Calmodulin Bidirectionally Regulates Evoked and Spontaneous Neurotransmitter Release at Retinal Ribbon Synapses

Abstract For decades, a role for the Ca2+-binding protein calmodulin (CaM) in Ca2+-dependent presynaptic modulation of synaptic transmission has been recognized. Here, we investigated the influence of CaM on evoked and spontaneous neurotransmission at rod bipolar (RB) cell→AII amacrine cell synapses in the mouse retina. Our work was motivated by the observations that expression of CaM in RB axon terminals is extremely high and that [Ca2+] in RB terminals normally rises sufficiently to saturate endogenous buffers, making tonic CaM activation likely. Taking advantage of a model in which RBs can be stimulated by expressed channelrhodopsin-2 (ChR2) to avoid dialysis of the presynaptic terminal, we found that inhibition of CaM dramatically decreased evoked release by inhibition of presynaptic Ca channels while at the same time potentiating both Ca2+-dependent and Ca2+-independent spontaneous release. Remarkably, inhibition of myosin light chain kinase (MLCK), but not other CaM-dependent targets, mimicked the effects of CaM inhibition on evoked and spontaneous release. Importantly, initial antagonism of CaM occluded the effect of subsequent inhibition of MLCK on spontaneous release. We conclude that CaM, by acting through MLCK, bidirectionally regulates evoked and spontaneous release at retinal ribbon synapses.


Introduction
At chemical synapses, fast, synchronous neurotransmitter release is evoked when membrane depolarization, usually in the form of action potentials, arrives at presynaptic axon terminals and opens voltage-gated calcium channels (VGCCs), permitting Ca 21 influx to trigger fusion of synaptic vesicles. Spontaneous neurotransmitter release also occurs independent from presynaptic activity, and understanding of its separate roles in synapse formation, synaptic maintenance and dendritic protein translation is emerging (Kavalali, 2015;Andreae and Burrone, 2018;Chanaday and Kavalali, 2018). Recent studies support the notion that there are distinct molecular mechanisms underlying evoked and spontaneous release: these two forms of neurotransmission may originate from separate synaptic vesicle pools , although evidence against this hypothesis also exists (Groemer and Klingauf, 2007;Hua et al., 2010;Wilhelm et al., 2010;Schneggenburger and Rosenmund, 2015).
Because they are designed to signal tonically over extended periods (Matthews and Fuchs, 2010;Cho and von Gersdorff, 2012), ribbon-type excitatory synapses experience sustained Ca 21 influx and elevations in intracellular [Ca 21 ] and appear to be a good model for understanding modulation of evoked and spontaneous release. Imaging of fish bipolar cells and salamander rod photoreceptors show that evoked release is concentrated near ribbontype active zones (AZs) and spontaneous release occurs at a higher rate away from ribbon sites (Midorikawa et al., 2007;Zenisek, 2008;Chen et al., 2013), and electrophysiological analysis of transmission at mouse rod bipolar (RB)!AII synapses reveals two putatively separate pools of synaptic vesicles: ribbon-associated and ribbon independent (Mehta et al., 2014). Deletion of RIBEYE, a ribbon AZ-specific protein, results in an absence of ribbon AZs, as expected, and severely impairs evoked, but not spontaneous, release at RB!AII synapses (Maxeiner et al., 2016).
Here, we used the RB!AII synapse in the mouse retina, a well-established model ribbon-type synapse (Singer, 2007;Singer et al., 2009;Pangrsic et al., 2018), to examine a role for calmodulin (CaM) in the regulation of evoked and spontaneous transmission at synapses that experience tonic elevations in [Ca 21 ] and release neurotransmitter in a sustained manner over long periods.
CaM, a Ca 21 -binding protein highly conserved among eukaryotes and expressed ubiquitously and abundantly in the brain, including the retina (Pochet et al., 1991), interacts with a large number of presynaptic targets, including, but not limited to, Ca 21 /CaM-dependent kinase II (CaMKII), myosin light chain kinase (MLCK), calcineurin, Munc13, and Ca channels, all of which are involved in multiple mechanisms regulating the vesicle cycle (Hoeflich and Ikura, 2002;Ben-Johny and Yue, 2014;Chanaday and Kavalali, 2018;Tarasova et al., 2018). We used a variety of experimental approaches to demonstrate that CaM, by acting through MLCK, bidirectionally regulates evoked and spontaneous release at retinal ribbon-type synapses.

