GABAergic Inhibition of Presynaptic Ca2+ Transients in Respiratory PreBötzinger Neurons in Organotypic Slice Cultures

Abstract GABAergic somatodendritic inhibition in the preBötzinger complex (preBötC), a medullary site for the generation of inspiratory rhythm, is involved in respiratory rhythmogenesis and patterning. Nevertheless, whether GABA acts distally on presynaptic terminals, evoking presynaptic inhibition is unknown. Here, we begin to address this problem by measuring presynaptic Ca2+ transients in preBötC neurons, under rhythmic and non-rhythmic conditions, with two variants of genetically encoded Ca2+ indicators (GECIs). Organotypic slice cultures from newborn mice, containing the preBötC, were drop-transduced with jGCaMP7s, or injected with jGCaMP7f-labeling commissural preBötC neurons. Then, Ca2+ imaging combined with whole-cell patch-clamp or field stimulation was obtained from inspiratory preBötC neurons. We found that rhythmically active neurons expressed synchronized Ca2+ transients in soma, proximal and distal dendritic regions, and punctate synapse-like structures. Expansion microscopy revealed morphologic characteristics of bona fide synaptic boutons of the en passant and terminal type. Under non-rhythmic conditions, we found that bath application of the GABAA receptor agonist muscimol, and local microiontophoresis of GABA, reduced action potential (AP)-evoked and field stimulus-evoked Ca2+ transients in presynaptic terminals in inspiratory neurons and commissural neurons projecting to the contralateral preBötC. In addition, under rhythmic conditions, network rhythmic activity was suppressed by muscimol, while the GABAA receptor antagonist bicuculline completely re-activated spontaneous activity. These observations demonstrate that the preBötC includes neurons that show GABAergic inhibition of presynaptic Ca2+ transients, and presynaptic inhibition may play a role in the network activity that underlies breathing.


Introduction
Synaptic inhibition plays an important role in most rhythmic networks, including the brainstem network that generates and pattern respiratory rhythm (Abdala et al., 2015;Anderson and Ramirez, 2017;Ikeda et al., 2017;Del Negro et al., 2018). The ventrolateral medulla contains physiologically and neurochemically identified groups of respiratory neurons in the ventral respiratory column (VRC), which comprise subsets of GABAergic and glycinergic neurons (Smith et al., 2009;Morgado-Valle et al., 2010;Koizumi et al., 2013). An estimated 50% of inspiratory neurons in the preBötzinger complex (preBötC), an essential site for inspiratory rhythm generation, contains GABA or glycine (Kuwana et al., 2006;Winter et al., 2009;Morgado-Valle et al., 2010;Baertsch et al., 2018), and some may be co-transmitting GABA and glycine (Hirrlinger et al., 2019). Expiratory neurons in the adjacent Bötzinger complex use GABAergic (Ellenberger, 1999) and glycinergic (Ezure et al., 2003) transmission to inhibit spinal and medullary neurons involved in respiratory pattern generation (Ausborn et al., 2018) and active expiration (Flor et al., 2020). Pharmacological blocking and optogenetic experiments suggest that synaptic inhibition is not obligatory in the network generation of inspiratory rhythm (Shao and Feldman, 1997;Janczewski et al., 2013) but has phase-dependent effects on the synchronization of preBötC neurons (Ashhad and Feldman, 2020), respiratory frequency and amplitude (Sherman et al., 2015;Cregg et al., 2017;Baertsch et al., 2018). Several in vivo and in vitro experiments point to a crucial role of synaptic inhibition in generating the threephase rhythm of eupnea (Richter and Smith, 2014;Marchenko et al., 2016), with phasic inhibition ensuring a stable rhythm generation, and integration of respiration with other ongoing motor programs. These effects of synaptic inhibition on respiratory rhythm generation and patterning are generally assumed to be because of somatodendritic inhibitory effects of GABAergic and glycinergic transmission (Liu et al., 2002). However, no studies have approached the question of whether GABA or glycine may affect presynaptic terminals directly, as is the case in presynaptic inhibition used in several other networks in the CNS, retina, and spinal cord (Kullmann et al., 2005). Here, we use organotypic slice cultures of the brainstem to investigate whether presynaptic GABAergic inhibition exists in inspiratory neurons in the preBötC, measuring the effect of GABA A receptor ligands on presynaptic Ca 21 transients.
