Single Ca2+ channels and exocytosis at sensory synapses


  • This review was presented at The Journal of Physiology Symposium entitled Size matters: formation and function of GIANT synapses, which took place at the Annual Meeting of the Society for Neuroscience, New Orleans, LA, USA on 12 October 2012. It was commissioned by the Editorial Board and reflects the views of the authors.

H. von Gersdorff: Oregon Health Sciences University, Vollum Institute L-474, 3181 SW Sam Jackson Park Rd, Portland, OR 97239-3098, USA. Email:


Abstract  Hair cell synapses in the ear and photoreceptor synapses in the eye are the first synapses in the auditory and visual system. These specialized synapses transmit a large amount of sensory information in a fast and efficient manner. Moreover, both small and large signals with widely variable kinetics must be quickly encoded and reliably transmitted to allow an animal to rapidly monitor and react to its environment. Here we briefly review some aspects of these primary synapses, which are characterized by a synaptic ribbon in their active zones of transmitter release. We propose that these synapses are themselves highly specialized for the task at hand. Photoreceptor and bipolar cell ribbon synapses in the retina appear to have versatile properties that permit both tonic and phasic transmitter release. This allows them to transmit changes of both luminance and contrast within a visual field at different ambient light levels. By contrast, hair cell ribbon synapses are specialized for a highly synchronous form of multivesicular release that may be critical for phase locking to low-frequency sound-evoked signals at both low and high sound intensities. The microarchitecture of a hair cell synapse may be such that the opening of a single Ca2+ channel evokes the simultaneous exocytosis of multiple synaptic vesicles. Thus, the differing demands of sensory encoding in the eye and ear generate diverse designs and capabilities for their ribbon synapses.

  • image

[ Mean-Hwan Kim received his PhD in Physics from the Pohang University of Science and Technology (POSTECH), Pohang, Korea, working with Dr Duk-Su Koh. He is currently a postdoctoral fellow at the Vollum Institute working on retinal ribbon synapses. Geng-Lin Li received his PhD in neurobiology from the Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, working with Dr Xiong-Li Yang. He is now an assistant professor in the Department of Biology, University of Massachusetts at Amherst. His main interest is in hair cell ribbon synapses. Henrique von Gersdorff received his BS from the Federal University of Rio de Janeiro and his PhD in theoretical physics from the University of Minnesota. He pursued postdoctoral studies with Dr Gary Matthews at Stony Brook University and with Dr Erwin Neher at the Max Planck Institute in Göttingen. He is now a senior scientist at the Vollum Institute in Portland, OR, USA.]

the voice of your eyes is deeper than all roses

— e. e. cummings

Sensory systems have exquisite sensitivity. Specialized detectors, synapses, and circuits were apparently fine-tuned by evolution to minimize cellular noise and amplify signal capture and transmission. When fully adapted, these systems can detect and transmit minute signals with energies and dimensions that reach the realms of quantum physics. Photoreceptors, for example, can detect single photons, olfactory epithelial cells can detect single molecules, and hair cells can detect nanometre deflections of their stereocilia. While the olfactory epithelial cells can generate spikes by themselves, which then depolarize their nerve endings, photoreceptors and mature hair cells do not generate spikes. Instead, photoreceptor and hair cell synapses display hypersensitivity to minute changes in presynaptic membrane potentials generated by small sensory stimuli. These receptor potentials, or graded membrane potential changes, constitute analog signals that can be as small as 0.1 to 1 mV, but which are nevertheless reliably encoded into all-or-none action potential spikes (or digital signals) with precise timing (or phase locking) in postsynaptic cells (Copenhagen, 2004). An analog-to-digital converter thus transmits information from the eye and ear to the brain as patterns of all-or-none spikes (Matthews & Fuchs, 2010; Safieddine et al. 2012).

Given this extreme sensitivity for detecting small signals, what is the corresponding lower limit of sensitivity for transmission of information at sensory synapses? One possibility is that a small depolarization opens a single Ca2+ channel that fluxes enough Ca2+ ions to trigger the exocytosis of one nearby vesicle, or quanta, from the nerve terminal. The released excitatory neurotransmitter produces a mEPSC that may be enough to trigger a spike on a high input resistance postsynaptic neuron. This mechanism would have high sensitivity to small receptor potentials but it may saturate easily (or have a small dynamic range). So it may be challenging for it to maintain good signal-to-noise ratios for sensory stimuli that vary in strength over several orders of magnitude, as occurs in the retina and cochlea. Therefore, adaptation mechanisms will be required to avoid saturating the flow of information. Alternatively, another possibility relies on the convergence of several small sensory synapses onto one postsynaptic neuron, which may lead to the summation of several small EPSPs, if they occur within a short time window. The summed EPSPs can then reach spike threshold. The postsynaptic neuron thus acts as a coincidence detector that rejects noise (non-coincident EPSPs) but accepts the signal (coincident EPSPs that trigger a spike). The retina and the cochlea presumably use variations of these mechanisms to enhance signal-to-noise ratios as they send information downstream to the brain.

