Introduction

Sensory synapses in the retina rely on the proper function of a specialized organelle, the synaptic ribbon ^1-5. Retinal bipolar cells serve as the major conduit for transmitting visual information across the vertebrate retina. Rod bipolar cells (RBCs) can release brief bursts of neurotransmitters to signal a change in contrast or sustain the continuous release of neurotransmitters in a graded manner to provide an analog read-out of luminance ⁶. To maintain this ability, the RBCs must exert dynamic control over neurotransmitter release rate and facilitate efficient recruitment of release-ready vesicles to fusion sites near the synaptic ribbon. It is established that the elevation of presynaptic Ca²⁺ in RBCs regulates both dynamic changes in the release rate and accelerates the rate of vesicle replacement ^6-14. However, the spatiotemporal properties of the Ca²⁺ signals that control neurotransmitter release and the molecular entities that regulate the interplay between Ca²⁺ signals and synaptic vesicle dynamics to sustain kinetically distinct neurotransmitter release components remain poorly understood. Here, we begin to address this lack of knowledge by measuring local Ca²⁺ signals at positions along the synaptic ribbon at different distances from the active zone in retinal RBCs.

Resolving local Ca²⁺ signals is technically challenging as it requires information about the spatiotemporal properties of Ca²⁺ signals specific to ribbon sites. Previous studies of Ca²⁺ dynamics in goldfish and mammalian RBCs focused primarily on the terminal as a whole and used quantitative methods to examine bulk Ca²⁺ levels, which are significantly lower and slower than those occurring at the active zone ^15-18. Our previous studies estimated that the Ca²⁺ signals at a single ribbon active zone in zebrafish RBCs likely exceed sub-micromolar concentration level within 3 milliseconds after the opening of voltage-gated Ca²⁺ (Cav) channels ¹⁹. However, these studies were based on estimates using soluble Ca²⁺ indicators to estimate the Ca²⁺ signals at the plasma membrane, which are free to diffuse and thus cause the spread of the signal. Measurements of Ca²⁺ signals along the axis of the ribbon have not been attempted, even though these signals likely control replenishment and priming of vesicles for slower phases of neurotransmitter release. To determine Ca²⁺ signals along the axis of zebrafish RBC ribbons, we establish a nanophysiology ratiometric approach that measures the spatial and temporal properties of Ca²⁺ transients by targeting the Ca²⁺ indicator to the ribbon via conjugation to a ribbon binding peptide (RBP) ²⁰. Resolving Ca²⁺ signals in the immediate vicinity of the active zone is not currently possible due to the finite spatiotemporal resolution of optical imaging. Thus, we employ a quantitative model using Ca²⁺ diffusion and buffering to resolve the RBC Ca²⁺ signals associated with the active zone and along the ribbon and provide a detailed description of the spatiotemporal Ca²⁺ dynamics across zebrafish RBC ribbons. Our nanophysiology approach for measuring local Ca²⁺ transients at a single ribbon with high spatiotemporal resolution provides the first evidence of heterogeneity of Ca²⁺ signals at zebrafish RBC ribbon synapses.

Results

Ca²⁺ concentrations at single ribbon locations measured using high- and low-affinity diffusible indicators

Synaptic vesicles in RBCs are distributed among at least four distinct pools based on their fusion kinetics, which are assumed to reflect the average proximity of vesicles to Cav channels and the state of vesicle preparedness for Ca²⁺-triggered fusion ^{1,6,8,11-13,21-25} (see Supplementary Fig. 1). To visualize and measure Ca²⁺ signals near the RBC synaptic ribbon controlling such kinetically distinct components of neurotransmitter release, we established a quantitative nanophysiology approach shown in Fig.1 (see also Materials and Methods). In this approach, both the ribbon location and the spatiotemporal Ca²⁺ signal profile were simultaneously measured by dialyzing the zebrafish RBC terminals with both TAMRA (tetramethyl rhodamine)-labeled RIBEYE-binding peptide (RBP) ²⁶ (to label synaptic ribbons; Fig. 1Ai) and a high-affinity Ca²⁺ indicator Cal-520^® (Cal520HA, effective K_D 795 nM; see Materials and Methods) using a whole-cell patch pipette placed directly at the cell terminal. A rapid x-t line scan was taken perpendicular to the plasma membrane across a ribbon, extending from the extracellular space to the cytoplasmic region beyond the ribbon (Fig. 1Aii). The spatial resolution was limited by the point spread function (PSF) of the microscope to approximately 270 nm (Supplementary Fig.2A). In our previous studies, line scans have been applied primarily to measure the temporal properties of Ca²⁺ signals ¹⁹. However, x-t raster plots obtained at a ribbon active zone labeled with RBP allow us to characterize the spatial localization of Ca²⁺ transients relative to the synaptic ribbon and plasma membrane (Fig.1C-D) in addition to characterizing its temporal aspects (Fig.1E-F). We previously used a similar approach for localizing and tracking single synaptic vesicles before and during fusion at a single ribbon ²⁷, and to measure the kinetics of clearance of fused synaptic vesicle membrane in zebrafish RBC ²⁸. We found that depolarization-evoked Ca²⁺ influx caused a rapid increase in Ca²⁺ signals at ribbon locations (Fig.1C, cyan, white horizontal arrowhead) and increased more slowly and less dramatically in the cytoplasm (Fig.1C, cyan, white vertical arrow) in zebrafish RBC. The Sigmoid-Gaussian function fitting of the x-t scans horizontal profile scans (see Materials and Methods) show that the local Ca²⁺ signals increased rapidly during stimuli (Fig.1D, cyan line) and then decreased immediately after the end of depolarization (Fig.1D, black lines) approaching the spatial profile corresponding to the resting Ca²⁺ levels (Fig.1D, gray lines). As expected, the centroid position of Ca²⁺ signals during depolarization (Fig.1D, cyan x₀) is closer to the plasma membrane (Fig.1D, magenta x_1/2) than the centroid position of RBP (Fig.1D magenta, x₀), since Cav channels are located at the membrane ^19,26,29-42.

Fig 1.

To characterize the temporal Ca²⁺ profile at different distances from the plasma membrane, we analyzed the x-t line scans along the time-axis at three distinct distances (Fig. 1E-F) based on the spatial profile of the ribbon described in Fig.1C-D (see also Materials and Methods). We will refer to the corresponding three measurements as ribbon-proximal, ribbon-distal, and cytoplasmic (see Materials and Methods). We found that at the onset of the stimulus (Fig. 1F, black arrow), the fluorescence at the location near the ribbon proximal to the membrane (ribbon-proximal) rose more rapidly and to a higher level (Fig. 1F, light cyan trace) than that at the ribbon-distal and cytoplasmic locations (Fig. 1F, dark cyan, and gray trace). To quantify the kinetics of ribbon-proximal and distal Ca²⁺ signals, we averaged multiple x-t scans acquired with Cal520HA under the same imaging conditions (Fig. 2A). The rise time of the Ca²⁺ signals produced in response to brief 10 milliseconds stimuli, defined here as the time for the averaged Ca²⁺ transient measured with Cal520HA to reach half-maximal peak fluorescence amplitude from the resting level (Fig. 2A), is faster for the proximal (∼2.81 milliseconds) than distal (∼3.44 millisecond) signals (Fig. 2A). The corresponding maximum values of trial-averaged ΔF/F_rest (changes in the Cal520HA fluorescence during depolarization normalized to background level before depolarization) was significantly higher at ribbon-proximal locations (nearest the Cav channels) than those at ribbon-distal locations, with mean ± SEM of 1.9 ± 0.3 and 1.5 ± 0.25, respectively (Fig. 2A; p< 0.001, N=24; 95% confidence limit). These findings suggest that the spatial resolution of our Ca²⁺ imaging using Cal520HA is sufficient to resolve differences in the smaller Ca²⁺ signals at ribbon proximal and distal locations.

Fig. 2.

Since Cal520HA will be saturated at the Ca²⁺ levels expected in Ca²⁺ microdomains relevant for vesicle exocytosis ⁴³, we repeated the x-t line scan analysis with a lower-affinity soluble Ca²⁺ indicator Cal520LA (K_D 90 μM; Fig. 2B) allowing us to better define the typical Ca²⁺ signals controlling distinct neurotransmitter release components corresponding to locations proximal and distal relative to the synaptic ribbon (Fig. 2B). We found that the ribbon-proximal signals detected with Cal520LA (Fig. 2B, light cyan) showed a sharper rise and decay at the onset and termination of the stimulus (Fig. 2B, black arrow), when compared to ribbon-distal signals (Fig. 2B, dark cyan), as one would expect for nanodomain Ca²⁺ elevations ^44-46. The relationship between rise times for the proximal (4.58 millisecond) and distal (7.10 millisecond) signals were similar to those we found with Cal520HA, with both proximal Ca²⁺ signals being faster than the distal signals. In twenty-one similar experiments, the peak ΔF/F_rest at the membrane after 10-ms depolarization was significantly larger for proximal than distal signals (ΔF/F_rest: 3.1 ± 0.4 and 1.9 ± 0.2 respectively P=0.001, N= 21, 95% confidence limit). As expected, the ratio of proximal to distal signals measured with Cal520LA (1.6) was significantly higher than that measured with Cal520HA (1.3).