Genetically encoded Ca 21 indicators (GECIs) have improved the measurement of Ca 21 in functioning neurons under in vitro and in vivo conditions (Mollinedo-Gajate et al., 2019). In particular, recent GECI variants show enhanced labeling and signal-to-noise ratio in neuronal populations and microcompartments (Dana et al., 2019). Thus, jGCaMP7 variants with slow and fast kinetics (jGCaMP7s and jGCaMP7f, respectively), label small neuronal processes, including presynaptic boutons and dendritic spines, and have been employed to report on Ca 21 fluctuations in intact networks in the fly and mouse brain. The use of GECIs has also benefited from improved viral vectors that can deliver the genetic material to neurons with specificity . In particular, adeno-associated virus (AAV) vectors have proved advantageous in effectively transducing neurons and labeling projection neurons retrogradely using the AAV retrograde serotype (Tervo et al., 2016).
Here, we use two jGCaMP7 GECI variants expressed in organotypic slice cultures from mice containing the preBötC, via AAV9 and AAV retrograde transductions, to investigate the effect of GABA A receptor ligands on presynaptic Ca 21 transients in inspiratory neurons. We show that, under non-rhythmic conditions, bath applied muscimol, a GABA A receptor agonist, and microiontophoretically applied GABA, reduce action potential (AP)-evoked and field stimulus-evoked presynaptic Ca 21 in inspiratory preBötC neurons, including commissural neurons projecting to the contralateral preBötC. This effect on presynaptic Ca 21 demonstrates that synaptic inhibition in the breathing oscillator may involve GABAergic presynaptic inhibition.
Field stimulation was performed with a sharp bipolar tungsten electrode, 1/À poles spaced ;400 mm apart, placed on the surface of the culture on one side encompassing a rhythmic preBötC. Trains of unipolar pulses (2ms duration, 4-to 9-V constant voltage) were used to stimulate fibers and somas between the two poles, using a stimulus isolation unit gated by a waveform generator (A.M.P.I. ISO-Flex isolator, RRID:SCR_018945, AMPI Master 8 generator, RRID:SCR_018889).

Experimental design and statistical analysis
Time series of field stimulus-evoked or AP-evoked Ca 21 transients were acquired in epochs of 20-50 frames at 10 frames/s, evoked every 15 s. Synaptic Ca 21 transients were calculated as the percent change in fluorescence relative to baseline values (DF/F 0 ). However, it is important to note that fluorescence from a whole-field imaged synapse, in AAV-GECI transduced cultures, is a combination of fluorescence from the synapse, fluorescence from labeled neuronal structures in the foreground and background, and reflected fluorescence from nearby neuronal structures. To correct for these added sources of fluorescence, the background was subtracted frame-byframe, taken as the equal-sized mean background fluorescence immediately adjacent to the synaptic region of interest (ROI). Since the background fluorescence could be higher or lower than synapse fluorescence in any given experiment, because of GECI labeling of nearby neuronal structures, the starting level of the background signal was set at the starting level of the synapse fluorescence before subtraction. In essence, this background correction algorithm subtracts the curve of nearby fluorescence from the curve of fluorescence over a synapse. Synaptic ROIs were drawn by hand as rectangles of 5 Â 5 mm for 20Â magnification and 3 Â 3 mm for 40Â magnification. The mean fluorescence over three initial frames in the synaptic ROI, before stimulus, was defined as F 0 . For each condition, 10-20 sweeps were acquired and averaged before calculation of DF/F 0 . Spontaneous rhythmic Ca 21 transients over the preBötC were acquired at 5-10 frames/s, and DF/F 0 was calculated by setting F 0 as the minimal fluorescent value over the time series (5-20 s). Optical data were analyzed offline using ImageJ 1.53d (ImageJ, RRID:SCR_003070), and Igor Pro 8 (IGOR Pro, RRID:SCR_ 000325). Electrophysiological data were acquired using pClamp 10.3 (pClamp, RRID:SCR_011323) and subsequently analyzed using custom scripts written in Igor Pro 8 (IGOR Pro, RRID:SCR_000325).