A variation on the first mechanism proposed above suggests that a small depolarization may open a single Ca2+ channel that is tightly coupled to the Ca2+ sensors for exocytosis of several docked vesicles. The open channel's nanodomain of Ca2+ could then trigger a multivesicular fusion event that produces a large and rapid EPSC, which is sufficient to elicit a postsynaptic spike. Indeed, large and stochastically occurring EPSC events have been observed from rat inner hair cell synapses that have a single synaptic ribbon as their hallmark for an active zone of glutamate release (Glowatzki & Fuchs, 2002). Large EPSCs have also been observed at hair cell synapses from the bullfrog amphibian papilla, an auditory organ for hearing sounds in the frequency range from 100 to 1200 Hz (Keen & Hudspeth, 2006). At this adult amphibian synapse, small depolarizations of the hair cell, which open only a few Ca2+ channels, are sufficient to trigger large EPSCs in a Ca2+-dependent manner (Li et al. 2009; see Fig. 1A and B). Just like the inner hair cell of mammals, one amphibian papilla hair cell is contacted by several afferent fibres (about 4 or 5 fibres), but unlike inner hair cells, one of the postsynaptic afferents of the bullfrog hair cell is a dominant claw-like afferent fibre that receives input from tens of ribbons (Graydon et al. 2011). Nevertheless, several lines of evidence indicate that the large EPSCs arise from a single ribbon-type active zone (Li et al. 2009).

Figure 1.

Multiquantal EPSCs triggered by the voltage-dependent opening of Ca2+ channels in hair cells 
A, paired recordings between a presynaptic bullfrog auditory hair cell and its afferent fibre. The hair cell was dialysed with 2 mm EGTA and held at −90 mV. The hair cell was then depolarized to various membrane potentials in 3 mV increments (shown in red). At −72 mV, only small-amplitude EPSCs are observed (presumably single vesicle or quanta events). However, a further small depolarization to −69 mV (at which Ca2+ current starts to activate) suddenly evokes large-amplitude EPSCs together with small amplitude events (marked with a blue asterisk). The large EPSC events (presumably multiquantal) become more frequent with further depolarization to −66 mV. These events occur in a stochastic manner and may reflect the highly synchronous release of several vesicles (modified from Li et al. 2009). B, paired recordings from a hair cell dialysed with 10 mm BAPTA and held at −90 mV. In this recording depolarizations to −69 mV and −66 mV did not evoke large EPSCs, presumably because the large concentration of BAPTA, a fast Ca2+ buffer, reduced Ca2+ nanodomain concentrations. However, hair cell depolarization to −57 mV evoked large-amplitude EPSCs. The Ca2+ current at −57 mV is substantial and thus saturates the BAPTA at the active zone. The Ca2+ concentrations can then reach levels sufficient to evoke release. Note that large-amplitude EPSCs occur together with small amplitude EPSC events (modified from Li et al. 2009). C, unitary voltage-gated CaV1.3 L-type Ca2+ channel currents from mouse inner hair cells (modified from Zampini et al. 2010). Representative unitary currents from cell attached recordings. The membrane patch potentials (red) are shown next to the traces. The pipette solution contained 5 mm Ca2+ and BayK 8644, an L-type Ca2+ channel agonist. The grey lines indicate the closed channel state. At hyperpolarized potentials of −67 mV relatively few channel openings occur (open probability Po low) and the open time is short (about 0.5–1 ms), but the single Ca2+ channel current is large. Note that Ca2+ channels can open and close quickly (flickers). At the more depolarized potentials of −57 mV and −37 mV, more openings occur (higher Po) and open times are longer, but the single Ca2+ channel current is smaller due to reduced driving force. The very brief Ca2+ channel openings may trigger the release of single synaptic vesicles that produces small amplitude EPSCs, while the longer Ca2+ channel openings may trigger the highly synchronous release of several synaptic vesicles that produces large EPSC events.

The opening of a few Ca2+ channels has been shown to be sufficient for the exocytosis of one vesicle in mouse inner hair cells (Brandt et al. 2005) and cone photoreceptors (Bartoletti et al. 2011). Understanding the properties of Ca2+ channels at membrane potentials slightly depolarized from resting membrane potentials will be critical for a better understanding of the performance levels reached by sensory systems in the limits of small signal detection and transmission (Demb & von Gersdorff, 2008). In addition, these data will provide insights into how single Ca2+ channels and clusters of Ca2+ channels form, respectively, nanodomains and microdomains of high Ca2+ concentrations near synaptic ribbons (Moser et al. 2006).