Electrophysiology
Retinal slices (200-mm thickness) were prepared from light-adapted retina isolated from either Pcp2-cre::Ai32 or wt mice. A retina was isolated into sodium bicarbonate buffered Ames' medium (Sigma) equilibrated with 95% O 2 and 5% CO 2 (carbogen) at room temperature. For slice preparation, the retina was embedded in low-melting temperature agarose (Sigma Type VIIA, 2-3% in a HEPES-buffered saline), and slices were cut on a vibrating microtome (Leica VT1200s). Slices were stored in carbogen-bubbled Ames' medium at room temperature until use.
During current-clamp recordings, pipettes were filled with the following: 110 mM K-gluconate, 5 mM NaCl, 10 mM HEPES, 8 mM Tris-phosphocreatine, 4 mM Mg-ATP, 0.4 mM Na-GTP, and 1 mM BAPTA. The pH value was adjusted to 7.2 with KOH and osmolarity to ;285 mOsm with sucrose. During voltage-clamp recordings, pipettes were filled with the following: 95 mM Cs-methanesulfonate, 20 mM TEA-Cl, 1 mM 4-AP, 10 mM HEPES, 8 mM Tris-phosphocreatine, 4 mM Mg-ATP, 0.4 mM Na-GTP, and 1 mM BAPTA. The pH value was adjusted to 7.2 with CsOH and osmolarity to ;285 mOsm with sucrose. Alexa Fluor 594 or 647 was included in the pipette solutions to visualize the cell morphology after recordings. Generally, RB holding potential was À60 mV and AII holding potential was À80 mV, and membrane potentials were corrected for junction potentials of ;À10 mV. Access resistances were ,25 MV for RBs and , 20 MV for AII amacrine cells and were compensated by 50-90%. Recordings were made using MultiClamp 700B amplifiers. Recorded currents were digitized at 10-20 kHz and low-pass filtered at 2 kHz by an ITC-18 A/D board (Heka/Instrutech) controlled by software written in Igor Pro 6 (WaveMetrics). Recorded RB Ca currents were leak-subtracted off-line (P/4 protocol). Analysis was performed in Igor Pro.

Optogenetics
ChR2 was activated by a high-power blue LED (Thor Laboratories; l peak 470 nm) directed through the 60Â or 100Â lenses to create a light spot (125 or 75 mm in diameter, respectively). The light intensities and durations were controlled by software written in Igor Pro.

Calcium imaging
Retinal slices (200-mm thickness) were prepared, as described above, from light-adapted retina isolated from Pcp2-cre::Ai38 mice. Imaging of GCaMP3 signals was performed using two-photon laser-scanning microscopy (2PLSM). GCaMP3 was excited using a pulsed infrared laser (Chameleon; Coherent) tuned to a 910-nm excitation wavelength, and emitted light was passed through a series of dichroic mirrors and filters and collected by GaAsP photomultiplier tubes (Thor Laboratories). Frames containing multiple RB axon terminals in each of which several varicosities were evident were collected at ;19 Hz and every nine frames were averaged to generate each calcium imaging picture shown below. Collected data were analyzed using ImageJ (Schneider et al., 2012).

Statistical analysis
Statistical analysis was performed with Prism 6 (GraphPad software). For better comparison among different groups, data acquired in each cell were normalized to the value under control condition. The Kolmogorov-Smirnov (KS) test was used to compare the cumulative distributions of data. Differences between experimental samples were assessed for significance using two-tailed Student's t test, Wilcoxon signed-rank test or Mann-Whitney test where appropriate. Significance was taken as p , 0.05. All data were illustrated as mean 6 SEM.

Optogenetic control of transmission at the RBfiAII synapse
To permit transmission at RB!AII synapses to be evoked reliably over extended experimental periods, we used an optogenetic approach in which ChR2 was expressed in RBs by cre-mediated recombination in Pcp2cre::Ai32 mouse retinas (Fig. 1A). As reported previously (Zhang et al., 2005;Ivanova et al., 2010), expression of ChR2-eYFP was observed in ON bipolar cells, largely RBs, in in vitro retinal slices prepared from these animals (Fig.  1B). With synaptic transmission between photoreceptors and bipolar cells blocked with the mGluR6 agonist L-AP4 (2 mM; to block photoreceptor!ON bipolar cell synapses, including rod!RB synapses) and the kainate receptor antagonist ACET [1 mM; to block photoreceptor!OFF cone bipolar (CB) cell synapses; Park et al., 2018Park et al., , 2020, 470-nm light induced ChR2-mediated currents in eYFP1 RBs and ON CBs, but not in eYFP-RBs and OFF CBs (Fig. 1C). Changes in light intensity and stimulus duration modulated ChR2-mediated current and membrane potential changes in RBs (Fig. 1D). Most importantly, brief flashes (e.g., 1-10 ms) evoked EPSCs in AII amacrine cells. The EPSCs recorded in AIIs were abolished almost completely by non-NMDA glutamate receptor antagonist DNQX (20 mM; Fig.  1E), which blocks transmission at RB!AII synapses (Singer and Diamond, 2003). In contrast, the EPSCs were only slightly affected by MFA (100 mM; Fig. 1F), which blocks the gap junctions between AII amacrine cells and ON CBs (Veruki and Hartveit, 2009). Overall, these results indicated that optogenetic control of transmission at RB!AII synapse was achieved reliably. Light-evoked EPSCs (termed eEPSCs hereafter) were quite stable over  Figure 1. Optogenetic study of transmission at RB!AII synapses. A, A diagram illustrating optogenetic study of synaptic transmission between RB cells and AII amacrine cells in the retinas of Pcp2-cre::Ai32 mice. After synaptic transmission between photoreceptors and bipolar cells is blocked with L-AP4 and ACET, 470-nm light flashes could stimulate light-sensitive ChR2 channels and thus directly activate ChR2-eYFP-expressing RBs and ON CB cells (green), and finally evoked responses in postsynaptic AII amacrine cells (magenta) could be recorded. B, A two-photon image showing ChR2-eYFP expression (green) in a retinal slice made from a Pcp2-cre::Ai32 mouse; an AII amacrine cell (magenta) was recorded and filled with Alexa Fluor 647 by a patch pipette (outlined by dashed lines). Scale bar: 10 mm. C, Representative traces showing ChR2-mediated currents recorded in an eYFP1 RB and an eYFP1 ON CB, but not in either eYFP-RB or eYFP-OFF CB, during brief flashes of 470-nm LED. V hold = À60 mV. D, Representative traces showing ChR2-mediated current (voltage-clamp mode; V hold = À60 mV) and membrane potential (voltage; current-clamp mode) changes recorded in an eYFP1 RB during brief (10 ms) and long (200 ms) light stimuli. E, During brief flashes, the eEPSCs recorded in AII amacrine cells postsynaptic to RBs were blocked almost completely by DNQX (20 mM). F, The eEPSCs recorded during brief flashes were only slightly influenced by the gap junction blocker, MFA (100 mM). long periods (.20 min), which allowed us to assess regulation of neurotransmitter release at RB!AII synapses.

