Individual synaptic vesicles mediate stimulated exocytosis from cochlear inner hair cells

Significance Synaptic transmission is codetermined by presynaptic and postsynaptic neurons. Therefore, to understand how the inner hair cell (IHC) signals to spiral ganglion neurons at the first synapse in the auditory pathway, here we directly studied individual membrane fusion events by making cell-attached membrane capacitance recordings from IHCs, for which the quantal size is debated. The observed fusion steps in membrane capacitance are consistent with the quantal hypothesis of synaptic transmission in which individual synaptic vesicles undergo exocytosis independently from each other. This finding, in conjunction with previous work, raises the exciting possibility that action potential generation can be triggered by the release of a single vesicle at the IHC synapse.


Supplementary Information Text
Signal Processing. Evaluating the Real part of the Lock-in signal for signs of a fusion pore conductance in the context of signal smoothing. The influence low-pass filtering and smoothing have on the Real (Y1) and Imaginary (Y2) parts of the Lock-in signal are in principle the same. However, the potential information carried in the Real and Imaginary signals offers distinct insight into vesicle fusion, and therefore the consequences of signal smoothing are not necessarily the same. Before examining how signal processing shapes fusion related signals, we first give a demonstration of how the Real and Imaginary parts were assigned to conductance and capacitance signals, respectively.
Three segments from an individual on-cell recording episode are presented in Figure S1. In the first example, the Imaginary part showed an up-step without accompanying changes in the Real or membrane current (I m ) traces (Fig. S1A). In the second example, the Real part fluctuated along with the I m (Fig. S1B), and the Imaginary part was steady throughout. Finally, at the end of the recording the system was calibrated by dialing in 2fF (peak-to-peak) into the fast capacitance compensation (C fast ) on the front panel of the EPC-7 amplifier. The adjustment to C fast only changed the Imaginary part (Fig. S1C). These examples support the designation of the Imaginary part as C m , and distinct from the Real part that represents an aggregate of conductance (G). The Real part includes membrane (G m ) and potentially pore (G p ) conductances. The example in Fig. S1B is considered to represent a change in G m , but not G pore , because the fluctuation in G is not associated with any obvious change in the C m that would represent membrane fusion (1, 2).
The transient nature of fusion pores makes them prone to being lost with smoothing (3)(4)(5). For this reason, the G and C m signals were inspected at full bandwidth (without smoothing) and after signal smoothing (32k-point Binomial smooth). Signal processing did not reveal changes in conductance at the onset of fusion steps that could be interpreted as a fusion pore signal, nor did fusion events display a steady G pore . For example, see Figure 1A and 1B to compare traces before and after smoothing. The C m events in these examples are not accompanied by a G that is suggestive of a pore conductance.
The fusion events reported here averaged 40 aF, which are almost 50 % smaller than all former studies using the on-cell method. The magnitude of the G pore signal is determined by the size of the C m step (1, 6, 7), and it has been predicted that to measure a G pore signal from a 80 aF fusion event, an rms-noise of 1 aF is needed (1). To measure a fusion pore from our small SVs would require an rms-noise < 1 aF, which is a noise level less than what we or others have achieved. Actual studies examining small vesicle fusion steps averaging ~70 aF, have found that full-fusion events (simple up-steps in C m ) lacked a detectable G pore (4) or only 1.4 % of the full-fusion events had a G pore signal (5) (nearly 94 % of our events were full-fusion). These previous studies confirm that pore conductance is difficult to measure from small vesicles.  . One possible explanation for the result is that the true rise time is slower than the limit imposed by smoothing, and an alternative scenario is that noise in the C m signal interferes with analysis of the rising phase.
