Zebrafish neuromuscular junction; The power of N

Abstract

In the early 1950s, Katz and his colleagues capitalized on the newly developed intracellular microelectrode recording technique to investigate synaptic transmission. For study they chose frog neuromuscular junction (NMJ), which was ideally suited due to the accessibility and large size of the muscle cells. Paradoxically, the large size precluded the use of next generation patch clamp technology. Consequently, electrophysiological study of synaptic function shifted to small central synapses made amenable by patch clamp. Recently, however, the unique features offered by zebrafish have rekindled interest in the NMJ as a model for electrophysiological study of synaptic transmission. The small muscle size and synaptic simplicity provide the singular opportunity to perform in vivo spinal motoneuron-target muscle patch clamp recordings. Additional incentive is provided by zebrafish lines harboring mutations in key synaptic proteins, many of which are embryonic lethal in mammals, but all of which are able to survive well past synapse maturation in zebrafish. This mini-review will highlight features that set zebrafish NMJs apart from traditional NMJs. We also draw into focus findings that offer the promise of identifying features that define release sites, which serve to set the upper limit of transmitter release. Since its conception several candidates representing release sites have been proposed, most of which are based on distinctions among vesicle pools in their state of readiness for release. However, models based on distinctions among vesicles have become enormously complicated and none adequately account for setting an upper limit for exocytosis in response to an action potential (AP). Specifically, findings from zebrafish NMJ point to an alternative model, positing that elements other than vesicles per se set the upper limits of release.

A. The early roots of N.

Pioneering studies by Katz and his colleagues on the frog NMJ led to the first quantitative description of chemical synaptic transmission (Fatt and Katz, 1952; del Castillo and Katz, 1954; Katz, 1966). A simple formulation was based on the key finding that release was quantal in nature and conformed to a binomial distribution. Katz proposed that release occurred from a fixed number of presynaptic reactive sites or release sites, collectively termed N. N serves to set the upper limit for transmitter release, which, at the functional level, reflected the theoretical maximum number of quanta released in response to a single motoneuron AP. Fluctuations in the amount of transmitter release associated with successive APs could be ascribed to failure of individual reactive sites to successfully release quanta. The total release then was proportional to N*Pr where Pr represented the probability of release for each release site. Inclusion of the mean size of a single quantum (N*Pr*q) yielded the mean amplitude of the recorded endplate response. The formulation represented a general framework for describing chemical synaptic transmission that was applied universally. The physical counterpart of q was quickly identified in form of synaptic vesicles (De Robertis and Bennett, 1954; Robertson, 1956). Confirmation that these vesicles represented the source of quantal release was provided by images of exocytosis, validating the Katz proposal (Heuser and Reese, 1973; Heuser et al., 1979). Identification of the structural entity representing a release site has proven much more challenging.

The purpose of this perspective is to draw into focus several issues related specifically to the original concept of N as it pertains to vertebrate synapses. This review is not intended to provide comprehensive coverage of NMJs and we have focused on vertebrate preparations. For comprehensive coverage the reader is directed to Slater, 2015 and Homan & Meriney, 2018. Findings from CNS synapses have pointed to greater complexity in transmitter release than provided for by the simple model developed for the NMJ by Katz, hampering exploration of the release site entity. However, the recent mining of the zebrafish NMJ returns us to the roots of the operationally defined release site in setting the upper limits for quantal release and recovery kinetics during repetitive bouts of stimulation (Wen et al., 2016). As such, we elaborate on the utility of zebrafish NMJ as a model for synapse function and for interrogation of the structural counterparts of release sites (sections D and E).

