation, AZ proteins are transported along axons to sites of newly forming presynapses where they interact with other AZ proteins, forming a macromolecular network of interactions. Furthermore, once synapses are established, the AZ undergoes molecular remodeling during the lifespan to support the requirements of synaptic activity and plasticity. Initially, due to their modular structure, AZ proteins were given the role of scaffolding; however, advances in live microscopy, super-resolution microscopy, electron microscopy, and genetic and molecular manipulations combined with studies in invertebrates have deciphered more specific and dynamic functions for some of these proteins in the various steps of neurotransmitter release. Therefore, AZ proteins are more than fixed scaffolds with a role in holding SV; they also participate in SV endocytosis/exocytosis and the maintenance, remodeling, and dynamics of the presynapses, working as a team with specific and shared functions (Fig. 1). The latter is thought to be part of a redundant and protective mechanism. Most AZ proteins have several common characteristics: they are multi-domain, interact with other AZ proteins, form temporal complexes with diverse proteins, and continue to be expressed at high levels in the adult brain, emphasizing their permanent role in synapse function. These characteristics situate AZ proteins as key molecular entities to modulate AZ content and efficacy during the lifespan of the synapse.

Fig. 1
figure 1

General functions of main active zone proteins. Vertebrate’s central presynapses contain seven functionally relevant proteins: Piccolo, Bassoon, ELKS, RIM, RIM-BP, liprin, and Munc13. The figure summarizes the specific and overlapping functions of these main active zone proteins suggesting that redundant mechanisms protect central synapse structure and function

Full size image

Vertebrate AZ Proteins

Piccolo

Piccolo was one of the first presynaptic molecules to be described in the vertebrate CNS [11]. There are several characteristics that make Piccolo a candidate for vertebrate CAZ assembly and organization: (a) early appearance, (b) large size, (c) multiple domains (two N-terminal zinc fingers, three coiled-coil, proline-rich region (Q domain), PDZ and two C2 domains), and (d) interaction with proteins of diverse function [11,12,13]. The multiple interactions described for Piccolo suggest that it is a very versatile molecule. Piccolo interacts directly with the AZ proteins Bassoon, ELKS, liprin-α, and Munc13. Piccolo also binds to GIT1, a GTPase-activating protein of the ADP-ribosylation factor family that participates in functions such as vesicle trafficking, adhesion, and cytoskeletal organization [13, 14]. GIT1 colocalizes with Piccolo at synapses and is part of a multi protein complex, suggesting a role in the organization of the CAZ [14]. Another interaction suggests that Piccolo has a role in SV movement and is with the prenylated Rab acceptor protein 1 (PRA1), which might control SV docking and fusion [12]. Piccolo is also postulated to have a role in synaptic vesicle clathrin-mediated endocytosis because of its interaction with Abp1, an F-actin-binding protein, and the GTPase dynamin [15]. Another possible role for Piccolo is in the scaffolding of voltage-gated calcium through its C2A domain, but its importance has not been further explored [16, 17]. Other recently described interactions of Piccolo, which will be discussed in the “Active zone and Wnt signaling” section, are Daam1 (Disheveled associated activator of morphogenesis 1) [18] and Rho-GEF Trio [19].

A specific shRNAi for Piccolo designed by Leal-Ortiz et al. [20] rendered non-Piccolo immunoreactive bands in western blot analysis of lysates prepared from rat hippocampal neurons [21]. The knock-down (KD) of Piccolo in rat hippocampal neurons did not affect synapse formation since presynaptic and postsynaptic proteins showed normal synaptic targeting [20] suggesting that this protein is not essential for glutamatergic synapse formation, although the loss of Piccolino, the major Piccolo isoform from mouse photoreceptor cells, causes defects in the maturation and ultrastructure of ribbons [22, 23]. When the presynaptic function was evaluated by styryl FM dyes [24] in the Piccolo KD model, no differences were found in the total reserve pool (TRP) of SV but the destaining kinetics of the TRP was more rapid in comparison with the control suggesting changes in the exocytosis of SV. This study showed that Piccolo negatively regulates the exocytosis of SV by modulating the synapsin 1 dynamic at the AZ by a calmodulin kinase II-dependent (CaMKII) mechanism [20] that involves presynaptic F-actin polymerization [21]. Interestingly, no neurotransmission defects were found in a study that used a mouse model with a targeted deletion of exon 14 for the Piccolo gene [25]; however, this Piccolo knock-out (KO) model continued to express presynaptic Piccolo isoforms, which most likely supports Piccolo function in neurotransmitter release [21]. Therefore, Piccolo will be necessary to restrain SV at the reserve pool by acting as the scaffold for several actin-binding proteins, which modulate presynaptic F-actin polymerization. Recently, an invertebrate homolog with conserved Piccolo functions in Drosophila, fife, has been described [26]. Fife mutants present a decrease in neurotransmitter release, abnormalities in the presynaptic membranes, and reduced SV clustering [26, 27].

Overall, Piccolo, in addition to having a clear function in scaffolding and organization of the CAZ, modulates SV dynamic and homeostasis of CAZ proteins (“Bassoon” section), suggesting a participation in synaptic plasticity (Fig. 2; Table 1).