A CaM antagonist affects evoked and spontaneous release differently
Physiologic depolarization of RB terminals results in Ca 21 influx sufficient to raise intracellular [Ca 21 ] to levels that saturate endogenous buffers (Mehta et al., 2014). Given that CaM is both a well-characterized Ca 21 sensor and a modulator of synaptic transmission , we postulated that were CaM expressed in RB terminals, it would play a significant role in modulating transmission at RB!AII synapses.
We therefore performed immunofluorescence double labeling of CaM and PKCa (a specific marker of RBs), and we found that CaM was strongly expressed in the axon terminals and, to a lesser extent, in the somata of RBs (Fig. 2A1,A2). Interestingly, RBs seemed to have the highest CaM expression level, especially in their axon terminals, among all retinal cell types, suggesting that CaM might play a prominent role in presynaptic functions (Pangrsic et al., 2018).
It has been reported that CaBP5, with ;58% sequence similarity to CaM, also is expressed in mouse RBs; because it interacts with Munc18-1 and myosin VI, both of which are involved in synaptic vesicle cycle, it has been suggested that CaBP5 may play a modulatory role in synaptic transmission (Haeseleer et al., 2000;Rieke et al., 2008;Sokal and Haeseleer, 2011). Preincubation of the anti-CaM antibody with the CaBP5 peptide, however, did not change the staining pattern for CaM as the CaM peptide did (Fig. 2A2,A3), indicating that the anti-CaM antibody used in this study did not cross-react with CaBP5 proteins.
We recorded miniature EPSCs (mEPSCs) and eEPSCs in AIIs, which reflect spontaneous and evoked neurotransmitter release from RBs, respectively, and assessed the effect of modulating CaM. Remarkably, bath application of a CaM antagonist, W-7 (50 mM) had strong, distinct effects on mEPSCs and eEPSCs: W-7 strongly increased mEPSC frequency (Fig. 2B1) while dramatically suppressing eEPSC amplitude (Fig. 2B1,B2). We also tested the effect of W-7 on mEPSCs in wt mice. Since there were no visible differences between the Pcp2-cre::Ai32 and wt mouse data, they were pooled. The effects of W-7 on eEPSCs (n = 9 for 50 mM; n = 6 for 25 mM; Fig. 2B3) and mEPSCs (n = 15 for 50 mM; n = 13 for 25 mM; n = 8 for 10 mM; Fig. 2C1) were both time and concentration dependent. W-7, however, had no obvious effect on mEPSC amplitude (n = 15 for 50 mM; n = 13 for 25 mM; n = 8 for 10 mM; Fig. 2C2), indicating that inhibition of presynaptic CaM strongly enhanced spontaneous release from RB terminals. Note that DMSO alone, at the same concentration as was present with 50 mM W-7, did not change the eEPSC amplitude ( Fig. 2B3), mEPSC frequency ( Fig. 2C1), or mEPSC amplitude (Fig. 2C2).
To strengthen our conclusion that suppression of eEPSCs by W-7 arose from a presynaptic mechanism, we recorded the postsynaptic AMPA receptor-mediated currents in AIIs evoked by glutamate (1 mM) applied directly onto the AII dendrites located at the border of the IPL and ganglion cell layer (GCL) when CoCl 2 (1 mM) was included in the external solution to block all Ca 21 -dependent synaptic transmission: 50 mM W-7 had no significant effect on the glutamate-evoked postsynaptic currents (n = 7; Figs. 2D1,D3). Notably, although CoCl 2 should have blocked all VGCCs, W-7 still increased mEPSC frequency (Fig. 2D2); this result is explored further, below.
Inhibition of CaM inhibits evoked transmission and stimulates spontaneous release at RB!AII synapses. Imaging exocytosis from bipolar cell terminals has led to the conclusion that evoked and spontaneous release might arise from physically separate presynaptic sites (Midorikawa et al., 2007;Zenisek, 2008), with spontaneous release occurring farther from the ribbon-type AZ than evoked release. The expression pattern of CaM in the axon terminals of RBs (Fig. 2A1) supports a role as a dual regulator of evoked and spontaneous release. Immunofluorescence double labeling of CaM and CtBP2/RIBEYE, a unique scaffolding protein of ribbons (Maxeiner et al., 2016), showed CaM to be expressed ubiquitously, at sites both near and away from ribbons (Fig. 3).