Ca²⁺ concentrations at single ribbon locations measured using ribbon-bound indicators

Although Ca²⁺-sensitive fluorescent chemical dyes have been used previously ^{7,20,29,36,42,43,45,46,49-59}, the visualization of signals within smaller domains using freely diffusible Ca²⁺ reporters is limited by the resolution of light microscopy and the spread of the indicator by diffusion ⁶⁰. Diffusible Ca²⁺ indicators report space-averaged Ca²⁺ concentrations, and their intracellular diffusion inherently broadens the spatial resolution of Ca²⁺ nanodomains. To partially overcome this problem, we targeted the Ca²⁺ indicators to the ribbon by fusing them to the RBP, as described previously ²⁰. RBP-conjugated Ca²⁺ indicators Cal520HA-RBP or Cal520LA-RBP were introduced to the RBC terminal together with fluorescently labeled RBP via whole-cell voltage clamp by placing the patch pipette directly at the terminal while imaging the terminal using laser scanning confocal microscopy. For two-color imaging, ribbons were labeled with TAMRA-RBP that did not interfere with Cal520-RBP fluorescence, and both channels were scanned sequentially to prevent possible bleed-through. Under these conditions, we used the spots detected by TAMRA-RBP to define the locus of the synaptic ribbon (Fig 3A&B, magenta). We also found punctate regions with both Cal520HA-RBP (Fig 3A, cyan) and Cal520LA-RBP (Fig 3B, cyan) at the same location, on a dimmer fluorescent background of the synaptic terminal, which correspond to the Ca²⁺ indicator-peptide complexes that are bound and not bound to the ribbon, respectively ²⁰. The overall changes in ΔF/F_rest were averaged over several trials for proximal and distal Ca²⁺ signals measured with Cal520HA-RBP (Fig. 3C, light cyan vs. dark cyan) and Cal520LA-RBP (Fig. 3D, light cyan vs. dark cyan). When compared to their distal location counterparts, the Ca²⁺ signals proximal to the membrane showed a sharper rise and decay at both the onset and termination of stimuli (Figs. 3C and 3D, light vs. dark cyan trace), as expected when comparing nano- and microdomain Ca²⁺ profiles. Our data shows that the amplitude differences between ribbon-proximal and ribbon-distal Ca²⁺ signals were well resolvable using the ribbon-bound Cal520HA-RBP indicator (Fig. 3C, light vs. dark cyan traces: ΔF/F_rest = 3 ± 0.4 vs. 1.9 ± 0.3, respectively, p=0.001) and Cal520LA-RBP indicator (Fig. 3D, light vs. dark cyan traces: ΔF/F_rest = 5.5 ± 0.9 vs. 3.3 ± 0.8, respectively, p=0.003). The amplitudes of ribbon-proximal Ca²⁺ signals were higher when measured with Cal520LA-RBP than with Cal520LA-free (Fig. 3E, Cal520LA-RBP (light cyan) vs. Cal520LA-free (gray): ΔF/F_rest = 5.5 ± 0.9, n=30 vs. 3 ± 0.4, n=19, p=0.04) but this was not the case for distal Ca²⁺ signals (Fig. 3F, Cal520LA-RBP (light cyan) vs. Cal520LA-free (gray): ΔF/F_rest = 3.3 ± 0.8, n=30 vs. 1.9 ± 0.2, n=21, p=0.15). Notably, the Ca²⁺ signals at distal sites measured with Cal520LA-RBP reached their peak amplitude earlier (Fig. 3F inset, dark cyan) than those measured with Cal520LA-free (Fig. 3F inset, light gray). Together, these findings strongly suggest that conjugating the Cal520LA indicator to RBP provides a more accurate, promising approach for measuring the distinct local Ca²⁺ signals at ribbon-proximal vs. ribbon-distal locations.

Fig. 3.

This method also enables ratiometric measurements with RBP-conjugated Ca²⁺ indicator by normalizing its fluorescence to that of RBP-peptides conjugated to Ca²⁺-insensitive fluorophores (TAMRA-RBP) to provide an estimate of Ca²⁺ concentration in RBC synaptic ribbons. We obtained the effective K_1/2 (K_eff) by measuring the Cal520HA-RBP/ TAMRA-RBP fluorescence ratio in buffered Ca²⁺ solutions and using the Grynkiewicz equation⁵⁵. For Cal520HA-RBP, we found the K_eff to be ∼795 nM, which is higher than the value of 320 nM reported by the manufacturer for Cal520HA. The differences between the in-cell measurements and the manufacturer’s values are likely to arise from differences in the cellular buffering capabilities, changes in dye properties due to interactions with the molecules inside the cell ^36,61-63, and possible differences in the binding properties of the peptide-conjugated Ca²⁺ indicators when bound to synaptic ribbons. Since the in-cell approach most closely matched the experimental conditions, we used this value (K_eff) for all further calculations.

We first measured the ribbon-proximal and ribbon-distal Ca²⁺ concentrations using Cal520HA-RBP with 0.2 mM EGTA in the patch pipette as it allowed to resolve the gradient between the two signals (Fig. 3C). Under these conditions, we found a maximum ribbon-proximal Ca²⁺ concentration produced in response to a brief 10-ms pulse of 3.7 µM, and a maximum ribbon-distal Ca²⁺ concentration of 0.7 µM. These apparent Ca²⁺ concentration amplitudes are well below levels required to trigger exocytosis of the ultrafast releasable pool (UFRP) and readily releasable pool (RRP) ⁴³ and, as discussed earlier, likely represent the lower bounds of the Ca²⁺ concentration at a single ribbon location due to local saturation of the high-affinity indicator and/or due to nanodomains that are smaller than the resolution attainable using light microscopy. It should also be noted that unlike freely diffusing Ca²⁺ indicators, RBP-conjugated indicators that slowly unbind from the ribbon are not readily replaced by free Ca²⁺ indicators, rendering them even more prone to saturation. Nevertheless, our data demonstrate that ribbon-bound indicators may better report local Ca²⁺ concentrations due to their localization, albeit still subject to the limitations of light microscopy.

To test the contribution of local saturation and to better estimate ribbon Ca²⁺ concentrations, we repeated our measurements using the low-affinity Ca²⁺ indicator Cal520LA conjugated to the RBP peptide. Because it was difficult to perform in-cell measurements to determine the K_eff for Cal520LA-RBP given the large amounts of Ca²⁺ required to calibrate the Cal 520LA indicator (see Methods and Materials), we used the K_1/2 value provided by the manufacturer for free Cal520LA (90 µM). However, we expect that in-cell measurements of K_eff, are likely to be different due to the cellular buffering properties reported previously for K_eff, measurements of Oregon Green BAPTA-5N in inner hair cells ³⁶.

Bipolar cells release neurotransmitters primarily from ribbon active zones, although some release also occurs at non-ribbon sites (referred to as NR, Fig. 4A) ^64-66. To reveal the Ca²⁺ signaling at and away from the ribbon, we performed whole-cell patch clamping and x-t line scans at ribbon (Fig. 4Ba, R) and non-ribbon (Fig. 4Bb, NR) sites perpendicular to the plasma membrane using TAMRA-RBP and Cal520LA-RBP. A brief (10-ms) depolarizing voltage-clamp pulse evoked rapid high Ca²⁺ signals (Fig. 4Ba, cyan raster plot) at ribbon locations (Fig. 4Ba, magenta raster plot) but not at non-ribbon locations (Fig. 4Bb, cyan raster plot). The amplitude of Ca²⁺ signals elicited by brief depolarization and detected with Cal520LA-RBP at the ribbon-proximal site (Fig. 4C, light cyan traces) were significantly higher than those at ribbon-distal (Fig. 4C, dark cyan traces) and non-ribbon (Fig. 4C, blue traces) sites. The Ca²⁺ concentration gradient along the ribbon is summarized in Fig. 4D. We found that the average Ca²⁺ concentration at the proximal side of the ribbon (26.4 ± 3.1 µM, n=26) was significantly different from that at the ribbon-distal (15.6 ± 1.5 µM, n=26) and non-ribbon (10.4 ± 0.4 µM, n=15) sites, and that Ca²⁺ concentrations at ribbon-distal sites is higher than those at non-ribbon sites (Fig. 4D). These measurements display large heterogeneity across distinct ribbons and distinct cells, with the coefficient of variation of about 60% for [Ca²⁺] measurements at locations proximal to the ribbon, compared to a 10% CV for multiple depolarizations for the same ribbon. The Ca²⁺ signals at distal sites also reached their peak amplitude earlier (Fig. 4C inset, dark cyan arrowhead) than those at the non-ribbon sites (Fig. 4C inset, blue arrowhead). These findings are consistent with non-ribbon vesicle release being governed by cytoplasmic residual Ca²⁺ or Ca²⁺ influx via clustered Cav channels at ribbon sites rather than non-ribbon active zones with clusters of Cav channels and are also in agreement with previous reports regarding the temporal delay and sensitivity of non-ribbon exocytosis to EGTA ^65,67.