Fig. 2
figure 2

Roles of Piccolo in immature and mature synapse. Piccolo is the largest protein and one of the most versatile molecules of the active zone. During synaptogenesis, Piccolo’s main function is the assembly of new forming presynapses and in mature synapses plays three main functions, some of them share by Bassoon: (i) CAZ organization through its multi-modular structure and interactions; (ii) the dynamic retention of the SV reserve pool acting as an adapter between SV synapsin1, actin-binding proteins, and F-actin; and (iii) stabilization of synaptic proteins by acting together with Bassoon on the ubiquitin-proteosomal system

Full size image

Bassoon

Bassoon is another large structural protein that interacts with Piccolo and other proteins of the AZ [11, 28]. Bassoon and Piccolo are structurally related proteins sharing ten highly conserved regions [12, 28, 29]. Bassoon contains two N-terminal zinc finger domains, three coiled-coil domains, and a glutamine repeats at its C-terminus [28]. Bassoon differs from Piccolo in that it does not participate directly in F-actin dynamic, but seems to be necessary for the synaptic architecture of other non-CNS synapses because the loss of Bassoon produces structural defects in retinal and cochlea ribbon synapses [30, 31] but not at central synapses. At the retinal photoreceptor synapses, the CAZ contains a specialization called the ribbon complex, which is enriched in the protein Ribeye, and the direct interaction of Bassoon and Ribeye links the ribbon to the AZ compartment, maintaining its integrity [30]. Additionally, Bassoon participates in the early formation of nascent ribbon synaptic sites during retinal ribbon synaptogenesis [32]. Other interactors for Bassoon include the dynein light chains DLC1 and DLC2, which function as a cargo adapter for the Piccolo-Bassoon transport vesicle (PTV), allowing its retrograde trafficking and the synaptic delivery of AZ proteins [10].

The loss of Bassoon at central synapses shows only a reduced number of fusion competent SV and a decrease in the RRP pool size of vesicles at glutamatergic synapses accompanied by an increase in short-term depression and a high number of silent synapses [33, 34]. The latter could be explained by the interaction of Bassoon with RIM-BP, which modulates the recruitment of CaV2.1 (P/Q-type) channels to SV release sites, thereby contributing to the molecular composition of the neurotransmitter release machinery [35]. On the contrary, loss of Bassoon from non-central sensory synapses alters the structure of the presynapses. These synapses contain a structure called ribbon that holds synaptic vesicles close to the AZ and present a tight vesicle-calcium channel coupling. In the absence of Bassoon, ribbons are loss from the synapses of retinal photoreceptors [36] and cochlear inner hair cells [31, 37, 38]. The loss of Bassoon both from the synapses of ribbon-type [37, 38] and Endbuld of Held synapses (a large synapse in the mammalian auditory CNS) [39] synaptic vesicle replenishment and the size of RRP are decreased suggesting an important role for Bassoon in the plasticity of neurotransmitter release.

A recent Bassoon interaction found in a two-hybrid screen is the protein Mover [40], which negatively regulates synapse release probability at the calyx of Held synapses [41]. It is postulated that Bassoon might recruit Mover into this specific type of synapse, which regulates the release probability [41].

Interestingly, two recently described functions shared by Piccolo and Bassoon are the regulation of the homeostasis of synaptic proteins [42] and the communication between synaptic activity and gene expression [43]. The regulation of homeostasis is performed in part by their zinc finger domain, which binds to the E3 ubiquitin ligase Siah1, inhibiting its function. Siah1 mediates ubiquitination and proteasome-mediated degradation of specific proteins. In neurons, the absence of Piccolo and Bassoon causes a decrease in the content of synaptic proteins due to an increase in ubiquitination and degradation of several AZ proteins and SV proteins [42]. In the absence of Piccolo and Bassoon, the loss of presynaptic proteins occurs through autophagy, but this intracellular degradative process is blocked in the presence of Bassoon, which interacts and inhibits Atg5, an E3-like ligase essential for autophagy [44]. Therefore, Piccolo and Bassoon stabilize presynaptic proteins and avoid premature synaptic degeneration, suggesting a role in presynaptic efficacy by regulating AZ protein turnover. The other shared function of Bassoon and Piccolo is mediated through their interaction with the C-terminal-binding protein 1/brefeldin A-ADP-ribosylation substrate (CtBP1/BARS) [43], called CtBP1, a transcriptional repressor [45] that is translocated between the presynapses and the nucleus carrying information on synaptic activity to modulate gene expression [30]. Hence, Bassoon and Piccolo anchor CtBP1 to the presynapses, allowing this protein to sense synaptic activity. Similar to Piccolo, Bassoon plays diverse functions at the CAZ that go beyond a simple scaffolding protein (Table 1).