Inhibition of CaM suppresses Ca currents in RB terminals
The similarity of the observations made in Pcp2-cre:: Ai32 and wt retinas indicated that the effect of W-7 on spontaneous release was not influenced by exogenous expression of ChR2-eYFP in RBs. To further exclude the possibility that the effect of W-7 on evoked release might be because of exogenous ChR2-eYFP expression, we evoked EPSCs in AII amacrine cells from wt mice by puffing the mGluR6 antagonist LY 341495 (LY; 5-10 mM) onto RB dendrites located at the OPL, and then assessed the effect of applied W-7. Not surprisingly, W-7 (50 mM) strongly reduced the LY-evoked EPSCs (Fig. 4A1) with a seemingly slower time course (Fig. 4A2). We reasoned that LY application elicited larger changes in presynaptic [Ca 21 ] than did brief light stimuli, as evidenced from the  Figure 2. CaM bidirectionally regulates evoked and spontaneous neurotransmitter release from RBs. A1, Confocal images showing immunofluorescence double labeling of CaM (green) and PKCa (magenta), a specific cell marker of RBs, in a frozen retinal slice. In the merged image (green 1 magenta), expression of CaM could be clearly seen in the axon terminals (arrow) and somata (asterisk) of RBs. Scale bar: 10 mm. ONL: outer nuclear layer; OPL: outer plexiform layer; INL: inner nuclear layer; IPL: inner plexiform layer; GCL: ganglion cell layer. A2, No labeling was observed in the negative control where the anti-CaM antibody was preincubated with the CaM immunopeptide. DAPI staining showed the three major cell body layers in the retina. Scale bar: 10 mm. A3, Preadsorption of the anti-CaM antibody with the CaBP5 immunopeptide did not change the staining pattern for CaM. Scale bar: 10 mm. B1, Twomillisecond flashes of 470-nm LED were presented to stimulate ChR2 in Pcp2-cre::Ai32 mice with L-AP4 and ACET in the bath to large difference in the integrals of the EPSCs evoked by the two (integral of LY-evoked EPSCs, 830 737 6 167 550, n = 7; integral of ChR2-driven EPSCs, 3929 6 419, n = 9).
Given that both LY-evoked EPSCs (Fig. 4B) and lightevoked EPSCs (Fig. 4C) were abolished, as expected, by removal of extracellular Ca 21 (0 Ca 21 ; with 2 or 5 mM EGTA in the external solution), we tested the hypothesis that inhibition of CaM reduced evoked release by suppressing presynaptic Ca 21 influx. Depolarization of an RB with brief voltage step from À60 to À20 mV for 50 ms elicited sustained, inward Ca current (I Ca ; 22.18 6 2.16 pA, n = 7; Fig. 4D1). Both the peak amplitude and integral of RB I Ca were reduced substantially by bath application of 50 mM W-7 (Fig. 4D2). The time course of suppression of RB I Ca was strikingly consistent with that of inhibition of eEPSCs by 50 mM W-7 (Fig. 4D2). Application of DMSO alone seemed to reduce slightly but insignificantly the peak amplitude of RB I Ca (Fig. 4D2).
To ensure that whole-cell recording from the RB did not alter modulation of presynaptic Ca channels, we imaged [Ca 21 ] in RB terminals of Pcp2-cre::Ai38 mice in which the fluorescent [Ca 21 ] indicator GCaMP3 was expressed (see Materials and Methods). Brief application (6-50 ms) of LY (5-10 mM) at the OPL depolarized RB dendrites, ultimately evoking strong fluorescence signals in RB axon terminals; these returned to baseline level after ;5 s ( Fig.  4E1, control). The LY-evoked Ca 21 signals, detected by GCaMP3, were strongly reduced by application of 50 mM W-7 ( Fig. 4E1-E3), consistent with our observations of Ca currents (Fig. 4D1,D2).
Taken together, these results proved our hypothesis that inhibition of CaM strongly reduced evoked release from RBs by suppressing Ca 21 influx into axon terminals.
Both Ca 21 -dependent and Ca 21 -independent spontaneous release are enhanced, with the latter to a greater extent, when CaM is inhibited The presynaptic mechanisms underlying spontaneous release of neurotransmitters have been studied for decades Katz, 1950, 1952;Kavalali, 2015). Spontaneous release at RB!AII synapses depends on extracellular Ca 21 concentration, indicating that it is driven in part by Ca 21 influx through Ca channels that open spontaneously at resting potential (Singer et al., 2004;Mehta et al., 2013;Maxeiner et al., 2016;Mortensen et al., 2016). But, because spontaneous continued block synaptic transmission between photoreceptors and bipolar cells; all the inhibitory connections were also blocked. The evoked responses (eEPSCs) and the small responses induced by spontaneous release before light onset (mEPSCs; see gray background area) in AII amacrine cells were recorded. V hold = À80 mV. Individual traces showed that the CaM antagonist, W-7 (50 mM) strongly increased mEPSC frequency and reduced eEPSC amplitude. B2, Average traces of eEPSCs recorded in the same AII in B1. B3, Statistics of the effects of 25 mM (n = 6) and 50 mM (n = 9) W-7 on eEPSC amplitude. The amplitudes were normalized to the amplitude at time 0 in each cell before averaging across cells. C1, Statistics of the effects of 10 mM (n = 8), 25 mM (n = 13), and 50 mM (n = 15) W-7 on mEPSC frequency. The frequencies were normalized to the frequency at time 0 in each cell before averaging across cells. C2, Statistics of the effects of 10 mM (n = 8), 25 mM (n = 13), and 50 mM (n = 15) W-7 on mEPSC amplitude. The amplitudes were normalized to the amplitude at time 0 in each cell before averaging across cells. D1, Individual traces showing that W-7 (50 mM) had no inhibitory effect on AMPA receptor-mediated currents recorded in an AII evoked by glutamate (1 mM) applied onto the AII dendrites at the border of the IPL and GCL. V hold = À80 mV. D2, Magnification of the traces in the dashed line frames of D1, showing increase of mEPSC frequency by W-7. D3, Statistics of the effects of 50 mM W-7 (n = 7) on the amplitude of glutamate-evoked currents. The amplitudes were normalized to the amplitude at time 0 in each cell before averaging across cells. E, Summary data showing the effects of 50 mM W-7 (circles), 100 mM CMZ (triangles), another CaM antagonist, and 1 mM CALP1 (squares), a CaM agonist, on eEPSC amplitude after bath application for 15 min. The amplitudes were normalized to the amplitude at time 0 in each cell before averaging across cells. The data were also illustrated as mean 6 SEM. Wilcoxon signed-rank tests were used (control vs W-7, n = 9, p = 0.0039; control vs CMZ, n = 12, p = 0.0005; control vs CALP1, n = 5, p = 0.6250); **p , 0.01, ***p , 0.001; ns: not statistically different. Note that CMZ reduced eEPSC amplitude too, but CALP1 did not enhance eEPSC amplitude under control conditions. F, Summary data showing the effects of 50 mM W-7 (circles), 100 mM CMZ (triangles), and 1 mM CALP1 (squares) on mEPSC frequency. The frequencies were normalized to the frequency at time 0 in each cell before averaging across cells. The data were also illustrated as mean 6 SEM. Wilcoxon signed-rank tests were used (control vs W-7, n = 15, p , 0.0001; control vs CMZ, n = 12, p = 0.6377; control vs CALP1, n = 9, p = 0.0273); *p , 0.05, ****p , 0.0001; ns: not statistically different. Note that CMZ did not change the mEPSC frequency, but activation of CaM by CALP1 slightly reduced mEPSC frequency.      (Mortensen et al., 2016), there appears to be a Ca 21 -independent component to the process. Given that W-7 (50 mM) increased mEPSC frequency after evoked transmission was blocked with CoCl 2 (1 mM; Fig. 2D2), it appeared that inhibition of CaM enhanced Ca 21 -independent spontaneous release. To validate this observation and to explore the potential underlying mechanisms, we tested the effects of 50 mM W-7 on mEPSCs under four experimental conditions: (1) removal of extracellular Ca 21 (0 Ca 21 plus 2 or 5 mM EGTA in the external solution; Fig. 5A); (2) 0 Ca 21 combined with bath application of 10 mM BAPTA-AM, a cell membrane permeable continued axon terminals (n = 21 terminals from 3 retinal slices), measured as areas under the curve (AUCs) of DF/F 0 traces, were strongly suppressed by 50 mM W-7 over time. All the data were illustrated as mean 6 SEM.  (Fig. 5D). For each condition, mEPSCs were recorded for at least 15 min before bath application of 50 mM W-7. We analyzed mEPSCs recorded under control, experimental (one of the four detailed above), and experimental 1 W-7 conditions. Under each of the four experimental conditions, mEPSC frequency was reduced strongly compared with control, but for each of the four, application of 50 mM W-7 increased mEPSC frequency substantially ( Fig. 5E; Table 1). The ratio of mEPSC frequencies before and after application of W-7 under each condition was also calculated ( Fig. 5F; Table 1). Additionally, we found that under experimental condition 1 (i.e., 0 Ca 21 condition), 1 mM CALP1 reduced average mEPSC frequency to ;53% of control (n = 8; p = 0.0078 by Wilcoxon signed-rank test; control condition vs experimental condition 1, p = 0.0027 by unpaired Student's t test). Interestingly, the relative effect on mEPSC frequency of either W-7 or CALP1 was stronger under experimental than control conditions, indicating that modulation of CaM influenced Ca 21 -independent spontaneous release more strongly than Ca 21 -dependent spontaneous release.