B. Estimating N through physiology

In theory, at synapses where Pr is near unity, N could be estimated on the basis of quantal content (Qc). Qc reflects the average number of quanta that are released in response to an action potential. At the NMJ Qc is best estimated under voltage clamp where the mean amplitude of the endplate current (EPC) is divided by the mean amplitude of the miniature endplate current (mEPC). Unfortunately, at the frog and mammalian NMJs, as well as at most central synapses estimates of Pr are highly variable and generally low, further complicating interpretations of N (Slater, 2015; Dittrich et al., 2018; Homan and Meriney; 2018). Estimation of N has relied principally on quantal analysis, which involves fitting experimental measurements of release, usually postsynaptic response amplitudes, with statistical models. Thus, the accuracy of N estimation is dependent on the validity of the critical assumptions associated with these models. For example, the simple binomial model holds that Pr is uniformly distributed among the release sites (Del Castillo and Katz, 1954). However, subsequent studies drew that assumption into question at certain central synapses (Walmsley et al., 1988; Rosenmund et al., 1993; Dobrunz and Stevens, 1997; Murthy et al., 1997) as well as the frog (D’alonzo and Grinnell, 1985; Bennett and Lavidis, 1989) and mammalian (Wang et al., 2010; Gaffield et al., 2009; Tabares et al., 2007) NMJs. In response to these findings it has been suggested that quanta are released with significantly higher probability from a restricted subset of N (Slater, 2015). Mechanisms causal to the differences in Pr likely involve differences in the density of calcium channels or calcium-binding proteins, and the coupling between them, central determinants shown to limit release at low probability release synapses (Pawson and Grinnell, 1984; Bennett et al., 1986; Dittrich et al., 2018). Another problem lies in a central tenant underlying the estimation of N among all synapses, which assumes that the amplitudes of the unitary events are normally distributed. At nearly all synapses tested the smallest unitary synaptic currents cannot be resolved, a problem exacerbated by incomplete voltage control. A lethal blow to this assumption followed identification of multi-vesicular release (MVR); a process ascribed to nearly simultaneous release of multiple quanta from individual release sites (Kriebel and Gross, 1974; Tong and Jahr, 1994; Wadiche and Jahr, 2001; Pulido and Marty, 2017). In such cases, the quantal amplitude distributions are typically skewed, compromising determination of mean unitary amplitude.

Measurements of quantal release are also fraught with postsynaptic complications. Saturation, desensitization and occlusion of postsynaptic glutamate and GABA receptors at central synapses often leads to large variability in amplitude of synaptic currents during bouts of repetitive stimulation. As a means of correction, deconvolution algorithms have been developed for analysis of synaptic currents at simple glutamatergic synapses (Malagon et al., 2016). By contrast, acetylcholine receptors (AChRs) at the NMJ are not prone to desensitization at low concentrations of ACh and the high postsynaptic density of receptors, along with fast binding and unbinding rates minimize saturation and persistent occlusion of receptors by ACh (Fertuck and Salpeter 1976; Adams, 1981). As a consequence of all of these problems the lens that was once centered on N has been considerably defocused with the result that most physiological characterization has relied simply on Qc as an estimate of release, despite its limitations.

Recently, the application of multiple-probability fluctuation analysis (MPFA) of synaptic currents has offered an effective means for estimating N (Silver et al., 1998; Clements and Silver, 2000; Silver, 2003). Specifically, the variance-mean plots generated over a range of Pr conditions offer solutions to some of the aforementioned limitations. The resultant plots generally conform to a parabolic relationship providing estimates of q and N. The decline in variance associated with the increased mean amplitude reflects the saturation expected on the basis of high Pr. Accordingly, N represents a fixed number of independent release sites that continues to call for exploration into the structural correlates.

C. Searching for structural correlates of N

The frog NMJ was the first vertebrate synapse to be deeply investigated at the level of electron microscopy (Birks et al., 1960; Heuser and Reese, 1973; Heuser et al., 1979). In particular, freeze fracture of presynaptic fingerlike projections revealed a repeating pattern of transmembrane particles called active zones. When the presynaptic rows were first observed it sparked the idea that N represented the number of active zones, an easily determined value for comparison to the physiologically determined estimates (Heuser and Reese, 1981). However, when the number of active zones outnumbered the N estimated from binomial theorem by 5 to 10 times, this simple idea was put to rest (Slater, 2015). A second opportunity for direct identification of N was offered by an inhibitory synapse in goldfish formed on the Mauthner neuron. The finding that estimates of N, based on a binomial fit, matched the number of presynaptic boutons led to the proposal that each bouton released only a single vesicle (Korn et al., 1981). This idea received additional support from synapses formed between hippocampal pyramidal cells and interneurons (Gulyas et al., 1993; Biro et al., 2005) and excitatory synapses at the barrel cortex (Silver, 2003) where there is one-to-one correspondence between the number of functional release sites and the synapses identified through electron microscopy. Should each bouton actually represent a single release site then there would be a straightforward means to identify the structural counterparts of N. However, subsequent identification of aforementioned complications, principally MVR, cast doubt on the interpretation (Korn et al., 2007). As set out in sections D and E, zebrafish NMJ provides renewed support for the assertion that each synaptic bouton releases a single vesicle, a promising test ground for identifying those physical elements that set the upper boundaries of quantal release.