ELKS

ELKSs are proteins with a high content of the amino acids glutamate (E), leucine (L), lysine (K), and serine (S). In vertebrates, two ELKS brain-specific isoforms have been described: ELKS1α [46, 47] and ELKS2α [47, 48]. The multi-coiled-coil nature of ELKS allows it to form oligomeric protein complexes with other CAZ proteins such as Munc-13, RIM1, Piccolo, and Bassoon [48, 49]. Both Piccolo and Bassoon bind directly to a central region of ELKS, and it binds to Munc13-1 indirectly through RIM1α. RIM1α binds through its domain PDZ to the carboxyl terminal of ELKS1α and ELKS2α. Liprin-α binds directly to Piccolo and ELKS. These multiple interactions and the fact that in cultured neurons ELKS might function to recruit or stabilize liprin-α and RIM [48, 50], position these proteins as candidates in the molecular organization of presynaptic AZ [50]. ELKS also interacts with Rab6 in a GTP-dependent manner that suggests a role in trans-Golgi network trafficking [51]. Additionally, the ELKS c-terminal domain also binds to a PDZ domain of syntenin-1, a protein involved in cytoskeletal-membrane organization and trans-membrane protein trafficking [52]. This interaction might also contribute to the molecular organization of the CAZ [52].

Overexpression and in vitro studies with cultured rat neurons suggested that ELKS2α is necessary for neurotransmitter release and that the interaction of ELKS2α with RIMs and Piccolo/Bassoon is required for the function and SV release of the AZ [48, 49]. Deletion of ELKS2α in mice causes an increase in inhibitory synaptic responses and the size of the RRP of SV of inhibitory synapses [53]. Interestingly, there were no changes in the overall structure of these synapses or a functional defect in excitatory synapses. The data suggest that ELKS2α is a negative regulator of inhibitory synapses [53]. Additionally, a scaffold function for ELKS2α has been described in ribbon synapses where ELKS2α KO induces reduced AZ size [54].

To avoid compensation mechanisms between the two brain ELKS isoforms, both genes were removed in hippocampal neurons in culture after synapses were established [55]. The simultaneous loss of ELKS1α and ELKS2α resulted in a 50% decrease in the release of neurotransmitters, accompanied by a 30% decrease in presynaptic Ca2+ influx along with a reduction in the probability of release (P) in inhibitory synapses [55]. The findings suggest that a normal influx of calcium into the nerve terminal of hippocampal inhibitory neurons requires direct interaction of ELKS with calcium channels, as was previously reported by Kiyonaka [56]. At excitatory synapses, the removal of ELKS1α and ELKS2α decreases the RRP and neurotransmitter release [57], but P and Ca2+ influx are not affected as occurs in inhibitory synapses when both isoforms are removed [55]. As both isoforms are present in both excitatory and inhibitory synapses, the functional differences observed in the depleted neurons might be explained by synapse specificity of other AZ proteins [57]. The removal of both ELKS isoforms did not affect the number of synapses or appearance by electron microscopy, and the deletion did not produce a decrease in levels of presynaptic calcium channels. However, because ELKSs were removed after the synapses were formed in that study, we cannot discard their participation in the formation of new synapses.

In Drosophila melanogaster, a crucial role in AZ assembly has been proposed for the ELKS homolog, bruchpilot [58, 59]. The bruchpilot N-terminal is homologous to vertebrate ELKS and has a C-terminal that is different from other AZ proteins. Mutants of bruchpilot lack dense projections (T-bars) and suffer from Ca2+ channel-clustering defects. The effect is dramatic because no other isoforms that compensate its function exist in that species. Interestingly, the first study of ELKS in Caenorhabditis elegans showed that this protein was a non-essential player in neurotransmitter release [60]. However, a syd-2 gain of function mutant was able to promote synapse formation in the absence of syd-1, which is essential for C. elegans synapse formation, but only in the presence of ELKS [61]. The mild effects on synaptic structure observed by the deletion of the two ELKS isoforms in the brain of vertebrates and the studies in C. elegans suggest the existence of redundant mechanisms for synapse formation to protect synapse integrity.

RIM

The AZ Rab3-interacting molecule (RIM) [62] is expressed from two RIM vertebrate genes, Rims1 and Rims2, which express five protein isoforms in the brain (RIM1α, RIM1β, RIM2α, RIM2β, and RIM2γ). The RIM molecule has five domains: a N-terminal zinc-finger motif, a central PDZ domain, two C2 domains, and a proline-rich sequence at the carboxy-terminal [63]. RIM1α is located both in the AZ of central synapses and in ribbon synapses and is evolutionary conserved among species. RIM1α, in addition to Rab3, interacts with Munc13-1, liprin-α, and ELKS and forms a protein scaffold at the presynaptic AZ [49]. RIMs also interact with RIM-BPs [64], Ca2+ channels [65, 66], and synaptotagmin 1 [65], which suggested a role in the regulation of SV exocytosis.