Inhibition of MLCK, but not other CaM targets, closely mimics the distinct effects of CaM inhibition on evoked and spontaneous release
Given that CaM modulates Ca channels directly (Ben-Johny and Yue, 2014), we wished to determine whether the effect of W-7 was mediated by CaM acting directly on target proteins or indirectly, via a downstream second messenger. Therefore, we recorded simultaneously mEPSCs and ChR2-driven eEPSCs in AII amacrine cells. Strikingly, bath application of a specific MLCK inhibitor, ML-9 (100 mM) had strong, distinct effects on   Figure 6. Inhibition of MLCK, but not other CaM targets, differentially regulates evoked and spontaneous release from RBs. A1, Five-millisecond flashes of 470-nm LED were presented to stimulate ChR2-expressing RBs in Pcp2-cre::Ai32 mice. The mEPSCs (in gray background area) and eEPSCs in AIIs were recorded. V hold = À80 mV. Individual traces showed that a specific MLCK inhibitor, ML-9 (100 mM) strongly increased mEPSC frequency and reduced eEPSC amplitude. A2, Average traces of eEPSCs recorded in the same AII in A1. A3, Statistics of the effects of 50 mM (n = 5) and 100 mM (n = 10) ML-9 on eEPSC amplitude. The amplitudes were normalized to the amplitude at time 0 in each cell before averaging across cells. The data of 50 mM W-7 (adapted from Fig. 2B3, superimposed in magenta) were also included for direct comparison. B1, Statistics of the effects of 25 mM (n = 7), 50 mM (n = 13), and 100 mM (n = 10) ML-9 on mEPSC frequency. The frequencies were normalized to the frequency at time 0 in each cell before averaging across cells. The data of 50 mM W-7 (adapted from Fig. 2C1, superimposed in magenta) were also included for direct comparison. mEPSCs and eEPSCs, reminiscent of the effects of W-7: ML-9 strongly increased mEPSC frequency (Fig. 6A1) while dramatically suppressing eEPSC amplitude (Fig.  6A1,A2). The effects of ML-9 on eEPSCs (n = 10 for 100 mM; n = 5 for 50 mM; Fig. 6A3) and mEPSCs (n = 10 for 100 mM; n = 13 for 50 mM; n = 7 for 25 mM; Fig. 6B1) were both time and concentration dependent. ML-9, also, seemed to have a weakly inhibitory effect on mEPSC amplitude, especially at high concentrations (n = 10 for 100 mM; n = 13 for 50 mM; n = 7 for 25 mM; Fig. 6B2). We conclude that inhibition of MLCK strongly enhanced spontaneous release from RB axon terminals.
To exclude the possibility that the suppression of eEPSCs by ML-9 was because of a postsynaptic effect, we recorded AMPA receptor-mediated currents evoked by pressure ejection of glutamate (1 mM) onto AII dendrites when CoCl 2 (1 mM) was included in the external solution to block synaptic transmission. As was the case with W-7, 100 mM ML-9 had little effect on the glutamate-evoked currents (n = 4; Fig. 6C1,C3). Notably ML-9 still increased mEPSC frequency even when Ca channels were blocked with CoCl 2 in the external solution (Fig. 6C2).
We also examined the effects of inhibitors of other downstream CaM targets such as the CaMKII inhibitor KN-62 (4 mM), the PDE1 inhibitor MMPX (40 mM), and the calcineurin inhibitor ascomycin (1 mM). None of these showed significant effects on eEPSCs (Fig. 6D) or mEPSCs (Fig. 6E) except ascomycin, which slightly reduced the amplitudes of eEPSCs ( Fig. 6D; Table 2). Ascomycin, however, did not affect mEPSC frequency (Fig. 6E), suggesting that it does not act in the pathway inhibited by W-7 and ML-9. Thus, we conclude that inhibition of MLCK, but not other CaM targets, results from inhibiting CaM in RB terminals.