Figure 1C shows examples of single Ca2+ channels obtained from immature mouse inner hair cells. Openings occur stochastically and for variable durations. Note that single channel currents occur even at very hyperpolarized membrane potentials of −67 mV. At these potentials the duration of the open channels is also shorter on average. Depending on their open time duration, these relatively large, but very brief, elementary Ca2+ currents may be the trigger for the EPSC events shown in Fig. 1A. Note that the large EPSC events can be triggered even with high concentrations of a fast Ca2+ buffer (10 mm BAPTA) in the hair cell at more depolarized potentials (Fig. 1B). This suggests that some Ca2+ channels and docked vesicles are located within a few nanometres of each other so that Ca2+ influx from voltage-dependent Ca2+ channels can saturate the Ca2+ buffers locally near a docked vesicle and, thus, lead to Ca2+ nanodomains and vesicle fusion (Fig. 3A). We propose that short duration events (perhaps flickers) trigger single vesicle fusions and small EPSCs, whereas the larger EPSCs are triggered by longer Ca2+ channel open times. However, in Fig. 1C, the pipette solution contained 5 mm Ca2+ and BayK 8644, which prolongs single Ca2+ channel open times. This allows single channels to be more easily observed, but BayK 8644 may affect the voltage dependence of Ca2+ channel open times. Therefore, the properties of adult hair cell L-type Ca2+ channels will need to be studied in more normal conditions to better understand their role in triggering large EPSCs. Moreover, Ca2+ channels in ribbon synapses of the retina and cochlear are clustered in one to three rows underneath the synaptic ribbon (Rodriguez-Contreras & Yamoah, 2001; Rutherford & Pangršič, 2012). This tight clustering may facilitate a nearly simultaneous and cooperative opening of several Ca2+ channels to trigger the large EPSC events (Dixon et al. 2012). In addition, a Ca2+-dependent facilitation of L-type Ca2+ channels may shorten the mean latency to first opening after a depolarizing step (Calin-Jageman & Lee, 2008). The resulting faster activation of Ca2+ currents may be in part responsible for the shorter synaptic delays observed at physiological resting membrane potentials (Cho et al. 2011; Goutman & Glowatzki, 2011).

Figure 3.

Presynaptic gain control by Ca2+ channel coupling to the fusion sensor of synaptic vesicles at the active zone 
A, for individual non-overlapping Ca2+ nanodomains the docked vesicles are tightly coupled with a single Ca2+ channel. Ca2+ influx through a single Ca2+ channel is able to evoke exocytosis reliably. In this case, the amount of release is linearly increased when the number of open Ca2+ channels is increased. Evoked release is insensitive to EGTA, a high-affinity but slow Ca2+ buffer. A small depolarization thus evokes a small number of open Ca2+ channels and reliable release. The synapse thus has high gain at low presynaptic membrane potentials (Vm). This is the modus operandi of mammalian rod bipolar cells for short depolarizing pulses (Jarsky et al. 2010) and for the squid giant synapse with presynaptic action potential depolarizations (Augustine et al. 1991). B, for microdomains individual docked vesicles are coupled to several presynaptic Ca2+ channels with overlapping Ca2+ domains. The overlapping Ca2+ domains cooperatively recruit vesicle exocytosis. In this case, the release rate is non-linearly increased when the number of open Ca2+ channels is increased. Release will increase in proportion to the third of fourth power of the increment in macroscopic Ca2+ current. Evoked release can be insensitive to EGTA. The synapse thus has low gain at small presynaptic Vm and high gain at high presynaptic Vm. This is the modus operandi of the squid giant synapse with long voltage-clamp depolarizing pulses (Augustine et al. 1991). C, a mixture of Ca2+ nanodomains and Ca2+ microdomains is also possible at some synapses. Some docked vesicles are tightly coupled with Ca2+ channels, while some docked vesicles are relatively far from a cluster of Ca2+ channels. This cluster of Ca2+ channels can produce a Ca2+ microdomain that can influence the release of distant vesicles. These Ca2+ microdomains can be reduced by EGTA. In this case the relationship between release and number of open Ca2+ channels can show both linear and non-linear components, and a pool of readily releasable synaptic vesicles will be insensitive to EGTA and a pool of vesicles will be sensitive to EGTA. This is perhaps the modus operandi of the Mb (mixed rod-cone) bipolar cell in goldfish retina (Mennerick & Matthews, 1996; Coggins & Zenisek, 2009). The dependence of release on presynaptic Vm may thus acquire a larger dynamic range, allowing the Mb bipolar cell terminal to transmit both rod and cone photoreceptor signals.