The elimination of the major RIM isoform in the mouse brain, RIM1α, did not produce major abnormalities either in the synaptic structure or in the protein composition, but Munc13-1 protein levels were decreased by 60% [67]. However, the RIM1α KO mice showed alterations in synaptic function, such as defects in short-term synaptic plasticity [67], lack of mossy fiber LTP in the hippocampus and the cerebellum [68], and deficits in learning and memory [69]. Additional studies have shown that RIM1α mediates synaptic vesicle docking and priming [3, 67, 70,71,72] and recruitment of Ca2+ channels to the AZ [73,74,75]. The role in SV docking for RIM1α has been supposed because of its interaction with Rab3 and the priming factor Munc13 [71, 74, 76]. RIMs interact directly with Ca2+ channels through its PDZ domain and indirectly through RIM-BPs [74, 77], localizing Ca2+ channels to the AZ. Therefore, the interaction of RIM with RIM-BP is necessary for proper localization of Ca2+ channels close to the synaptic vesicle release machinery [73, 74, 78]. At the functional level, the depletion of the five isoforms caused, in addition to the defect in priming and neurotransmitter release observed in the single gene deletion, a decrease in Ca2+ influx, and diminution of responsiveness and synchronization of release were observed [74]. Elimination of the two RIM-BP isoforms does not affect neurotransmitter release but is necessary for high-fidelity coupling of synaptic transmission [78]. However, simultaneous deletion of RIMs and RIM-BPs affects synapse function by blocking SV priming, delocalizing Ca2+ channels, and altering postsynaptic organization, suggesting a redundant function for these two presynaptic proteins [79].

A function worth mentioning, although observed in the Drosophila neuromuscular junction, is the role of RIM and RIM-BP in homeostatic plasticity. In homeostatic presynaptic plasticity, different levels of postsynaptic receptor perturbation induce compensatory mechanisms at the presynapses. Both RIM and RIM-BP perform this role by modulating the RRP of synaptic vesicles [80, 81], and RIM-BP in addition regulates the presynaptic Ca2+ influx [81].

Hence, the main functions for RIM are SV priming and synaptic plasticity (Table 1).

RIM-BP

Vertebrate RIM-BPs consist of three isoforms containing three SH3-domains, which bind to voltage-gated Ca2+ channels and RIM1α, and two-three fibronectin III repeats [64, 82]. RIM-BP1 and RIM-BP2 present different expression pattern in the brain. About the RIM-BP’s functions mentioned in previous sections, we can highlight the coupling of voltage-gated Ca2+ channels to RIM and Bassoon proteins in order to regulate the strength of synaptic transmission [77]. Hence, brain RIM-BPs, although not essential for synaptic transmission, they are important in the tight coupling between voltage-gated Ca2+ channels and the release machinery [78]. Loss of the main hippocampal isoform RIM-BP2 induces an increase in the distance between Bassoon and the voltage-gated Ca2+ channel subunit CaV2.1, which explain the decrease in both the vesicular release probability and the defect in short-term plasticity [83].

Liprin-α

The liprin-α family of proteins was identified by their interaction with LAR-RPTPs (LAR family of receptor protein tyrosine phosphatases) [84, 85]. In vertebrates, there are four liprin-α genes, liprin-α1, -α2, -α3, and -α4, all of which are expressed in the brain, but the α1 and α4 isoforms are also expressed in non-neuronal tissues. Liprin-α amino acid organization suggests the presence of coiled-coil at the N-terminal and three C-terminal SAM domains [84]. The liprin-α N-terminal binds to itself, forming homodimers, or binds to the AZ proteins, RIM, and ELKS [86,87,88]. The C-terminal of liprin-α binds to liprin-β [89], CASK [90], and LAR-type receptor phosphotyrosine phosphatases [84].

In mature hippocampal synapses, liprin-α2 was found to be a very dynamic protein in comparison with Munc13 and Bassoon, which are very stable [91, 92], and through its interactions with RIM1 and CASK, it regulates presynaptic organization and hence SV release in response to network activity [93, 94]. Elimination of liprin-α2 by knockdown in mature hippocampal neurons does not affect the number of active synapses but does alter the efficiency of SV release by regulating RRP size. An ultrastructural analysis shows lengthening of the synapse and a reduced number of docked vesicles. The presence of liprin-α2 at synapses does not depend on depletion of several AZ proteins [95]. Depletion of liprin-α2 decreases the levels of its direct interactors CASK and RIM and other AZ proteins, such as Bassoon, Rab3, Munc18, VAMP2, and synapsin, and vesicular glutamate transporter VGlut and P/Q voltage-gated Ca2+ channel Cav2.1 are partially diminished [95]. Furthermore, in the absence of liprin-α2, the remaining synaptic CASK and RIM become more stable, suggesting a role in the dynamics of these proteins and consequently SV release efficacy.

In C. elegans and Drosophila loss of liprin-α produced an increase in the size of AZ and affected SV accumulation [61, 96]. Liprin-α participation in presynaptic assembly will be modulated by its interaction with the LAR-type receptor phosphotyrosine phosphatase PTP-3, thereby stabilizing the active zone, Ca2+ channels and SV by linking synaptic cell adhesion to core AZ proteins [88].

In summary, vertebrate liprin-α2 association with the CAZ is regulated by synaptic activity and is a key organizer of mature presynapses and modulates the dynamics of RIM and CASK, which regulate synaptic plasticity. In vertebrates, there are no studies of liprin-α2 depletion before synapses are formed, and hence, a role in the assembly of new presynapses cannot be ruled out.

