Inter-relationships among physical dimensions, distal–proximal rank orders, and basal GCaMP fluorescence levels in Ca²⁺ imaging of functionally distinct synaptic boutons at Drosophila neuromuscular junctions

Abstract

GCaMP imaging is widely employed for investigating neuronal Ca²⁺ dynamics. The Drosophila larval neuromuscular junction (NMJ) consists of three distinct types of motor terminals (type Ib, Is and II). We investigated whether variability in synaptic bouton sizes and GCaMP expression levels confound interpretations of GCaMP readouts, in inferring the intrinsic Ca²⁺ handling properties among these functionally distinct synapses. Analysis of large data sets accumulated over years established the wide ranges of bouton sizes and GCaMP baseline fluorescence, with large overlaps among synaptic categories. We showed that bouton size and GCaMP baseline fluorescence were not confounding factors in determining the characteristic frequency responses among type Ib, Is and II synapses. More importantly, the drastic phenotypes that hyperexcitability mutations manifest preferentially in particular synaptic categories, were not obscured by bouton heterogeneity in physical size and GCaMP expression level. Our data enabled an extensive analysis of the distal–proximal gradient of GCaMP responses upon genetic and pharmacological manipulations. The results illustrate the conditions that disrupt or enhance the distal–proximal gradients. For example, stimulus frequencies just above the threshold level produced the steepest gradient in low Ca²⁺ (0.1 mM) saline, while supra-threshold stimulation flattened the gradient. Moreover, membrane hyperexcitability mutations (eag¹ Sh¹²⁰ and para^bss1) and mitochondrial inhibition by dinitrophenol (DNP) disrupted the gradient. However, a novel distal–proximal gradient of decay kinetics appeared after long-term DNP incubation. We performed focal recording to assess the failure rates in transmission at low Ca²⁺ levels, which yielded indications of a mild distal–proximal gradient in release probability.

Keywords:

• Proximal-distal gradient
• bouton size
• Ca²⁺ clearance
• ion channels
• GCaMP baseline fluorescence
• mitochondria
• octopamine
• tonic
• phasic
• type Ib
• type Is
• type II synapses
• frequency response
• facilitation

Introduction

Regulation of presynaptic Ca²⁺ is critical for transmitter release as well as short-term and long-term synaptic plasticity (Atwood & Karunanithi, 2002; Brown, Chapman, Kairiss, & Keenan, 1988; Rusakov, 2006; Zucker, 1996; Zucker & Regehr, 2002). The Drosophila larval body-wall neuromuscular junction (NMJ) is an ideal system to contrast Ca²⁺ dynamics in synapses of different functional categories, as it contains in close proximity both tonic and phasic (type Ib and Is, respectively) glutamatergic synapses, as well as modulatory octopaminergic (type II) synapses, all of which can be imaged in the same microscopic field (Atwood, Govind, & Wu, 1993; Hoang & Chiba, 2001; Jia, Gorczyca, & Budnik, 1993; Johansen, Halpern, Johansen, & Keshishian, 1989; Koon et al. 2011; Monastirioti et al., 1995; Schmid, Chiba, & Doe, 1999; Zhong & Wu, 2004). Our previous work has demonstrated characteristic frequency dependence of GCaMP signals for type Ib, Is and II synaptic boutons, indicating their distinct Ca²⁺ dynamics, i.e. type II, Is and Ib synapses being responsive to low, medium and high stimulus frequencies, respectively (Xing & Wu, 2018).

Our studies over the past years have accumulated a large body of single bouton records of GCaMP responses, along with quantifications of bouton sizes and GCaMP baseline fluorescence intensities. Despite the nomenclature implying size-related distinctions among these synapses (type Ib also known as ‘I big’, Is as ‘I small’), we observed significant heterogeneity with overlapping bouton sizes between different synaptic categories. Furthermore, for each synaptic category, the level of GCaMP expression (as indicated by baseline fluorescence intensity F) varied significantly in our database. We set out to determine the ranges of variation in these parameters and examined whether such high levels of heterogeneity confound interpretations from GCaMP measurements for salient physiological properties of the distinct synaptic bouton types.

Our database also allowed a re-examination of the previously reported distal-to-proximal gradient in GCaMP response (ΔF/F, cf. Guerrero et al., 2005; He, Singh, Rumpal, & Lnenicka, 2009; Lnenicka, Grizzaffi, Lee, & Rumpal, 2006). Analyses based on genetic and pharmacological manipulations further revealed the various conditions that can obscure or optimize the gradient. Lastly, we carried out simultaneous electrophysiological focal recording (Xing & Wu, 2018) to map local synaptic transmission events to investigate the physiological significance of such distal–proximal GCaMP response gradient along the motor terminal branch.

Materials and methods

Fly stocks

In this study, two stable fly strains carrying the Gal4 driver and UAS-GCaMP on the same chromosome were made by recombination: c164Gal4-UAS-GCaMP1.3 (on the 2nd chromosome, abbreviated hereafter c164-GCaMP1.3, see Torroja et al., 1999 for c164-Gal4, a motor neuron driver) stock was used to carry out most of experiments, and nSynaptobrevinGal4-UAS-GCaMP6m (nSyb-GCaMP6m, nSyb-Gal4 is a pan-neuronal driver on the 3rd chromosome) was used to confirm the results. The UAS-GCaMP1.3 stock was a generous gift from Drs. Yalin Wang and Yi Zhong of Cold Spring Harbor Laboratory (Cold Spring Harbor, NY) (Nakai, Ohkura, & Imoto, 2001; Ueda & Wu, 2006; Wang et al., 2004). UAS-GCaMP6m was from the Bloomington Stock Center (Bloomington, IN, see also Chen et al., 2013). nSyb-Gal4 was a gift from Dr. Toshihiro Kitamoto of the University of Iowa. In addition, K⁺ channel mutational line eag¹ Sh¹²⁰/Y; c164-GCaMP1.3 and Na⁺ channel mutational line para^bss1/Y; c164-GCaMP1.3 were created for analysis (cf. Xing & Wu, 2018). For simplification, eag¹ Sh¹²⁰ is abbreviated as eag Sh, and para^bss1 is abbreviated as bss. OK371-mCD8GFP was constructed from OK371-Gal4 (2nd) and UAS-mCD8GFP (2nd) for the use in electrophysiological recording.

