Abstract
The neurotransmitter glutamate is used by most neurons in the brain to activate a multitude of different types of glutamate receptors and transporters involved in fast and relatively slower signaling. Synaptic ribbons are large presynaptic structures found in neurons involved in vision, balance, and hearing, which use a large number of glutamate-filled synaptic vesicles to meet their signaling demands. To directly measure synaptic vesicle release events, the ribbon-type presynaptic terminals of goldfish retinal bipolar cells were coaxed to release a false transmitter that could be monitored with amperometry by placing the carbon fiber directly on the larger synaptic terminal. Spontaneous secretion events formed a unimodal charge distribution, but single spike properties were heterogeneous. Larger events rose exponentially without interruption (τ ∼ 30 μs), and smaller events exhibited a stammer in their rising phase that is interpreted as a brief pause in pore dilation, a characteristic commonly associated with large dense core granule fusion pores. These events were entirely Ca2+-dependent. Holding the cells at −60 mV halted spontaneous release; and when the voltage was stepped to >−40 mV, secretion ensued. When stepping the voltage to 0 mV, novel kinetic phases of vesicle recruitment were revealed. Approximately 14 vesicles were released per ribbon in two kinetic phases with time constants of 1.5 and 16 ms, which are proposed to represent different primed states within the population of docked vesicles.
Introduction
Serving as the principle cell type linking the outer and inner retina, bipolar cells (BPCs) must transmit all varieties of signals, and their presence is absolutely required for the retina to function (Burmeister et al., 1996). Most types of BPCs possess only a few terminals, and a single terminal communicates with dozens to hundreds of dendrites in the inner retina (Dacheux and Raviola, 1986; Marc and Liu, 2000) by triggering the fusion of synaptic vesicles that release packets of glutamate into the extracellular space.
BPCs build active zones referred to as “synaptic ribbons,” which are easily identified at the ultrastructural level as electron-dense masses that protrude into the cell and are decorated with a dense layer of synaptic vesicles. Ribbon synapses typically consist of a single synaptic ribbon that is juxtaposed to two or more dendrites. In the case of a rodent (rod On-BPC), a single ribbon creates a synapse with an AII amacrine dendrite on one side and an A17 amacrine dendrite at its other face (Raviola and Dacheux, 1987). This arrangement likely allows a ribbon to communicate with two targets either through the coordinated release from both faces of the ribbon or through cross-talk where release from one face of the ribbon diffuses to the other side. Transmitter spillover from a single release event onto different BPC dendrites has been reported to occur at the cone photoreceptor ribbon synapse by DeVries et al. (2006). Spillover from a ribbon synapses in the inner plexiform has not been explored as extensively, but Matsui et al. (1998) provided evidence that BPC ribbons signal to local non-NMDA receptors and to more distant targets that express NMDA receptors.
A classic BPC preparation suited for studying exocytosis are the isolated goldfish, Mb1 BPCs (for review, see Tachibana, 1999) that possess large, singular synaptic terminals with ∼50 ribbons dispersed across their surface (Marc and Liu, 2000). Mb1 cells secrete the excitatory neurotransmitter glutamate, which has been monitored from dissociated Mb1 cells using a glutamate “sniffer-cell” (Sakaba et al., 1997; von Gersdorff et al., 1998), yet single release event properties were not resolved. A more common method for monitoring synaptic activity is to make paired recordings from presynaptic and postsynaptic cells, but in principle variability in postsynaptic measurements of release may arise from differences in retinal circuitry, heterogeneity in glutamate receptors (for review, see Lisman et al., 2007), morphology of the synapse (Coggan et al., 2005), glutamate clearance from the synapse (Clements et al., 1992), and the electrotonic structure of the postsynaptic cell (Bekkers and Stevens, 1996), which together can mask features of exocytosis. To directly follow the time course of release with a continuous and sensitive assay, isolated BPCs were coaxed to accumulate norepinephrine (NE) so that release could be monitored with amperometry.
The results presented here demonstrate that amperometric detection of false transmitter release from BPCs can faithfully resolve features of ribbon mediated secretion that have not been captured with conventional methods.
Materials and Methods
Tissue culture.
Retina were taken from large comet goldfish (5–7 inches, of either sex) in accord with the animal protocol approved by Institutional Animal Care and Use Committee at Yale University as previously described (Coggins et al., 2007). Eye cups were prepared by removing the cornea and then digesting the vitrea for 20 min at room temperature in 20 units/ml hyaluronidase. Next, the neural retina was removed from the eyecup and digested for 35 min at room temperature in papain (9 mg/ml) that was preactivated for 10 min in cysteine (0.5 mg/ml). The digested material was rinsed several times and stored in a 12°C–14°C, oxygenated refrigerator. And before each recording, one-fourth of the retina was mechanically dispersed and plated. The following low Ca2+, HEPES-buffered goldfish saline was used in the digestion steps, tissue storage and plating (in mm) as follows: 0.5 Ca2+, 120 NaCl, 2.5 KCl, 1.0 MgCl2, 0.5 CaCl2, 10 glucose, and 10 HEPES, adjusted to a pH 7.4.
Electrophysiology.
All whole-cell recordings were made in goldfish saline containing 2.5 mm Ca2+ in addition to the following (in mm): 120 NaCl, 2.5 KCl, 1.0 MgCl2, 10 glucose, and 10 HEPES, adjusted to a pH 7.4. The intracellular solution contained the following ingredients (in mm): 120 cesium methanesulfonate, 4.0 MgCl2, 10 HEPES, 10 TEA-Cl, 0.5 EGTA, 0.5 Na2GTP, 4.0 Na2ATP, adjusted to pH 7.3 with CsOH. Glass recording electrodes were prepared from thick-walled pipettes that were pulled to a resistance of ∼8.5 MΩ, using a Narishige vertical puller, and the tips were wax coated to reduce noise. However, to determine the amount of release over a range of voltages (data in Fig. 3J), four recordings were made with a perforated, whole-cell configuration as these recordings tended to last longer. For these recordings, the perforated pipet solution contained the following (in mm): 60 CsMeSO4, 60 CsCl, 10 HEPES, 10 CsEGTA, pH 7.3, and 266 mOsmol. Nystatin stocks were made at 10 mg/ml in MeOH and diluted to ∼2% in the pipet solution, mixed vigorously before each recording to create a final Nystatin conc. of 250 μm.