The structural counterparts for N have turned increasingly towards vesicle-based candidates. The identification of vesicles differing in their state of readiness for release has its roots at frog NMJ where direct imaging of exocytosis and physiology led to the identification of distinct vesicle pools, reserve, recycling and readily release (RRP) (Elmqvist and Quastel, 1965; Richards et al., 2003). Currently, the physical identification of ‘docked vesicles’ among the RRP takes on specific importance as it relates to N (Imig et al., 2014; Miki et al., 2016; Stanley, 2016; Kaeser and Regehr, 2017; Pulido and Marty, 2017). At nearly all synapses examined, electron microscopy has revealed a subset of vesicles that are juxtaposed with the plasma membrane in what appears to be fusion ready condition. Where each vesicle is positioned is operationally called a “docking site” based on electron microscopic study. Now, a picture is emerging with vesicle docking sites setting the upper limits of release and with occupancy of docking sites setting Pr (Ruiz et al., 2011; Trigo et al., 2012; Pulido et al., 2015; Miki et al., 2016). Models using docking site number and time-dependent site occupancy adequately describe the vesicular release statistics at single GABAergic synapses (Pulido et al., 2015). However, at mammalian NMJs the number of docked vesicles greatly exceeds the Qc (Slater, 2015). This discrepancy is even more obvious at the zebrafish NMJ (Helmprobst et al., 2015; Wen et al., 2016). A potential solution is that not all docked vesicles are fusion ready, thereby calling for an additional layer of heterogeneity (Biederer et al., 2017). This idea has ushered in a further distinction among docked vesicles in their speed of recruitment (Kaeser and Regehr, 2017) reflected in the proposed existence of ‘primed’ and ‘super-primed’ subclasses (Neher, 2017).

The proposition that intrinsic molecular based differences between vesicles serve as determinants of the readiness of release receives, to date, no direct experimental support. Thus, in their present form, vesicles per se do not constitute a viable definition of N. As a means of addressing the structural counterpart for N we consider that, for a release site to function, it requires three components; a population of vesicles in a state that is fusion ready, calcium channels precisely located near the site of fusion and in sufficient density to serve as the trigger and canonical exocytotic machinery to carry out the membrane fusion. In this scenario the coincidence and proximity of these three elements would define a functional release site, the total number of which set the upper boundary of N.

This picture of release site receives support from recent ultrastructural analysis of presynaptic molecules coupled with functional assay at individual synapses. A number of studies have shown that key active zone proteins such as Rim, Munc13 and P/Q type calcium channels are not distributed randomly but rather form Nano-clusters (Holderith et al., 2012; Tang et al., 2016; Sakamoto et al., 2018). Furthermore, at the hippocampal synapses these Nano-clusters tend to align with each other and with optical markers for exocytosis, serving as an attractive candidate for the physical correlate of a functional release site (Tang et al., 2016; Biederer et al., 2017; Sakamoto et al., 2018). Presynaptic multi-domain proteins such as Bassoon, Piccolo and ELKs are capable of direct binding to vesicle, calcium channels and SNAREs (reviewed by Sudhof, 2012; Hallermann and Silver, 2013). It is conceivable that these cytomatrix proteins play critical roles in aligning the three components representing a release site, providing additional regulation of N. With the advent of new fixation methods that preserve the native cytomatrix structures at the active zone it may soon become possible to view directly how the release site components are linked together to support vesicle fusion.

D. Zebrafish NMJ as a simple model synapse

Alongside the historical frog NMJ now stands a zebrafish counterpart that offers greater simplification in terms of both anatomy and function, while still sharing the key functional features (Wen and Brehm, 2010; Wen et al. 2016). Each tail segment contains 4 large primary neurons and over 40 secondary motoneurons. The primary types are responsible for the powerful bend that initiates swimming and secondary types are recruited to swimming according to the size principle, with smaller diameter neurons causal to slow swimming speed and larger diameter neurons regulating faster speeds (McLean et al., 2007). The size of secondary motoneuron is also related to both the Qc and the number of muscle cells that form synaptic contacts with the motoneuron (Wang and Brehm, 2017). By contrast, The 4 primary motoneurons have non-overlapping muscle quadrants such that each innervates all of the fast axial muscle cells in the relevant quadrant. These anatomical and functional distinctions between primary and secondary motoneurons have only recently come to light for zebrafish and will not be a subject of this review. Instead, we focus entirely on the synapses formed by the CaP primary motoneuron and fast-twitching axial ventral muscles because it has greatly facilitated the use of paired recordings (Wen et al., 2005; Wen and Brehm, 2010). Consequently, we use the zebrafish CaP NMJ as a metric for comparison to frog and mammals, citing the following advantages.