Ca²⁺ imaging and focal recording

All equipments, procedures and protocols have been described previously (Xing & Wu, 2018). In brief, 3rd instar Drosophila larvae were dissected and laid open in HL3.1 saline (0.1 or 0.5 mM Ca²⁺, 7 mM glutamate was added to stop muscle contraction in case of 0.5 mM Ca²⁺, cf. Feng, Ueda, & Wu, 2004; Stewart, Atwood, Renger, Wang, & Wu, 1994 for details about HL3.1), with segmental nerve bundles severed from the ventral ganglion and stimulated using a glass suction electrode (about 10 μm in diameter, filled with HL3.1). To monitor synaptic transmission, a second extracellular glass electrode with an opening of 7–8 μm (fire-polished and bent to ‘L’ shape by heat using a microforge, De Fonbrune, and filled with HL3.1) was employed to probe individual synaptic boutons. The signals picked up was fed to an AC amplifier (GRASS model p15, Warwick, RI) to register the extracellular focal excitatory junction potentials (efEJP) (cf. Xing & Wu, 2018). Distal and proximal boutons of muscle 4 type Ib synaptic terminals were recorded at 0.1 and 0.133 mM Ca²⁺, and transmission failure rates of the designate stimulus frequencies were examined.

The effects of inhibiting mitochondrial proton gradient were studied with the proton ionophore 2,4-dinitrophenol (DNP, Kodak, ‎Rochester, NY). Dissected larval preparations were first imaged in HL3.1 (0.1 mM Ca²⁺) to obtain control data and the saline was then replaced with HL3.1 containing DNP (0.2 mM). The effect of DNP incubation was monitored up to 60 min (Xing & Wu, 2018).

Image capture and recording was carried out with the RedshirtImaging NEUROCCD-SM256 system, which includes a CCD camera by SciMeasure Analytical Systems, and the data acquisition control system (RedshirtImaging, Decatur, GA). The excitation light was generated with a 75 W xenon lamp (Ushio, CA). Data compilation and first-order analyses were performed using the computer software NeuroPlex. The digital images (256 × 256 pixels per frame) were sampled at a frame rate of 25 Hz.

Data analysis

The fluorescence intensity (F) was measured at the brightest spots of each bouton, with a fixed kernel size of 3 × 3 pixels (Xing & Wu, 2018). The GCaMP signal amplitude ΔF/F was calculated using the equation ΔF/F = (F_t – F)/F, in which F represents the baseline fluorescence intensity (25 frames right before the stimulation), and F_t represents the fluorescence intensity at any time point during the imaging. Max ΔF/F (determined with a running-average bin of five frames) was used as the amplitude of GCaMP signals. Half-decay time was obtained as the time point when ΔF/F falls to 50% from the end of stimulus train. Bouton size (measured as cross-sectional bouton area directly obtained from fluorescent images, in µm²) and absolute nerve branch length (from distal bouton to branch point, in µm) measurements were carried out using ImageJ (NIH), calibrated to a standard microscopic ruler (micrometer level). For each bouton, its relative position (normalized distance) from the distal end was determined by dividing its absolute distance from the distal bouton with absolute total nerve branch length. The bouton rank order was defined as the serial number of boutons counted from the distal most end within the image. The pseudocolor maps (Figure 1), ΔF/F traces, amplitudes (maximum ΔF/F) and kinetics (half-decay time) of GCaMP fluorescence signals and electrophysiological recording traces were processed using a Python-based program developed by the author (see also Xing & Wu, 2018). Failure rates of synaptic transmission for each bouton were determined as the number of stimuli that failed to evoke efEJPs being divided by the total number of stimuli that were delivered to the same bouton. All other plots and analysis, such as Pearson’s linear regression and Student’s t-tests, were generated and compiled using Microsoft Excel.

Figure 1. Distinct frequency response patterns of GCaMP Ca²⁺ signals in glutamatergic type Ib (tonic) and Is (phasic), and aminergic type II synapses. (A–C) GCaMP signals (ΔF/F) evoked at 10-, 20- and 40-Hz stimulation, respectively. Stimulation duration: 2 s. External Ca²⁺ concentration: 0.1 mM. Left panels: maximum response intensity shown in monochromatic maps, with the lookup color map shown in bottom right of the figure. Right panels: kinetic waveform traces from representative boutons. Type Ib, Is and II motor terminals are encircled with orange (thick grey), black and cyan contours (thin grey), respectively; corresponding color-coded waveform traces are shown in the right panels. (D) GCaMP baseline fluorescence (F) of type Ib, Is and II boutons (inverted fluorescence image). Genotype type: +/Y; c164-GCaMP1.3, which was used throughout this study and served as the wild type control in Figures 5 and 9.

Results

Distinct presynaptic Ca²⁺ dynamics in type Ib, Is and II boutons independent of GCaMP baseline fluorescence and bouton size

The Drosophila NMJ contains three major types of synapses, i.e. the excitatory glutamatergic type Ib and Is, and modulatory octopaminergic type II synapses, each of which is derived from a different type of motor neurons (Hoang & Chiba, 2001). Our study has focused on the larval body-wall muscles M12 and M13, in which the motor terminals of type Ib (tonic), Is (phasic, Bradacs, Cooper, Msghina, & Atwood, 1997; Kurdyak, Atwood, Stewart, & Wu, 1994; Lnenicka & Keshishian, 2000) and II synapses can be readily identified and imaged simultaneously in the same microscopic field (Figure 1). As previously described (Xing & Wu, 2018), a salient feature that distinguishes the three synaptic types is the characteristic frequency responses of their GCaMP signals. At a low external Ca²⁺ level (0.1 mM), sizable response amplitude (peak ΔF/F) could be observed at 10 and 20 Hz nerve stimulation for type II and Is boutons, respectively, whereas type Ib responses became detectable only at frequencies beyond 40 Hz (Figure 1(A–C)). However, it was also apparent that the basal GCaMP fluorescence levels (F) was highest in type Ib, and lowest in type II boutons (Figure 1(D)).

Our database of large populations of boutons indicates wide ranges of variation in both the basal GCaMP level (Figure 2) and the bouton size (Figure 3). An immediate question is whether such high levels of heterogeneity could degrade the salient frequency responses which are characteristic of each bouton type. Furthermore, it is not known whether variable bouton sizes could modify the frequency responses (through varying the surface area to volume ratio and other more obscure factors).