All whole-cell recordings, including membrane capacitance (Cm) measurements, were performed with a EPC-10 amplifier and Pulse software (HEKA Instruments). The liquid junction potential was nulled with the electrode in the bath when minimal positive pressure was being applied just before contacting the cell. The electrode's capacitance was subtracted in cell-attached configuration after achieving a seal typically >10 GΩ. Once intracellular access was gained with suction, the cell was dialyzed for 2 min before stimulating. Series resistance was typically between 18 and 25 MΩ. Whole-cell capacitance transients were subtracted, and membrane leak currents were estimated in advance of each stimulation by applying voltage steps equal to 10% of the stimulus amplitude. The whole-cell currents were sampled at 50 kHz and digitally filtered online at 2.9 kHz corner frequency using a Bessel low-pass filter provided in the Pulse software. To assure better control of the terminal's voltage, the whole-cell electrode was placed on the terminal of the intact cell when measuring stimulated changes in Cm and rapid, evoked release of NE, as presented in Figures 3 and 6, whereas the recordings presented in Figure 2 accessed the cell's interior from the soma. The cell's Cm was measured by applying a 500 Hz sine wave at an amplitude of ±15 mV around Vhold, sampled at 50 points per cycle, and computed with the “sine + dc” routine in the EPC-10 Lockin software (Lindau and Neher, 1988). The average change in Cm resulting from a 30 ms step depolarization was made by subtracting the Cm baseline averaged 500 ms before stimulating from the period 100–500 ms after stimulation.
Electrochemistry.
The method used for fabricating carbon fiber electrodes is essentially the same as described by Grabner et al. (2005), with one exception. The carbon fibers used in this study were 5 μm in diameter and purchased from Goodfellow. The sides of the carbon fiber electrode (CFE) protruding from the glass pipet remained insulated with epoxy after cutting the electrodes by hand with a scalpel. Every electrode was visually inspected with a microforge, set to 900× magnification, and only electrodes with flat, disc-shaped endings and intact epoxy on the sides of the cylinder were used in experiments. The low-noise recordings, such as those in Figures 4 and 7, were made with CFEs that had fibers protruding only ∼10–20 μm from the glass because longer fibers that are coated with epoxy still gain excess capacitive noise that is proportional to the surface area of the coated fiber extending beyond the glass. Electrodes were used only once. Amperometric recordings were made with an EPC-7 amplifier. The CFE was held at +750 mV relative to the silver-chloride, ground electrode. The signal was initially filtered through a 4-pole Bessel in the EPC-7 amplifier at 10 kHz and then fed into an 8-pole analog/Bessel filter (Analog Devices) set to a corner frequency of 8.5 kHz, which was collected at 50 kHz via the A/D of an EPC-10 (HEKA Instruments). All recordings were made at an inverted Zeiss Axiovert200, equipped with a 100×, 1.3 numerical aperture, oil-immersion objective and a 2× magnifying lens before the 10× eyepieces. Such high magnification was necessary to clearly view the side of the synaptic terminal when contact was being made with the CFE, and without causing injury to the cell as judged by a lack of spontaneous release at Vhold −60 mV. The CFE was set to an angle of 45 degrees relative to the plane of the coverslip, which made it possible to rest the CFE's entire disc shaped surface that was 2–3 times smaller in diameter than the cell (10–15 μm in diameter). Therefore, the path from the cell to the active surface of the CFE can be described as two opposing planes, which minimizes the diffusional distances and dilution of NE from peripheral origins.
Electron microscopy.
All of the EM reagents were purchased from Electron Microscopy Sciences. Dissociated retinal cells were prepared and plated on glass coverslips as described above. Once the cells attached to the coverslip, ∼20 min after plating, the cells were carefully rinsed in 2.5 mm Ca2+ recording media (described below) and equilibrated for 10 min at room temperature. Fixation, embedding, and sectioning have been described in detail previously (Coggins et al., 2007). Sections were viewed on a Tecnai 12 Biotwin electron microscope (FEI) that is maintained by the Center for Cell and Molecular Imaging at the Yale School of Medicine. Vesicle diameter was estimated as described previously.
Evoked release with 60 mm KCl.
To assess whether naive BPCs, not loaded with NE, released oxidizable compounds, amperometric recordings were made from the terminals of intact BPCs that were incubated in a solution identical to the NE loading solution, but absent NE, and the cells were plated as described above. The coverslip with attached cells was continually perfused in standard extracellular media that contained 2.5 mm Ca2+, and a multibarrel puffer controlled by a piezo stepper (Warner Instruments) was positioned ∼0.5 mm away from the cell to allow rapid and complete solution exchange with a delay of ∼250 ms. The puffer switched between the standard 2.5 mm Ca2+ extracellular media and a stimulating media with 60 mmol of KCl that was made by substituting 60 mmol of NaCl to keep the solutions isotonic. Each cell was stimulated with the high potassium solution for 2.5 or 4.5 s and then rinsed for 20 s before being stimulated one or two additional times.
Loading cells with NE and cell plating.
NE was selected as the false transmitter over other compounds for many reasons. Unlike dopamine (DA) and serotonin, NE is not known to be bioactive in neuronal retinal cells. Additional advantages of NE include that it is more resistant to air oxidation than DA and less expensive than 5-HT. NE, like DA, has a secondary amine that is more likely to become protonated in the acidic lumen of a synaptic vesicle, and the positive charge renders the molecule less membrane permeate, trapping it in the vesicle (Kim et al., 2000; Pothos et al., 2000). To load the cells, the papain-digested tissue was incubated in a hypertonic, low Ca2+ media plus NE and additional HEPES (in mm: 120 NaCl, 2.5 KCl, 1.0 MgCl2, 0.5 CaCl2, 10 glucose, 20 NE, and 100 HEPES, pH 7.4). Incubation typically lasted for ∼2.5 h (range, 1.5–3 h) and always at a temperature between 12°C and 14°C. The NE-loaded tissue was triturated and plated on glass coverslips in the NE loading solution, similar to the plating procedure for unloaded tissue as described above. At this time, the CFE and patch pipette were positioned above the recoding chamber, and ∼20 min was given for the cells to adhere to the coverslip. The cells were then carefully rinsed for 2 min in low Ca2+ media that contained 100 mm HEPES without NE, and slowly returned to low Ca2+ media (∼2 min). The cells were next equilibrated in the standard extracellular goldfish media containing 2.5 mm Ca2+ (in mm: 120 NaCl, 2.5 KCl, 1.0 MgCl2, 2.5 CaCl2, 10 glucose, 10 HEPES, adjusted to a pH 7.4) just before lowering the recording electrodes into the bath. A few features were key to producing healthy cells and good secretion of NE. First, the tissue had to be kept cool (∼14°C) before and during the loading, which we find is also true for tissue not loaded with NE. Second, a high HEPES concentration was found to be essential to keep the cells stable when NE was being removed. Third, the dissociated cell had to adhere to the glass, which allowed the cells to be washed, and eliminated movement, drift of the cell during dual-electrode contact.
