RIM1 and RIM2 redundantly determine Ca²⁺ channel density and readily releasable pool size at a large hindbrain synapse

Abstract

The localization and density of voltage-gated Ca²⁺ channels at active zones are essential for the amount and kinetics of transmitter release at synapses. RIM proteins are scaffolding proteins at the active zone that bind to several other presynaptic proteins, including voltage-gated Ca²⁺ channel α-subunits. The long isoforms of RIM proteins, which contain NH₂-terminal Rab3- and Munc13-interacting domains, as well as a central PDZ domain and two COOH-terminal C2 domains, are encoded by two genes, Rim1 and Rim2. Here, we used the ideal accessibility of the large calyx of Held synapse for direct presynaptic electrophysiology to investigate whether the two Rim genes have redundant, or separate, functions in determining the presynaptic Ca²⁺ channel density, and the size of a readily releasable vesicle pool (RRP). Quantitative PCR showed that cochlear nucleus neurons, which include calyx of Held generating neurons, express both RIM1 and RIM2. Conditional genetic inactivation of RIM2 at the calyx of Held led to a subtle reduction in presynaptic Ca²⁺ current density, whereas deletion of RIM1 was ineffective. The release efficiency of brief presynaptic Ca²⁺ “tail” currents and the RRP were unaffected in conditional single RIM1 and RIM2 knockout (KO) mice, whereas both parameters were strongly reduced in RIM1/2 double KO mice. Thus, despite a somewhat more decisive role for RIM2 in determining presynaptic Ca²⁺ channel density, RIM1 and RIM2 can overall replace each other's presynaptic functions at a large relay synapse in the hindbrain, the calyx of Held.

neurotransmitter release takes place at active zones, which contain docked vesicles with their molecular fusion machinery, voltage-gated Ca²⁺ channels, and presynaptic scaffolding proteins (Schoch and Gundelfinger 2006). The immediate vicinity of readily releasable vesicles to Ca²⁺ channels on the range of a few tens of nanometers is essential for fast transmitter release, because it allows short Ca²⁺ diffusion times (Neher 1998). However, the molecular mechanisms that determine the enrichment of Ca²⁺ channels at active zones, and the exact Ca²⁺ channel-vesicle colocalization, are only beginning to be understood (Südhof 2012). RIM proteins are presynaptic scaffolding proteins specifically localized to the active zone and found to bind several presynaptic proteins, like Munc13-1, Rab3a, and voltage-gated Ca²⁺ channels (Betz et al. 2001; Kaeser et al. 2011; Schoch et al. 2002; Wang et al. 1997). RIM proteins are encoded by four genes (Rims1–4); the Rim1 and 2 genes give rise to five RIM isoforms, called RIM1α, RIM1β, RIM2α, RIM2β, and RIM2γ (see Fig. 1, A and B; Kaeser et al. 2008b; Wang and Südhof 2003). Studies using RIM1α constitutive knockout (KO) mice and RIM mutants in C. elegans found roles for RIM1 in transmitter release, presumably via determining readily releasable vesicle pool (RRP) size (Calakos et al. 2004; Koushika et al. 2001; Schoch et al. 2002). Constitutive genetic inactivation of both RIM1α and RIM2α in mice leads to postnatal lethality not visible in the individual KO mice, indicating redundant functions of the two long isoforms for mouse survival (Schoch et al. 2006). This redundancy has initially hindered a more detailed understanding of the presynaptic functions of RIM proteins.

Fig. 1.

Expression of RIM1/RIM2 isoforms in the cochlear nucleus as detected by quantitative (q)PCR analysis. A: scheme of the mouse Rim1 gene with numbered exons (top) and protein domains of the resulting RIM1α and β isoforms (bottom). The isoform-specific 5′ exons are shown in blue, and the first common exon shared by all isoforms is shown in red. The exon-exon boundaries targeted by isoform-specific and generic RIM1 PCR probes were marked by blue and red brackets, respectively. Zn, Zinc-finger domain; PxxP, proline-rich region. B: schemes of mouse Rim2 gene and protein domain structure of RIM2α/β/γ. C: example of cycle threshold (CT) values analyzed in triplicate from the cochlear nuclei of a C57/Bl6 mouse at postnatal day 14 (P14). D: quantification of the relative expression level, normalized to the β-actin levels, and averaged for n = 3 mice at P14.

More recently, floxed alleles of both the Rim1 and Rim2 genes (also called Rims1 and Rims2 genes) have been produced to allow conditional removal of both genes (Kaeser et al. 2011, 2008b). Experiments with conditional RIM1/2 double KO synapses of cultured neurons have shown that the central PDZ domain in RIM1/2 confers a high extracellular Ca²⁺ sensitivity on release and mediates a high action-potential-driven presynaptic Ca²⁺ influx; furthermore, the PDZ domain was shown to bind to P/Q- and N-type Ca²⁺ channel α-subunits (Kaeser et al. 2011). Conditional inactivation of Rim1 and Rim2 at the calyx of Held, a synapse accessible to direct presynaptic recordings, has shown that RIM proteins determine a high Ca²⁺ current density in the nerve terminal (Han et al. 2011). Together, these data showed that the central PDZ domain and a more COOH-terminal proline-rich motif (PxxP; Fig. 1) of the long RIM1 and RIM2 isoforms are critical for anchoring and enriching Ca²⁺ channels at the active zone. In addition, the RIM1/2 double KO studies revealed a role for RIM proteins in vesicle docking (Han et al. 2011; Kaeser et al. 2011). The reduced RRP size at the calyx of Held might be largely caused by the docking deficit (Han et al. 2011).