Figure 2. Lack of correlation between GCaMP baseline fluorescence (F) and response amplitude (ΔF/F) within type Ib, Is and II boutons. (A) ΔF/F at 0.1 mM Ca²⁺, with stimulus frequency of 10, 20, and 40 Hz. and (B) same conditions but at 0.5 mM Ca²⁺. Note the large range of response amplitude and baseline fluorescence level for each synaptic category. Regardless, the distinct frequency response for each population remained evident (see the collapsed mean ± SD for max ΔF/F). Separate populations of larval preparations were sampled for 0.1 (8–11 NMJs) and 0.5 mM Ca²⁺ (5 NMJs). Mostly, only one NMJ was imaged from each larva. Student’s t-tests with Bonferroni’s corrections. *p < .05, **p < .01, ***p < .001 for this and following figures.

Figure 3. Lack of correlation between GCaMP signal amplitudes (ΔF/F) and bouton size within type Ib, Is and II boutons. Max ΔF/F vs. bouton sizes of type II, Is and Ib synapses with stimulation frequency at (A) 10 Hz and (B) 20 Hz. 8–11 NMJs.

The GCaMP baseline fluorescence intensities (F) in type Ib, Is and II synaptic boutons were individually measured and plotted against the corresponding maximum ΔF/F, i.e. peak GCaMP Ca²⁺ transient response (Figure 2(A,B), at 0.1 and 0.5 mM Ca²⁺, respectively, with stimulus frequencies at 10, 20 and 40 Hz, covering GCaMP Ca²⁺ responses from threshold up to saturation levels). This treatment delineates a large range of variation in baseline GCaMP fluorescence intensities (F), which spanned about 20 times for type Ib (from about 20 to nearly 400 mV, measured by the CCD camera, Figure 2), and in somewhat narrower ranges for types Is and II (about 20–200 mV, Figure 2). However, it was evident that the distinct frequency response for each of the three synaptic types was independent of the absolute levels of basal GCaMP expression. For example, among type Is boutons at both 0.1 and 0.5 mM external Ca²⁺, there was a clear overall elevation in ΔF/F across the entire population as the stimulus frequency increased from 10 Hz to 20, and then 40 Hz (see population means and SDs of max ΔF/F on the right of each panel in Figure 2(A,B)). Further, there is no indication of preferential increase among boutons with either higher or lower baseline fluorescence intensity. Consistent results were obtained using both c164-GCaMP1.3 (Figure 2) and nSyb-GCaMP6m (not shown). For type Ib boutons, this became more evident when external Ca²⁺ was increased to 0.5 mM to reach the threshold level and beyond (Figure 2(B), cf. Xing & Wu, 2018). Lastly, the GCaMP response of type II synapses approached saturation at stimulus frequencies above 10 Hz (Xing & Wu, 2018), without showing a trend in or dependence on baseline fluorescence intensity (Figure 2(A,B)). Type Is approached saturation at 40 Hz (Figures 1(C) and 2), whereas type Ib beyond 80 Hz (Xing & Wu, 2018). Therefore, the distinct frequency dependence of GCaMP Ca²⁺ signals appeared to be intrinsic to each type of synaptic terminals, regardless of the absolute levels of basal GCaMP expression.

Although it is generally conceived that type Ib synapses are the biggest in bouton sizes, type Is intermediate, and type II the smallest (Figure 1, cf. Budnik, Zhong, & Wu, 1990; Johansen et al., 1989; Kurdyak et al., 1994), our database indicates wide ranges of heterogeneity in bouton sizes within each synaptic category, with overlapping bouton sizes between different synaptic categories. Measurements of bouton areas showed considerable variation, with type Ib ranging from 2 to 20 µm², type Is from 2 to 14, type II from 2 to 11 (Figure 3). However, such overlap did not obscure the distinction in the frequency response of GCaMP signals from the three bouton types. When plotting bouton areas against GCaMP Ca²⁺ signal amplitude (max ΔF/F), there is no clear trend of preferential ΔF/F enhancement among either larger or smaller boutons (Figure 3).

Correlation between GCaMP baseline fluorescence and bouton size

In principle, for boutons of a uniform shape, the bouton area may be correlated with the thickness of individual boutons, which determines the optical path through which the incident light interacts with the GCaMP indicator to emit fluorescent signal. To examine this possibility, we plotted GCaMP baseline intensity (F) against corresponding bouton area, and indeed found the trends indicating that boutons with larger sizes tended to be associated with stronger baseline fluorescence F (Figure 4). Linear regression analysis produced the correlation with r² values of 0.38 for type Ib, 0.40 for type Is, and 0.19 for type II boutons (Figure 4, statistically significantly different from the null hypothesis of zero correlation, p = .0011 or less). The result clarifies that the sizes of boutons are indeed correlated with the absolute readings of GCaMP baseline fluorescence (F, cf. Figure 1(D)), but does not appear to be a confounding factor when determining the relative GCaMP response amplitude (ΔF/F, cf. Figure 1(A–C)), which is normalized to the basal reading F.

Figure 4. Correlation between GCaMP baseline fluorescence intensity with bouton area. (A) Baseline fluorescence intensity (F) vs. bouton size measured by maximum cross-sectional area for type Ib, Is and II boutons. Each symbol is color coded for type Ib (crosses), Is (circles) and II (squares). Dashed lines indicate the mean sizes. Linear regression between baseline fluorescence (F) and bouton area: (B) type Ib, (C) type Is, and (D) type II. The significant correlation may be accounted for by increasing fluorescence levels due to thickening of optical paths in larger boutons. 8–11 NMJs.

Preferential alterations in GCaMP Ca²⁺ signals in hyperexcitable mutants largely independent of bouton size and baseline fluorescence intensity

K⁺ and Na⁺ channels control action potential firing of axons, which leads to the opening of Ca²⁺ channels in presynaptic terminals. Thus, in principle, hyperexcitable K⁺ and Na⁺ channel mutations can enhance Ca²⁺ influx, subsequently increasing GCaMP Ca²⁺ signal amplitudes.