Large EPSCs at ribbon synapses: multivesicular or univesicular?

The small EPSCs in Fig. 1A obtained by depolarizing the hair cell to −72 mV from a holding potential of −90 mV were interpreted as being due to the fusion of single synaptic vesicles (univesicular release), whereas the larger EPSCs at more depolarized hair cell potentials were interpreted as multivesicular (or multiquantal) events (Li et al. 2009). The dependence of multiquantal EPSC amplitudes on presynaptic membrane potential has so far been reported only in the bullfrog hair cell synapse (Li et al. 2009), but recently similar results have also been observed in the mouse rod bipolar cell ribbon synapse (Mehta et al. 2013). One mechanism for the generation of large EPSCs is the highly synchronous fusion of several docked vesicles underneath a synaptic ribbon (coordinated release; Singer et al. 2004; Graydon et al. 2011). However, an alternative interpretation is that the large EPSCs arise from a single compound fusion event, originating from the pre-fusion of several synaptic vesicles attached to the ribbon (Matthews & Sterling, 2008). On the other hand it is also conceivable that the large EPSC events are due to the fusion of single vesicles that contain higher amounts of glutamate and a lower release probability. These vesicles may require higher amounts of hair cell depolarization, and the resulting higher levels of Ca2+ concentration, for their fusion sensors to be activated. Therefore, it is still possible that the large EPSC events are univesicular, especially given that they have the same kinetics as small EPSCs (Li et al. 2009). Moreover, even at saturating Ca2+ concentrations, vesicles fuse with different delays and the distribution of these delays would have a characteristic time of about 0.5 ms (see Beutner et al. 2001; Neher & Sakaba, 2008). Very fast multiquantal EPSCs have been observed at cerebellar granule cells (10–90% rise time of 0.4 ms; Silver et al. 1992), but action potential spikes trigger these EPSCs. It thus seems unlikely that three or four vesicles could fuse nearly simultaneously to produce the large EPSCs without the benefit of action potentials. Nevertheless, the specialized proteins involved in hair cell exocytosis may coordinate a highly synchronous form of release (Safieddine et al. 2012; Rutherford & Pangršič, 2012). Interestingly, multiphasic (complex) EPSCs were also observed from rat inner hair cell synapses (Glowatzki & Fuchs, 2002), suggesting that an asynchronous form of multivesicular release can also occur at single synaptic ribbon synapses. Further studies are clearly needed to confirm the multivesicular release hypothesis for the large EPSC events at hair cell synapses and to establish the detailed mechanism of vesicle fusion.

Tonic and phasic transmitter release

Exocytosis and glutamate release has also been studied at retinal ribbon synapses formed by bipolar cell terminals. Figure 2A shows AMPA receptor-mediated currents in cultured horizontal cells evoked by step depolarizations in a single Mb (mixed rod-cone) bipolar cell terminal from goldfish retina (von Gersdorff et al. 1998). The acutely isolated and patch-clamped bipolar cell terminal was manoeuvred and pushed against the membrane of a horizontal cell that was patch clamped by a second pipette and held at −70 mV. The horizontal cell was thus used as a ‘sniffer cell’ or detector of glutamate release. The bath contained cyclothiazide to block AMPA receptor desensitization. Note that weak depolarizations produce a small amplitude and slowly activating L-type Ca2+ current that produces a glutamate triggered current (IGlu) with a long synaptic delay (arrows in Fig. 2A). Further depolarization of the step pulse produces larger IGlu with shorter synaptic delays, although the amplitude of IGlu appears to saturate. Repolarization from +52 mV to −68 mV evokes a large Ca2+ tail current and a large IGlu with a submillisecond synaptic delay (bold traces in Fig. 2A). The release has both linear and non-linear dependence on the presynaptic membrane voltage and Ca2+ current charge. This allows the synapse to transmit small and sustained presynaptic depolarizations, via a slow tonic release that evokes small and sustained EPSCs, and also to transmit rapid and large depolarizations, via a phasic short-delay release that evokes a transient EPSC (Palmer & von Gersdorff, 2002). This synapse exhibits depletion of its readily releasable pool of vesicles, which is manifested by strong paired-pulse depression that takes seconds to recover (von Gersdorff & Matthews, 1997; Gomis et al. 1999; Palmer et al. 2003). It thus appears to have properties that make it a good encoder of both contrast and luminance, because the phasic component of the EPSC can encode Weber contrast between two different light levels, while the sustained EPSC component may encode luminance (Jarsky et al. 2011; Oesch & Diamond, 2011). A similar ability for phasic and tonic release has also been reported for cone photoreceptor terminals (Jackman et al. 2009), whereas a more tonic and linear input–output synaptic release curve was observed at the rod photoreceptors (Thoreson et al. 2004). Given the stark differences in the kinetics of their light responses, rod and cone photoreceptors may have synaptic ribbons tuned for diverse release rates (Heidelberger et al. 2005).