Fig. 4.

Sensitivity of microdomain Ca²⁺ to exogenous buffers

Previous work has shown that the rate of vesicle replacement in RBCs is accelerated by elevated Ca²⁺ levels at sites along the ribbon, and that while millimolar levels of EGTA have little effect on the fast exocytosis component near Cav channels, they selectively block sustained exocytosis, likely by preventing Ca²⁺ from reaching the distal locations on the ribbon ^11,68-71. These findings raise the possibility that Ca²⁺ has two sites of action, one near the Cav channels that trigger vesicle release and one further away that replenishes the supply of releasable vesicles. Burrone et al. (2002) proposed that endogenous Ca²⁺ buffers regulate the size of the RRP by limiting the spatial spread of Ca²⁺ ions and could suppress vesicle release at the periphery of the active zone in bipolar cells ⁷. However, the Ca²⁺ gradient at different locations along the ribbon controlling RRP release has never been examined. Thus, we used our nanophysiology approach (Fig.1) to measure the Ca²⁺ gradient along the synaptic ribbon under buffering conditions that differentially modulate vesicle release and resupply. To determine the spatiotemporal properties of Ca²⁺ signals under these conditions, we performed rapid x-t line scans at a single ribbon location in the presence of ribbon-bound Cal520LA-RBP and under varying concentrations of exogenous buffer in the patch pipette solution, including (EGTA at 0.2 (Fig. 5A), 2 (Fig. 5B), and 10 mM (Fig. 5C) or BAPTA at 2 mM (Fig. 5D)). We found that the ratio between proximal and distal Ca²⁺ signal amplitudes was similar in 0.2 and 2 mM EGTA (proximal-to-distal ratio ∼1.7, Supplementary Table 1) but was enhanced with 10 mM EGTA (proximal-to-distal ratio ∼1.9, Supplementary Table 1) and further enhanced with 2 mM BAPTA (proximal-to-distal ratio ∼4.5, Supplementary Table 1). These findings once again emphasize the reliable measurement of resolving ribbon-proximal vs. ribbon distal Ca²⁺ signals using our nano-physiology approach. Our experimental findings of the increase in the proximal-to-distal Ca²⁺ concentration ratio with increasing EGTA concentration are consistent with our simulation results (Fig. 7), although the corresponding ratios are greater in the simulation than in the experiment due to the large size of the microscope’s point-spread function.

Fig. 5.

Though BAPTA significantly abolished the spread of Ca²⁺ signals, it is impressive that a substantial amount of ribbon-proximal Ca²⁺ concentration measured with ribbon-bound indicators was still present with 2 mM BAPTA. One reason for this observation could be that ribbon-bound Ca²⁺ indicators are likely to measure Ca²⁺ signals very close to Cav channels without diffusing away, which makes it impossible for BAPTA to intercept Ca²⁺ ions. If so, similar experiments conducted with free Ca²⁺ indicators should report a significantly lower proximal Ca²⁺ signal since the measurement by a diffusible dye would effectively spread the signal over a larger volume. Indeed, this is what we found. As shown in Fig. 5E, the apparent proximal Ca²⁺ signal in response to a 10-ms brief pulse measured with Cal520LA-free in the presence of 2 mM BAPTA was 0.9 ± 0.2 (n=9), 4-fold lower than proximal Ca²⁺ signals measured with Cal520LA-RBP (3.6 ± 1, n=20). The latter value is closer to the true measure of the Ca²⁺ concentration in the vicinity of the ribbon base. These findings emphasize that the reliable measurement of ribbon-proximal Ca²⁺ signals in the vicinity of Cav channels greatly benefits from the increased spatial resolution of our nano-physiology approach. We also measured the spatiotemporal properties of local Ca²⁺ signals using Cal520HA-RBP as we have done above for Cal520LA-RBP (Supplementary Fig. 3). We found that the ratio between proximal and distal Ca²⁺ signals was similar in 0.2 and 2 mM EGTA (proximal-to-distal ratio ∼1.5, Supplementary Table 2) but was enhanced with 10 mM EGTA (proximal-to-distal ratio ∼2.6, Supplementary Table 2) and further enhanced with 2 mM BAPTA (proximal-to-distal ratio ∼2.8, Supplementary Table 2). Cal520 acts as a buffer that binds Ca²⁺ and carries it away by diffusion. Thus, the higher affinity indicator Cal520HA will bind and shuttle away more Ca²⁺ than Cal520LA. If the Cal520HA indicator is saturated near the Cav channels, it may be underreporting the fast Ca²⁺ transients that occur in those locations. However, Cal520 LA and Cal520 HA show similar increase in the ribbon proximal-to-distal ratios with increasing concentrations of exogenous buffer.

Computational models of Ca²⁺ signals along the axis of the RBC synaptic ribbon

Since we could identify the position of the ribbon and we used RBP-fused Ca²⁺ indicator-RBP, we were able to measure and distinguish ribbon-proximal vs. ribbon-distal Ca²⁺ signals that drive the release of UFRP and RRP. However, several factors, including finite spatiotemporal resolution of optical imaging, spatial diffusion, dye saturation, and dye binding kinetics, may limit our ability to achieve optimal resolution. These limitations, however, can be overcome by the use of quantitative models. Thus, we developed a model based on our data with published information on Ca²⁺ diffusion and buffering to estimate more accurately the [Ca²⁺]_I gradient along the ribbon at distinct distances from the plasma membrane.

Combining whole-terminal Ca²⁺ current measurements and our estimate for the number of synaptic ribbons per terminal allowed us to infer the magnitude of Ca²⁺ current per single ribbon, which we used in our model to determine the spatiotemporal [Ca²⁺] dynamics near the ribbon. Fig. 7 shows the results of the simulation of [Ca²⁺] at various distances from the ribbon during and after a depolarizing pulse of 10ms duration in the presence of different concentrations of exogenous and endogenous buffers, replicating the conditions used in our experiments. The geometry of the simulation domain box is shown in Fig. 6; it represents the fraction of total synaptic terminal volume per single ribbon (see Methods, Supplementary Table 3 and Supplementary Fig.4A.). Fig. 7 shows the time-dependence of [Ca²⁺] during and after the depolarization pulse at 5 specific locations both near and distant from the ribbon (left-hand panels in each subplot). These five spatial locations are marked by circles of different colors in the right-hand panel of each subplot, which shows [Ca²⁺] at the end of the Ca²⁺ current pulse in pseudo-color in the entire planar cross-section of the simulation domain cutting through the middle of the ribbon, as shown in Fig. 6. We assumed a highly simplified Cav channel arrangement into four clusters forming a square with a side length of 80nm. The closest of the five spatial locations (red curves and circles in Fig. 7) is X=20 nm away from the base of the ribbon center-line, Z=10 nm above the ribbon, about 45nm away from the closest Cav channel cluster (see Fig. 6). We assumed that this location was within the Ca²⁺ microdomain that triggered the release of the UFRP. Vesicles were not included in the simulation since the total exclusion volume attributed to them represented only a small fraction of the total inter-terminal volume and, therefore, did not significantly impact [Ca²⁺] at the qualitative resolution level we are interested in. A couple of simulations with vesicles included were performed to confirm this statement, requiring much finer spatial resolution and longer computational time. We note also that [Ca²⁺] for short pulse durations considered here is relatively insensitive to our assumptions on the membrane Ca²⁺ extrusion mechanisms, listed in Supplementary Table 3.

Fig. 6.

Fig. 7

Fig. 7A1-B2 reveals that [Ca²⁺] could rise above 50 µM within the microdomain near the base of the ribbon in the presence of 0.2 mM up to 10 mM EGTA. Given the size of the microscope point-spread function, this is in qualitative agreement with the 26 μM estimate of ribbon-proximal [Ca²⁺] that we recorded using the ribeye-bound low-affinity indicator. In addition, this concentration level is reached very soon after the channel opening event due to the rapid formation of the Ca²⁺ microdomain ^72,73, and therefore, this estimate is expected to hold for shorter pulses as well. However, [Ca²⁺] decayed rapidly with distance, with the rate of decay significantly increasing when EGTA concentration was increased to 10mM (Fig. 7B1, B2). Note that a 7-fold change in the capacity of the immobile endogenous Ca²⁺ buffer (cf. Fig. 7A1 vs. 7A2) had a relatively modest effect on the [Ca²⁺] level up to about 150 nm distance from the ribbon base. This effect of increasing endogenous buffer concentration was further reduced in 10 mM EGTA (cf. Fig. 7B1 vs. 7B2), due to strong competition of endogenous buffer with large concentrations of EGTA. This agrees with the expectation that immobile buffers primarily slow down Ca²⁺ signals but do not affect the rapidly forming quasi-steady-state Ca²⁺ microdomains in the immediate vicinity of the channel ^72-74. Finally, Fig. 7C1-C2 shows that 2 mM BAPTA had a much greater effect on localizing the Ca²⁺ signal to the immediate vicinity of the ribbon base.