All statistical tests were performed using GraphPad Prism 8 (RRID:SCR_002798). Data were verified for normal distribution with Kolmogorov-Smirnov test, and then analyzed using one-way ANOVA with Tukey's post hoc test, unpaired or paired Student's t test as appropriate. Results are reported as mean 6 SEM, and as N = number of cultures, and n = number of presynaptic terminals. Statistical significance was set at p , 0.05 based on the culture number for system and neuronal activity data.

Axonal arborizations and Ca 21 imaging from presumed presynaptic terminals
Considering that preBötC neurons have extensive projections to different brainstem regions implicated in the control of breathing (Tan et al., 2010), we next performed whole-cell patch-clamp recordings with biocytin-filled pipettes from rhythmically active preBötC neurons to visualize axonal projection patterns within the plane of single organotypic slice cultures. In a sample of 25 labeled neurons, 16% showed ipsilateral local projections in the surrounding preBötC, 36% projected to the contralateral preBötC region crossing the midline in the ventral third of the slice culture, 20% projected to the ipsilateral dorsal regions, 24% projected to the midline, and a single neuron (4%) had projections to multiple of the above-mentioned areas (Fig. 2). Importantly, high-magnification imaging of the terminal projection regions showed synapse-like structures associated with axons in all neurons analyzed (Fig. 2, insets). Rhythmically active preBötC neurons showed trains of spikes riding on top of summating synaptic potentials in synchrony with Ca 21 transients from the soma and surrounding preBötC (Fig. 3A). To corroborate the identity of the synapse-like structures in rhythmically active preBötC neurons labeled with GECIs, the rhythm was stopped by exchanging high-aCSF to low-aCSF. Neurons were then activated by trains of brief depolarizing current pulses (10 ms) evoking 20 APs. This resulted in Ca 21 transients in somatodendritic compartments and in discrete synapse-like structures near and far from the neuron (N = 4; Fig. 3B,C). Moreover, the same presynaptic terminals could be identified after processing for biocytin using streptavidin-Alexa Fluor 594, and importantly following further processing using expansion microscopy on the same neurons (N = 4 reconstructed and expanded neurons; Fig. 3D). The expanded neurons displayed morphologic characteristics of bona fide synaptic boutons of the en passant and terminal type (Fig. 3E). When correcting for an expansion factor of 3.5, the maximal diameter of the synaptic boutons was 1.5 6 0.6 mm (n = 83, N = 4 neurons). Together, these data suggest that jGCaMP7 sensors may be a useful tool to detect Ca 21 activity at the level of individual presynaptic terminals with high reliability.

AP-dependent and stimulus-dependent changes in presynaptic Ca 21 transients
To characterize presynaptic Ca 21 transients in rhythmically active preBötC neurons two labeling paradigms were used. First, entire slice cultures were jGCaMP7s drop-transduced. After 14-35 d, simultaneous whole-cell patch-clamp and Ca 21 imaging were performed on rhythmically active preBötC neurons (Fig. 4A, top diagram). Network rhythmic activity was reduced by hyperpolarizing neuronal membrane potentials and increasing the threshold for Na 1 channel activation with low-aCSF (Phillips et al., 2018). APs were then evoked by trains of brief intracellular current pulses (10-ms duration) with an increasing number of pulses (1-60 with a set time interval of 10 ms; Fig. 4; Movie 2). This protocol resulted in a near-linear rise in the amplitude of presynaptic Ca 21 transients (single AP gave ;0.4% DF/F), reaching a saturation maximum above 30 pulses (n = 30, N = 3 cultures; Fig. 4B).
In the second labeling paradigm, cultures were jGCaMP7f unilateral-injected in the ventrolateral preBötC area (Fig. 5A).