Munc13

The Munc13 family comprehends the three specific brain isoforms, Munc13-1, bMunc13-2, and Munc13-3 [97]. Additionally, there is a ubiquitously expressed Munc13-2 splice variant called ubMunc13-2 [98]. Munc13-1 and ubMunc13-2 have the same domain structure: (a) N-terminal C2 domain (C2A) and a Ca2+/CaM-binding site, (b) a central diacylglycerol and phorbol ester–binding C1 domain and a second C2 domain (C2B), and (c) a C-terminus with a Munc13 homology or minimal Munc13 priming domain and a third C2 domain (C2C). The isoform Munc13-1 is the most abundant in the brain, and only 10% of cortical and hippocampal synapses contain both Munc13-1 and bMunc13-2 [99]. Both bMunc13-2 and Munc13-3 differ from the other two isoforms at the N-terminus [100,101,102].

The two main functions described for Munc13 in neurotransmitter exocytosis are SV priming and modulation of presynaptic plasticity. The first function is accomplished by acting on the SNARE/SM protein fusion machinery, resulting in SV competent for exocytosis. Munc13s prime SVs by acting on syntaxin, which is a t-SNARE (acronym for SNAP Soluble NSF Attachment Protein REceptor). A SNARE complex is formed on the vesicle side by SNAP-25, and syntaxin and synaptobrevin proteins are located at the target synaptic membrane. During SV priming, syntaxin-1 shifts from a closed state that binds Munc18-1 toward an open state conformation that is able to form part of the SNARE complex; this last step is accelerated by Munc13s [103]. The isoforms Munc13-1 and ubMunc13-2 bind to the Zn2+ finger region of αRIMs via their conserved N-terminal region [97, 104, 105], thereby forming Munc13-RIM1α-Rab3 complexes which are a requisite for Munc13s-mediated SV priming [70]. Although the recruitment of the primary brain Munc13-1 isoform to AZ requires its interaction with RIM1, the anchoring of bMunc13-2 is mediated by ELKS1α in a small subset of synaptic terminals in hippocampal neurons [99]. This synaptic specific anchoring explains the molecular and functional heterogeneity of presynaptic AZs.

As Munc13 KO hippocampal neurons show normal AZ structure [97], a role in AZ assembly is discarded, but the N-terminal region of Munc13-1 may be the hub for the AZ proteins Piccolo, Bassoon, ELKS, and RIM1, which serve as a core for the physical and functional integrity of the protein machinery at the AZ, thereby orchestrating SV priming [106]. However, all AZ proteins seem to have that characteristic according to their multi-interactions.

Mechanisms Modulating Recruitment of AZ Proteins to Sites of Newly Forming Synapses

AZ Proteins, Cell Soma Packing, and Traffic along the Axon.

The generation and maintenance of functional presynaptic sites require time and site-specific delivery of AZ and SV components. In the last two decades, sustained advances have been made in understanding the mechanisms involved in the transport of presynaptic proteins. Those studies have shown that presynaptic proteins are not transported as individual units, but they travel along axons in groups linked to vesicles originating at the Golgi apparatus. In 1998, Nakata et al. [107] used GFP-tagged proteins and laser scan microscopy to show that putative precursors of SV were transported by tubulovesicular organelles, suggesting that SVs are not transported as a mature unit but are synthesized locally by recycling at the nerve terminal [107]. Ahmari et al. strengthened the idea of precursor vesicles as sources of presynaptic specialization [5]. In their study, they used the synaptic vesicle protein VAMP tagged with GFP and time lapse microscopy combined with DIC imaging to study the dynamics of VAMP-GFP in young hippocampal neurons in culture before synapses were formed [5]. They observed mobile packets that stop at sites of axon-dendrite contacts and analysis by electron microscopy showed that the contact areas contained tubulovesicular structures, dense core vesicles and small pleomorphic vesicles with no resemblance to mature synaptic vesicles [5]. Those contact areas apparently became functional presynaptic recycling sites, as evidenced by FM 4-63 uptake, soon after contacts were formed [5, 108]. Other proteins such as SV2, synapsin I, and calcium channel subunit α1 were found on those packets, suggesting that the AZ in bulk might be transported in the VAMP-GFP labeled packets [5] (Fig. 3). Thereafter, a specific AZ precursor vesicle immunoisolated from young rat brains was identified to transport Piccolo and Bassoon [8]. This vesicle with an 80-nm diameter had a dense core suggesting the transport of secreted synaptogenic factors, and analysis by western blot identified the presence of a plethora of other AZ proteins [8, 48]. However, no synaptic vesicle proteins were detected, suggesting that this factor was a specific AZ protein transport vesicle [8]. This AZ precursor vesicle, named PTV, has an origin at the trans-Golgi network (TGN) where Piccolo, Bassoon, and ELKS are recruited [6, 109, 110]. Additional evidence for the hypothesis of a multivesicular mechanism for presynapse formation came from the studies of Tao-Cheng, who used an ultrastructural analysis to show that AZ and SV proteins are transported together in large aggregates, but they are carried in different types of vesicles [7]. Interestingly, not all AZ proteins seem to be associated with the same vesicle (e.g., Munc13-1α is transported in a different TGN-derived vesicle than Piccolo and Bassoon) [6], and RIM1α only seems to associate with PTVs during axonal trafficking [6]. RIM has also been associated with a vesicle that transports neurexin, CASK, and voltage-dependent Ca2+ channels [111] (Fig. 3). Hence, as golgi-derived PTVs travel along the axon, they are thought to suffer further maturation before reaching nascent presynapses [6] (Fig. 3).