Since a native unpatched cell may well contain mobile rather than immobile buffers, it was interesting to examine the effect of increasing the concentration of mobile endogenous buffer on Ca²⁺ dynamics in the ribbon vicinity, as compared to the corresponding effect of immobile buffer. Supplemental Figure 4B-C shows that a 7-fold increase in mobile buffer concentration reduces the microdomain Ca²⁺ by a factor of 2 and greatly reduces the size of the microdomain. This contrasts with the effect of increasing the concentration of the immobile buffer, which has a much more subtle effect (cf. Fig. 7A1, A2). We note that 0.2 mM EGTA was absent in the simulation with mobile buffer shown in Supplemental Figure 4B-C; in general, the effect of modest concentrations of EGTA is expected to be negligible due to its slow Ca²⁺ binding speed compared to the endogenous buffer.

Variability of local Ca²⁺ signals across the RBC synaptic ribbons

Bipolar cells release neurotransmitters primarily from ribbon active zones ^{1,6,8,11-13,21-25}. However, the factors that shape the synaptic ribbon microdomains at retinal ribbon synapses have not been examined. We wondered whether all of the 30∼50 synaptic ribbons at an RBC terminal release glutamate in a similar fashion or whether there is some heterogeneity in glutamate release from different ribbon active zones. In particular, we asked what underlying mechanisms could differentiate the Ca²⁺ signals between ribbons of the same RBC terminal and across different RBCs. We first found evidence for such variability in local Ca²⁺ signals at single-ribbon locations of different cells using the freely diffusible high-affinity indicator Cal520HA, observing in particular high variability of the Ca²⁺ transient amplitudes, even for RBCs with similar depolarization-evoked Ca²⁺ current amplitudes (cf. Supplementary Figs. 5A vs. 5B. However, we did not observe such variability when we obtained multiple line scans across the same ribbon (Supplementary Figs. 5A vs. 5B, gray traces). We wondered whether the observed variability between cells could be due to distinct subtypes of zebrafish RBCs ⁴⁷, which might have different subtypes or numbers of Cav channels near the ribbon. Recent work in zebrafish retina identified two distinct RBC subtypes: RBC1 and RBC2 ⁴⁷. The wiring pattern of amacrine cells postsynaptic to RBC1 closely resembles the circuitry of mammalian RBCs, whereas RBC2 forms distinct pathways. Furthermore, RBC1 specifically labels for the known marker of mammalian RBCs, PKC-α ⁴⁷. Due to these similarities, RBC1 in zebrafish is classified as analogous to mammalian RBCs. Moreover, RBC1s have expected morphological characteristics, in particular, the shape and size of soma and the presence of a single synaptic terminal ⁴⁸. Immunolabeling of isolated RBC preparation used in these experiments showed primarily intact RBCs, which are PKC-α-specific RBC1 (Supplementary Fig. 2B). For simplicity, and because this study focuses exclusively on RBC1, we will refer to RBC1 as ‘RBC’ throughout this work. Thus, we attribute the variability in the RBC Ca²⁺ transients to the variability between the Ca²⁺ microdomains within the same cell type.

To identify mechanisms that contribute to the heterogeneity in the local Ca²⁺ signals we have reported in this study, we began by asking whether differences in local Ca²⁺ buffering could account for this heterogeneity. To examine the role of Ca²⁺ buffering in the variability of local Ca²⁺ signals across different cells and different ribbons (Figs. 8A and 8D, respectively), we compared proximal (Figs. 8B and 8E) and distal (Figs. 8C and 8F) Ca²⁺ signals in 10 mM EGTA by averaging the Ca²⁺ recordings from all the ribbons of a given cell (Supplementary Figs. 6 and 7). When the Ca²⁺ signals were restricted to the ribbon sites with 10 mM EGTA in the pipette solution (Supplementary Tables 1 and 2), the amplitude of the ribbon-proximal and ribbon-distal Ca²⁺ signals averaged over all ribbons of a given cell exhibited significant variation across different cells (Figs. 8B, 8C, and Supplementary Fig. 6). We next compared variability in Ca²⁺ signals in the presence of 10 mM EGTA at individual ribbons within the same cell (Fig. 8D and Supplementary Fig. 7) at proximal (Fig. 8E) and distal (Fig. 8F) locations. For the cell described in Figs. 8D-8F, the proximal Ca²⁺ signals were significantly different across all ribbons examined, and there were considerable differences in distal Ca²⁺ signals between ribbons, with the exception of ribbons numbered 2 and 5. These findings suggest that the heterogeneity in proximal and distal Ca²⁺ signals at distinct ribbons within the same cell may result from different underlying mechanisms, for example, heterogeneity in Cav expression or subtype and Ca²⁺ handling mechanisms. These findings suggest that exogenous Ca²⁺ buffering has a negligible effect on experimentally observed heterogeneity and variability of the proximal Ca²⁺ signals and that local Ca²⁺ signals at RBC ribbons are dominated by Ca²⁺ in regions close to the ribbon base where Cav channels are located.

Fig. 8.

The ultrastructure of the zebrafish RBC terminal reveals diversity in the size of the synaptic ribbons across the terminal

In hair cells, Ca²⁺ microdomain signaling varies with ribbon size, reflecting larger patches of Cav channels aligned with larger ribbons ⁵³. However, the number, size, and shape of ribbon active zones per terminal have not been established in our experimental system, the synaptic terminals of zebrafish RBCs. To provide quantitative measurement of ribbon microdomains, we examined the ultrastructure of zebrafish RBC synaptic ribbons by serial block face scanning electron microscopy (SBF-SEM) and reconstructed three RBCs from serial section electron micrograph (Fig. 9A). We analyzed the bipolar cell terminals that are closest to the ganglion cell layer with a morphology similar to mammalian RBCs and zebrafish RBC1s ⁴⁷. Our reconstruction of three RBCs revealed 30-41 ribbons within RBC terminals (Fig. 9A), similar to what was previously reported in goldfish giant ON-type mixed RBCs (range 45-60 ⁷⁵).

Fig. 9.

We examined the distribution of these 30–41 synaptic ribbons within the zebrafish RBC terminals. The distribution of RBC ribbons as estimated by the distance between ribbon sites (Fig. 9B and Video 1; see Materials and Methods) revealed a wide distribution in synaptic ribbon distance to its nearest neighbor across the three reconstructed RBCs (means of 2.48 ± 0.12 µm, 1.84 ± 0.06 µm and 1.98 ± 0.08 µm; Fig. 9C). The distance between RBC synaptic ribbons ranged from 141 nm to 6 µm, with a mean of 2 ± 0.05 µm (Fig. 9D; across all three reconstructed RBCs). Of the 510 zebrafish RBC synaptic ribbons, only 4 (0.8%) were separated by distances smaller than our confocal microscope resolution limit of 270 nm.

To reveal whether heterogeneity in synaptic Ca²⁺ signals correlates with active zone size, we compared the shape and size of the ribbon active zones across the three reconstructed RBCs. Our results show significant variations in the shape and size of zebrafish RBC synaptic ribbons in RBCs and associated active zones (Fig. 10). On average, individual ribbons spanned 2–5 consecutive sections, with some located within the axons or closer to the terminal. The SBF-SEM images and 3D projections of ribbon structures (colored in magenta) revealed considerable variability in shape and size across all dimensions, as illustrated in Figs. 10A–C. The number of Cav channels per active zone in hair cells from chicken, frog, and turtle varies with the size of the synaptic ribbon ⁷⁶. We thus measured the area of the ribbon facing the plasma membrane where the Cav channels are known to be located to estimate the number of Cav channels per active zone and to determine whether variation in active zone size might plausibly contribute to the heterogeneity in Ca²⁺ signaling. The EM images and 3D projection of the plasma membrane (yellow) and the area of the ribbon facing the plasma membrane (cyan), representing the active zone, are shown in Figs. 10A-C. We observed large variability in the area of the ribbon facing the plasma membrane or active zone within the RBCs, with substantial variability in their average distribution across the three RBCs (Figs. 10C, cyan, 10E, Video 2 and Supplementary Fig. 9), suggesting that the number of Cav channels per RBC active zone could plausibly be heterogeneous.

Fig. 10.