Inhibition of CaM occludes the potentiating effect of MLCK inhibition on spontaneous release
To determine whether W-7 and ML-9 exerted their effects via the same intracellular signaling pathway (i.e., the CaM-MLCK pathway), we performed an occlusion experiment: following inhibition of CaM with W-7 (.20-min preincubation), we antagonized MLCK with ML-9. If inhibition of CaM reduces MLCK activity, then we expected the effect of ML-9 to be reduced in the presence of W-7. Indeed, the experimental result is in line with this expectation.
Specifically, in the presence of 50 mM W-7 (.20-min preincubation), application of 100 mM ML-9 did not increase, but instead reduced, AII mEPSC frequency (control 1 vs ML-9, 1.00 vs 0.12, n = 13, p = 0.0002; Fig. 9A,C). As a control for the prolonged recording period, we measured mEPSC frequencies over time in the absence of any drug application and noted that mEPSC frequency declined over time (control 2 vs no drug, 1.00 vs 0.41, n = 8, p = 0.0078; Fig. 9B,C), although the decline was less pronounced than in the W-7 1 ML-9 condition (ML-9 vs no drug, p = 0.0001; Fig. 9C). DMSO alone did not have any additional effect on mEPSC frequency compared with the control condition (control 3 vs DMSO, 1.00 vs 0.45, n = 7, p = 0.0156; DMSO vs no drug, p = 0.7337; Fig. 9C). We conclude, then, that the CaM-MLCK pathway is involved in regulating spontaneous release from RBs.

Discussion
We monitored the dynamics of transmission at RB!AII synapses to examine the effects of modulating presynaptic CaM and its target proteins and made two major findings. One, we found that inhibition of CaM strongly reduced evoked release by suppressing presynaptic Ca currents while simultaneously potentiating both Ca 21 -dependent and Ca 21 -independent spontaneous release . Two, the effect of inhibiting CaM seemed to be mediated by inhibition of MLCK but not other CaM downstream targets (including CaMKII,PDE1 and calcineurin;. Thus, CaM, via activation of MLCK, suppresses spontaneous release and promotes evoked release at retinal ribbon synapses.