Fig. 3
figure 3

Schematic representation of the precursor vesicles model of active zone formation. The figure shows the mechanism of axonal transport of several presynaptic proteins during synapse formation. Piccolo, Bassoon, and ELKS exit the trans-Golgi network associated to Golgi-PTV (gPTV), with Munc13 using a different Golgi-derived vesicle. RIM and Munc13 also associate to a soluble pool. During its traffic along the axon, RIM and Munc13 are loaded by an unknown mechanism into gPTV, which turns into mature PTV (mPTV). Synaptic vesicle proteins and other presynaptic protein use different Golgi-derived precursor vesicles with pleomorphic shapes

Full size image

The traffic of presynaptic precursor vesicles involves motor proteins that transport cargo bidirectionally along actin and microtubule cytoskeletal tracks. Actin employs myosin, and microtubules use kinesin and dynein as motors. The specificity of the transport is provided by a molecular adapter that is part of the vesicle. In the case of PTVs, Bassoon interacts with DLC1 and DLC2 and functions as a cargo adapter for retrograde trafficking of this vesicle [10]. Although PTVs move both anterograde and retrograde, their net movement is anterograde; however, the retrograde movement regulated by Bassoon is necessary for synaptic delivery of AZ proteins [10]. Furthermore, syntabulin, a kinesin-1 family member 5B motor adaptor protein [112], mediates the transport of AZ components through an unknown vesicle adapter during synapse formation and during synaptic plasticity [113].

Thus, the assembly of the presynapse seems to occur by the simultaneous deposition of SV proteins and AZ proteins, which are carried in different types of vesicles (Fig. 3). Precursor vesicles for synaptic vesicles (STVs) and AZ proteins will then be transported in axonal aggregates [7, 114], which co-pause at common axonal sites, probably responding to unidentified local signals specific for each type of vesicle [115]. Therefore, the sites in the axon where these vesicles deliver their content seem to be predefined and independent of the existence of a neuronal contact [114]. The signals determining the stop and clustering of AZ proteins at newly forming synapses remain to be identified.

Homo and Hetero-Oligomerization as a Mechanism of AZ Trapping/Assembly

After biogenesis and axonal trafficking, presynaptic proteins have to be trapped and then maintained at the AZ of mature synapses. The conformation state of the protein might be relevant for its correct trapping and posterior proteome assembly. Several years ago, we proposed a theoretical model of AZ protein trapping [116] that postulates that particular presynaptic proteins undergo a prion-like concentration-dependent conversion, adopting a conformation that will stimulate their own aggregation and aggregation of other proteins. Generally, domains that aggregate in prion proteins are rich in the amino acids glutamine (Q) and asparagine (N). Two vertebrate presynaptic proteins with this characteristic are Piccolo and Bassoon, which have Q-rich areas. Interestingly, Bassoon and Piccolo are known to be homo- and heterodimerizing binding partners [28, 110, 117] and form large aggregates when overexpressed in neurons and heterologous cells [15, 110, 118], and in the case of Bassoon, segments of the protein that do not contain Q domain are unable to interact with the AZ. These presynaptic proteins, with prion-like domains, seem to be under the control of a specific enzyme that modulates their state of aggregation. That is the case with the Drosophila AZ protein, bruchpilot, the homolog of ELKS in vertebrates that contains regions rich in Q or Q/N and is essential for the structural integrity of the Drosophila AZ. Bruchpilot travels along the axon and is associated with a protein complex that contains the motor adaptor protein Aplip1 [119]. The presence of Aplip1 allows proper transport in axons and avoids premature aggregation of the bruchpilot molecular complex [119], an event regulated by the Serine Arginine Protein Kinase (SRPK)79D. SRPK79D was identified in Drosophila by two parallel studies [120, 121] and was found to colocalize with the T-bar-associated protein bruchpilot in both axons as synapses. A mutated SRPK79D causes nerve bruchpilot aggregates in motoneurons [120]. Curiously, although ELKS does not have Q domains, it has a tendency to aggregate if expressed in heterologous cells. Therefore, the presence of coiled-coil domains in its structure, the region homologous to the N-terminus of bruchpilot, might also be the key in its self-clustering and binding to other AZ proteins [48, 49].

A similar oligomerization mechanism has been observed in C. elegans because a mutation in a small protein called Arl-8 produces abnormal clustering of Rab-3, UNC-10, and SYD-2 (the last two are homologous to the mammalian RIM and liprin, respectively) close to the cell body, suggesting a premature and abnormal delivery of AZ proteins. Therefore, Arl-8 would be necessary to avoid ectopic aggregation of presynaptic proteins in this nematode [122]. The abnormal accumulation in these mutants was partially suppressed by mutation in a JNK MAP kinase pathway. Hence, in C. elegans AZ proteins, aggregation is regulated by the interplay between Arl-8 and the JNK MAP kinase pathway.

Therefore, a conserved and specialized mechanism exists in vertebrate and invertebrate neurons that regulates the temporo-spatial aggregation of particular presynaptic proteins to avoid premature AZ protein interactions, trapping, and assembly, and the regulating molecules need to be identified.