Data processing and analysis.
To determine single spike parameters, amperometric traces were analyzed with MiniAnalysis software (Synaptosoft). Throughout the manuscript, the spike detection threshold was set to 5 × rms noise using MiniAnalysis, and additional filtering of the amperometric traces used the low-pass Gaussian filter in the MiniAnalysis program. Overlapping events were excluded from the analysis of single-event parameters. MiniAnalysis's Analysis Window was used for averaging events within cell by aligning events to the point immediately before 50% maximal amplitude (abbreviated as RT-50). In Figure 7, an additional alignment of spikes was performed, which used a cubic spline function, available in Origin software, to smooth and interpolate the data to 2.4 μs intervals over the following 3 points: (1) RT-50, (2) 20 μs before RT-50, and (3) 20 μs after RT-50. Individual fits were realigned in time to their half-maximal amplitude of the fit, and the average was then made within the cell and spliced into the parent cell's average and linearly interpolated over areas preceding and following the realigned rise. The temporal shift between the original average and the fit realigned data was measured within cell and then averaged across cells (see Fig. 7).
Describing response time and corner frequency.
Full bandwidth frequency was calculated as the effective corner frequency (efc) using the following formula (f in kHz): 1/(fc2) = 1/(f8.52) + 1/(f102), which gives an fc of 6.47 kHz for all amperometric records that were filtered at fc settings of 10 and 8.5 kHz (see above). Corner frequency was calculated in a similar fashion when the amperometric traces were further filtered with the Gaussian filter. When the Gaussian filter was set to fc values >50% of the original, full bandwidth, the rms noise was not sufficiently removed as illustrated in Figure 4E, which is a known property of this filtering procedure (Sakmann and Neher, 1983, chapter by Sigworth and Colquhoun).
To examine the frequency range of the amperometric data, Power Spectral Density plots were made (see Fig. 4D) from five cells while the CFE was in contact with the cell, but the cells were not voltage-clamped. One second segments containing spontaneous amperometric spikes were compared with recordings from neighboring stretches without spikes (baseline traces) (see Fig. 4D). Fast Fourier Transform calculations were performed with Origin (Microcal Software), and the power (amplitude2/Hz, abbreviated as Irms2/Hz) is plotted in Figure 4D, Di. Spectra calculated from baseline traces increased in power from 100 Hz to the fc, but spectra from traces with spikes are relatively flat over this lower range of frequencies. The actual fc estimated from baseline Power Spectral Density plots were measured as the point where the linear, ascending power intersected the initial taper in power. The more conventional S(0)/2 measure of fc (see below) was not possible to make because baseline spectra were not flat in the lower frequency range. The square root of the integral of power over Hz, which gives Irms, was made for the amperometric traces from 5 cells in Figure 4Di and yielded an average Irms = 0.71 ± 0.1 (mean ± SE). This value is comparable to the average Irms ∼0.65 pA made from these and additional cells (see Figs. 4E and 7) that was calculated in MiniAnalysis as the SD of the amperometric current. Figure 4Dii presents “Normalized Power,” which was calculated by taking the spectra and dividing them by their power at 0 Hz, so that S(0 Hz) = 1. In Figure 4Dii, the flatness of the difference spectra and spectra made from the average amperometric spike (see Fig. 4I) made it possible to estimate fc as S(0)/2.
The response time (tr) for the Gaussian filter has the following relationship: tr ∼ 0.340 s/fc, which estimates the time needed for the signal to reach maximal value. Because spike rise time data presented in Figure 4 are given as 10–90% RT, the relevant measure of the filter's rise time can be calculated as 10–90% RT = 0.8 × tr.
Quantifying the amount of evoked NE release.
To estimate the number of quanta within the period of a 30 ms step depolarization (see Fig. 6) or longer steps (see Fig. 2D), the following procedures were used: (1) the prestimulation baseline was averaged 100 ms before depolarization and set to zero, (2) the transient current artifacts appearing in the amperometric traces that arose from the onset and offset of the voltage step applied to the whole-cell electrode were subtracted from the amperometric trace, (3) the amperometric current was integrated to give the amount of charge over time, and (4) the trace of charge over time was divided by the average quantal size determined within cell. Finally, all of the recordings used to estimate the kinetics and quantal content of a 30 ms depolarization used only the first stimulation (the naive condition) to assure that rundown was kept at a minimum. The cumulative number of events during long step depolarizations were made by plotting the cumulative quanta from the first 30 ms depolarization, and beyond 30 ms the spikes were counted, and when spikes were partially overlapping, each peak in the cluster was counted as an individual event. However, two cells used to estimate the relative amount of NE released over extended periods were simply integrated from 0 to 500 ms (related to Fig. 2).
Results
BPCs can concentrate and spontaneously secrete NE as a false transmitter
Because glutamate is not appreciably oxidized at the surface of a bare CFE, this type of electrode cannot monitor secretion of glutamate directly. If by chance goldfish Mb1 BPCs concentrate and secrete oxidizable materials that have yet to be characterized, then CFE amperometry may sense the release of such compounds. To test this possibility, amperometric recordings were made by gently pressing the CFE (5 μm diameter) against the large terminal (∼10 μm diameter) of an isolated, intact Mb1 cell. Inspection of amperometric traces measured from several cells (n = 18) bathed in standard extracellular media failed to show secretion events (Fig. 1A). When cells were intentionally ruptured with the CFE to release their cytosolic contents, there was no change in the amperometric current, demonstrating that these cells lack appreciable levels of oxidizable materials.