Pharmacological agents
Concerns about the specificity of pharmacological agents are common and vex experimentalists. We , an mGluR6 antagonist, onto the dendrites of RBs located at the OPL, were strongly reduced by application of 100 mM ML-9. V hold = À80 mV. A2, LY-evoked EPSCs decreased over time with bath application of 100 mM ML-9 (n = 5). The peak amplitudes and integrals of EPSCs were normalized to the amplitude and integral at time 0, respectively, in each cell before averaging across cells. All the data were illustrated as mean 6 SEM. B1, Average traces showing that ML-9 (100 mM) strongly suppressed the voltage step-generated calcium currents (I Ca ) in an RB. B2, Statistics of the effects of 100 mM ML-9 (n = 10) on the peak amplitude and integral of RB I Ca . The suppression of I Ca recorded in RBs was closely related to the inhibition of eEPSCs recorded in AIIs (adapted from Fig. 6A3, superimposed in gray). All the data were illustrated as mean 6 SEM.  Fig. 6B1, empty and full triangles, respectively) were also presented for direct comparison. The frequencies were normalized to the frequency under control condition in each cell before averaging across cells. The data were also illustrated as mean 6 SEM. Wilcoxon signed-rank tests were used (control vs ML-9, n = 10, p = 0.0020; 0 Ca 1 BAPTA-AM vs 0 Ca 1 BAPTA-AM 1 ML-9, n = 9, p = 0.0039) except for comparison of 0 Ca and 0 Ca 1 ML-9 data by paired Student's t test (n = 10, p = 0.0001); **p , 0.01, ***p , 0.001. D, Summary data for changes of AII mEPSC frequency after bath application of 100 mM ML-9 for 15 min under control and two experimental conditions. The change in each cell was calculated as the ratio of mEPSC frequencies before and 15 min after application of ML-9. The data were also illustrated as mean 6 SEM. Mann-Whitney tests were used (ML-9 vs 0 Ca 1 ML-9, p = 0.0232; ML-9 vs 0 Ca 1 BAPTA-AM 1 ML-9, p =0.0030); *p , 0.05, **p , 0.01. Non-normal distribution 1.00 6 0.00 10 100 mM ML-9 Normal distribution 7.19 6 1.32 10 a Control vs ML-9 Wilcoxon signed-rank test made three observations that should assuage such concerns. One, our immunohistochemical analysis revealed extremely high expression of CaM in RB axon terminals. Even so, application of CaM antagonists such as W-7 might exert off-target effects. Therefore, we tested a structurally dissimilar CaM antagonist, CMZ (Gietzen et al., 1982), in our preparation and found that at concentrations of 50-100 mM, it too reduced evoked release, although to a lesser extent (Fig.  2E). We also tested the effects of CALP1, a CaM agonist (Villain et al., 2000). CALP1 reduced spontaneous release, especially after removal of extracellular Ca 21 , whereas it had no effect on evoked release (Fig. 2E,F), which might be explained by the almost full activation of CaM by Ca 21 sufficient to saturate endogenous buffers (Mehta et al., 2014) under our experimental conditions. Three, inhibition of MLCK also bidirectionally regulated evoked and spontaneous release and the effect could be occluded by preinhibition of CaM, which would in turn support the conclusion that neurotransmitter release at RB!AII synapse is modulated by CaM. Although CaBP5 is expressed in mouse retinal RBs and is suggested to be involved in neurotransmitter release since it interacts with Munc18-1 and myosin VI (Haeseleer et al., 2000;Rieke et al., 2008;Sokal and Haeseleer, 2011), to our knowledge, there is no evidence showing that W-7 could antagonize CaBP5. It has been shown that CaBP5 has only a relatively weak effect on inactivation of Ca currents when cotransfected with calcium channels in HEK293T cells (Rieke et al., 2008). In contrast, we found that W-7 inhibited RB I Ca strongly. This would indicate that, even if W-7 might inhibit CaBP5, it exerted its inhibitory effect on RB I Ca by primarily targeting CaM, rather than CaBP5.

CaM and its downstream targets at ribbon synapses
It has been proposed that CaMKII may be the downstream target of CaM in regulation of evoked release (Llinás et al., 1985;Pang et al., 2010). CaMKII has been found located close to synaptic ribbons (Uthaiah and Hudspeth, 2010;Kantardzhieva et al., 2012), and it has been shown to phosphorylate syntaxin 3B, the retinal isoform of syntaxin and an essential component of the core SNARE complex mediating vesicle fusion (Liu et al., 2014). We were surprised, then, that the CaMKII inhibitor KN-62 only had a small (not significant) effect on either evoked or spontaneous release at RB!AII synapses. In contrast, the MLCK inhibitor ML-9 exerted strong effects on both two forms of release.
MLCK has been suggested to be involved in regulation of neurotransmitter release, with opposing effects observed at different synapses (Mochida et al., 1994;Mochida, 1995;Ryan, 1999;Polo-Parada et al., 2005;Srinivasan et al., 2008), although there is a study arguing that MLCK is not a regulator for synaptic vesicle trafficking in hippocampal neurons (Tokuoka and Goda, 2006). In addition, MLCK accelerates vesicle endocytosis at the calyx of Held and hippocampal synapses (Yue and Xu, 2014;Li et al., 2016) and enhances ribbon replenishment in cone photoreceptors (Van Hook et al., 2014). Our findings support the role that MLCK plays in regulating neurotransmitter release. We could imagine that inhibition of MLCK slowed down the replenishment of vesicle pools and reduced subsequent evoked release by making the readily-releasable pool (RRP) of vesicles smaller. Potentially, reducing the RRP size could shunt vesicles into a spontaneously-releasing pool.