To simulate the influence baseline noise has on event rise time, a hybrid trace was constructed by adding a series of small voltage steps (taken from the voltage traces as described above; Fig. S2E) to segments of C m baseline that were processed with 32kpoint Binomial smoothing (Fig. S2F,G). The results show that baseline noise lengthened the average hybrid steps rise by 2-fold. The original steps had rt20-80 values of 4.9 ± 0.2 ms versus the hybrid steps: 10.8 ± 3.4 ms (mean ± s.e.). The difference was not statistically significant (unpaired Two-Sample t-Test p: 0.24; n=16 steps in original and hybrid steps), because the hybrid steps had too variable. The hybrid steps had rt20-80 values that ranged from 2.7 to 45 ms (CV = 1.3), and voltage steps without added noise had rt20-80 values of 4.4 to 5.8 ms (CV = 0.1). The step amplitudes were similar for the voltages step with and without added noise, 50 ± 2 aF and 47 ± 0.2 aF, respectively (p > 0.2) (the original step was set to 47 aF, Fig. S2E).
The rt20-80 was measured for IHC. Though the signals were overly attenuated, the values are presented as a check on the analysis. When fusion step height was measured, as presented in Figure 4 and in Table 1, the rt20-80 for these events had the following median values (mean ± s.e.), wild-type: 8.7 ± 0.8 ms (n: 7 patches) and ko: 9.1 ± 0.9 ms (n: 5 patches). There was no significant difference between the two groups (p: The analysis shows that estimation of rt20-80 is susceptible to interference from baseline fluctuations (noise), which can attenuate the true rise time by > 6 msec. Thus, using C m as a indicator for vesicle fusion flickers or pore expansion is unlikely to offer insight into the sub-millisecond transitions in transmitter release (8-10) that have been proposed to generate the monophasic and multiphasic transitions in the EPSC signals (11, for review see: 12).
Kiss-and-run events have been described by others as C m steps with similar rising and falling phase kinetics (4,5,13). Such events were not apparent in our recordings.
Interestingly, a fraction of the fusion events were succeeded by a relatively slower decaying C m that may represent a membrane retrieval process. Examples of individual C m up-steps followed by C m decays are presented in Figure S3. Some of the events showed a decay in C m that was not possible to discern from a drift towards baseline (Fig.   S3A), while others showed a relatively faster decay phase that could be quantified (Fig.   S3B, C). Inspection of 3 wild type cells with low noise levels (~4 aF rms-noise) yielded 16 up-steps followed by a C m decay transition (8 % of total up-steps, 16/201). The event half-width (HW) was measured to summarize event duration (Fig. S3C), and yielded a mean HW = 471 ± 67 ms. This estimate of duration falls within the range of kiss-andrun event durations that have been reported for SVs and granules: 270 ms (5) and 560 ms (4), respectively. Next, the up-and down-step amplitudes were averaged and found to have similar mean amplitudes: 41 ± 4 aF and 40 ± 4 aF ( n=16), respectively, and when the ratio of pairs of up and down step amplitudes were computed, the mean ratio was close to unity: 1.01 ± 0.09. In contrast, the kinetics of event rise and fall were very distinct. The average fall time constant was 66.3 ± 18.2 ms and universally slower than the average rise time constant: 3.3 ± 0.7 ms (p: 0.002; n = 16).
These transient steps were not associated with a change in G that could be argued to represent a fusion pore conductance. In total, the relatively slow decay in C m and the lack of a pore conductance fit the description of 'pseudo-flicker' events. Such events have previously been described in a study that made on-cell C m recordings from human neutrophils (14), and therein is ample discussion on what these events represent.  (15)). Finally, the bath solution was lowered to the minimal amount permitted, and then the recording electrode was sealed onto the basolateral end, and modiolar side of the IHC.

Methods
Rapid seal formation was achieved without drawing a tongue of membrane into the pipette. Only one patch was made per cell, and typically one cell per cochlea. Variability in event frequency was witnessed between patches, and this is summarized in Figure 3.
Patches without events were not observed, though some patches had few events. A few wild type cells were too active to analyze.