Cells that typically secrete electrochemically inert compounds, like glutamate, can be coaxed to concentrate and exocytose oxidizable transmitters (Zhou and Misler, 1996; Kim et al., 2000; Zhang and Zhou, 2002). Such exogenously applied molecules are commonly referred to as “false transmitters,” and they have been used in a variety of electrochemical (Kim et al., 2000) and electrophysiological (Pan et al., 1993) studies. The false transmitter method was adapted here to study secretion from Mb1 cells. NE was chosen to serve as the false transmitter because it is readily oxidized at a CFE, less susceptible to air oxidation than dopamine (Kim et al., 2000), and importantly BPCs have not been reported to express noradrenergic receptors; yet, vasculature within the eye expresses adrenergic receptors (Ferrari-Dileo et al., 1990; Dal Monte et al., 2012). The recording in Figure 1B shows several spikes measured from the terminal of an intact BPC bathed in standard extracellular media and preloaded with NE. A closer look at single spikes highlights the near-instantaneous rise to peak, and then exponential decay back to baseline (Fig. 1C), which is the expected profile for vesicular discharge of transmitter from a nearby source (Chow et al., 1992; Schroeder et al., 1992; Haller et al., 1998; Bruns et al., 2000).
Some of the cells produced strong bouts of release without any overt attempt to stimulate the cells (Fig. 1D). Previous studies of Mb1 BPCs reported spontaneous, rhythmic elevations in intracellular Ca2+ that were dependent on VGCCs (Burrone and Lagnado, 1997; Zenisek and Matthews, 1998). To assess whether Ca2+ was mediating release, cells were contacted by a patch-clamp electrode and their secretory activity was examined before and after filling the cell with an intracellular solution that buffered Ca2+ with 0.5 mm EGTA. Almost all of the cells (10 of 11) showed spontaneous release before gaining whole-cell access; and of the active cells, 70% produced strong bursts of release (Fig. 1E). When the cells were monitored with the patch-clamp electrode in on-cell configuration and voltage-clamp mode (Vhold: −60 mV), deflections in membrane current preceded and persisted through bouts of secretion (Fig. 1E). However, after gaining entry and establishing whole-cell, voltage clamp (Vhold: −60 mV), the cell release activity was silenced, and only one of 11 cells produced an amperometric event during a 1 min period of monitoring before the first stimulation (Fig. 1F). When these same cells were depolarized, robust release began within milliseconds (described in the next sections; Fig. 2), which demonstrates that Ca2+-stimulated exocytosis remained intact, but spontaneous secretion was quelled when Ca2+ entry was restricted at a hyperpolarized membrane potential.
Step depolarizations reveal different kinetic phases of release
Figure 2B shows an amperometric trace measured from a cell loaded with NE and stimulated for 4 s with 60 mm KCl. The depolarization triggered what appears to be the release of many vesicles at the onset of stimulation, and spikes continue throughout the remainder of the KCl application. The time course of NE exocytosis is reminiscent of the rapid and sustained kinetic phases described by Sakaba et al. (1997) and von Gersdorff et al. (1998). The large current offset in the baseline that accompanies KCl application (Fig. 2B) is likely to arise from the distant release of NE, which is delivered to the CFE via the perfusion system (convective mass transport; cells not loaded failed to exhibit such an offset; Fig. 2A). To eliminate convective transport of NE to the electrode and to more precisely control the moment of depolarization, paired amperometric and whole-cell voltage-clamp recordings were made from NE-loaded cells.
Mb1 cells express L-type VGCCs that allow maximal Ca2+ entry in the range of −10 to +10 mV (Mennerick and Matthews, 1998). NE-loaded, intact Mb1 cells depolarized with a voltage ramp produced a maximal Ca2+ current of −110 ± 8 pA at +3.6 ± 2.4 mV (n = 4 cells; Fig. 2Ci). As illustrated in Figure 2C, secretion is silent at a resting potential of −60 mV; but at the onset of Ca2+ entry, stimulated with a Vm of −10 mV, a large spike in the amperometric current is generated. After this burst are desynchronized events that decline in frequency and then rebound at an accelerated rate from 50 to ∼350 ms (Fig. 2C,D). A steady, linear sustained rate persisted for the remaining depolarization and even a couple hundred milliseconds after stepping back to −60 mV. The cumulative number of events is plotted in Figure 2E for two cells with different total numbers of release events but similar kinetic phases. On average, the first 30 ms of a prolonged depolarization to −10 mV amounted to 22 ± 2% of the total number of vesicles released by 500 ms (n = 3 cells).
Two kinetic components of release exist within the rapidly releasable phase
Numerous studies have investigated rates of exocytosis in response to brief depolarizations and concluded that hundreds of vesicles fuse within the period of a 30 ms step depolarization to ∼0 mV (Mennerick and Matthews, 1996; von Gersdorff et al., 1998; Burrone and Lagnado, 2000; Heidelberger et al., 2002; Coggins and Zenisek, 2009). To assess whether intact cells loaded with NE are able to incorporate vesicle membrane at a similar rate, cells were simultaneously monitored with amperometry and membrane capacitance techniques. Figure 3A shows an example of an evoked response stimulated with a 30 ms depolarization to 0 mV, which shows an abrupt, transient spike in the amperometric signal that occurs during a steady Ca2+ current (for the average response, see Fig. 3B), and Cm measured after stimulation rose by ∼40 fF (Fig. 3C). Of the cells examined, an increase in Cm was always accompanied by a burst of NE secretion, and the average change in Cm was 40 ± 2 fF (n = 4 cells; Fig. 3C), which is similar to previous measurements of capacitance increases from isolated Mb1 terminals with a similar intracellular Ca2+ buffering and stimulation protocol (Mennerick and Matthews, 1996; von Gersdorff et al., 1998).