The size of the active zone and maximal Ca²⁺ influx correlate with the size of the synaptic ribbon

To reveal whether larger ribbons display stronger maximal Ca²⁺ influx, we measured Ca²⁺ signals in response to a series of 200-ms depolarizations. We first imaged TAMRA-RBP and Cal520LA-RBP in sequential scans to obtain ribbon location and resting Ca²⁺ levels. To maximize the capture of Ca²⁺ signals during brief stimuli, we image the Cal520LA-RBP channel, followed by TAMRA-RBP and Cal520LA-RBP sequential scans to confirm the ribbon locations. We found that the maximum amplitude of depolarization-evoked Ca²⁺ signals increased with an increase in ribeye fluorescence (Fig. 11A & B and Video 3; r = 0.51, N = 121 ribbons, p< 0.001), consistent with findings reported in cochlear inner hair cells ^53,58. Since ribeye fluorescence correlates with the number of ribeye molecules per ribbon, and, therefore, ribbon size ^26,53,58,77, our findings suggest that larger ribbons display stronger Ca²⁺ signals as active zone size scales with ribbon size ⁵⁸. Given the diverse shapes of RBC synaptic ribbons (Fig. 10A & B, magenta), we measured the longest length and width of ribbons from EM images and compared this with the active zone size. We found that RBC ribbon length and width have a positive correlation (Fig. 11C, r = 0.47, N = 102 ribbons, p< 0.001), and both dimensions have a positive correlation with active zone size, albeit the ribbon width (Fig. 11D, r = 0.52, N = 102 ribbons, p< 0.001) has a stronger correlation with active zone size than the ribbon length (Fig. 11D, r = 0.32, N = 102 ribbons, p< 0.001). These results suggest that heterogeneity in synaptic Ca²⁺ signals correlates with ribbon dimensions and that active zone and the number of Cav channels ^53,58 scale with ribbon size.

Fig. 11.

Discussion

A deep look at the active zone Ca²⁺ microdomain in RBC terminals

In this study, we developed a quantitative nanophysiology approach using ribeye-bound Ca²⁺ indicators and fluorescently labeled RBP, combined with ratiometric dual-color imaging, to measure changes in Ca²⁺ concentration at and along the ribbon axis. By targeting Ca²⁺ indicators to the ribbon combined with fluorescently labeled RBP, we determined the Ca²⁺ concentration along the ribbon at different distances from active zone, and found that near the ribbon base the Ca²⁺ concentration could increase to an average of 26 μM upon depolarization. Given the localization of our indicator on the ribbon and minimal sensitivity to exogenous buffers, we believe that our ribbon-proximal signal represents Ca²⁺ concentrations on the ribbon in locations very near release sites, likely some of those that contribute to vesicle resupply. Notably, the [Ca²⁺] _i levels obtained from our nanophysiology approach using confocal microscopy are similar to those that were reported for measurements in hair cells using Stimulated Emission Depletion (STED) lifetime measurements ³⁶. Previous work on RBCs found slowing of recovery from paired-pulse depression by EGTA, suggesting that Ca²⁺ levels at distal sites on the ribbon might be important for the recruitment of new vesicles following depletion of the RRP and UFRP ^11,24. Here, we measured directly the effect of EGTA on the spread of Ca²⁺ signals and found that the Ca²⁺ signal at the distal half of the ribbon was two-fold lower than that at the location proximal to the membrane under conditions of low Ca²⁺ buffering. It should be noted, however, that diffraction of light during the optical imaging procedure is likely to underestimate spatiotemporal [Ca²⁺] along the ribbon. Therefore, we used our measurement of the Ca²⁺ current per ribbon to develop a detailed computational model that allowed us to resolve the expected profile of Ca²⁺ microdomains governing UFRP fusion. The model results shown in Fig. 7 reveal, on a finer spatial scale, the distribution of Ca²⁺ around a ribbon in the presence of various quantities of immobile endogenous buffer and added exogenous buffers. Our finding of high proximal [Ca²⁺] levels on the order of 26 µM recorded using a ribeye-bound low-affinity indicator dye is in qualitative agreement with the simulated high concentration of Ca²⁺ in the microdomain at the base of the ribbon. Simulations also show that, as expected, the peak Ca²⁺ microdomain amplitude is relatively insensitive to the concentration of endogenous immobile buffer and added EGTA, but its decay with distance is accelerated by increasing the concentrations of exogenous and endogenous buffers. In contrast, the addition of 2 mM BAPTA causes a profound reduction of the Ca²⁺ microdomain amplitude and localizes the signal to a small volume around the ribbon. We also compared the impact of mobile vs immobile endogenous buffers on the Ca²⁺ distribution, showing the much greater effect of mobile buffers on the microdomain amplitude and distance dependence. Although these results are generally expected^72-74, the simulation results estimate the [Ca²⁺] levels in the vicinity of the ribbon with greater spatial and temporal resolution. As expected, the ratios of Ca²⁺ concentration values at the base of the ribbon vs. further away from the ribbon are much greater in the model simulation than our experimentally determined proximal-to-distal Ca²⁺ ratio values due to the large size of the microscope PSF, which precludes precise localization of the corresponding Ca²⁺ signals. Further, simulations show the effect of exogenous buffers with greater spatial resolution. However, the simulations fully agree with the overall increase of the proximal-to-distal Ca²⁺ concentration ratios by the application of exogenous Ca²⁺ buffers, especially BAPTA. Despite the technical challenges posed by the limited spatiotemporal resolution of optical imaging, the spatial resolution attained with the nanophysiology approach developed in our study is sufficient to differentiate Ca²⁺ signals associated with distal sites on the ribbon from those localized near the plasma membrane. Conceivably, the nanophysiology approaches established here could generally be applied to other classes of ribbon-containing cells, such as rods, cones and hair cells.

Heterogeneity of Ca²⁺ microdomains across RBCs terminals

Our quantitative nanophysiology provides evidence for the heterogeneity in synaptic Ca²⁺ signals across zebrafish RBCs, suggesting variability in the Ca²⁺ microdomains as has been previously reported for hair cells ^53,58 and cone photoreceptors. Retinal bipolar neurons contain many ribbon active zones. This ranges from 45-60 in adult goldfish, as quantified by EM ⁷⁵, or 25-45 in RBCs of adult living zebrafish, as estimated from laser scanning confocal microscopy⁷⁸. Our analyses of ribbon number across zebrafish RBCs through SBF-SEM arrived at an estimate of 31-40. The differences in the number of ribbons as determined through confocal microscopy compared to SBF-SEM could be explained by the resolution limit encountered by the confocal image analyses. Our analyses of zebrafish RBC terminal SBF-SEM images and 3D projections provide the first evidence of diversity in synaptic ribbon shape and sizes across zebrafish RBC terminals. Further, our quantitative measurements of SBF-SEM images reveal that synaptic ribbon width and length correlate with the active zone size, consistent with previous findings in hair cell ribbon synapses ^76,77 and cone photoreceptor ribbon synapses⁵⁴. Our observation that maximal active zone Ca²⁺ influx correlates with ribeye fluorescence in live imaging is consistent with findings previously reported in hair cells ⁵⁸. Thus, we expect larger active zones, with more Ca²⁺ influx, to have more Cav channel expression, as demonstrated for hair cell ribbon synapses ⁵⁸. Heterogeneity among ribbon active zones has been previously reported in hair cells where synaptic Ca²⁺ microdomains varied substantially in amplitude and voltage-dependence within individual inner hair cells ^53,58. Studies in aquatic tiger salamander cone photoreceptors revealed that the amplitude and midpoint activation voltage of Ca²⁺ signals varied across individual ribbons within the same cone. Additionally, local Ca²⁺ signal dynamics at cone photoreceptor ribbons were found to be independently regulated ⁵⁴.