AZ Proteins and Its Relationship with the F-actin Dynamic

Actin is globular and is the most abundant protein in the majority of eukaryotic cells, playing several roles in cells, such as cellular movement, scaffolding, and intracellular trafficking. Actin has the capacity to polymerize and form filamentous-actin (F-actin) by a highly dynamic process that is under the control of diverse known and unknown clues that determine the function of actin in a specific time and space manner. During CNS development, neurons migrate and develop axons and dendrites to build an intricate network of communication. Axon outgrowth occurs until a contact with the appropriate postsynaptic partner triggers formation of a synapse. During all these processes, the actin cytoskeleton and mainly F-actin participates actively. It is well known that actin is present in young presynaptic terminals and is involved in the assembly and development of presynaptic specializations, participating as a primordial scaffold [123,124,125,126]. Accordingly, AZ assembly and structure in young neurons but not in old neurons is F-actin dependent, as disrupting agents such as latrunculin block presynapse formation [125]. However, in mature presynapses, F-actin plays primarily a structural role holding SV and preventing its premature non-regulated fusion [124, 127], thereby modulating synaptic transmission and efficacy [128]. Interestingly, AZ and SV precursor vesicles that deposit at newly forming presynapses utilize different and not well-understood mechanisms of F-actin dynamics, indicating the existence of two structurally separate F-actin pools at this location early during synaptogenesis [129].

There is not much knowledge about which AZ proteins participate in presynaptic F-actin assembly. In C. elegans, the AZ protein NAB-1/neurabin mislocalizes if F-actin is disassembled by latrunculin [130]. NAB-1/neurabin is an actin-binding protein that recruits AZ proteins SYD-1 and SYD-2 (liprin-α) (core proteins in C. elegans AZ assembly) acting as a bridge between F-actin and AZ proteins during synapse development [130]. In the case of vertebrates, Piccolo is the only AZ protein known to be required for the activity-dependent assembly of presynaptic F-actin through its interaction with actin-binding proteins. These proteins are Profilin2 [29], Epac2 [131], Abp1 [15], GIT1 [14], and Daam1 [18]. The role of these interactions has been described in mature synapses where they regulate the delivery and recycling of SVs at the presynaptic terminal [18]. In this respect, it is postulated that Piccolo serves as a platform coordinating the activity of Profilin2, GIT1, and Daam1 with the spatial assembly of F-actin, which is necessary for the recruitment of CaMKII and the regulation of the kinetics of Synapsin1a during activity-dependent exocytosis [18, 20, 132]. In other words, the interaction of Piccolo with these proteins will hold SV at the reserve pool by modulating synapsin I through F-actin assembly [20, 132].

Active Zone and Wnt Signaling

Wnt signaling plays diverse functions in the development of the mature nervous system. During brain development, Wnt proteins play critical roles in cell differentiation, migration, neurite polarization, and synapse assembly and plasticity [133,134,135]. In the adult nervous system, Wnt signaling is required for synapse maintenance, synaptic activity, and plasticity [136,137,138,139]. There are 19 Wnt ligands [140] that activate three alternative signaling pathways: (1) the canonical Wnt/β-catenin pathway, (2) the Wnt-Planar Cell Polarity (Wnt/PCP) pathway, and (3) the Wnt/calcium pathway [134, 141, 142]. All three pathways are activated by the binding of Wnt ligand to a Frizzled (Fz) receptor, which activates intracellular dishevelled (Dvl) protein. In the canonical pathway, the Wnt ligand signals through β-catenin, which enters the nucleus to activate Wnt target genes. In the Wnt/PCP pathway, the Wnt ligand binds to its receptor Fz, thereby activating Dvl, which signals through two independent and parallel pathways activating the small GTPases Rho and Rac. The activation of Rho GTPase occurs through Daam1 leading to the activation of the Rho-associated kinase Rock and consequently cytoskeletal organization [143]. The other pathway signaling through Rac, which in turn activates c-Jun N-terminal kinase (JNK) targeting gene transcription that culminates with the reorganization of the cytoskeleton [144, 145]. In the Wnt/Ca2+ pathway, the ligand binds to Fz receptors, activating classical G protein pathways and phospholipase C (PLC), which acts on phosphatidylinositol 4,5-biphospahte (PIP2) and produces diacylglycerol (DAG) and inositol triphosphate (IP3). This process generates an increase in intracellular Ca2+ that activates Ca2+-dependent proteins and the transcription factor nuclear factor associated with T cells (NFAT) to promote the transcription of target genes [139, 146].

Wnt signaling has shown to play a role not only in axon guidance and remodeling but also participates in presynaptic assembly [136] (Table 2). In the cerebellum, granular cells secrete Wnt7a, which induces mossy fiber axonal spreading and branching accompanied by an increase in the clustering of synapsin I [147, 148]. Similar results were observed in rat hippocampal neurons where Wnt7a stimulates the clustering of synaptophysin and induces recycling and exocytosis of SV [149]. The ligand Wnt7b induces clustering of VAMP2 both in mossy fibers and hippocampal neurons as early as 15 min of treatment [150]. The AZ protein Bassoon clustering is increased in 10 DIV hippocampal cultures treated with Wnt7b [150], and Dvl was found to be necessary for the clustering of this AZ protein [150]. Additionally, Wnt3a/Fz1 in hippocampal neurons in culture stimulates the clustering of Bassoon and increases the number of functional presynaptic sites [151]. Interestingly, the Wnt-mediated effect of presynaptic clustering of both synaptic proteins and AZ proteins are observed at 15–30 min, suggesting that immediate local changes induce clustering [149,150,151].