Mennerick and Matthews (1996) demonstrated that a strong step depolarization stimulates the fusion of an “ultra-fast” pool of vesicles that empty with an estimated time constant of ∼1.6 ms (also see Heidelberger, 1994). The protocol used to measure membrane capacitance was achieved by applying stimulations of varying length to generate plots of dCm over time and the period needed to measure Cm increases the sampling time. Thus, these measurements are discontinuous and unable to distinguish between vesicle fusion during the stimulation from asynchronous events happening after the calcium channels have closed. Furthermore, changes in membrane capacitance are not always simultaneous or even correlated with secretion (Haller et al., 1998; Yamashita et al., 2005). To examine this initial kinetic phase of release with the continuous amperometric readout, NE-loaded cells were given a brief voltage step to 0 mV. Figure 3D shows an individual amperometric trace that exhibits synchronous release immediately after the onset of stimulation, and this was trailed by desynchronized events that are presented on an expanded time scale in Figure 3E, and the cumulative charge is plotted in Figure 3F. From the average response, release is shown to start after a delay of 1.4 ms after the onset of depolarization, peaked at 2.3 ms with a 10–90% rise time of 1.0 ms, reaching 5.1 pA, and forming a half-width of 2.0 ms (the average ICa was −97.7 ± 13.6 pA). The normalized cumulative charge can be nicely fit as a biexponential process with time constants of 1.5 and 16.4 ms (Fig. 3G). When accounting for the time needed to change the voltage of an intact Mb1 cell (∼0.7 ms) (Mennerick et al., 1997) and activate the calcium channels (∼0.2 ms) (Mennerick and Matthews, 1998), which on average required 0.94 ms to reach 50% ICa in this study when stepping to 0 mV, the delay between Ca2+ elevation and release onset is estimated to be <0.5 ms (also see Fig. 3K). Hence, the amperometric approach reveals that the first kinetic phase of release is synchronized to Ca2+ entry and a second novel trailing phase is an order of magnitude slower even though Ca2+ continued to enter at a steady rate. Both phases of release are likely arising from the membrane directly opposite the electrode because the fast up and down strokes in the amperometric current demonstrate the nearness of the source and the rapid oxidation of NE (Zhou et al., 1996).
The synchronous burst of release often has a split peak (Fig. 3A,D,H), and spikes or rounded transitions in the rising and falling phases are also common (Fig. 3H), which suggests the contribution of individual vesicles. To examine whether the burst could be decomposed into smaller components, the following two approaches were taken to minimize the synchrony of vesicle fusion: (1) the cells were given a series of 30 ms steps to different depolarizing potentials; and (2) release was stimulated with very transient 0.5 ms steps. In the first approach, the amount of release produced in response to a −20 mV depolarization for 30 ms reached 83% of the maximal amperometric charge resulting from 30 ms steps to more positive potentials between −10 to +20 mV (Fig. 3J). However, the time taken for steps to −20 mV to reach the same initial 20% of amperometric charge that was attained by steps to 0 mV was on average threefold greater (an additional 4.8 ± 1.3 ms; paired Student's t test: p = 0.04; n = 4 cells), which illustrates that the synchronous phase can be dispersed over the 30 ms period by lowering the stimulus intensity. The second approach used steps to +40 mV for 0.5 ms that generated small Ca2+ tail currents upon return to −60 mV, which were able to trigger amperometric spikes with delays of ∼0.5–1.5 ms (Fig. 3K). The tail-triggered spikes appear similar to individual spikes recorded during the desynchronized phase triggered during steps to 0 mV (Fig. 3Ki,Kii); furthermore, the largest amplitude of a tail-triggered spike was only a fraction of the peak amplitude of the synchronous response measured within the same cell (22 ± 5%; tail-triggered, 1.8 ± 0.6 pA; and synchronous bursts, 8.0 ± 2.4 pA; n = 3 cells), yet the time of release was tightly coupled to the start of the stimulation under both conditions, and no spontaneous amperometric events were measured over these periods. Together, these results argue that the burst of release is derived from the highly synchronized fusion of multiple vesicles.
Detection of unitary events
To estimate the number of vesicles fusing during the 30 ms stimulations, the amount of NE forming a single amperometric spike had to be determined. This first required developing a rational approach to separating the amperometric events from the noise. In Figure 4A–C, it is apparent that the removal of noise with low-pass Gaussian filtering unmasks smaller amplitude events (Hochstetler et al., 2000). Figure 4D describes the influence of low-pass filtering on the amperometric traces in the frequency domain (also see Materials and Methods).
Figure 4E plots the rms noise over fc, and the relationship between the number of events exceeding 5 times the rms noise for a given fc is presented in Figure 4F. The threshold crossings at 6.5 kHz fc were only 30% of the total measured at 1 kHz fc. Over the lower range of fc settings where noise reduction and fc are approximately linearly related (Fig. 4E), 80% of the events detected at 1.8 kHz were also captured at an fc of 2.7 kHz, but an insignificant difference of 6% was witnessed between 1.8 and 1 kHz (Fig. 4F). From these results, the advantage of noise removal is a gain in event detection frequency, but low-pass filtering also attenuates high-frequency currents arising from exocytosed NE (Fig. 4D–Dii), which can introduce sluggish response times that affect the spike's kinetics and amplitude. In contrast, a spike's charge is not distorted with the Gaussian filter routine used in this analysis (Fig. 4I) (for review, see Sakmann and Neher, 1983, chapter by Colquhoun and Sigworth) and spike charge is an additional, independent parameter for evaluating the types of events.
The relationship between spike charge and fc is positively correlated, with spikes measured from traces filtered to 2.7 kHz fc being significantly different from neighboring filter settings of 1.8 and 6.5 kHz (Fig. 4G); however, spike charge measurements made at 1 kHz fc were similar to those analyzed at 1.8 kHz fc (5% difference; p = 0.38). The plateau in event charge at lower fc settings suggests that the majority of events are detected at a bandwidth of 1 kHz, which is in accord with the finding that the change in event number is insignificant over this range of fc settings. Low-pass filtering in this context has unearthed the smallest spikes.
What useful spike kinetics can be derived from the low-pass filtered data? The average spike analyzed at 6.5 kHz fc has a 10–90% RT of ∼140 μs, which is slower than the filter's 10–90% response time of ∼42 μs (Fig. 4H); therefore, by this measure, the rising phase is not limited by the filter (this is examined further in Fig. 7), but these are only a minority of the total events that are detected at 1 kHz fc. As the fc is lowered to 1 kHz, the filter's response time is limited to 340 μs, which shifts the events detected at higher bandwidths to longer rise times and half-width values and suppresses their amplitudes (Fig. 4H). Thus, more low-pass filtering creates a false sense of linearity between amplitude and charge (Fig. 4J, 1 vs 6.5 kHz fc) and reduces the slope, but it remains possible that the smaller amplitude events are actually kinetically distinct as can be inferred from the analysis performed at the intermediate fc of 2.7 kHz that shows a clustering of smaller charge events <0.5 fC that are not linearly associated with larger amplitude events (Fig. 4K). Plots of amplitude versus charge when made from chromaffin granule release events are also not linearly related over all size ranges (Elhamdani et al., 2001; Grabner et al., 2005).