1-. Highly ordered structure with spatially segregated boutons (Fig. 1-1)

Figure 1.

(1) Reconstruction of CaP motoneuron branches (blue) and synapses (cyan and pink spheres). A single muscle cell (yellow) with the associated synapses (pink spheres) is indicated. The inset shows a single dissociated muscle with its individual synaptic puncta labeled with abungarotoxin (a-btx, red). The right panels show co-localized labels for postsynaptic receptor (a-btx, cyan), synaptic cleft acetylcholinesterase (fasII, red) and presynaptic vesicle marker (FM1–43, green). (2) Cartoon (left) and experimental image (middle) depicting the paired recording configuration with a CaP neuron (green) and target muscle (red). Right panel shows a sample CaP current clamp recording of an AP (black) and associated EPC (red) with exponential fit of the decay phase (grey). (3) Sample recordings of spontaneous and asynchronous mEPCs with muscle held at −50mV. Amplitude distributions for spontaneous (black) and asynchronous (gray) mEPCs fit individually by Gaussian functions (red). The scatterplot of mean mEPC amplitudes for individual recordings along with mean +/− S.D. (4) Sample APs from the CaP soma and associated EPCs from target muscle during 0.2Hz stimulation. The amplitude and Qc estimates for individual recordings determined by mean EPC/mean mEPC amplitudes. (5) The relationship between mean EPC amplitude and external calcium concentrations (color indicated) for individual recordings fit with a Hill function (black). (6) The relationship between current amplitude at those different calcium concentrations and associated variance were fitted to a parabolic function using data obtained using 0.2 Hz stimulation. A scatterplot of estimated Pr in panel 5 and N in panel 6 are shown for all recordings. Data was adapted from Wen and Brehm, 2010; Wen et al., 2016; Wang and Brehm, 2017.

The CaP branches extensively to innervate all of the ventral-most fast skeletal muscle cells through formation of ~15 well segregated contacts on each muscle. cf: In mammals and frogs each muscle cells receives innervation by one of many candidate motoneurons. The most commonly studied vertebrate NMJs are compressed boutons into a linear array of fingerlike projections (frog) and pretzel shapes (mammalian) that are structurally complex (Slater, 2015; Homan and Meriney, 2018).

2-. Accessibility to simultaneous paired patch clamp recording (Fig. 1-2)

Zebrafish provides the opportunity to perform current clamp stimulation of the CaP motoneuron and simultaneous voltage clamp recording from target muscle (Wen et al., 2005; Wen and Brehm, 2010). cf: The large size of most muscle types in mammals and frogs precludes postsynaptic patch clamp recordings. The inability to identify the target of individual motoneurons precludes paired recordings. The only other vertebrate NMJ to permit simultaneous pre and postsynaptic recordings to date has been Xenopus spinal neuron-myotomal muscle co-culture (Feng and Dai, 1990; Hulsizer et al., 1991; Thaler et al., 2001;Yazejian et al., 2013).

3-. Full resolution of single quantal events (Fig. 1-3)

The high input resistance of muscle cells renders zebrafish NMJ unique in the ability to use patch clamp technology to resolve all of the quantal events termed miniature endplate currents (mEPCs). Shown for comparison, is the amplitude distribution for the spontaneous mEPCs recorded in the absence of stimulation and the amplitude distribution of asynchronous mEPCs recorded during and after 10s/100Hz stimulation. cf: Nearly all mammalian and frog muscles studied to date are low resistance and synaptic responses are subject to loss of amplitude due to cable properties. The two electrode clamp measurements of mEPCs suffer from poor signal to noise and small amplitude, stemming from a low input resistance.