Figure 2.

Non-linear and linear dependence of exocytosis on Ca2+ currents at retinal bipolar cell ribbon synapses 
A, paired recording from an acutely dissociated goldfish Mb (mixed rod-cone) bipolar cell terminal and a cultured horizontal cell that expresses AMPA receptors. The horizontal cell is used to detect the glutamate release from the bipolar cell terminal. Superimposed responses were evoked by 10 ms pulses to +52 (in bold), −45, −43, −38 and −18 mV (from top to bottom) that elicited respectively larger calcium current (ICa) and glutamate-gated current (IGlu), except for the IGlu response to the +52 mV pulse (in bold), which occurred only during the repolarization from +52 mV to −68 mV, when a tail ICa is elicited (membrane potentials were corrected for a 8 mV liquid junction potential; von Gersdorff et al. 1998). Note that the synaptic delay (arrows) is considerable for weak depolarizations, which produce slow ICa activation and small ICa amplitudes. B, a dual-component IGlu is revealed by longer depolarizing pulses. Asynchronous release occurs after Ca2+ channel closure. Responses evoked by 50 ms and 100 ms pulses to −10 mV elicited, respectively, the bold and thin IGlu traces. The arrowhead marks the truncated second component of release evoked by the 50 ms pulse (von Gersdorff et al. 1998). A fast releasing pool of vesicles triggers the phasic first component of IGlu, whereas a delayed and slow second pool of released vesicles triggers a more sustained and tonic component of IGlu. C, paired recording from the mouse rod bipolar cell synapse. Exocytosis varies linearly with the number of open Ca2+ channels. Presynaptic voltage steps (1.5 ms) from −60 mV to potentials between −47 and −31 mV (in 4 mV increments; grey-scale from −47 mV (dark) to −31 mV (light)) elicit Ca2+ currents (top) and EPSCs (bottom). EPSC quantal content is plotted against QCa, illustrating a near-linear relationship between Ca2+ influx and exocytosis when the number of open presynaptic Ca2+ channels is altered. Data are normalized to the condition of the smallest voltage step in which 44% of responses are failures of transmission ([Ca2+]e= 1 mm; modified from Jarsky et al. 2010).

When longer duration depolarizing steps are applied to the Mb bipolar cell terminal a dual-component EPSC is observed (Fig. 2B). This component is immediately interrupted when the bipolar cell is hyperpolarized (arrowhead in Fig. 2B) suggesting that it is strictly due to the release of a second pool of vesicles and not due to non-linear properties of the AMPA receptors. Moreover, the total charge of the IGlu current correlated well with changes in membrane capacitance (a purely presynaptic measure of exocytosis), suggesting that this second component was not due to slow glutamate diffusion from distal synaptic ribbons which would lead to some glutamate molecules being lost and going undetected (von Gersdorff et al. 1998). Thus, during a prolonged depolarization, goldfish bipolar cell terminals appear to have a fast releasing pool of vesicles that is depleted quickly, followed by a larger and slower pool that contains thousands of vesicles (Mennerick & Matthews, 1996; Heidelberger et al. 2005). These releasable vesicles may be tethered to the synaptic ribbon and/or docked near to the ribbon (Midorikawa et al. 2007). At rat bipolar cell synapses a smaller and more sustained second component of release follows the phasic EPSC (Singer & Diamond, 2003). However, recently a more prominent second component has been observed at mouse rod bipolar cell synapses (Mehta et al. 2013).