Single events are uniformly distributed in size
The spike charge distributions derived from events detected with progressively stronger filtering are presented in Figure 5A–C. Clearly, the distribution of event sizes broadens as the noise is removed, growing toward small quantal events, and leading to a increased coefficient of variation (fc: CV; 6.5 kHz: 0.36; 2.7 kHz: 0.46; 1 kHz: 0.51). Several lines of evidence argue that the high CV in spike charge measured at 1 kHz is not attributable to recording, or between cell, variability. First, the rms noise and likewise the amplitude threshold varies little across recordings, and this kept the amplitude detection threshold constant (Fig. 4E). Second, the ensemble averages for these cells are highly uniform (Fig. 5D). And finally, the spike charge distributions (Fig. 5A–C) and the average events (Fig. 5D) were created by incorporating a similar number of events from each cell to give them equal influence on the distribution. These findings support the interpretation that the spread in charge values measured in the 1 kHz dataset arises from variation in secretion events common to most cells.
Another approach often used to inspect the population of amperometric events is to plot the cube-root of the charge (Q1/3) (Wightman et al., 1991; Bruns et al., 2000). Figure 5E–G plots Q1/3 for the same events presented in Figures 5A–C, and this gives the 1 and 2.7 kHz datasets a more Gaussian profile that is centered at ∼0.92 fC1/3 (Fig. 5H). The far left side of each distribution appears to lack small events, and this is more pronounced in the 2.7 kHz dataset than the 1 kHz dataset. Assuming the geometry of the Q1/3 distribution is truly Gaussian, then the left half of the 1 kHz distribution, which appears to end prematurely at ∼0.6 fC1/3 (∼700 molecules), is taken to represent the system's limit of detection rather than the actual smallest event. If the smallest events were missed, the central mode of the Q1/3 distribution (0.92 fC1/3) will provide the most reliable estimate of the average charge, and from the plot in Figure 5G the average spike charge is 0.78 fC, which equates to 2429 molecules of NE released per event.
NE-loaded synaptic vesicles are uniformly distributed in size
In a previous study, we reported that the majority of synaptic vesicles in dissociated BPCs are closely grouped in size (Coggins et al., 2007), which has also been described by other groups (Holt et al., 2003; Paillart et al., 2003). The cytoplasm of cells loaded with NE show numerous small-sized SVs, and their ribbons display a rich halo of tethered vesicles (Fig. 5K). An examination of vesicle morphologies common to BPCs in intact tissue was also made (Fig. 5I). Plots of vesicle sizes measured from dissociated, NE-loaded cells, and intact, untreated retinal preparations (Fig. 5J,M) have abundances of vesicles between 40 and 45 nm (outer vesicle diameter). Dissociated cells show slightly more variation in vesicle size (CV = 0.20) than SVs in intact tissue (CV = 0.17), arising from the addition of more smaller- and larger-sized vesicles. Other studies have described a greater degree of variation in vesicle sizes in the dissociated condition (Holt et al., 2003; Paillart et al., 2003; Matthews and Sterling, 2008) than experienced in intact tissue (Lagnado et al., 1996; von Gersdorff et al., 1996). This is likely an outcome of higher spontaneous activity in dissociated cells (unclamped), which lack the inhibitory feedback that is present in situ (Matthews and Sterling, 2008).
Amperometry studies commonly compare the cube-root of charge distribution to vesicle diameter sizes to assess the influence of vesicle morphology on secretion (Finnegan et al., 1996; Pothos et al., 1998; Bruns et al., 2000; Grabner et al., 2005). These studies have proposed that the releasable transmitter content is proportional to the volume of the vesicle; therefore, the distribution of Q1/3 should be influenced by the distribution in vesicle diameters. Both the secretion and vesicle distributions reported here are Gaussian in shape, and events analyzed at 1 kHz fc (the most inclusive filter setting) have a CV equal to 0.17, which is similar to the CV estimated for all vesicle diameters. Given the amount of transmitter released per event (<2500), the concentration of NE for a 40 nm vesicle equates to 0.25 m, which is similar to previous estimates derived from cells that normally exocytose catecholamine and indolamine compounds (Wightman et al., 1991; Bruns et al., 2000; Grabner et al., 2005).
The quantal content of primed vesicles at a ribbon
The number of vesicles fusing during a depolarizing stimulus has been measured at specialized synapses, such as the calyx of Held, where it is possible to control Ca2+ entry while simultaneously monitoring the EPSC. A complication in this approach is the need for extensive algorithms to sort out the quantal size, which is a prerequisite for estimating the quantal content (e.g., Neher and Sakaba, 2001). Amperometry is well suited for measuring quantal size of dopaminergic (Pothos et al., 1998) and serotonergic (Bruns et al., 2000) SV release events, but these previous amperometry studies from neurons stimulated release through bath application of secretagogs or current-clamp techniques. Unique to this study is the ability to voltage-clamp the cell and tightly control calcium entry, making it possible to then estimate the quantal content of the rapid evoked response.
The integrated oxidative current measured over a brief stimulation was divided by the average single-event charge (determined within cell), which yielded on average 27 ± 3 quanta per recording site (active surface area of electrode, 19.6 μm2) at the end of the 30 ms depolarization (one recording site/cell; n = 11 cells). Plots of individual evoked responses are overlaid and presented as cumulative quantal content in Figure 6A. Inspection of the entire set of charge over time traces reveals that the number of quanta released in 5 ms (a period ∼3-fold > the fastest kinetic component of evoked release; see Fig. 3G) appears to predict the final outcome at 30 ms. The magnitude of the fast and slow phases are plotted separately (Fig. 6B,C), and the relative contribution from the first 5 ms and remaining 25 ms is 57% and 43%, respectively (Fig. 6D). Figure 6E shows that the quantal content of the fast and slow kinetic components are correlated with one another, which is taken as evidence that both release phases originate from the same type of structure. Assuming these events originate from vesicles docked at ribbons, and a previous study has reported that dissociated Mb1 BPCs express ∼1.9 ribbons per 20 μm2 of membrane surface area (Zenisek et al., 2004), an average output of 26 quanta/1.9 ribbons leads to an estimate of 14 vesicles/ribbon in 30 ms.