4-. Small quantal content (Fig. 1-4)

The ability to fully resolve quantal events along with a normal distribution of amplitudes provides direct estimates of quantal content which, during low frequency stimulation, form a small range that centers on ~15. cf: This average quantal content is much smaller than the 20–200 range reported for other muscle types tested in frog or mammals (Slater, 2015).

5-. High Pr (Fig. 1-5)

At physiological levels of extracellular calcium the EPC amplitude is nearly maximal, reflecting a Pr close to 1. cf: This very high Pr runs counter to both frog and mammalian NMJs (Slater, 2015; Homan and Meriney, 2018).

6-. Tractable N (Fig. 1-6)

MPFA using mean EPC amplitudes and associated variances measured under conditions of different release probability yields estimates of N ~15. cf: Where available, estimates of N for NMJs range from 30 to 100, which are much higher than zebrafish NMJ (Slater, 2015).

Because of its unique features, zebrafish NMJ has provided a platform for new discoveries, most notably mechanisms underlying asynchronous transmission. At most synapses synchronous transmission, reflecting time locked release with the action potential, is the predominant form. However, a non-time locked asynchronous mode, is also observed at many synapses (Kaeser and Regehr, 2014) and at certain inhibitory synapses it represents the predominate mode of transmission (Best and Regehr, 2009). At zebrafish NMJ, as with frog NMJ (Feng, 1941), where it was first discovered, asynchronous release is reflected in continued vesicle release after stimulation is terminated. Studies, using paired recordings from zebrafish NMJs provided the first identification of synaptotagmin-7 (syt-7) as the calcium sensor mediating asynchronous release (Wen et al., 2010). A central role played by syt-7 in asynchronous release has been subsequently confirmed for many central synapses. Additionally, zebrafish NMJ revealed an unexpected off-bouton source of calcium as causal to triggering asynchronous release (Wen et al., 2013a). The latter finding challenges the widely held idea that accumulation of calcium through the same calcium channels controlling synchronous release is solely responsible for asynchronous release as well. Other noteworthy discoveries were provided by motility mutant lines characterized by our lab and others, due largely to the fact that, unlike their mammalian counterparts, even full paralytics are able to survive well past the maturation of the NMJs (Ono et al., 2001). This provides the unique opportunity among vertebrate NMJs to test, individually, the roles of key synaptic proteins on synapse development and function. Examples include mutations that block receptor expression (Ono et al., 2001), clustering (Ono et al., 2002), and function (Mongeon et al., 2011; Walogorsky et al., 2012a) as well as presynaptic mutations affecting calcium channel function (Wen et al., 2013b) and ACh synthesis (Wang et al., 2008). Additionally, mutant lines have been identified for the enzyme acetylcholinesterase which terminates synaptic transmission through ACh hydrolysis following release (Behra et al., 2002; Downes and Granato, 2004). In nearly every case the mutants served as a model for study of congenital human myasthenic disorders, thereby providing translational insights into the disease manifestations (Ono et al., 2001; Wang et al., 2008; Walogorsky et al., 2012a; Walogorsky et al., 2012b). Section E provides a detailed account for the most recent discovery, which forms the major impetus behind this mini-review.