When short 1.5 ms depolarizing pulses with increasing strength are applied to a rod bipolar cell a progressively larger EPSC is observed in the postsynaptic AII amacrine cell (Jarsky et al. 2010). The dependence of EPSC amplitude on the Ca2+ current charge is nearly linear (Fig. 2C). This synapse is also insensitive to large concentrations of EGTA (Singer & Diamond, 2003). Together these results suggest that this synapse operates with non-overlapping Ca2+ nanodomains so that the opening of one Ca2+ channel leads to the exocytosis of one vesicle (Fig. 3A; Eggermann et al. 2011; Matveev et al. 2011). A linear input–output relationship is also present in the squid giant synapse when the nerve terminal is depolarized by progressively wider action potential waveforms (Augustine et al. 1991). In the squid giant synapse only 10% of the available Ca2+ channels are opened by a presynaptic action potential (Pumplin et al. 1981). This promotes non-overlapping Ca2+ nanodomains. However, with longer depolarizing step pulses the input–output relationship is non-linear (4th power law) in the squid giant synapse (Fig. 3B; Augustine et al. 1991). A 2nd to 3rd power law is found in the goldfish Mb bipolar cell (Fig. 2A; von Gersdorff & Matthews, 1997; Coggins & Zenisek, 2009). The Mb bipolar cell terminal release has a pool of vesicles that is insensitive to EGTA (Mennerick & Matthews, 1996) and a pool that is sensitive to EGTA (Coggins & Zenisek, 2009). This synapse might thus operate with both non-overlapping Ca2+ nanodomains, which are insensitive to EGTA, and Ca2+ microdomains, which are EGTA sensitive (Fig. 3C; Beaumont et al. 2005). This may allow the goldfish mixed bipolar cell, which receives both rod and cone input, and thus needs to operate in different light levels, a larger dynamic range (or a broader bandwidth) for the transmission of visual signals with different intensities and temporal frequencies (Fig. 3C).

Ca2+ nanodomains and microdomains

One N-type Ca2+ channel opening is sufficient to release one nearby vesicle at some synapses (nanodomain; Fig. 3A; Weber et al. 2010), whereas the opening of several Ca2+ channels may be necessary for the fusion of distant vesicles at other synapses (microdomain; Fig. 3B; Borst & Sakmann, 1996). Recently, even spontaneous vesicle fusion has been shown to require the opening of two or more Ca2+ channels at some cortical bouton-type synapses (Williams et al. 2012). During early postnatal development synapses may also mature from a microdomain to a nanodomain-controlled mode of vesicle fusion (Fedchyshyn & Wang, 2005; Leão & von Gersdorff, 2009). Interestingly, at rod bipolar cell synapses synchronous, sustained, asynchronous, and even spontaneous release appear to depend on intact synaptic ribbons (Mehta et al. 2013).

Hair cell synapses may contain an extreme example of a nanodomain-type active zone with a specialized bauplan for highly synchronous multivesicular release. Here the opening of one L-type Ca2+ channel may be sufficient for the exocytosis of three or more vesicles within 0.2 ms (the rise time of the large EPSC event; Li et al. 2009). Figure 4 shows results from a Monte Carlo simulation that suggest that the opening of one Ca2+ channel can produce a free intracellular Ca2+ concentration, [Ca2+]i, of about 310 μm near the mouth of the Ca2+ channel. Indeed, recent imaging of Ca2+ nanodomains of single Ca2+ channels (using a Ca2+ indicator tethered to the channel intracellular mouth) indicate that [Ca2+]i near the mouth of the channel follows a relationship of 700 μm per picoamp of single Ca2+ channel current (Tay et al. 2012). So a Ca2+ channel current of 0.5 pA (as we estimate for frog hair cells at −70 mV at room temperature and with external calcium [Ca2+]e= 2 mm; Graydon et al. 2011) will produce [Ca2+]i= 350 μm near the channel mouth, which is in excellent agreement with our simulations (Fig. 4). Note also in Fig. 4 that [Ca2+]i reaches high levels of 50–100 μm near about three or four docked vesicles. These three or four vesicles may thus have enough Ca2+ to fuse nearly simultaneously and thereby produce the large EPSCs observed in Fig. 1A. Note that the ribbon and the docked vesicles restrict the volume available for Ca2+ diffusion. So the ribbon may act as an impermeable ‘mirror’ for Ca2+ (or a diffusion barrier) that boosts the [Ca2+]i underneath the ribbon (Roberts, 1994). We propose that the ribbon, and its coterie of vesicles docked on the plasma membrane, create a small and crowded volume around the Ca2+ channel cluster that acts as a Ca2+‘trap’. This topography boosts Ca2+ concentrations near the docked vesicles thereby promoting highly synchronous multivesicular release.

Figure 4.