As previously demonstrated, certain hyperexcitable K⁺ and Na⁺ channel mutations can result in striking GCaMP signal enhancement (Xing & Wu, 2018). We investigated whether such striking phenotypes are preferentially displayed by boutons of particular sizes and baseline fluorescence levels. We first examined the K⁺ channel double mutant eag Sh (Ganetzky & Wu, 1983, 1985, 1986), which enhanced Ca²⁺ dynamics preferentially in type I synapses, especially in type Is (Xing & Wu, 2018). As shown in Figure 5, even at 2 Hz stimulation, the striking effect on type Is was evident (see Figure 5(E), ensemble population mean ± SD for ΔF/F, WT: 0.064 ± 0.038, eag Sh: 0.43 ± 0.35), with robust GCaMP signals triggered by each stimulus. Significant enhancement of GCaMP signals was also observed in type II at 2 Hz (WT: 0.11 ± 0.05, eag Sh: 0.17 ± 0.10), and type Ib at 40 Hz (WT: 0.15 ± 0.11, eag Sh: 0.59 ± 0.42). In general, the ranges of bouton size (in µm², WT: 4.7 ± 2.2, eag Sh: 4.3 ± 1.6) and baseline fluorescence intensity (in mV, WT: 108.8 ± 61.9, eag Sh: 120.9 ± 59.3) of WT and eag Sh roughly overlap, and we did not detect a clear trend of GCaMP response enhancement preferentially associated with particular ranges of bouton size or baseline fluorescence intensity in eag Sh mutant larvae.

Figure 5. Robust differential effects of hyperexcitability mutations on type Ib, Is and II boutons independent of bouton size and GCaMP baseline fluorescence intensity. (A–C) Representative ΔF/F traces for type Ib (40 Hz), Is (2 Hz) and II (2 Hz), respectively. (D–F) Max ΔF/F vs. bouton area. (G–I) Max ΔF/F vs. GCaMP baseline fluorescence intensity. Differential effects of Na⁺ (bss, 6–9 NMJs) and K⁺ (eag Sh, 6–12 NMJs) channel mutations, against WT (8–11 NMJs) are shown along individual columns for type Ib, Is and II boutons. Individual symbols are color-coded (D–I) for genotypes in correspondence to the traces shown in A–C. Note that bss preferentially affects type II while eag Sh confers hyperexcitability in type I synapses, most strikingly on type Is. To the right of each panel (D–I), the ensemble population mean ± SD is presented for WT, eag Sh, and bss. Student’s t-tests with Bonferroni’s correction.

In contrast, our earlier study indicates that a gain-of-function Na⁺ channel mutant bss (Ganetzky & Wu, 1982; Parker, Padilla, Du, Dong, & Tanouye, 2011) preferentially enhances type II synapses (Xing & Wu, 2018). Figure 5 shows that bss enhanced GCaMP response in type II synapses even at 2 Hz stimulation, exceeding the effects of eag Sh (WT: 0.12 ± 0.07, bss: 0.39 ± 0.26 vs. eag Sh: 0.14 ± 0.07). This enhancement again had no clear association with bouton size (in µm², WT: 9.9 ± 6.6, eag Sh: 9.7 ± 4.7, bss: 10.1 ± 4.8). The GCaMP response was minimally increased in type Is at 2 Hz (see population mean ± SD in Figure 5(E,H)), and again, the range of type Is bouton size overlaps with that of WT.

It should be noted that type Ib boutons in bss displayed an unusual skewed distribution of GCaMP response with a high proportion of NMJs with low responsive boutons, as compared to other genotypes and bouton types (seven out of nine NMJs in bss, vs. one out of six in eag Sh, three out of eight in WT, Figure 5(D,G)). In addition, a larger proportion of type Ib boutons in bss showed high basal GCaMP fluorescence levels but exhibited small responses that were not clearly above the noise level (0.05, Figure 5(G)). This led to a negative correlation of GCaMP signal amplitude (ΔF/F) vs. baseline fluorescence (F) for type Ib in bss. A milder tendency was seen in type Is boutons as well (Figure 5(H)). The functional implication of this finding is unknown.

Correlation of GCaMP signal amplitude and distal–proximal rank order of boutons along motor terminals

Several reports have documented a distal–proximal gradient of Ca²⁺ signal amplitude along the larval motor terminals, which are sometimes coupled with strength of postsynaptically registered optogenetic quantal responses release (Guerrero et al., 2005; He et al., 2009; Lnenicka et al., 2006). Most of previous studies focused on type Ib, but did not establish distinctions among type Ib, Is and II synapses. We took advantage of our samples from muscles 12 and 13 to document any existing gradients in type Ib, Is and type II synaptic terminals. To characterize the variation of the GCaMP signal amplitude along the type Ib and Is synaptic terminals, the rank order of boutons was designated, ranking from the distal most (end of the terminal) to the proximal (the entry point of muscle innervation). Type II terminals traverse far beyond the microscopic field, which is suitable for imaging type Ib and Is boutons. Therefore, when simultaneously recording all three types of boutons, we could observe entire type Ib and Is terminal branches, but only a portion of the type II boutons close to the entry point were collected. Therefore, the most distal type II boutons were not included in the analysis.

We confirmed that the GCaMP signal amplitudes displayed a distal–proximal gradient in the majority of type Ib and Is terminal branches, given appropriate stimulus frequencies and Ca²⁺ concentrations (Figure 6). For type Ib, a high frequency was needed to reveal the gradient (40 and 80 Hz, 0.1 mM Ca²⁺, Figure 6(A)). For type Is, the gradient was more pronounced when stimulated at 20 Hz in 0.1 mM Ca²⁺ saline. However, the gradient diminished at higher frequency of stimulation (40 Hz), presumably due to saturation of the GCaMP response or differences in frequency-dependent properties between distal and proximal boutons (Figure 6(B)). As expected, subthreshold low frequency stimulation (10 Hz for type Is, 20 Hz for type Ib, see Figure 6(A,B), respectively) did not evoke any GCaMP signals to build up any gradient. Notably, within the field imaged, type II boutons displayed large variability and no hint of any reliable gradient from 10 to 40 Hz (Figure 6(C)).

Figure 6. Detection of gradients of max ΔF/F following the sequential order of boutons along the motor terminals. (A) Type Ib, (B) Type Is, and (C) Type II synaptic terminals. Amplitudes of max ΔF/F at different stimulus frequencies were measured for individual boutons and plotted against their distal–proximal rank order (20–80 Hz for type Ib, 10–40 Hz for type Is and II synapses). Boutons were ranked from distal end (1st bouton) to proximal along the motor terminal branch. Representative ΔF/F traces from a distal bouton superimposed on a proximal bouton of the same terminal branches are shown above each plot. Linked grey dots represent boutons along individual terminal branches. Large black circles are the mean values for the gray dots at each sequential rank order (5–11 NMJs). At least three boutons from different NMJs were needed for averaging. Therefore, proximal boutons of some long terminals were not included for averaging. Student’s t-tests were performed between the most distal sets of boutons against all the boutons beyond the 3rd rank (*p < .05).