Hardware. An EPC-7 amplifier (HEKA; Lambrecht, Pfalz; Germany) was configured with a Lock-in amplifier model SR830 from Stanford Research Systems (Sunnyvale, CA; USA) as described previously (16) to monitor changes in membrane electrical impedance. Briefly, a 58.5 kHz sine wave with a V rms between 130 to 200 mV was fed into the EPC-7 V input with the input time constant set to 0.2 µs. The EPC-7 gain was set to 50 mV/pA (using the 100 GΩ feedback resistor), and the signal output was sent at fullbandwidth to the Lock-In amplifier. The Y1 (Real, G) and Y2 (Imaginary, C m ) outputs were filtered to a 100 µs time constant with a 24 dB roll-off (SRS830 internal filter), and sampled at an interval of 20 µs using a connector box and A/D card (SCB-68A and NI PCIe-6321, respectively, from National Instruments, Austin, TX; USA) that was run by software from NI and Igor NIDAQ Tools (Wavemetrics, Portland, OR; USA). The unfiltered membrane current was passed through a 5 kHz Low-pass Bessel filter (Frequency Devices, Haverhill, MA, USA) and then collected along with the V applied (V monitor from EPC-7) as well as with Y1 and Y2. Recording epochs were collected over 120 to 300 sec, ~1 to 4 epochs per patch. A C m calibration procedure was made per epoch, which consisted of ±1 fF steps introduced via a well calibrated C fast circuitry on the front panel of the EPC-7 (see Fig. S1C). The Y1 (Real, G) and Y2 (Imaginary, C m ) Lock-in signal outputs were 90° out of phase as evidenced by 1) the exclusive sensitivity of Y2 to a charged body (hand) approaching the open headstage input, and 2) adjustments to C fast on the EPC-7 only altered Y2 (C m ; see Fig. S1C). Further phase adjustment to actual C m and G traces from patch recordings were made off-line using software provided by Manfred Lindau's group, which is described elsewhere and available online (16).
Most recordings required a phase adjustment of ± 5° from the original recording phase (0°) (as in Fig. S1). Once the C m signal was phase adjusted and deemed not to project into the G m signal, fluctuations in membrane current (I m ) (which appeared to be K + channels since E rev was ~ E K ) created a corresponding change in G m without altering C m (as in Fig. S1). were entered into the analysis program: event threshold: 20 aF (which was 5-8 x rmsnoise), time for a local maximum: 100 ms, time before a peak for baseline: 300 ms, period to average a baseline: 100 ms, and an event area threshold: 1 fF. In segments where the steps were crowded, the local max and time before peak for baseline were reduced to 50 ms and 100 ms, respectively, which created a ~100 ms baseline for the next step in the sequence. In most instances, especially during low event frequency, the capacitance steps made a smooth uninterrupted ascent with a rt20-80 % in under 10 ms (see Fig. S2). Overlapping events were not analyzed since a baseline could not be established. After the events were detected, they were aligned to their rising phase, baseline subtracted, and then the rising phase was fit with a sigmoid function ('Logistics 5' function; OriginLab; Northampton, Mass., USA). The following parameters were obtained from the sigmoid fits: 1) rt20-80 % (ms) and 2) step amplitude (aF): the difference between the pre-and post-step levels. OriginLab was used for statistical comparisons of unpaired samples, using the Two-Sample t-Test, and equal variance was not assumed (Welch correction).  with progressive smoothing and Logistic fits made to each of the three C m traces. D, Zoomed view of (C). E, Voltage step smoothed with 32k-pt Binomial smooth and its amplitude arbitrarily scaled to 47 aF. F and G, Hybrid steps made from voltage step in (F) plus baseline noise. The Logistic function was fit to steps in E-G.
. Figure S3. Pseudo-flicker fusion events have a rapid rise and slow decay. A and B, the up-steps in C m (black trace) of a wild-type IHC appear in the absence of a correlated change in G (red trace) or I m (green trace). A, The small fusion event is followed by a slow decay towards baseline, while in (B) the two steps show relatively faster and more obvious decays towards baseline, and are referred to here as pseudo-flicker events. C, Boltzmann fits to the rising (red line fit) and falling phase (blue line fit) were made to the first event in (B) to illustrate how event amplitude, and kinetics were measured. Half-width determination for these asymmetrically shaped events were measured as the time between the EC 50 rise and EC 50 fall (EC 50 is designated with grey boxes). D, Plot of up-versus down-step amplitude for each fusion event (linear fit to data in red).
The points are scattered around unity (dashed line). Scale bars are the same for (A-C).