Table 2 Wnt ligands inducing synaptic protein clustering
Full size table

As mentioned above, there would be two distinct pools of F-actin at new forming presynapses, one that participates in the recruitment of AZ proteins and another that participates in the recruitment of SV proteins during synapse formation [129]. As clustering of AZ proteins and SV proteins is induced both by Wnt3a and Wnt7a and requires Dvl, the Wnt pathway diverges after Dvl, acting specifically on those F-actin pools. Piccolo might act as a link between Wnt signaling and the cytoskeleton because it interacts with AZ proteins, diverse actin-binding proteins (see “Piccolo” section) and with two proteins of the Wnt pathway: (1) Daam1, which has been postulated to modulate actin dynamics through Wnt/PCP signaling [18, 143, 152, 153], and (2) Rho-GEF Trio, which also interacts with Bassoon [19] and is an activator of the Rho family of GTPases [154] and F-actin dynamics. Interestingly, Piccolo interacts with Daam1 only when it is in its open activated conformation [18]. The latter suggests that presynaptic assembly in young neurons and/or synaptic efficacy in mature neurons mediated by Piccolo might be regulated by the Wnt/PCP signaling (Fig. 4).

Fig. 4
figure 4

Piccolo could be an important link between Wnt signaling and presynaptic assembly. In the left panel is represented Piccolo as the hub between active zone proteins, F-actin dynamic, and Wnt signaling molecules. The panel on the right represents a simplified model of how the Wnt signaling pathway would mediate the assembly of the active zone through Piccolo. Piccolo through its interaction with actin-binding proteins (ABP), and molecules of the Wnt signaling would allow communication of the Wnt signaling with the organization of the active zone

Full size image

Nervous system integrity requires that synapse assembly be tightly coordinated with synapse maturation and maintenance. Fulfillment of this requisite depends on precise control of both protein and organelle synthesis and degradation. Some of the postulated mechanisms that account for control of neuronal protein and organelle half-life include involvement of the ubiquitin/proteosomal system [155,156,157] and autophagy [158, 159]. Another likely requirement includes tight regulation of local calcium levels in order to support neuronal life and synaptic stability [160, 161]. A defect in either of these systems leads to neuronal degeneration. Although it is not the focus of the present work to review the mechanisms that regulate the intracellular processes involved in synapse maintenance the evidence suggests a role for the Wnt pathway. In fact, Wnt signaling, in addition to playing a role in the development of the central nervous system, is one of the mechanisms postulated to regulate synaptic stability. It has been shown that neuronal activity promotes the stability of synapses by modulating the levels of endogenous-secreted Wnts [162,163,164,165]. Incubation of neurons with Dickkopf, a secreted Wnt pathway antagonist, delocalizes pre- and postsynaptic components in mature and stable hippocampal synapses inducing disassembly of synapses in mature neurons [166]. Also, it was recently reported that Wnt5a is necessary for maintaining dendritic arbor and spines in the adult hippocampus [167]. Future studies are needed to decipher the downstream signals of the Wnt receptors that would control the assembly and maintenance of the synapse.

Concluding Remarks

In vertebrate CNS synapses, there is no protein to which a strict role in AZ assembly can be attributed, mainly because studies of loss of function do not show an evident abnormal presynaptic structure. However, this lack of evidence does not mean that an alteration in one of the AZ proteins could produce an abnormal synaptic function over time. Three presynaptic mechanisms that will contribute to stability of vertebrate central synapses are the following: (1) AZ proteins with overlapping functions, (2) additional isoforms, and (3) the existence of proteins with related structural domains.

In the vertebrate brain, the AZ proteins Piccolo, Bassoon, RIM, RIM-BP, Munc13, liprin-α, and ELKS form an interaction network that gives the structural framework to the CAZ and allows regulated communication between them to respond efficiently to synaptic demands. These proteins also interact temporarily with other molecules that regulate F-actin dynamics and consequently synaptic plasticity according to the prevailing physiological situation. This group of proteins participates directly in the recruitment of Ca2+-channels in which RIM and Munc13 have a more specialized role in docking and priming of SV, and Bassoon with Piccolo will be key in the stability of several presynaptic proteins. Furthermore, among AZ proteins, Piccolo seems to be one of the most versatile component of the CAZ due to its participation in multiple functions at the presynapses. Interestingly, a newly emerging function of some AZ proteins is to communicate with the transcription apparatus in the nucleus to inform the actual synaptic activity.

Therefore, although substantial information has been accumulating about the AZ interactome and functions of its constituents, there are many questions that remain to be answered, such as the cell signaling pathways regulating the dynamics of these interactions during synapse formation and plasticity, and how a defect in any of these interactions is translated to cognitive impairment in childhood and adulthood.