The faster, smaller, and more spatially confined Ca²⁺ signals that are insensitive to the application of high concentrations of exogenous Ca²⁺ buffers, referred to here as ribbon proximal Ca²⁺ signals, could be due to Ca²⁺ influx through Cav channel clusters beneath the synaptic ribbon ^19,26,29-42. However, the variability in Ca²⁺ signals at the distal ribbon, away from the plasma membrane, may result from the spread between two closely spaced microdomains; thus, if proximal Ca²⁺ signals are variable, the distal Ca²⁺ signals are also variable between ribbons. Alternatively, the variability in distal Ca²⁺ signals could be due to additional mechanisms, for example, Ca²⁺ influx from internal stores. Future studies should aim to elucidate the specific mechanisms underlying the observed heterogeneity in distal Ca²⁺ signals. For instance, exploring differences in Ca²⁺ release from intracellular stores or variations in Ca²⁺ sequestration, as reported in goldfish, could reveal key contributing factors ⁷⁹. Although the potential impact of these intracellular Ca²⁺ stores on Ca²⁺ microdomain heterogeneity in hair cells has not been reported,⁵³ previous studies in goldfish bipolar cell terminals have documented spatially restricted Ca²⁺ oscillations in voltage-clamped retinal bipolar cells, occurring independently of membrane potential ²⁰. To our knowledge, the sources of Ca²⁺ for these oscillations remain unknown. However, previous studies in goldfish bipolar cells suggest that the endoplasmic reticulum⁷⁹ and mitochondria¹⁸ may act as internal Ca²⁺ stores. Notably, IP₃ receptors have been localized to retinal bipolar cell terminals in some species.^80,81

In the current study, we investigated possible mechanisms for heterogeneity in proximal Ca²⁺ signals by measuring the ribbon size, particularly the area of the ribbon adjacent to the plasma membrane, where Cav channel clusters are located. Our SBF-SEM shows substantial variability in synaptic ribbon size, shape, and the area of the ribbon facing the plasma membrane where Cav channel clusters are tethered. Since both Ca²⁺ signals and the area of the synaptic ribbon facing the plasma membrane are heterogeneous, we propose that the larger ribbons may anchor more channels, leading to larger Ca²⁺ microdomains. Since local Ca²⁺ signals control kinetically distinct neurotransmitter release components, heterogeneity in local Ca²⁺ signals may alter the rate of vesicle release and allow them to function independently. Indeed, heterogeneity in neurotransmitter release kinetics has been proposed for the observed diversity in excitatory postsynaptic current amplitudes and kinetics reported from paired recordings of goldfish bipolar cells⁸² and ON-type mixed RBCs ⁸³. The heterogeneity in the RBC Ca²⁺ microdomains, synaptic ribbon shape, size, and active zone area reported in this study may contribute to regulating the dynamic range of RBC ribbon synapses^22,84,85 - a hypothesis that needs to be tested in future studies.

Methods

Rearing of zebrafish

Male and female zebrafish (Danio rerio; 16 ∼20 months) were raised under a 14 h light/10 h dark cycle and housed according to NIH guidelines and the University of Tennessee Health Science Center (UTHSC) Guidelines for Animals in Research. All procedures were approved by the UTHSC Institutional Animal Care and Use Committee (IACUC; protocol # 23-0459).

Isolation of zebrafish retinal RBCs

Dissociation of RBCs was performed using established procedures ⁸⁶. Briefly, retinas were dissected from zebrafish eyes and incubated in hyaluronidase (1100 units/ml) for 20 minutes. The tissue was washed with a saline solution containing 120 mM NaCl, 2.5 mM KCl, 0.5 mM CaCl₂, 1 mM MgCl₂, 10 mM glucose, and 10 mM HEPES, pH = 7.4] before being cut into quadrants. Each quadrant was incubated at room temperature for 25 to 40 min in the same saline solution, to which was added DL-cysteine and papain (20-30 units/ml; Sigma Millipore, St. Louis, MO) and triturated using a fire-polished glass Pasteur pipette. Individual cells were transferred to glass-bottomed dishes, allowed to attach for 30 minutes, and washed with saline solution before being used for experiments.

Ribeye binding peptides

As a means of localizing the ribbons, custom peptides containing the ribbon binding sequence fused to TAMRA (tetramethylrhodamine; TAMRA: GIDEEKPVDLTAGRRAG) dye were synthesized, purified, and purchased from LifeTein (>95% purity, LifeTein LLC, NJ).

Ca²⁺ indicators

Free Ca²⁺ indicators. The potassium salts of high and low-affinity Ca²⁺ indicators Cal-520^® (high affinity, K_D 320 nM) and Cal-520N™ AM (low affinity, K_D 90 µM), referred to as Cal520HA and Cal520LA, respectively were purchased from AAT Bioquest.

Direct conjugation of Ca²⁺ indicator to cysteine-containing ribbon-binding peptides. To target Ca²⁺ indicators to the ribbon, custom-made cysteine-containing ribeye binding peptides NH₂-CIEDEEKPVDLTAGRRAC-COOH were synthesized and purchased from LifeTein to directly fuse with the fluorogenic 520^® maleimide (purchased from AAT Bioquest) for high affinity (HA) and low affinity (LA Ca²⁺ indicator dyes). Each peptide at one mM concentration was mixed and incubated with two mM Cal-520^® maleimide (20 mM stock solution in DMSO purchased from AAT Bioquest) for one hour at room temperature, then overnight at 4°C. Calibration of Ca²⁺ indicator dyes was performed as described ¹⁹ and as detailed below. The conjugated Ca²⁺ indicators were stored at -20°C in smaller aliquots, each sufficient for a single day’s experiments.

Measurement of dissociation constants (K_D) for Ca²⁺ indicator peptides

The effective K_D (K_eff) was obtained by measuring the fluorescence of Cal520HA-RBP and TAMRA-RBP in buffered Ca²⁺ solutions, determining the ratio between them ¹⁹, and using the Grynkiewicz equation to determine the Ca²⁺ concentration [Ca²⁺] from this ratio ⁵⁵. However, this was not possible for the low-affinity indicator Cal520LA due to the large Ca²⁺ levels required to calibrate it. Thus, K_eff for Cal520LA-RBP could be larger than the K_1/2 provided by the manufacturer for Cal520LA (K_D, 90 μM), as reported previously with K_eff, measurements of OB-5N in inner hair cells ³⁶. Thus, our estimates of local Ca²⁺ concentrations obtained using Cal520LA represent the lower bounds of the underlying true values. As noted in the Results section, the same may be true for the high-affinity Cal520HA despite its accurate K_eff estimate, due to potential dye saturation effects.

RBC voltage clamp recording

Whole-cell patch-clamp recordings were made from isolated RBCs, as described previously ^19,27,87. Briefly, a patch pipette containing pipette solution (120 mM Cs-gluconate, 10 mM tetraethyl-ammonium-Cl, 3 mM MgCl₂, 0.2 mM N-methyl-d-glucamine-EGTA, 2 mM Na₂ATP, 0.5 mM Na₂GTP, 20 mM HEPES, pH = 7.4) was placed on the synaptic terminal, as described previously. The patch pipette solution also contained a fluorescently-labeled RBP peptide (TAMRA-RBP) to mark the positions of the ribbons and either 1) free Ca²⁺ indicators Cal520HA-free (Figs. 1 and 2) and Cal520LA-free (Fig. 2, 3 and 5) to demonstrate our nanophysiological approach or 2) ribeye-bound Ca²⁺ indicators Cal520HA-RBP (Figs. 3 and Supplementary Figs. 3 and 5) and Cal520LA-RBP (Figs. 3, 4, 5, and 8) to measure local ribbon-associated Ca²⁺ signals. Current responses from the cell membrane were recorded under a voltage clamp with a holding potential (V_H) of -65 mV that was stepped to 0 mV (t₀) for 10 milliseconds. These responses were recorded with a patch clamp amplifier running PatchMaster software (version v2x90.4; HEKA Instruments, Inc., Holliston, MA). Membrane capacitance, series conductance, and membrane conductance were measured via the sine DC lock-in extension in PatchMaster and a 1600 Hz sinusoidal stimulus with a peak-to-peak amplitude of 10 mV centered on the holding potential ⁸⁸.

Acquisition of confocal images

Confocal images were acquired using an Olympus model IX 83 motorized inverted FV3000RS laser-scanning confocal microscopy system (Olympus, Shinjuku, Tokyo, Japan) running FluoView FV31S-SW software (Version 2.3.1.163; Olympus, Center Valley, PA) equipped with a 60 X silicon objective (NA 1.3), all diode laser combiner with five laser lines (405, 488, 515, 561 & 640 nm), a true spectral detection system, a hybrid galvanometer, and a resonant scanning unit. Fluorescently labeled ribeye binding peptide (RBP) ²⁶ and Ca²⁺ indicator were delivered to RBC via a whole-cell patch pipette placed directly at the cell terminal. We waited for 30 s after break-in to allow Cal520HA to reach equilibrium with the patch-pipette before obtaining the first fluorescence image. Rapid x-t line scans at the ribbon location were performed to localize synaptic ribbons (Fig.1Aii) and to monitor local changes in Ca²⁺ concentration at a single ribbon, as we demonstrated previously, to estimate the Ca²⁺ levels at the plasma membrane and to track a single synaptic vesicle at ribbon locations ^19,20. The z-projection from a series of confocal optical sections through the synaptic terminal (Fig. 1Ai) illustrates ribbon labeling (magenta spots). RBP fluorescence was used to localize a synaptic ribbon and to define a region for placing a scan line perpendicular to the plasma membrane, extending from the extracellular space to the cytoplasmic region beyond the ribbon to monitor changes in the Ca²⁺ concentration along the ribbon axis (Fig. 1Aii). The focal plane of the TAMRA-RBP signal was carefully adjusted for sharp focus to avoid potential errors arising from the high curvature near the top of the terminal and the plane of membrane adherence to the glass coverslip at the bottom of the terminal. Sequential dual laser scanning was performed at rates of 1.51 milliseconds per line. Two-color laser scanning methods allowed observation of Ca²⁺ signals (Fig.1C, cyan) throughout the full extent of the ribbon in voltage-clamped synaptic terminals, while the ribbon and cell border were imaged with a second fluorescent label. The Exchange of TTL (transistor-transistor logic) pulses between the patch-clamp and imaging computers synchronized the acquisition of electrophysiological and imaging data. The precise timing of imaging relative to voltage-clamp stimuli was established using PatchMaster software to digitize horizontal-scan synced pulses from the imaging computer in parallel with the electrophysiological data (Fig.1B). Acquisition parameters, such as pinhole diameter, laser power, PMT gain, scan speed, optical zoom, offset, and step size were kept constant between experiments. Sequential line scans were acquired at 1-2 millisecond/line and 10 μs/pixel with a scan size of 256 × 256 pixels. Bleed-through between the channels was confirmed with both lasers using the imaging parameters we typically use for experiments. To test bleed-through from the RBP channel (TAMRA-RBP) to the Ca²⁺ indicator (Cal-520HA and Cal-520LA), whole-cell recordings from RBC terminals were performed with patch pipette solution that contained TAMRA-RBP or the aforementioned Ca²⁺ indicators, and line-scan images were collected and analyzed using the same procedures used for experimental samples.