Amperometric spikes are preceded by a brief foot signal
The profile of catecholamine released from individual neuroendocrine granules, and detected with amperometry, is well known to deviate from the profile of release expected for an instantaneous point source (Chow et al., 1992; Schroeder et al., 1992; Albillos et al., 1997). Individual granule exocytotic events often exhibit an elevation in amplitude that precedes the larger, main spike, and this prespike feature is considered to represent a transition from a low to high conducting pore. To look for such a feature in synaptic vesicle fusion events, amperometric spikes were inspected at full bandwidth (6.5 kHz fc) to avoid loss of transient features in the spike's rising phase. Because an event's intersection with the baseline is complicated by noise, it is difficult to evaluate this portion of the rising phase on an individual spike basis; therefore, single events were aligned to their rising phase at half-maximal amplitude (RT-50) and averaged. Figure 7A presents superimposed, aligned events for two separate recordings to highlight the event's rapid ascent and exponential decay back to baseline, and Figure 7B shows an ensemble of averages within cell; the grand average from these cells is shown in Figure 7C.
The rising phase of the average spike elevates above a threshold of 2 × the rms noise at ∼120 μs before RT-50 and progresses at a relatively slower rate before rapidly jumping to peak amplitude (Fig. 7D). By grouping these events into two amplitude size ranges, it was possible to identify heterogeneity in the early portion of the rising phase, with the average small event (<4.5 pA) exhibiting a greater delay before the onset of the main spike (Fig. 7E). If the transmitter were diffusing from an instantaneous point source to the much larger surface of the CFE, the rising phase of the spike should have an exponential rise until it reaches an inflection point before stepping to peak amplitude (Chow et al., 1992; Schroeder et al., 1992). An exponential curve fit to the early portion of the small and large events, and then extrapolated further, shows that the small events pause momentarily at ∼40 μs before RT-50, whereas the larger events more closely conform to the exponential rise (Fig. 7E).
The events detected at full bandwidth are closer to the noise, putting them at a higher susceptibility to contamination from random signal fluctuations, which could produce an erroneous feature at the foot of the spike. To rule out this possibility, square-wave steps were repeatedly immersed into actual stretches of amperometric traces that lacked secretion events, and then the hybrid traces were analyzed at full bandwidth to test whether the noise could add a foot signal to the ideal step. Square-waves with similar amplitude and the same sampling rate as the actual average spike set to two different corner frequencies (6.5 and 4.0 kHz) failed to acquire a prespike foot from the noise (Fig. 7F,G). This demonstrates that the ensemble averaged, foot signal apparent in small events is neither diminished nor born from random noise.
As shown in Figure 7D–G, the spikes transit most rapidly through their rising phase, depositing only 1 point between 50 and 80% of the spike's upstroke. To improve on the kinetic description of the rising phase, the data points immediately before and after RT-50 were digitally smoothed and interpolated through RT-50 using a cubic spline function (see Materials and Methods). The interpolated data points were set to an interval of 2.4 μs, improved from the original sampling rate of 20 μs, and the new fits were realigned to their RT-50. This procedure shifted the rising phase leftward by 6 ± 1.4 μs from the original RT-50, and the original points before and after RT-50 were leftward shifted to a lesser extent (2.3 ± 2.7 μs and 4 ± 1.9 μs). By normalizing the amplitudes of the realigned small and large events, a high degree of overlap above RT-50, along the steepest, linear portion of the rising phase, is nicely illustrated in Figure 7H,I, and Table 1. Exponential fits to the small and large events initial rise give similar values (τ ∼ 30 μs; Table 1), but the small events deviate much earlier from the exponential fit compared with large events, pausing for an additional ∼50 μs before rejoining the fast upward accent. The resultant average response is reminiscent of a prespike foot that has been observed in release from dense core granules (Chow et al., 1992), suggesting a brief small opening of the fusion pore precedes its dilation.
Discussion
The false transmitter method extends our understanding of vesicle priming and single-event fusion properties
Several studies exploring exocytosis from Mb1 BPCs have described a very rapid evoked response (Heidelberger, 1994), and there is also ample evidence that prolonged depolarizations can stimulate late phases of release with time constants of a few hundred milliseconds (Mennerick and Matthews, 1996; Palmer, 2010). The data presented in this study recapitulate these well-established rapid and late phases of release; and by virtue of the amperometric signal's sensitivity and superior temporal readout, novel features of release were resolved at the level of single events and evoked responses.
Vesicles in different primed states coexist at the ribbon
Separate research groups studying exocytosis from Mb1 BPCs have proposed that 17–20 vesicles are physically tethered to the base of the ribbon where it joins the plasma membrane (von Gersdorff et al., 1996; Llobet et al., 2003), and these vesicles are thought to be the first to fuse with a τ ∼ 1.6 ms (Mennerick and Matthews, 1996), which was referred to as the “ultrafast” component of release. Here we directly measured the quantal content from a circumscribed area of membrane and found an additional rapid, kinetic stage of release. The first phase formed a multiquantal spike, synchronized to the onset of Ca2+ entry, and expired with a τ = 1.5 ms. The second component involved multiple fusion events that were distributed over a greater period of time, τ = 16 ms, and referred to as the desynchronized phase. The quantal content measured for each phase was correlated in size per recording site (Fig. 6), which is interpreted to reflect a common ribbon structure gives rise to both phases of release. As described in Results, a ribbon is estimated to release 14 vesicles within a 30 ms step to 0 mV. This estimate is 3–6 vesicles less than the number of docked vesicles at a Mb1 ribbon's base (von Gersdorff et al., 1996; Llobet et al., 2003), which could represent a population of “unprimed” vesicles. It is also possible that the morphological estimates of docked vesicles are overestimates, or alternatively a few of the fusing vesicles lack NE and are not captured with amperometry. In summation, our results demonstrate a temporal splitting of the “ultrafast” component of release that could not be resolved using other techniques, and we propose that two pools of primed vesicles are formed at the ribbon synapses.