GABA B receptors are well-established modulators of presynaptic terminals in nearly all areas of the brain, including the preBötC (Bongianni et al., 2010;Ghali, 2019). Thus, the effect of the GABA B agonist baclofen (10 mM) on presynaptic Ca 21 transients in jGCaMP7f preBötC-injected preBötC neurons was further explored. No significant differences in field-evoked presynaptic Ca 21 were found in baclofen treated preBötC neurons compared with controls (Ctr, 3.5 6 0.4% DF/F, n = 28, N = 3; baclofen, 3.5 6 0.3% DF/F, n = 28, N = 3, paired Student's t  Figure 5. Stimulus-dependent Ca 21 transients and Ca 21 signal stability in presynaptic terminals in field stimulated preBötC neurons. A, top, Diagram of the experimental paradigm, with a unilateral injection of jGCaMP7f in the ventrolateral area containing the preBötC (PBC), followed by field stimulation in the contralateral preBötC area (Field. stim.) after 14-35 d. The contralateral preBötC region with rhythmic activity was identified in high-aCSF, followed by field stimulation in low-aCSF plus NBQX (20 mM) and CPP (20 mM) to inhibit glutamate-driven group activity. Left images, Averaged Ca 21 transients in soma and presynaptic terminals at low and high magnification (insets) in response to field stimulation with 5-and 10-pulse trains (stimulus-triggered average, color inverted, smart filter, identical contrast settings). Right diagrams, Averaged presynaptic Ca 21 transients with 5-and 10-pulse trains (n = 60 presynaptic terminals, N = 6 cultures). The graph shows number of pulses in the pulse train versus peak amplitude of the Ca 21 transient in single presynaptic terminals (n = 60 presynaptic terminals, N = 6 cultures, line is a spline fit). B, Stability of the Ca 21 signal in presynaptic terminals over 0, 15, and 30 min (n = 39, 39, and 29, respectively, N = 4 cultures) in a culture with a unilateral injection of jGCaMP7f, and contralateral preBötC field stimulation with 10 pulses (left images) and group data (right diagrams). Note that the Ca 21 transient amplitude does not change significantly over 30 min in this experimental paradigm. test, t (27) = 0.46, p = 0.65). Thus, a possible involvement of GABA B receptors on the presynaptic Ca 21 transients was unlikely to have contributed to the presynaptic inhibition observed here.

Discussion
Recent improvements in GECIs allow accurate measurements of Ca 21 in neuronal cell bodies and dendrites with high spatial and temporal resolution. The present study demonstrates that it is possible to target novel GECIs, such as jGCaMP7s and jGCaMP7f, to individual presynaptic terminals in inspiratory neurons in organotypic slice cultures containing the preBötC with high reliability. This novel approach, combined with whole-cell patchclamp and field stimulation, provides new evidence that GABAergic presynaptic inhibition exists in the neurons comprising the breathing oscillator. This evidence includes reduced presynaptic Ca 21 transients in inspiratory neurons and commissural neurons projecting to the contralateral preBötC, in response to bath applied muscimol and microiontophoresis of GABA under non-rhythmic conditions. These observations indicate a role of GABAergic presynaptic inhibition in rhythmic breathing activity.
The use of small-molecule fluorescent Ca 21 indicator dyes (Tsien, 1981;Tsien et al., 1982) revolutionized neuronal Ca 21 imaging since these indicators are easily loaded into the intracellular space by membrane-permeant ester (acetoxymethyl, AM) forms (Tsien, 1981;Tsien et al., 1982;Grynkiewicz et al., 1985), or through a pipette containing cell-impermeant salt forms (Tank et al., 1988;Müller and Connor, 1991). Loading distal subcellular compartments, such as presynaptic terminals is difficult but has been achieved in some systems using dextran-conjugated indicators (Brenowitz and Regehr, 2012). However, recent advances in the design of GECIs may provide a solution to the challenge of recording Ca 21 transients in small subcellular compartments. Thus, earlier variants of GECIs (GCaMP6 fused to synaptophysin) can report on presynaptic Ca 21 in neurons in primary culture (Brockhaus et al., 2019;Ferron et al., 2020), and in the calyx of Held synapse following viral injection of GCaMP6 variants three weeks prior (Singh et al., 2018). Biolistic transfection, electroporation, or AAV transduction of hippocampal, hypothalamic, and cortical organotypic slice cultures with plasmids encoding GCaMP3/6 variants or RFPbased GECIs, result in strong somatodendritic labeling, but no apparent presynaptic labeling (Podor et al., 2015;Shen et al., 2018;Gasterstädt et al., 2020;Kim et al., 2020). Interestingly, jGCaMP7 variants show labeling of presynaptic boutons of the Drosophila larval neuromuscular junction (Dana et al., 2019). This prompted us to combine these latest jGCaMP7 variants with AAVbased expression in organotypic slice cultures to enable direct visualization of presynaptic Ca 21 in preBötC neurons. Indeed, we find that transduction with AAV9 and AAV-retro encoding jGCaMP7s and jGCaMP7f, respectively, driven by the Syn promoter, results in strong labeling of soma, dendrites, and notably in synapse-like structures that were identified as synaptic boutons of the en passant and terminal type using expansion microscopy. Morphologic reconstruction of biocytin-labeled preBötC neurons in our brainstem cultures shows that the projection pattern of these neurons, and synaptic terminal fields, match that of in vivo-labeled neurons at the rostrocaudal level of the preBötC (Yang and Feldman, 2018  E, F, left images, Averaged Ca 21 transients in presynaptic terminals at low and high (insets) magnification in response to field stimulation in presence or absence of (E) muscimol or (F) bicuculline (10-and 5-pulse trains, respectively, stimulus-triggered average, color inverted, smart filter, identical contrast settings). Right upper, Averaged presynaptic Ca 21 transients with 5-and 10-pulse trains. Right bottom, Summary scatter graphs show peak amplitudes of presynaptic Ca 21 transients before and after (E) muscimol and (F) bicuculline application. E, n = 69 presynaptic terminals, N = 7 cultures. F, n = 49 presynaptic terminals, N = 5 cultures. Paired Student's t test; *p , 0.05, **p , 0.01.