The spatial profile of a Ca2+ nanodomain 
Left panel, a spherical synaptic ribbon is shown above an array of docked vesicles (red) and two clusters of Ca2+ channels (blue stripes). Right panel, the Ca2+ concentration underneath the synaptic ribbon in a bullfrog hair cell active zone of exocytosis after the opening of a single Ca2+ channel. The white circles show the locations of the docked spherical vesicles with a diameter of 30 nm. A Monte Carlo simulation was created to describe the diffusion of Ca2+ ions and the spatial extent of local Ca2+ nanodomains (Graydon et al. 2011). A 30 × 30 grid of 10 nm boxes was used to monitor free Ca2+ concentration at the presynaptic membrane surface over time while a single Ca2+ channel was opened. Shown are free Ca2+ concentrations averaged over 1 ms time bins beginning 5 ms after channel opening. The hair cell contained 2 mm EGTA as a mobile Ca2+ buffer. The single open channel current i=−0.54 pA at Vm=−70 mV. This current resulted in a widespread, large free Ca2+ nanodomain that peaked at about 300 μm near the open mouth of the Ca2+ channel. The presence of the docked vesicles and the ribbon as diffusion barriers and restrictors of available space led to an enhancement in Ca2+ concentration of tens of additional micromolars for tens of square nanometres (see Graydon et al. 2011). The simulations suggest that docked vesicles and the synaptic ribbon effectively ‘trap’ Ca2+ ions in a small and crowded volume. The synaptic ribbon may thus function as an aggregator of Ca2+ channels and docked vesicles, and if it is impermeable to Ca2+ ions, as a booster of Ca2+ nanodomain concentrations.

Synaptic ribbons: the emerging central role of charged phospholipids in exocytosis and endocytosis

How do sensory cells build a synaptic ribbon from its molecular constituents? Recent proteomic analyses have revealed a multitude of proteins located within the synaptic ribbon active zone (Uthaiah & Hudspeth, 2010; Kantardzhieva et al. 2012), but the main building block of the synaptic ribbon appears to be a protein called RIBEYE (Schmitz, 2009; Rutherford & Pangršič, 2012). This protein binds to NADH and thus may be able to sense changes in cell metabolism as occur, for example, during the day/night cycle and during hibernation when the synaptic ribbon changes size and shape (Hull et al. 2006; Emran et al. 2010; Mehta et al. 2013). Interestingly, the degree of Ca2+ influx through open CaV1.3 channels during development affects ribbon size and shape at zebrafish lateral line hair cell synapses (Sheets et al. 2012). The exact mechanisms by which Ca2+ and NADH control ribbon size and function remain to be elucidated, but several Ca2+ binding proteins are located on or near the synaptic ribbon (Schmitz, 2009; López-del Hoyo et al., 2012).

One recent and unexpected discovery is that RIBEYE is also directly involved in the synthesis of phosphatidic acid, a cone-shaped phospholipid that promotes negative membrane curvature (Schwarz et al. 2011). This negative curvature in the lipid bilayer is needed for both fusion and fission pore formation (Chernomordik et al. 2006). Phosphatidic acid is also a biosynthetic precursor for diaclyglycerol (DAG), which is a precursor for PIP2, a phosphoinositide that is abundant at cone photoreceptor terminals (Brockerhoff, 2011). Both phosphatidic acid and PIP2 are negatively charged lipids, which allows them to interact with positively charged proteins. Indeed, after Ca2+ binding, the exocytosis Ca2+ sensor synaptotagmin interacts with negatively charged lipids to trigger vesicle fusion (Striegel et al. 2012). Moreover, the endocytosis-involved phosphatase synaptojanin dephosphorylates PIP2 and synaptojanin is required for proper hair cell synapse function (Trapani et al. 2009). Phosphatidic acid and PIP2 thus play crucial roles in the coupling of exocytosis to endocytosis (Koch & Holt, 2012). Under normal conditions cells maintain a very low level of phosphatidic acid, so the fact that the ribbon is involved in the synthesis of phosphatidic acid is highly intriguing. It is also suggestive that the ribbon may play an important role in promoting endocytosis. Indeed, ‘hot spots’ of endocytosis have been discovered near to synaptic ribbons at mouse bipolar cells (LoGiudice et al. 2009). The ribbon may thus actively prime vesicles for exocytosis, as well as facilitate endocytosis. Physiological evidence for priming of exocytosis has been obtained recently using the FALI technique that acutely damages the ribbon with a bright pulse of light (Snellman et al. 2011).

Vesicles pools and synaptic ribbons

Synaptic ribbons have been proposed to act as a conveyor belt for high throughput exocytosis by first binding vesicles from the cytoplasm, and subsequently guiding them to sites on the plasma membrane where they are docked within nanometres of Ca2+ channels (Heidelberger et al. 2005). The high mobility of synaptic vesicles in the cytoplasm may be key to the high release rates of some ribbon synapses (Holt et al. 2004; Griesinger et al. 2005; Graydon et al. 2011). Three distinct kinetic components of release have been uncovered in retinal bipolar cells and these components may correspond to (1) vesicles docked on the plasma membrane near the ribbon (the rapidly releasable pool), (2) vesicles attached to the ribbon but not docked to the membrane (the readily releasable pool), and (3) vesicles in the cytoplasm and near to ribbons (the reserve pool that is recruited during prolonged depolarizations; Mennerick & Matthews, 1996; von Gersdorff & Matthews, 1997). Recently, three distinct vesicle pools have also been identified in turtle hair cells (Schnee et al. 2005), frog sacculus hair cells (Rutherford & Roberts, 2006), and frog amphibian papilla hair cells (Graydon et al. 2011). Three distinct vesicle pools have also been identified at conventional active zone synapses (Denker & Rizzoli, 2010). However, only two distinct vesicle pools have been identified in mammalian inner hair cells (Moser & Beutner, 2000; Johnson et al., 2005). The ribbon is thus emerging as a complex and dynamic structure that can change its molecular components on a time scale of hours and may perform several tasks that facilitate indefatigable exocytosis (Griesinger et al. 2005). Nevertheless, we also emphasize that the synaptic ribbon is not strictly necessary for prolonged high frequency exocytosis that is resistant to synaptic depression since this feat can also be accomplished by some conventional active zone synapses (Kushmerick et al. 2006; Saviane & Silver, 2006). So the specific function of the synaptic ribbon still remains poorly understood.