Despite the significant ΔF/F gradient along the terminal branches, the gradients in type Ib and Is synaptic terminals were not associated with bouton size and GCaMP baseline fluorescence, since we did not detect any distal–proximal differences in these two parameters (Figure 7). As the numbers of boutons along each branch could vary considerably (Figures 6 and 7), we measured the physical distance of individual boutons from the distal end. We replotted all data points in Figure 6(A,B), from bouton rank order to its normalized distance from the distal end (see Materials and Methods). The results confirmed the presence of a gradient along the motor terminal, and revealed there was significant linear correlations for both type Ib and Is boutons, according to their physical distance from the end (regression values 0.40 and 0.41, respectively, Figure 8).

Figure 7. Lack of longitudinal gradient in bouton size and GCaMP baseline fluorescence. (A) Bouton size measured by area, and (B) GCaMP baseline fluorescence intensity. Both were plotted against the sequential order of boutons. 8–11 NMJs.

Figure 8. Regression of GCaMP signal gradients (max ΔF/F) along the normalized physical distance of type I bouton locations from the motor terminal end. (A) Type Ib (40 Hz) and (B) Type Is (20 Hz). The relative distances of individual boutons were expressed as a fraction, determined by dividing the absolute distance from the distal end over the entire length of the terminal nerve branch. Ib: N = 5, Is: N = 10.

Disruption of distal–proximal gradient of GCaMP response by hyperexcitability mutations

As demonstrated in Figure 5, hyperexcitable voltage-gated Na⁺ and K⁺ channel mutations could greatly enhance presynaptic GCaMP signals by increasing Ca²⁺ influx. However, it remains unclear whether these mutations exert preferential effects on boutons at different locations along the synaptic terminal branches. An analysis of the longitudinal distribution of GCaMP ΔF/F signals indicates that the enhancement was associated with increased variability, but such effects appeared uniform along synaptic terminal branches. No particular hot spots or longitudinal gradient of excitability was observed (Figure 9). Importantly, previously reported preferential mutational effects (Xing & Wu, 2018) were clearly preserved among distinct types of synaptic terminals, regardless of the longitudinal positions of the boutons, i.e. the striking effects of eag Sh on type Ib (40 Hz) and type Is (2 and 20 Hz) boutons (Figure 9) and bss on type II (2 Hz) and Is (20 Hz) synaptic boutons. The flattening of the distal–proximal gradient in the hyperexcitable mutants may be due to enhanced GCaMP response well above the threshold level, where the distal–proximal gradient was more clearly discerned (compare to Figure 6(A,B)).

Figure 9. Lack of longitudinal gradients of GCaMP signals in motor terminals of hyperexcitable mutants of Na⁺ and K⁺ channels. GCaMP signal amplitudes (max ΔF/F) are plotted against the bouton distal–proximal rank order. (A) Type Ib, 40 Hz. (B) Type Is, 20 Hz. (C) Type Is, 2 Hz. (D) Type II, 2 Hz. Same samples from Figure 5. Note the preferential effects of bss on type II at 2 Hz (D) and on type Is at 20 Hz (B), and the pervasive effects of eag Sh, most strikingly on type Is (even at 2 Hz, C). One-way ANOVA and Tukey’s post hoc tests for boutons in the same longitudinal position were performed (N = 5–12 for each genotype).

Enhanced distal–proximal gradient of GCaMP signal amplitudes and decay kinetics following prolonged inhibition of mitochondrial metabolism

Actions of cytosolic free Ca²⁺ produced GCaMP signals, a dynamic process determined by both Ca²⁺ influx and clearance. A distal–proximal differential in either of these processes can contribute to the GCaMP signal gradient. Clearance of cytosolic Ca²⁺ in synapses is dominated by energy-dependent mechanisms such as extrusion by plasma membrane Ca²⁺-ATPase (PMCA), and sequestration by mitochondria and ER (Chouhan, Zhang, Zinsmaier, & Macleod, 2010; David, Barrett, & Barrett, 1998; Dipolo & Beaugé, 1979; Klose, Boulianne, Robertson, & Atwood, 2009; Nguyen, Marin, & Atwood, 1997; Rusakov, 2006; Shutov, Kim, Houlihan, Medvedeva, & Usachev, 2013). Previous studies demonstrate that prolonged inhibition of mitochondria using DNP to gradually deplete the ATP reserve over tens of minutes can lower the frequency response range of type Ib and Is below 10 Hz, and drastically slow the decay kinetics of the GCaMP signals (Xing & Wu, 2018). Therefore, we ask how mitochondrial inhibition impacts the longitudinal GCaMP signal gradient in type Ib and Is synapses along the motor terminals.

After GCaMP imaging under control condition (HL3.1, 0.1 mM Ca²⁺, stimulated at 10, 20, and 40 Hz), we incubated the preparations with 0.2 mM DNP in the same saline and observed enhanced GCaMP signals (max ΔF/F) became evident at 20 min, revealing a clear distal–proximal ΔF/F gradient in type Ib boutons (Figure 10(C)). Before DNP incubation, type Ib response was mostly too small to clearly resolve a ΔF/F gradient, whereas, a gradient in type Is boutons was already evident without DNP treatment (20 Hz, Figure 10(D)). Further DNP incubation (60 min) saturated the GCaMP response and impaired ΔF/F gradient detection (Figure 10(A–D)).

Figure 10. Effects of mitochondrial inhibition by DNP on GCaMP signal gradients in amplitude and decay kinetics. (A, B) Representative ΔF/F traces of type Ib (40 Hz) and Is (20 Hz) boutons, respectively, with the traces of the bouton pairs from the distal and proximal ends superimposed. (C, D) Comparisons of distal–proximal pairs of max ΔF/F from individual nerve branches, encoded with different symbols. Blue (grey) and black circles represent the pairs of ΔF/F traces displayed above. (E, F) Comparisons of the half-decay time of GCaMP signals in the distal–proximal pairs shown in (C, D). 6 NMJs from six larvae.