comprising the central pattern generator (CPG) network, and has several advantages in the current embodiment, but also some disadvantages. Here, using conventional widefield illumination microscopy, labeled synapses were only readily visible in the top ;100-mm layer of the cultures, because of diffraction of light coming from deeper structures. However, we surmise that the use of confocal and multiphoton imaging may resolve presynaptic terminals located deeper. Transducing organotypic cultures from Cre transgenic animals with AAVs encoding for Cre-dependent GECI variants may allow for visualization of presynaptic Ca 21 in genetically specified neurons. The GECIs used here are non-ratiometric Ca 21 indicators, which represents a disadvantage, since bleaching, concentration differences, and optical path problems make quantification difficult. Careful control experiments and corrective calculations allowed for comparisons of drug effects performed over ;15-min experimental windows, but this time window was also constrained by small movements of the tissue. Time-lapse imaging (data not shown) reveals mobile cells, presumed to be microglia, push their way through the cultures, which combined with small flow-dependent movement of the membrane-attached cultures, and slight thermal movement of the optical system induced a drifting focus in some experiment. This could be corrected for, if it was in the x-y-plane, but invalidated quantification if it was in the z-plane.
Breathing movements are very complex, and need to change constantly during, e.g., exercise, coughing, swallowing, and vocalizing, fine-tuning respiratory frequency and depth. These state-dependent changes require inhibitory neural mechanisms that can temporarily stop or modulate the activity of premotor neurons driving respiratory motor pools. However, the role of synaptic inhibition in the breathing CPG is controversial, since pharmacological experiments, applying GABAergic and glycinergic antagonists to the VRC in rodents, have given conflicting results (Paton and Richter, 1995;Shao and Feldman, 1997;Pierrefiche et al., 1998;Bongianni et al., 2010;Janczewski et al., 2013;Marchenko et al., 2016). Some studies show a critical role of synaptic inhibition in rhythm generation (Schmid et al., 1996;Pierrefiche et al., 1998;Marchenko et al., 2016), whereas others show that rhythm can persist following blockade of synaptic inhibition in the preBötC and BötC (Janczewski et al., 2013). Nonetheless, the available data shows that synaptic inhibition plays a role in shaping the temporal sequence and amplitude of the neural outputs from the respiratory controller to motor pools driving respiratory muscles. An important insight into the role of synaptic inhibition in the preBötC comes from optogenetic experiments selectively activating inhibitory preBötC neurons (Vgat1 putative GABAergic and glycinergic neurons), demonstrating that phasic inhibition is critical for maintaining a normal rapid preBötC rhythm (Cregg et al., 2017;Baertsch et al., 2018). The mechanism that mediates this effect may involve curtailing the refractory period following inspiratory bursts, allowing rebound excitation to activate subsequent bursts. Activity of GABAergic neurons also appears to regulate synchronization of rhythmogenic preBötC neurons by changing the conductance state of the neurons (Ashhad and Feldman, 2020). GABAergic and glycinergic ionotropic receptor mechanisms localized at the somatodendritic membranes might mediate these effects by hyperpolarizing and shunting the membrane, and thereby reduce the effect of excitatory input and affect the balance of inhibition and excitation in the entire network of glutamatergic, GABAergic, and glycinergic neurons. Under low-excitability conditions, GABA A receptor blockade can restart spontaneous preBötC rhythm in acutely prepared brainstem slices (Ashhad and Feldman, 2020). Here, we find that muscimol blocks rhythm, and bicuculline, in Right, Voltage response to depolarizing current steps in preBötC neurons, previously rhythmically active in high-aCSF, before (dark line) and after GABA application (red line). B, Representative images of averaged Ca 21 transients in presynaptic terminals and soma, belonging to the neuron in focus, at