E. Tracking N during synaptic activity at zebrafish NMJ

The combined advantages provided by the zebrafish NMJ permitted application of MPFA to obtain estimates of N (Wen et al., 2016). At 0.2 Hz, the estimated N agreed well with the estimated Qc, which is expected on the basis of the estimated near unity Pr. It indicates that N exists in finite numbers and serves to set an upper limit of quantal release. It is also supportive of the idea that each individual release site is restricted to release of a single vesicle in response to a motoneuron AP. Furthermore, the close match between the estimated N and the number of synaptic boutons associated with each muscle cell further points to a relationship wherein each bouton is the functional equivalent of a single independent quantal release site. As with synapses with a high Pr, zebrafish NMJ undergoes depression in EPC amplitude during repetitive stimulation, reaching a steady state level that reflects the balance between vesicle release and replenishment. The use of MPFA was extended to the steady state transmission in order to test the widely held assumption that the drop in Pr among all release sites accounts for the reduced responses. The drop in Pr is thought to result from rapid depletion of the readily releasable pool and reliance on recruitment of a separate pool that is unable to reload and release with fidelity at high frequency stimulation (Schneggenburger et al., 2002; Zucker and Regehr, 2002; Alabi and Tsien, 2012). Most models of depression have assigned this process to kinetic distinctions among vesicles with the untested assumption that all release sites are equivalent in terms of vesicle availability and mobilization. However, our analysis pointed to a scenario where changes in N are involved in setting levels of steady state synaptic depression, requiring non-uniformity among release sites (Fig. 2E). Moreover, our findings pointed to fixed differences among release sites with slow sites requiring seconds to reload a vesicle for release and fast sites reloading within tens of milliseconds. The central evidence supporting this conclusion is reflected in the relationship between variance and mean steady state EPC amplitudes over a range of stimulus frequencies. Instead of a distribution that was fit by a single parabola, ours requires the sum of two parabolas suggesting two subgroups of N are involved in release at different frequencies (Fig. 2A). N estimates based on MPFA averaged ~15 sites at low frequency stimulation with about half of those sites participating in post-depression steady state transmission (Fig. 2B). In fact, the variance for steady state levels was ascribed principally to a drop in N and could not be accounted for on the basis of a modest drop in Pr associated with post-depression activity (Fig. 2C). Also, recovery from depression followed a bi-exponential time course with time constant >50 fold apart, a phenomenon observed at many depressing synapses. Our model involving two kinetically distinct classes of release sites offered a simple explanation to the biexponential recovery (Fig. 2D). At low frequency stimulation, all release sites are able to follow, but only the fast sites are able to follow the higher frequencies causal to synaptic depression (Fig. 2E).

Figure 2.

Analysis of steady state levels over a range of stimulus frequencies using MPFA at the zebrafish CaP motoneuron muscle synapse. (A) Variance as a function of steady state EPC amplitudes normalized to the maximum (mean +/− SEM with n values indicated). The red line reflects dual parabolic fit to a dual N site model and the blue and green represent fits to a single N site model with or without consideration of non-uniform Pr distribution. (B) The overall dependence of N estimates versus frequency with recording numbers indicated along with mean +/− S.D. **** p<0.0001; n.s. not significant. (C) Overall data showing steady state Pr estimates as a function of stimulus frequency ** p<0.01. (D) Fractional recovery from depression as a function of stimulus intervals. Each symbol is an average of 8–20 trials. The time course is fitted to a bi-exponential function with time constants indicated. (E) Cartoon illustrating the key difference between the conventional uniform release model (left) and the non-uniform release site model (right). In this simplified schematic 3 individual muscle cells are shown (pink) which are each contacted by 6 presynaptic boutons (yellow circle), each bouton containing 7 vesicles (blue open and filled circles). For illustration the vesicle that is released from each bouton in response to an action potential is indicated by a blue fill. Both models predict release of a single vesicle (blue filled) from every bouton during low frequency stimulation without failure (Upper panels left and right). However, the models differ in response to high frequency stimulation (lower panels left and right). In the uniform site model the release failures (shown by the absence of filled blue vesicles) are stochastically distributed among the 6 release sites between trials. By contrast, in the non-uniform site model the failures are associated with a subset of fixed release sites (indicated as slow), representing about half of the total sites in number (outlined by a rectangle). Adapted from Wen et al., 2016.

The central distinction setting this new model apart from previous vesicle based models is the requirement for fixed differences among individual release sites. Functional heterogeneity among synapses has been extensively documented at both NMJ and central synapses, but primarily in terms of non-uniform Pr at low frequency activity (del Castillo and Katz, 1954; Rosenmund et al., 1993; Murthy et al., 1997; Holderith et al., 2012). Evidence for kinetically non-equivalent release sites with respect to vesicle reloading at higher activity level might have been overlooked partly due to limitations in the ability to define and track N. Indeed, an unexpected low variance after initial depression, similar to ours, has been observed at vertebrate NMJs (Slater, 2015). Moreover, a dropout of release sites during repetitive stimulation has been at the root of proposals positing that site-clearance limits reloading and release (Stevens and Wesseling, 1999; Neher, 2010). Finally, recent studies at hippocampal synapses have led to the proposal that nanomodules contain multiple release sites that are functionally non-equivalent in a way so as to contribute to synaptic depression (Gustafsson et al., 2019).