CaM regulates evoked and spontaneous release differentially
Recent studies have showed selective molecular regulation of evoked and spontaneous release. Proteins such Control (with 50 μM W-7) Figure 9. Inhibition of CaM occludes the effect of MLCK inhibition on spontaneous release. A, Representative traces showing that, after preincubation with 50 mM W-7, 100 mM ML-9 did not increase, but instead reduced, AII mEPSC frequency. B, Representative traces showing that, after preincubation with 50 mM W-7, AII mEPSC frequency decreased significantly after 10-min recording. C, Summary data for AII mEPSC frequency under three experimental conditions in A (empty and full circles), B (empty and full squares), and DMSO control (empty and full triangles). The frequencies were normalized to the frequency under control condition in each cell before averaging across cells. The data were also illustrated as mean 6 SEM. Wilcoxon signed-rank tests (control 1 vs ML-9, n = 13, p = 0.0002; control 2 vs no drug, n = 8, p = 0.0078; control 3 vs DMSO, n = 7, p = 0.0156) or unpaired Student's t test (ML-9 vs no drug, p = 0.0001; no drug vs DMSO, p = 0.7337) were used for comparison; *p , 0.05, **p , 0.01, ***p , 0.001; ns: not statistically significant.
It has been reported that W-7 can inhibit VGCCs in some non-neuronal systems such as ciliary membrane of Paramecium and smooth muscle from rat vas deferens (Hennessey and Kung, 1984;Nakazawa et al., 1993). CMZ inhibits VGCCs in different smooth muscle cells (Klöckner and Isenberg, 1987;Nakazawa et al., 1993;Sunagawa et al., 1999), but it has no effect on Ca currents in Paramecium (Ehrlich et al., 1988). The inhibitory effects of W-7 and CMZ on VGCCs are suggested to be CaM-independent and likely because of direct actions of these drugs on VGCCs, based on the limited evidence that exogenous CaM has no effect on VGCCs and that CaMKII antagonists, when applied either extracellularly or intracellularly, do not block the effect of CMZ on VGCCs (Klöckner and Isenberg, 1987;Ehrlich et al., 1988;Sunagawa et al., 1999).
Similar results have also been observed in our study: activation of CaM by CALP1 did not enhance evoked release, and neither CaMKII nor PDE1 seemed to be involved in regulating neurotransmitter release from RBs. Note, however, that W-7 and CMZ likely have distinct effects on different CaM-dependent pathways. For example, CMZ, at the concentration of 1 mM, dramatically inhibits the activity of CaM-dependent PDE, while W-7, even at the concentration as high as 100 mM, only has a very small effect (Ehrlich et al., 1988). By contrast, it may be possible that W-7 has a stronger effect on other downstream targets of CaM, such as MLCK, than CMZ. Indeed, we found that W-7 inhibited evoked release from RBs more strongly than CMZ, and MLCK was likely the mediator of the effects observed. It has been shown that ML-9 (and also its structural analog, ML-7) inhibits VGCCs in hippocampal neurons, and this effect may be independent of MLCK since it is not mimicked by wortmannin, a relatively non-specific MLCK inhibitor (Tokuoka and Goda, 2006). We could not exclude the possibility that both W-7 and ML-9 inhibit VGCCs directly. But it is unlikely to be true since ML-9 not only closely mimicked the effects of W-7 on VGCCs and evoked release but also on Ca 21 -independent spontaneous release (Figs. 2,5,6,8), which is not related to VGCCs. Further, preincubation of W-7 completely occluded the potentiating effect of ML-9 on spontaneous release (Fig. 9), indicating that these two drugs exerted their effects via the same (CaM-MLCK) pathway.
Evidence for direct interactions between MLCK and Ca channels in RB terminals is not apparent in the literature, and therefore it will be interesting to explore how MLCK controls the activity of Ca channels and Ca 21 -dependent exocytosis in the future. But generally, our present observations support the notion that CaM promotes evoked release, which is consistent with other studies (Chamberlain et al., 1995;Chen et al., 1999;Junge et al., 2004;Pang et al., 2010).
Unique mechanisms of spontaneous neurotransmitter release have received significant attention recently (Kavalali, 2015). Spontaneous release is largely dependent on Ca 21 influx through VGCCs or Ca 21 efflux from internal stores (Kaeser and Regehr, 2014;Schneggenburger and Rosenmund, 2015;Williams and Smith, 2017). Indeed, we observed that spontaneous release was reduced by ;70% (Figs. 5, 8) when the extracellular Ca 21 was removed and the intracellular Ca 21 was buffered by the membrane permeable Ca 21 chelator BAPTA-AM. Our observations, however, indicate that there is a Ca 21 -independent component of spontaneous release that is poorly understood. Given that spontaneous energy fluctuations that overcome the energy barrier for vesicle fusion can trigger spontaneous exocytosis (Schneggenburger and Rosenmund, 2015), we believe it possible that inhibition of CaM or MLCK somehow reduces the energy barrier to exocytosis, much as Ca 21 binding to synaptotagmin does.