The impairment of Ca²⁺ clearance by DNP treatment drastically slowed decay kinetics, and revealed a gradient of another parameter, half-decay time of the GCaMP signals, most evident at 60 min (Figure 10(A,B)). Before the DNP treatment, there was no strong indications of distal–proximal difference in half-decay time. Throughout the DNP incubation, a progressive increase in GCaMP signal half-decay time occurred among the type Ib and Is terminals, slowed to about 10 times by 60 min treatment (Figure 10(E,F)). It became evident that distal boutons displayed slower decay kinetics than proximal boutons, even though the distal–proximal gradient for GCaMP signal amplitude (ΔF/F) became degraded by this time (Figure 11). Therefore, retarding ATP-dependent active Ca²⁺ clearance through inhibition of mitochondrial metabolism had differential impacts on the Ca²⁺ dynamics of distal and proximal boutons. The results also indicate that Ca²⁺ clearance processes contribute to the distal–proximal gradient of GCaMP signals.

Figure 11. Comparison of transmission efficacy in distal–proximal pairs of boutons. Failure rates were assessed to indicate release probability at low stimulus frequency in 0.1 and 0.133 mM Ca²⁺. (A) Representative efEJPs from five consecutive stimuli to indicate transmission failure events, recorded at 1 Hz (total 60 stimuli). Staggered blue (grey) and black traces indicate distal and proximal bouton pairs from muscle 4 type Ib terminals. (B) Release failure rates of the bouton pairs. Genotypes: +/Y; c164-GCaMP1.3 (squares, N = 6, left panel, N = 9, right); OK371-mCD8GFP/+ (circles, N = 6). Paired Student’s t-test, p = .0508 (left) and 0.677 (right).

Electrophysiological correlates of the distal–proximal GCaMP response gradient

We examined the electrophysiological significance of the distal–proximal differences in GCaMP ΔF/F responses by employing focal recording to simultaneously assess the transmission efficacy of different boutons along type I synaptic terminals. In order to confidently monitor the extracellular field EJP (efEJP) from individual boutons at the proximal end, we chose to use type Ib synapses in muscle 4, in which a single isolated type Ib terminal on the inner surface of the muscle fiber is readily accessible to focal recording. Focal recording from type Is synaptic terminals on muscle 4 is more challenging because their locations are too close to the muscle edge. In addition, for the majority of muscle fibers, type Ib and Is motor terminals usually overlap in the proximity of nerve entry point and it is difficult to isolate the responses from particular proximal boutons with certainty.

Our efEJP recording registers field potentials of synaptic events but does not monitor absolute synaptic current amplitudes. However, it could readily detect transmission events to determine whether each of the nerve stimuli led to a ‘success’ or ‘failure’ in transmission. The release of individual vesicles generates quantal miniature EJPs (mEJPs), which are more readily encountered at low Ca²⁺ levels, when the release rate is low and failure rate is more substantial. Therefore, transmission efficacy can be quantified by using failure rates of efEJPs during repetitive stimulation. The efEJP recording was carried out in the range of 0.1–0.133 mM Ca²⁺ at 1 Hz. Under this condition, the stochastic quantal release will have a failure probability over a larger range to facilitate the detection of more subtle differences. Higher Ca²⁺ concentration tends to produce nearly 100% success, whereas Ca²⁺ concentration lower than 0.1 mM would have too few transmission events to quantify. Both are undesirable for our purpose.

At low Ca²⁺ levels, 1-Hz stimulation did not lead to detectable GCaMP ΔF/F signals (cf. Xing & Wu, 2018). Even though on average the combined transmission failure of distal boutons were not significantly lower than proximal boutons, the pair-wise distal–proximal readouts suggested a weak trend of transmission efficacy gradient at 0.1 mM Ca²⁺ (four out of six from +/Y; c164-GCaMP1.3, and six out of six from OK371-GFP/+, Figure 11 left panels), and at 0.133 mM Ca²⁺ (six out of eight Figure 11). The relatively small differences in failure rates (about 10% but no larger than 20%) found in these distal–proximal pairs indicate only subtle gradients in transmission efficacy at low Ca²⁺ concentrations examined here.

To correlate physiological recording with detectable GCaMP signals, we increased the stimulus frequency to 40 Hz, just above the threshold level for type Ib bouton GCaMP signals. Under this condition, clearly detectable GCaMP signals (e.g. top pair of traces in Figure 12(A)) were obtained in five of the nine type Ib terminals, which displayed strong distal–proximal differences in GCaMP signal amplitude (Figure 12(A,C), squares). Nevertheless, simultaneous focal recording from the same distal–proximal bouton pairs on these terminal branches showed no clear distal–proximal gradient in transmission failure (Figure 12(B,D), squares). In addition, no prominent distal–proximal trend of failure rates was observed in the remaining 4 NMJs that lacked GCaMP signals (circles, below the noise level, ΔF/F < 0.05, Figure 12).

Figure 12. Transmission efficacy in distal and proximal boutons assessed at high frequency stimulation and correlated with simultaneous GCaMP signals. Two-second trains of 80 stimuli were delivered at 40 Hz in 0.1 mM Ca²⁺ saline. (A) GCaMP ΔF/F signals from four representative pairs of distal and proximal boutons. The first five and last five stimulus pulses are indicated using darker stimulus bar. (B) The initial 20 efEJPs to indicate the failure events, from the bouton pairs shown in (A). (C) ΔF/F amplitudes of the distal and proximal bouton pairs, N = 9. (D) Failure rates of synaptic transmission for the same paired distal–proximal boutons, measured from the 2-s, 40 Hz trains of efEJPs. The square symbols indicate the boutons with clearly detectable GCaMP signals (C). The circles indicate the boutons displaying no clear GCaMP signals (e.g. the bottom two traces in panel A). Low ΔF/F signals (circles, C) correspond to higher transmission failure rates (circles, D). This color and shape scheme of data symbols is shared with the other panels. (E) Failure rate vs. max ΔF/F for boutons recorded in panels C and D. Cut-off level 0.05 indicated by the broken line. Note the drastic drop of failure rates above 0.05. (F, G) Failure rates of the efEJPs at the beginning, and the end of the 2-s, 40 Hz trains, measured from the first 5 (1–5) and the last 5 (76–80) stimuli, recorded from distal (F) and proximal (G) boutons. Each pair represents data measured from the efEJP traces of one bouton.