The neural code for sensory information

What is the relationship between the synaptic ribbon and the generation of action potential spikes in the afferent fibres? The rapid and precise encoding of sound onset is degraded from hair cell synapses without synaptic ribbons (bassoon knock-out mice) presumably because they have lost a small pool of rapidly releasable vesicles and also have reduced Ca2+ currents (Buran et al. 2010). Remarkably, this experimental result had been anticipated by a computer modelling study that indicated that synaptic delays should become shorter for active zones with more primed docked vesicles (Wittig & Parsons, 2008). The depletion rate of the rapidly releasable pool depends on the strength of the depolarizing pulses (Edmonds et al. 2004; Brandt et al. 2005), and this pool depletion may cause spike adaptation in the auditory nerve (Moser & Beutner, 2000; Spassova et al. 2004; Goutman & Glowatzki, 2007), an idea that follows from the classic work on sound-evoked EPSPs and spikes in the afferent fibres of goldfish sacculus hair cells (Furukawa & Matsuura, 1978). Depolarization with short pulses produces a linear exocytotic output from mature mouse and gerbil hair cell ribbon synapses (Johnson et al. 2005, 2009), from adult bullfrog hair cells (Keen & Hudspeth, 2006; Graydon et al., 2011), from vestibular hair cell synapses (Dulon et al. 2009), and from turtle hair cells (Schnee et al. 2005). Moreover, hair cells have pools of vesicles that are EGTA insensitive and thus are thought to operate via nanodomains (Fig. 3A; Moser & Beutner, 2000; Graydon et al. 2011). Recovery from vesicle pool depletion appears to be Ca2+ dependent and sensitive to intracellular Ca2+ buffer concentrations at hair cells (Spassova et al. 2004; Cho et al. 2011; Schnee et al. 2011), bipolar cells (Mennerick & Matthews, 1996; Gomis et al. 1999) and cone photoreceptor terminals (Babai et al. 2010).

The large amplitude of the multiquantal EPSCs at hair cell synapses probably makes them ideal for triggering postsynaptic spikes in the afferent fibres. Moreover, the rapid rise time and decay kinetics of the EPSCs makes them suitable for phase locking to sound stimuli of different frequencies and levels. Whether uniquantal release is sufficient, or muliquantal release is necessary, for triggering spikes in the afferent fibres is still not clear (Yi et al. 2010), although at some afferents single vesicle release can trigger a spike and larger EPSC events reduce its latency and jitter (Rutherford et al. 2012). The mechanisms that produce EPSC and spike phase locking in the auditory afferent fibres also remain uncertain, although this probably involves short-term facilitation and depression of transmitter release (Cho et al. 2011; Goutman, 2012). But the specialized properties of the hair cell synaptic ribbon will likely be critical for phase locking in the auditory system.

In summary, the properties of photoreceptors, bipolar cells and hair cell ribbon synapses may be finely tuned to maximize the efficient processing of sensory information. Just like different conventional synapses in the CNS, auditory and visual system ribbon synapses appear to have several similarities, but also some key differences in the size of their readily releasable pool of vesicles, in their release probability and kinetics, in their quantal mEPSC size, and rate of recovery from vesicle pool depletion. Further in vitro studies of synaptic vesicle fusion at ribbon-type active zones from adult animals at physiological temperatures, as well as further in vitro and in vivo probing of the mechanisms of postsynaptic firing of spikes in retinal ganglion cells, whose axons form the optic nerve, and of spiking in spiral ganglion cells, whose axons form the auditory nerve, will be necessary for a better understanding of the analog-to-digital neural encoders located in our eyes and ears.



We thank NIH-NIDCD (DC004274) and NIH-NEI (EY014043) for RO1 grant support to H.v.G. and for a K99 grant to G.-L.L.