Point spread function

The lateral and axial point spread function (PSF) was obtained as described previously ^19,28,86. Briefly, an XYZ scan was performed through a single 27 nm bead and the maximum projection in the xy-plane was fit to the Gaussian function using Igor Pro software. We obtained the full width at half maximum (FWHM) values for x and y-width of 268 and 273 nm, respectively, in the lateral (x-y plane) and for y-z width of 561 nm in the axial (y-z axis) resolution.

Photobleaching

We minimized photobleaching and phototoxicity during live-cell scanning by using fast scan speed (10 microsec/pixel), low laser intensity (0.01–0.06% of maximum), and low pixel density (frame size, 256 × 256 pixels). We estimated photobleaching using x-t line scans of Cal520HA or Cal520LA and TAMRA-RBP in the absence of stimulation, with the same imaging parameters used for experimental samples.

Data analysis

Quantitative FluoView x-t and x-y scans were analyzed initially with ImageJ software (imagej.nih.gov) and subsequently with Igor Pro software (Wavemetrics, Portland OR) for curve fitting and production of the figures. Data from PatchMaster software were initially exported to Microsoft Excel (Version 16.81) for normalizing and averaging and exported from MS Excel to Igor Pro (Version 9.05) for curve fitting and production of the l figures.

Analysis of x-t scan data

X-axis profile. To determine the Ca²⁺ signals along the ribbon axis, we spatially averaged the x-axis profile intensity of RBP (i.e., a horizontal row of pixels, see Fig.1C, top) to determine the position of the center of the ribbon and estimate the location of the plasma membrane. The parameter x₀ is the peak of the Gaussian fit, giving the x-position of the center of the ribbon. The x-axis profile is also used to obtain the spatial profile of Ca²⁺ signals before, during, and after with respect to the ribbon profile. To identify the Ca²⁺ signals specific to the ribbon location, we fit x-axis intensity profiles with the equation f(x) = s(x) + g(x), where s(x) is a sigmoid function that describes the transition from intracellular to extracellular background fluorescence at the edge of the cell, given by s(x) = b – c / (1–exp((x_1/2 – x)/d)), and g(x) is a Gaussian function that represents the fluorescence of RBP, given by g(x) = a exp(-(x-x0)²/w²), as described ²⁷. The parameters x_1/2 and x₀ were taken as the x-axis positions of the plasma membrane and the fluorescence emitter, respectively. While parameter b is intracellular background fluorescence, c is extracellular background fluorescence, d is the slope of the sigmoid, a is the peak amplitude of emitter fluorescence, and w is √2* the standard deviation of the Gaussian function, in practice, the latter parameters were highly constrained by the data or by the measured PSF, leaving only x_1/2 and x₀ as free parameters in the fitting. Fig.1D, demonstrates that the peak of the Ca²⁺ signals (x₀, cyan; Fig.1D) during the stimuli is proximal to the ribbon center (x₀, magenta; Fig.1D) towards the plasma membrane (x_1/2, magenta; Fig.1D), as expected for Ca²⁺ influx originating from Ca²⁺ channels localized in the plasma membrane ^44-46.

T-axis profile. The temporal profiles of the Cal520, Cal520-RBP, and TAMRA-RBP signals were determined by analyzing the time-axis profile of the x-t line scan to obtain the kinetics of the Ca²⁺ transient with respect to the ribbon. We determined the baseline kinetics by averaging the fluorescence obtained immediately before depolarization. The timing of depolarization and the amplitude of the Ca²⁺ current was obtained from the PatchMaster software. The rising phase of the Ca²⁺ transient was fit with the sigmoid function, the peak of which is referred to as the peak amplitude Ca²⁺ transient.

We used this baseline profile of ribbon-proximal Ca²⁺ signals to distinguish signals proximal or distal to the ribbon despite the distance between these two signals being within the PSF. For example, the ribbon-proximal signals were obtained by averaging 5 pixels of the temporal profile of the Ca²⁺ signals (obtained with Cal520 or Cal520-RBP) between the x_1/2 and x₀ values obtained in the x-axis profile of TAMRA-RBP. The distal profile was obtained similarly but was 5-pixels after x₀ towards the cytoplasm.

Quantifying the rise-time of the Ca²⁺ transients. The rise-time of the fluoresence transients (Figs. 2A and 2B) was defined as the time to reach half-maximal value from the start of the stimulation, after subtracting its baseline. It was determined by fitting a hyperbolic function with an argument shifted by a constant value β, which serves as one of two free parameter of the fit, along with the argument scale factor α characterizing the steepness of the rising phase of the signal:

The constant C = tanh(α*β) ensures that f(0) = 0 at the start of the stimulation, t = 0. This fitting function reaches its peak value F_peak at t = t_peak, as determined from the averaged fluorescence trace. The sign of the argument shift β controls the shape of the signal near t = 0: β < 0 corresponds to a signal which is concave downward (i.e. very rapid onset), while β > 0 corresponds to a signal which is concave upward near t = 0 (i.e. gradual rise). Therefore, this function provides a more flexible functional relationship than the Hill function or the standard sigmoidal function.

Analysis of x-y scan data

We analyzed the rate of loading the TAMRA-RBP and Ca²⁺ indicator into the RBC terminal with ImageJ software by placing a square region of interest (ROI; 5 x 5 μm) and using a Plot z-axis profile function to obtain spatially averaged TAMRA-RBP and Ca²⁺ indicator fluorescence as a function of time. The rising phase of TAMRA-RBP and Ca²⁺ indicator fluorescence was obtained by fitting the rising to the peak to the single exponential function using the curve fitting function in Igor Pro 9 software. The rate of the exponential function is defined as the rate of fluorescence loading to the terminal.

For analysis comparing the fluorescence intensity of ribbons and Ca²⁺ influx elicited by 100-ms stimuli (Fig. 11A & B), 10 images of both TAMRA-RBP and Cal520LA-RBP were collected in sequential, followed by imaged Cal520LA-RBP only during 200-ms depolarization, and the sequence ended with 10 images of both TAMRA-RBP and Cal520LA-RBP in sequential to confirm the ribbon locations. Two color imaging were obtained sequentially with a frame interval of 407 ms, and Cal520LA-RBP only during 100ms-stimuli was 205 ms. Images before and after were then averaged to obtain the ribbon location. Synaptic ribbon fluorescence was quantified as the ratio of TAMRA fluorescence to the fluorescence of the nearby RBC cytoplasm, measured approximately eight to nine pixels away (F_ribbon/F_{near by}), measuring the pixel with the highest intensity. The change in Cal520LA-RBP fluorescence, used as a proxy for active zone Ca²⁺ influx, was estimated as ΔF/F_rest, where F_rest is the fluorescence intensity before depolarization at −65 mV, and ΔF is the difference when depolarized to 0 mV. Ca²⁺ indicator intensity was calculated as the average fluorescence of nine pixels centered on the pixel with the greatest fluorescence increase.

Experimental design and statistical methods

We did not use any statistical methods to determine the sample size prior to experiments. Mean value variances were reported as ± the standard error of the mean (SEM). The statistical significance of differences in average amplitudes of Ca²⁺ current, capacitance, synaptic ribbon size and number, and Ca²⁺ transients were assessed using unpaired, two-tailed t-tests with unequal variance using R Studio (Version 2023.09.0+463) and Igor Pro software.