In general, the terminal branches that displayed clear GCaMP signals tended to have lower failure rates in focal recording (compare Figure 12(C,D), squares vs. circles). We plotted the amplitude of GCaMP signals (maximum ΔF/F), vs. the efEJP failure rates (Figure 12(E), based on the responses to the 2 s, 40 Hz stimulus trains) from the same set of boutons. There was an apparent cut-off level of maximum ΔF/F at 0.05, above which a drastic increase in release probability (decrease in failure rate) was observed. Below this level, the GCaMP signals did not stand out above the noise level (e.g. bottom two traces of Figure 12(A)). We further separately measured the failure rates of the initial 5, and the last 5 in each train of efEJPs for indications of activity-dependent plasticity, such as facilitation. The last 5 efEJPs usually correlated with the peak of GCaMP ΔF/F signals, whereas the initial 5 efEJPs occurred in the period before GCaMP signal had any significant increase (Figure 12(A), note that the initial and the last 5 efEJPs are indicated by darker stimulus bar). By plotting these failure rates in pairs, it was clear that the last 5 efEJPs had much lower failure rates vs. the initial 5 efEJPs in both distal and proximal boutons (Figure 12(F,G)), consistent with well-established synaptic facilitation phenomenon. The above comparisons suggest that the larger GCaMP signals above the cut-off level tend to display more robust release probability determined by efEJP recording (Figure 12(E)), also coupled with a more striking trend of facilitation in transmission within each stimulus train (Figure 12(F,G), compare the pairs of squares vs. circles). This also supports our previous conclusion that GCaMP signals reflect the dynamics of cytosolic residual Ca²⁺ (Xing & Wu, 2018), which participates in facilitation (Zucker & Regehr, 2002). However, no striking differences in the extent of facilitation was observed between distal and proximal boutons (Figure 12(F,G)).

Discussion

Among different cell types, neurons exhibit the most complex cellular morphology, with many specialized subcellular compartments, including presynaptic boutons and postsynaptic spines in different parts of soma and along axons and dendrites. Ca²⁺ dynamics are crucial in the regulation of synaptic development, function, and plasticity (Dason et al., 2009; Lee, Ueda, & Wu, 2014; Lohmann & Bonhoeffer, 2008; Rusakov, 2006; Zucker, 1996; Zucker & Regehr, 2002). However, these structures of micrometer scale are highly variable and plastic in their size and location. Furthermore, synapses of different categories of neurons in the nervous systems, e.g. ionotropic and metabotropic, are distinct in morphology and distribution. It is therefore important to determine whether and how variation in geometric factors, such as physical size and location of synapses, sets constraints on local Ca²⁺ dynamics, and interferes with the readout of Ca²⁺ indicators in the investigation of intrinsic regulation mechanisms in functionally distinct synapses.

Inter-relationships among bouton size, baseline GCaMP fluorescence, and GCaMP signal amplitude

Although a plethora of Drosophila GCaMP imaging studies has been published on CNS neuronal activity and peripheral synaptic function (Akerboom et al., 2012; Chen et al., 2013; Reiff et al., 2005; Tian et al., 2009; Wang, Wong, Flores, Vosshall, & Axel, 2003; Wang et al., 2004), the inter-relationships between synaptic parameters, such as bouton size, location, GCaMP baseline fluorescence F, and GCaMP signal amplitude ΔF/F, have not been fully established (cf. earlier work on type Ib boutons, Guerrero et al., 2005; Lnenicka et al., 2006).

The Drosophila NMJ provides a unique opportunity to contrast ionotropic synapses (type Ib and Is) with metabotropic synapses (type II) in close proximity (Johansen et al., 1989; Monastirioti et al., 1995) for simultaneous imaging within the same microscopic field. Previous results show that type Ib, Is and II synapses manifest distinct frequency responses of GCaMP Ca²⁺ signals controlled by different ion channels and clearance mechanisms (Xing & Wu, 2018). In the present study, correlation analysis of large samples of boutons clarifies that neither bouton size nor GCaMP baseline fluorescence represents a confounding factor, when determining the intrinsic distinctions of frequency-dependent responses among type Ib, Is and II synapses.

In principle, bouton size variation can be a contributing factor in determining the dynamics of GCaMP Ca²⁺ membrane properties, larger boutons should have a slower rate in accumulating cytosolic Ca²⁺ to produce detectable GCaMP signals. Thus, it can be argued that the size differences could be a possible explanation for the characteristic higher frequency response for type Ib bouton in contrast to the lower frequency ranges for type Is and even lower for type II synapses (cf. Figure 1 and Xing & Wu, 2018).

In fact, the bouton size effect has been demonstrated in type Ib boutons to exist only for single-stimulus evoked Ca²⁺ transients, but not for plateau levels of Ca²⁺ signals evoked by trains of repetitive stimuli, using the fast indicator OGB-1 (Lnenicka et al., 2006). Theoretically, the rise of Ca²⁺ transients to a steady plateau level in response to a prolonged stimulus train is a process involving both Ca²⁺ influx and clearance, kinetically distinct from the measured amplitude of a single-stimulus evoked Ca²⁺ signal (Lnenicka et al., 2006; Pérez Koldenkova & Nagai, 2013; Tank, Regehr, & Delaney, 1995). In our case, GCaMP signals are generally slow and do not resolve Ca²⁺ transients evoked by single action potentials. Instead, the plateaus of GCaMP signals report integration of Ca²⁺ influx and clearance over repetitive stimulus responses (Xing & Wu, 2018). In this study, our results demonstrate that the large overlaps of bouton sizes among the three synaptic types did not obscure their distinct frequency-dependent GCaMP signal characteristics. Therefore, physical dimension of synaptic boutons is not a practical predictor for the dynamics of GCaMP signals for boutons either within or between synaptic categories (Figure 3). Furthermore, it cannot account for the striking preferential effects of hyperexcitability mutations on different synaptic categories (Figure 5, see also Xing & Wu, 2018). Instead, intrinsic mechanisms such as membrane excitability and Ca²⁺ clearance capacity play far more important roles in the regulation of presynaptic Ca²⁺ dynamics.

However, our data indeed confirm that GCaMP baseline fluorescence is significantly correlated with bouton size (Figure 4), as expected from their longer optical path in which GCaMP indicators interact with excitation light. Nonetheless, neither of these two absolute measures (in µm² and mV) plays a significant role in determining GCaMP signal amplitudes, which are normalized, unit-less quantities (ΔF/F). A direct measurement of bouton width may be an alternative method to estimate bouton thickness (Lnenicka et al., 2006) and correlate with bouton fluorescence intensity, which can be investigated in further studies.