Independently fusing vesicles are likely to form the two rapid kinetic phases of release
The synchronous phase is realized by stepping to 0 mV and is proposed to originate from the fusion of vesicles in less than perfect unison (Isaacson and Walmsley, 1995), as can be inferred from the small spikes decorating the burst phase as it rises, peaks, and then falls (Fig. 3A,D,H). The synchrony was disrupted by stepping to less depolarizing voltages for 30 ms (Fig. 3H–J); and at the other extreme, stepping to +40 mV for 0.5 ms elicited small amplitude, single events that were tightly synchronized to the brief tail currents (Fig. 3I,K). The desynchronized events and tail-triggered spikes were rapid to rise and fall (Fig. 3E,H,J), demonstrating their nearness to the electrode. The recording configuration was optimized to collect events in direct opposition to the CFE (see Materials and Methods).
Another possibility is that the two phases of release originate from a greatly oversized vesicle forming either before or after Ca2+ entry (Matthews and Sterling, 2008). If a large transmitter-filled vesicle were present before the onset of stimulation and then triggered by Ca2+ to empty through a flickering pore (Henkel et al., 2000; Staal et al., 2004), a rundown in spike amplitude would be expected as the vesicle's transmitter concentration fell. In this study, the first phase released the majority of NE as a large spike with jagged edges, and then an abrupt diminution in spike amplitude marked the desynchronized phase, but without any obvious trend toward a reduction in amplitude (Fig. 3D,E,H). Release frequency tapered off with time (Figs. 3G, 6A); but when the depolarization was extended beyond 30 ms, the spike frequency returned to desynchronized release events that were much smaller in amplitude than the original synchronous spike (Fig. 1C–E). But what directly addresses the challenging question at hand is the fact that the single events are kinetically much faster than the synchronous burst phase. Single events rose faster than they fell and required only ∼1 ms from start to finish (Figs. 3Ki, 7A). This same amount of time was consumed by the burst phase as it scaled its 10–90% RT. If two kinetic phases were arising from an oversized vesicle that was flickering over several milliseconds, then the flux of transmitter would have to follow the same kinetic profile even though the amplitude/concentration was progressively diminishing (Staal et al., 2004). This is not congruent with what we see; rather, jitter in spike timing broadens the synchronous phase and a distinct prime state leads to the desynchronized events.
Why is multivesicular release used at a BPC ribbon?
Coupling the release of multiple vesicles to an abrupt depolarization will improve the chances of the ribbon signaling to both proximal and distal receptors (Matsui et al., 1998; Midorikawa et al., 2007), which may be needed to engage certain features of the inner plexiform's network (Vigh and von Gersdorff, 2005; Vigh et al., 2011). When considering the nonuniform, anisotropic orientation of a ribbon-docked vesicle with the numerous postsynaptic targets (Marc and Liu, 2000; Sikora et al., 2005) and the variability inherent in unitary events (Fig. 5), single-event singling at this synapse would likely fail to activate a coordinated postsynaptic response.
The foot signal
It is a widely held view that synaptic vesicles open instantly to a high conducting state and give rise to the peak of transmitter in the synapse (Stiles et al., 1996). By contrast, our results suggest that some spikes are preceded by a prespike signal (Fig. 7). This observation is novel in that a SV fusion event has a prespike foot, a characteristic only previously described in larger, granule fusion events. However, other groups have argued that the SV's opening pore is subject to modulation in ways that alter transmitter efflux (Liu et al., 1999; Zhang et al., 2007), and there is evidence that SVs can maintain their pore throughout exocytosis (Pawlu et al., 2004; He et al., 2006), and even enter a “flickering” state that is characterized by the repetitive, opening and closing of the exocytotic pore within a millisecond (Staal et al., 2004).
When release is monitored via the postsynapse, the static, electrotonic structure of the cell and its active conductances can limit the temporal readout and contaminate the single events, respectively (Bekkers and Stevens, 1996; Franks et al., 2003). For instance, a cell with a capacitance of 10 pF and series resistance of 10 MΩ creates a time constant of 100 μs (cutoff frequency ∼ 1.6 kHz; fc = 1/2πτ), which is too slow to resolve the foot signal described here. On the other hand, measurements from the neuromuscular junction have provided the bulk of high temporal resolution single-event data (Del Castillo and Katz, 1954; Stiles et al., 1996).
Stiles et al. (1996) estimated that mEPC measured at the lizard's nmj have RT 20–80% values ∼100 μs, with the fastest events predicted to take only ∼30 μs (Stiles et al., 1996). The large events described in Figure 7 and Table 1 match their fastest miniature RT 20–80% perfectly, and the amperometric events with a foot signal described here have RT 20–80% of 100 μs. The small and large amperometric events have similar kinetic properties, except for the small events slow for just enough time to gain a foot. The amperometric foot is simply interpreted as a “stammer” in the fusion pore's dilation. Because evidence for postfusion modulation of the pore has been described to underlie single event properties measured at the fly nmj that uses glutamate as an NT (Pawlu et al., 2004), and complex fusion events have been described at the DA synapse (Staal et al., 2004), it is not unreasonable to suggest a distinction between the different preparations. In time, the SV fusion pores may be characterized as having stable subconducting states, referred to as stand-alone foot signals in the context of secretory granule fusion events (Albillos et al., 1997; Poberaj et al., 2002; Vardjan et al., 2007).
Assuming that the stammer in pore dilation can lead to trapping or premature closing of the vesicle, such a subconducting state could impact synaptic transmission in numerous ways. For instance, a stand-alone foot may stimulate the fastest activating AMPA or kainate receptors (Li et al., 2003) and render them desensitized before the next spike of transmitter enters the synapse (Trussell and Fischbach, 1989); or as reported previously, the smaller amplitude spikes may selectively activate the highest-affinity NMDA receptors and bypass AMPA/KA receptors (Liu et al., 1999; Choi et al., 2000).
Footnotes
This work was supported by National Eye Institute Grants EY014990 and NRSA EY07115-15. We thank Dr. Fred Sigworth and Victor Pantani for their helpful discussions on signal processing and low noise recordings and the following laboratories in the Department of Cellular and Molecular Physiology for use of their equipment: the late Dr. Steven Hebert's laboratory for loaning the EPC-7 amplifier and vertical electrode puller; Dr. Sigworth for use of the Narashige Microforge; and the late Dr. Marc Pypaert for use of excellent diamond knives.
The authors declare no competing financial interests.
- Correspondence should be addressed to Dr. Chad P. Grabner, Saarland University, Department of Anatomy and Cell Biology, 66421 Homburg, Germany. chadgrabner{at}gmail.com