The contribution of the individual Rim1 and Rim2 gene products to presynaptic Ca²⁺ channel density and RRP size has not been addressed in direct presynaptic recordings. A redundancy of RIM1 and RIM2 function has been found recently with more indirect methods at inhibitory synapses of cultured hippocampal synapses (Kaeser et al. 2012). Another recent study in mouse cerebellar slices found that genetic inactivation of RIM1α causes a reduced release probability and decreased axonal Ca²⁺ influx and suggested that RIM1α has a predominant role in setting presynaptic Ca²⁺ channel density at many synapses (Kintscher et al. 2013). Using the ideal presynaptic accessibility of the calyx of Held synapse, we here investigated the effect of conditional inactivation of the single Rim1 or Rim2 gene for presynaptic Ca²⁺ current density and RRP size. Our results suggest redundant roles for RIM1 and RIM2 isoforms in determining both the presynaptic Ca²⁺ channel density, as well as enabling a large RRP.

MATERIALS AND METHODS

Conditional knockout of RIM1, RIM2, and RIM1/2 in the calyx of Held.

Procedures of mouse breeding, handling, and the scarification of mice before slice preparation were approved by the Veterinary Office of the Canton of Vaud, Switzerland (authorization no. 2063.3). We used the Krox20::Cre mouse line, which drives expression of Cre-recombinase in neurons descending from rhombomeres 3 and 5 (Han et al. 2011; Maricich et al. 2009; Renier et al. 2010; Voiculescu et al. 2000), to recombine the floxed RIM1 allele and the floxed RIM2 allele, either separately or in combination. For this purpose, we crossed the Krox20^+/Cre mouse line with RIM1^lox/lox mice (Kaeser et al. 2008b); or with RIM2^lox/lox mice (Kaeser et al. 2011); or with both RIM1^lox/lox and RIM2^lox/lox mouse lines, respectively. Because of germline recombination in the Krox20Cre mouse line, we obtained RIM1^lox/Δ,Krox20^+/Cre mice (called RIM1 cKO for conditional KO); RIM2^lox/Δ,Krox20^+/Cre mice (called RIM2 cKO); or else RIM1^lox/Δ,RIM2^lox/Δ,Krox20^+/Cre mice (called RIM1/2 double cKO for conditional double KO); the “Δ” symbols indicate removal of the allele in the germline. The Cre negative littermates were used for control experiments; the detailed breeding schemes have been described before (Han et al. 2011). Since the data from all three control groups did not significantly differ, we pooled the control data and refer to those data as “control mice.”

Slice preparation and electrophysiology.

Postnatal day 9–11 (P9–P11) mice were decapitated, and transverse brainstem slices of 200-μm thickness were prepared with a LEICA VT1000S slicer. Slices were kept at 37°C in a bicarbonate-buffered extracellular solution containing the following (in mM): 125 NaCl, 25 NaHCO₃, 1.25 KCl, 2 CaCl₂, 1 MgCl₂, 25 glucose, 3 myo-inositol, 2 Na-pyruvate, and 0.4 ascorbic acid, pH 7.2 when bubbled with 95% O₂-5% CO₂. During recordings, slices were perfused with the same solution at room temperature (21–23°C). For paired recordings (Figs. 2 and 3), tetraethylammonium chloride (TEA; 10 mM), tetrodotoxin (TTX; 1 μM), d-2-amino-5-phosphonopentanoic acid (d-AP5; 50 μM), γ-d-glutamylglycine (γ-DGG; 2 mM), and cyclothiazide (CTZ; 100 μM) were added to the extracellular solution. For fiber stimulation experiments (Fig. 4), excitatory postsynaptic currents (EPSCs) were evoked by stimulating presynaptic axons with a custom-made bipolar stimulation electrode placed medially to the medial nucleus of the trapezoid body in the presence of d-AP5 (50 μM), bicuculline (10 μM), and strychnine (2 μM) in the extracellular solution. In these experiments, we reversibly applied 500 μM γ-DGG as a rapid-off antagonist at AMPA receptors to probe whether the cleft glutamate concentration was different between control synapses and RIM1/2 double cKO synapses (Clements et al. 1992; Wadiche and Jahr 2001). The choice of the relatively low concentration of γ-DGG was dictated by practical considerations. Since EPSCs are small in RIM1/2 double cKO mice (∼1 nA), we aimed for a γ-DGG concentration that would block EPSCs down to approximately one-third of their initial amplitude.