Longitudinal gradient of presynaptic GCaMP Ca²⁺ signal amplitude

Earlier Ca²⁺ imaging studies based on different Ca²⁺ indicators have led to somewhat different pictures on the distal–proximal gradient along synaptic terminals in the Drosophila NMJ. A relatively weak distal–proximal gradient of presynaptic Ca²⁺ dynamics has been detected by back-filling type Ib and Is synaptic terminals with the synthetic indicator Oregon Green BAPTA-1 (OGB-1, He et al., 2009; Lnenicka et al., 2006) in muscles 4, 6 and 7 in high Ca²⁺ saline (1 mM). A different study using presynaptic expression of genetically encoded Ca²⁺ indicators (Cameleon2.3) has also demonstrated a presynaptic Ca²⁺ gradient of similar magnitude (Guerrero et al., 2005).

Postsynaptic expression of Ca²⁺ indicators, including GCaMP and derivatives of Cameleon, has revealed a stronger distal–proximal gradient of nerve-evoked optical signal representing Ca²⁺ influx through postsynaptic glutamate receptors along type Ib synaptic terminals (1.5 mM Ca²⁺, Guerrero et al., 2005; Peled & Isacoff, 2011). However, a more recent study reports relative uniform distribution of evoked postsynaptic myrGCaMP5 signals (1.5 mM Ca²⁺, Melom, Akbergenova, Gavornik, & Littleton, 2013). It is not known whether the differences in the indicator types and their cellular localization could account for the discrepancy between these observations.

Our work indicates that several factors and conditions may be manipulated to either enhance or obscure the detection of a gradient of GCaMP signals along the synaptic terminals. These include external Ca²⁺ concentration, stimulus frequency, and larval genotype. The initial choice to use low Ca²⁺ saline in our earlier GCaMP imaging studies was to optimize the striking effects of hyperexcitable ion channel mutations and to avoid muscle contraction without relying on the usage of glutamate for postsynaptic receptor desensitization (Xing & Wu, 2018). Fortuitously, we found that low Ca²⁺ saline could better reveal the distal–proximal gradients (a proximal/distal ratio about 30%, Figures 6, 8 and 12(C)), in contrast to the previously reported gradient (type Ib: in the range of 60–80%, Guerrero et al., 2005; Lnenicka et al., 2006; type Is: 80–90%, He et al., 2009) at high Ca²⁺ concentrations (1–1.5 mM). To sum up, the optimal detection of the distal–proximal gradient falls in the range just above the threshold of effective stimulus frequency (20 Hz for Is, 40–80 Hz for Ib) for GCaMP imaging experiments at 0.1 mM Ca²⁺.

For both type Is and type Ib, longitudinal ΔF/F gradients could be demonstrated corresponding to the bouton rank order (Figure 6), or their physical distance from the distal end of the terminals (Figure 8). It should be noted, however, that there were no corresponding longitudinal gradients in bouton sizes or GCaMP baseline fluorescence (Figure 7).

This gradient was disrupted by hyperexcitable mutations (Figure 9), or DNP incubation (Figure 10). Interestingly, after prolonged DNP treatment, which saturated the ΔF/F response and obscured its gradient, a novel gradient in decay kinetics of the GCaMP signals was revealed in both type Ib and Is, i.e. slower half-decay time in distal boutons (Figure 10), suggesting a possible role of mitochondria in gradient formation.

To investigate the functional significance of the GCaMP signal gradient, we took advantage of the focal recording of efEJP events from distal and proximal boutons to correlate with the optical imaging results from type Ib boutons in muscle 4 (see Figure 12). With this approach, the failure rate of transmission could be readily quantified at low Ca²⁺ levels. There were indications for a relatively weak distal–proximal difference in release probability with repetitive 1-Hz stimulation (Figure 11). However, to assess the GCaMP response amplitude, a higher simulation frequency was applied (40 Hz), at which a clear distal–proximal gradient of GCaMP signals could be observed in all terminals that displayed ΔF/F above the noise level (0.1 mM Ca²⁺, Figure 12). In this case, we found that GCaMP ΔF/F signals beyond the cut-off level (0.05) were correlated with drastically higher release probability (Figure 12(E)). A clear trend of facilitation was observed during the 2 s stimulus train in both distal and proximal boutons (Figure 12(F,G)). However, we did not observe any striking distal–proximal difference of facilitation properties under this condition (Figure 12(F,G)).

It should be stressed that GCaMP signals reflect cytosolic residual Ca²⁺ accumulation, which takes place in a time scale of hundreds of milliseconds to a few seconds, and involves both Ca²⁺ influx and clearance mechanisms. In contrast, Ca²⁺ entry and subsequent vesicle fusion and transmitter release occur in milliseconds. Therefore, depending on the stimulus protocols and local physiological conditions, the two measurements may yield rather different readouts that reflect the states of two steps in the chain of cellular Ca²⁺ dynamics, from influx, local actions, cytoplasmic accumulation and clearance (cf. Xing & Wu, 2018). Thus, in type Ib synaptic boutons at low Ca²⁺ condition, the distal–proximal gradient of GCaMP signal gradient does not necessarily imply a similar gradient in transmission strength.

Since GCaMP signals reflect the dynamics of cytosolic residual Ca²⁺ accumulation, more investigations are needed to fully understand its relationship with short-term synaptic plasticity properties. Furthermore, a more definitive assessment of any gradient in transmitter release, and transmission efficacy of individual boutons along the synaptic terminal requires direct quantification of local synaptic currents at various Ca²⁺ levels under voltage-clamp conditions. This awaits future studies employing the focal loose-patch clamp, first pioneered in this preparation by Prof. Harold Atwood and associates (Kurdyak et al., 1994; Millar, Zucker, Ellis-Davies, Charlton, & Atwood, 2005; Renger, Ueda, Atwood, Govind, & Wu, 2000; Ueda & Wu, 2009, 2012).

Acknowledgements

We thank Drs. Yalin Wang and Yi Zhong for providing GCaMP fly stocks. We thank Mr. Timothy Patience and Dr. Yu Li for help in proofreading.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by Department of Health and Human Services National Institutes of Health Grants GM88804, AG047612, and AG051513.