The dependence of release on calcium would render differences in calcium channel density among boutons a leading candidate for release site distinction. However, in response to low frequency stimulation Pr is near unity for all release sites at zebrafish NMJ at physiological calcium concentrations. The lack of distinction suggests that calcium channel density and coupling are sufficient at all release sites, making it difficult to account for the faithful drop out of certain release sites during depression (Wen et al., 2016). For the same reasons calcium channel inactivation is an unlikely mechanism. Further dampening this idea are measurements of heterologously expressed zebrafish P/Q type calcium channels showing an absence of inactivation on the time base corresponding to depression (Naranjo et al., 2015).

A newly developed means for direct test of fixed heterogeneity among release sites is provided through optical monitoring of individual release sites. Until recently most optical measurements relied on bulk exocytosis following repetitive stimulation as the sensitivity of neither FM1–43 nor the ecliptic synaptopHluorin, a GFP-containing fluorescent reporter of vesicle fusion was sufficient to monitor fusion of single vesicles (Ribchester, 2009). Recently, however, measurements involving more sensitive fluorescence indicators, coupled to high-resolution imaging have resulted in reliable detection of single fusion events. At central synapses nanoscale imaging has provided release profiles for single vesicles through use of vGlut1-pHluorin (Balaji and Ryan, 2007, Leitz and Kavalali, 2011, Leitz and Kavalali, 2014). A complementary means for detecting single vesicle release has been developed through monitoring of postsynaptic responses. The expression of the genetic encoded calcium indicator GCaMPs in fly muscle provides sufficient sensitivity to detect responses associated with release of single vesicles. The permeability of glutamate receptors to calcium provides robust signals in response to receptor activation (Peled and Isacoff, 2011; Newman et al., 2017). The ability to interrogate spatially separate synapses using high-resolution microscopy further identified differences in Pr among the many distinct release sites. Through combined use of immunohistochemistry and mutant lines, differences in Pr among release sites at fly NMJs was assigned to differences in the levels of synaptotagmin and calcium channel density (Akbergenova et al., 2018). Applying such approaches to the zebrafish NMJ now offers the potential to determine the molecular bases for functional distinction among release sites (Wen and Brehm, 2018).

F. The challenges facing N

Placed in the context of fixed release sites, N collectively sets the upper boundaries for Qc. However, after decades of anatomical investigation the release site still lacks physical definition. Without physiology to guide the mission, structural investigations will likely continue to fail. Physiological estimates of N have been severely hampered by presynaptic complications such as MVR as well as postsynaptic complications involving receptor occlusion, desensitization and saturation. Consequently, physiologists have instead, elected to focus on Qc as a functional means to quantitate and compare synapses. Now, through identification of simple synapses such as those with high Pr like zebrafish NMJ, it is possible reinvestigate the physical entity corresponding to a release site. To date, no molecular mechanisms have been identified that account for differences in reloading and release. Attractive candidates for distinguishing slow and fast release sites lie in the clustered distribution of key active zone proteins such as RIM, and Munc13 and bassoon, which have all been linked to vesicle recruitment and depression. As such they could not only limit the overall release but could also contribute to the functional diversity among release sites, a concept that is deserving of greater attention. In particular, the collective advantages for physiology, imaging and genetics offered by the zebrafish NMJ offers a powerful platform for probing the structural bases for setting release limits at individual boutons and for functional diversity among boutons. Combined with the recent application of high-resolution tomographic reconstruction, zebrafish NMJ may crack the code, revealing the identity of release sites (Helmprobst et al., 2015).

Highlights.

• Highlighting features that set zebrafish apart from the traditional NMJs.

• Zebrafish NMJ provides functional tracking of individual release sites.

• Release sites fall into two classes, differing in speed of vesicle reloading

• Slow reloading release sites are causal to frequency dependent synaptic depression.

• Zebrafish NMJ offers possible physical identification a release site.

List of abbreviations:

cf

to compare

N

total number of release sites

Qc

mean number of quanta released in response to an action potential

Pr

mean release probability

NMJ

neuromuscular junction

AP

action potential

q

single quanta

MVR

multivesicular release

AChR

acetylcholine receptor

FasII

fasciculin II, a toxin binding to acetylcholinesterase

MPFA

multiple-probability fluctuation analysis

RRP

readily releasable pool of vesicles

α-btx

α–bungarotoxin

EPC

endplate current

mEPC

miniature endplate current

Syt-7

synaptotagmin 7

GCaMP

genetically encoded calcium indicator

FM1–43

fluorescent styryl dye used to visualize synaptic vesicles exocytosis