Amplitude analysis for Figure 8. Amplitude data were processed and analyzed using R Studio software (Version 2023.09.0+463). The data were divided into two groups, that of individual cells and of multiple cells, and analyzed as follows. Multiple individual cells from each condition (proximal 10 mM EGTA Cal520LA-RBP and distal 10 mM EGTA Cal520LA-RBP) were analyzed to compare the variability between the Ca²⁺ amplitude measurements of the ribbons from each cell. Having verified the normality of the data to ensure the validity of the statistical test, we used one-way ANOVA and Tukey’s honest significant difference test to assess the specific statistical differences between ribbons. To compare the variability between the amplitude measurements of different cells, we compared readings for various cells. If there were multiple measurements for a single ribbon in a specific cell, those readings were averaged. In this case, the normality of the data was again verified and Welch’s ANOVA was employed to assess the between-group differences as it does not assume homogeneity of variance, with the Games-Howell post-hoc test being used to understand the specific differences between cells.

Computational modeling of Ca²⁺ dynamics

Spatio-temporal Ca²⁺ dynamics is simulated in a volume shown in Fig. 9, which is a 3D box of dimensions (1.28 x 1.28 x 1.1) µm³. Assuming an approximately spherical bouton of diameter 5µm and an average of 36 ribbons per terminal, one obtains synaptic volume per ribbon of 1.81 µm³, which matches the volume of our box domain. Ribbon serves as an obstacle to Ca²⁺ and buffer diffusion, and has an ellipsoidal shape with semi-axis of 190nm along the Y- and Z-axes (for a total height of 380nm), and a semi-axis of 70 nm along the X-axis. It is attached to the plasma membrane by a thin ridge (arciform density) of dimensions 60x30x30 nm. Ca²⁺ ions enter through 4 channels (or clusters of channels) at the base of the ribbon indicated by black disks in Fig. 6, forming a square with a side length of 80 nm; this highly simplified channel arrangement is sufficient to capture the level of detail in spatial [Ca²⁺] distribution that we seek to resolve. Each of the four clusters admits a quarter of the total Ca²⁺ current. Ca²⁺ current values and current pulse durations are listed in Fig. 7. Although the actual distribution of Cav channels is more heterogeneous, channels close to the ribbon play a greater role in neurotransmitter release ².

The simulation includes a single dominant Ca²⁺ buffer with molecules possessing a single Ca²⁺ ion binding site. We assume that only immobile buffer is present in a patch-clamped cell, since any mobile buffer would diffuse into recording pipette. Simulations with mobile buffer in an un-patched cell are also provided (Supplementary Fig. 4B&C). Buffer parameter values were set to values reported by ⁷, namely dissociation constant of 2 µM, and total concentration of 1.44 mM (total buffering capacity at rest of 720). We also examine the impact of the assumption on buffer concentration by repeating the simulation with a lower buffering capacity of 100 (total buffer concentration of 200 µM). The Ca²⁺ ion diffusivity is set to 0.22 µm²/ms ⁸⁹.

To minimize the number of undetermined parameters, we assumed the same parameters for the Ca²⁺ clearance on all surfaces, with one high-capacity but lower affinity process simulating Na⁺/Ca²⁺ exchangers, and one lower capacity but higher affinity process simulating Ca²⁺ extrusion by ATPase PMCA and SERCA pumps. We used the Hill coefficient of m=2 for the Ca²⁺ sensitivity of the SERCA pumps on all surfaces except the bottom surface ⁹⁰, whereas Hill coefficient of m=1 corresponds to PMCA pumps on the bottom surface ⁹¹. Thus, the flux of [Ca²⁺] across the surface (ions extruded per unit area per unit time) is described by

where C is the Ca²⁺ concentration, D_C is the Ca²⁺ diffusivity, and n ⋅ ▽C is the gradient of [Ca²⁺] normal to the boundary. The constant leak term Φ_Leak ensures that the flux is zero at the resting value [Ca²⁺]_rest = C₀ = 0.1 µM. The Ca²⁺ clearance rate and affinity parameters for the pumps and exchangers are listed in Supplementary Table 3. We note however that the density of Na⁺-Ca²⁺ exchangers and PMCA/SERCA pumps are rough estimates and have a significant effect on results only for longer stimulation durations of over 100-200 ms. Reaction-diffusion equations for [Ca²⁺] and buffer are solved numerically using Calcium Calculator software version 6.10.7 ^92,93.

Serial block-face scanning electron microscopy (SBF-SEM)

The eyecups of adult (16 ∼20-month-old) zebrafish were dissected in bicarbonate-buffered Ames’ medium and, after removing the lens, the eyecups were fixed in 4% glutaraldehyde in 0.1 M sodium-cacodylate buffer at RT for one hour, followed by overnight fixation at 4°C. Thereafter, samples were treated to generate blocks for serial block face scanning electron microscopy (SBF-SEM) according to the published protocol ⁹⁴, in this case using Zeiss 3-View SBF-SEM. Multiple 35 µm montages were acquired at a 5 nm X-Y resolution and a 50nm Z resolution across the entire retina from the outer (photoreceptor) layer to the inner (ganglion) cell layer. Images were inspected visually and structures of interest were traced using the TrakEM plug-in in ImageJ software (NIH/Fiji), as described below.

Analysis of serial block-face scanning electron microscopy images. SBF-SEM image stacks of the zebrafish retina were imported, aligned, visualized, and analyzed using the TrakEM plugin (version 1.5h) in ImageJ software (NIH). RBC1 were identified based on their characteristic morphology and 4-7 μm terminal size, as previously described ⁴⁸. RBCs, synaptic ribbons, pre-synaptic membranes, and the area of the ribbon touching the plasma membrane were traced, painted using the TrakEM brush features, and rendered in 3D for structural visualization. The area of the synaptic ribbon touching the RBC membrane was quantified using the measuring tools in ImageJ and images were prepared using Adobe Photoshop software (version 25.9). The distance between each ribbon from three different RBCs and its nearest five ribbons (Fig. 9B and Video 1) was measured as follows: Ribbons on the same plane had their linear distance measured, whereas the distance between ribbons that were close to each other but in different layers was calculated using the Pythagorean Theorem.

Acknowledgements

This research was funded by the National Institutes of Health (NIH) National Eye Institute (NEI), award numbers R01EY030863, 3R01EY030863-02S1, and 3R01EY030863-03S1, and a UTHSC College of Medicine Faculty Research Growth Award (to TV), R01 EY032396, P30EY026878, R01NS122388 (DZ), and a University of Wisconsin-Madison (UW2020) WARF discovery award (to MH) which funded the acquisition of the 3D view serial block-face electron microscope. We would like to acknowledge the UW Madison School of Medicine electron microscopy facility for processing and imaging samples and thank Dr. Alex Dopico, the Van Vleet Chair of Excellence and Professor in the Department of Pharmacology, Addiction Science, and Toxicology, UTHSC, for critical reading, and Dr. Kyle Johnson Moore in the UTHSC Office of Research for editing the manuscript.

Additional information

Author Contributions

Conceptualization, D.Z. and T.V.; Methodology and Investigation, N.R., A.S.; Data analysis, N.R., A.S., J.B., M.H., T.V.; Computational modeling, V.M.; Visualization, J.B., M.H., V.M., T.V.; Writing - Original Draft, N.R., T.V.; Writing-Review & Editing, all authors; Funding, M.H., D.Z., T.V.; Supervision and Project administration, T.V.

Additional files

Supplemental figures and legends

Video1. 3D volume movie of the synaptic terminal of a zebrafish rod bipolar cell terminal showing the distribution of included synaptic ribbons and measurements. RBC terminal is shown in light green with ribbons in magenta. Illustrations show how the measurements were obtained for ribbons on the same vs. different layers.

Video2. A 3D reconstruction of the RBC synaptic terminal and ribbon from SBF-SEM stacks is provided in Video 2. A. 3D reconstruction of the RBC terminal (green) with included synaptic ribbons (colored magenta). B. 3D reconstruction of the RBC synaptic ribbons shows different shapes and sizes of ribbon (magenta) and the AZ, the area of the ribbon facing the plasma membrane (cyan). AZ, active zone.

Video3. A time-lapse movie of the synaptic terminal of a zebrafish bipolar neuron during Ca²⁺ influx is provided in Video 3. Left: RBC terminal labeled with TAMRA-RBP shows the locations of synaptic ribbons (magenta spots). Middle: RBC terminal filled with Cal520LA-RBP shows the Ca²⁺ influx in 0.02 s. looped two times to see Ca²⁺ influx in 0.07 s. Right: Superimposed TAMRA-RBP and Cal520LA-RBP shows the Ca²⁺ influx as spot-like maxima near the membrane during depolarization. TAMRA-RBP 10 images before depolarization, and stacks of Cal520LA-RBP 5 images before depolarization and 5 images during depolarization were compiled to demonstrate the locations and magnitude of the Ca²⁺ influx. The interval between frames is 407 ms, Total duration of the movie 0.9 s. Each frame is an individual, unaveraged image.