low and high (insets) magnification during 20 APs (stimulus-triggered average, color inverted, smart filter, identical contrast settings). (1) Dendrites proximal to the GABA-pipette (solid black arrow: no GABA; solid red arrow: GABA application); presynaptic terminals (2) and soma (3) distal to the GABA-pipette. C, upper, Averaged presynaptic and somatic Ca 21 transients with 20 Aps. Bottom, Summary scatter graphs show peak amplitudes of synaptic and somatic Ca 21 transients before and after GABA application; n = 35 presynaptic terminals, 5 somas. Paired Student's t test; *p , 0.05. low-aCSF, can restart rhythm in organotypic cultures containing the preBötC, suggesting indeed that GABAergic inhibition might play a role in rhythmogenesis. However, when excitability was lowered with low-aCSF, and glutamatergic synaptic transmission was blocked with NBQX and CPP, there was no effect of GABA A receptor blockade on presynaptic Ca 21 transients. This implies that baseline release of GABA under these conditions is too low to induce presynaptic inhibition, and that the GABAergic presynaptic inhibition requires a rhythmic network, or neurons with heightened excitability. When GABA was applied locally to presynaptic terminals, a small decrease was also noted in the somatic Ca 21 transient amplitude. This is surprising, and could result from diffusion of GABA to nearby dendrites since separate experiments showed that dendritically applied GABA reduces somatic Ca 21 transient amplitudes. Together, these observations suggest that GABA acting on the somatodendritic membrane may shunt, and profoundly change the electrotonic compactness of preBötC neurons, affecting AP-induced Ca 21 influx. The GABAergic inhibition of presynaptic Ca 21 transients discovered here in inspiratory preBötC neurons, including commissural preBötC neurons, adds to the ways that synaptic inhibition can influence the network activity in the neural kernel of inspiratory rhythm generation. We propose that the GABAergic reduction in presynaptic Ca 21 transients leads to a reduced transmitter output from the synaptic terminal, as less Ca 21 is available for the release machinery. How activation of presynaptic GABA A receptors couple to a reduced Ca 21 influx is unknown from these experiments, but likely result from hyperpolarization and shunting of the presynaptic membrane reducing the depolarizing effect of arriving APs, and thereby reducing the activation of presynaptic voltage-gated Ca 21 channels. It is also unknown whether somatodendritic and ) and local microiontophoresis ( 500 mm) of GABA to subsets of presynaptic terminals (A) or to soma (D) of rhythmic preBötC neurons. Right, Voltage response to depolarizing current steps in preBötC neurons, previously rhythmically active in high-aCSF, before (dark line) and after GABA application (red line). B, E, Representative images of averaged Ca 21 transients in soma and presynaptic terminals, belonging to the neuron in focus, at low and high (insets) magnification during 20 APs (stimulus-triggered average, color inverted, smart filter, identical contrast settings).
presynaptic GABAergic inhibition arises from the same presynaptic GABAergic neurons, or whether there might be more focused projection patterns relying on subsets of presynaptic and subsets of somatodendritic projecting neurons. This series of experiments did not determine whether the neurons that show presynaptic inhibition where themselves glutamatergic, GABAergic, or glycinergic. Thus, an open question is still whether GABAergic presynaptic inhibition targets a subset of these classes of neurons. An important caveat of this series of experiments is that the GABAergic presynaptic inhibition was observed under nonrhythmic conditions, and it remains to be determined whether it occurs cyclically under normal breathing rhythm, and thus contribute to the synaptic dynamics of the breathing CPG.
In summary, we demonstrate that jGCaMP7 variants label presynaptic terminals in organotypic slice cultures of the brainstem, and that the breathing CPG includes neurons that show GABAergic inhibition of presynaptic Ca 21 transients. This approach may be used to investigate the dynamics of presynaptic Ca 21 in other functioning networks of the CNS.