Fig. 2.

Presynaptic Ca²⁺ currents and readily releasable pool are not strongly affected in RIM1 and RIM2 single knockout (KO) mice. A: transmitter release in response to prolonged presynaptic depolarizations (0 mV for 50 ms) in representative RIM1 single cKO (blue), RIM2 single cKO (green), RIM1/2 double cKO (red), and control synapses (black). Presynaptic Ca²⁺ currents, excitatory postsynaptic currents (EPSCs), release rates, and traces of cumulative release are shown from top to bottom. The time constant of the fast release component, as analyzed by double-exponential fits to the cumulative release traces, is indicated (τ_f). cDKO, conditional double KO. B: quantification of presynaptic Ca²⁺ current properties. Peak Ca²⁺ current amplitude at 0 mV (B1), presynaptic membrane capacitance C_m (B2), Ca²⁺ current density (B3), and Ca²⁺ current activation time constant (τI_Ca; B4) are shown. In B and in C and D, the numbers of paired recordings for each observation are indicated below each bar. C: quantification of EPSC and release parameters. The average values and individual data points for maximal EPSC amplitude (C1), peak release rate (C2), number of vesicles released in the fast component (A_fast; C3), 20–80% rise time of EPSCs (C4), and time constant of the fast release component (τ_f; C5) are shown. Note that none of the EPSC and release parameters were significantly different in RIM1 or RIM2 single KO mice with respect to control mice, whereas all parameters were significantly different between RIM1/2 double cKO and control mice. D: average values and individual data points for the slow release time constant (D1) and the number of slowly released vesicles (D2). *P < 0.05, **P < 0.01, and ***P < 0.001.

Fig. 3.

Ca²⁺ channel-release coupling as measured by tail current protocols is unaffected in single RIM KOs but strongly impaired in the RIM1/2 double cKO mice. A: presynaptic voltage-clamp protocol (top), Ca²⁺ tail currents (middle; time scale of bottom applies), and the corresponding EPSCs (bottom) upon presynaptic depolarizations to +60 mV of varying durations. A representative paired recording is shown for each genotype. Insets: peak of the Ca²⁺ currents on expanded scales. B: double-logarithmic plots of the EPSC amplitudes as a function of the presynaptic Ca²⁺ charge (Q_Ca) for RIM1 cKO (B1, blue), RIM2 cKO (B2, green) and control (black) group. A line fit to the logarithmized data set is superimposed for each data set from a given cell. For clarity, the results from only n = 3 representative paired recordings out of larger data sets (see C) are shown for each group. C: slope values, or Ca²⁺ current cooperativity, obtained from line fits to log-log plots of EPSC amplitude vs. Q_Ca. Note the large values of 4–5 for the single RIM1- and RIM2 cKO synapses, and for the control data set. The number of paired recordings was RIM1 cKO mice, n = 5; RIM2 cKO mice, n = 4; and control mice, n = 11. D: presynaptic Ca²⁺ charge, Q_Ca, needed to evoke an EPSC of 2 nA, as a measure of the absolute efficiency of small Ca²⁺ currents in evoking release.

Fig. 4.

The fast AMPA-receptor antagonist γ-d-glutamylglycine (γ-DGG) reveals a slightly lower cleft glutamate concentration in RIM1/2 double cKO mice. A: fiber stimulation-evoked EPSCs were suppressed by 500 μM γ-DGG in a representative control calyx. Left: individual (gray) and average (black) EPSC traces before (“baseline”) and after applying γ-DGG. Right: time course of reversible block of EPSCs by γ-DGG. The black horizontal bar indicates the time of γ-DGG application. Individual EPSC amplitudes and the average value for baseline, steady-state block, and washout are shown as gray circles and black lines, respectively. B: as in A, but for a RIM1/2 double cKO mouse. Note the smaller baseline EPSC amplitude and the stronger steady-state block of EPSCs by γ-DGG compared with the control cell in A. C: quantification of EPSC amplitudes under baseline conditions (left bars) and during steady-state block by γ-DGG (right bars) in control mice (left panel), and in RIM1/2 double cKO mice (right panel). P < 0.05, **P < 0.01. D: relative γ-DGG block, normalized to the baseline EPSC amplitude measured in each cell, in control mice (left bar) and in RIM1/2 double cKO mice (right bar; *P < 0.05).

Calyceal terminals and medial nucleus of the trapezoid body principle cells were visually identified with an upright microscope (Zeiss Axioskop 2 FS) equipped with a ×60 objective (Olympus LUMPlanFI; NA, 0.9) and gradient contrast infrared visualization (Luigs and Neumann). Pre- and/or postsynaptic whole cell recordings at the calyx of Held synapse were made with an EPC10/2 double patch-clamp amplifier (HEKA). The intracellular (pipette) solutions contained the following (in mM): 140 Cs-gluconate, 20 TEA, 10 HEPES, 5 Na₂-phosphocreatine, 4 MgATP, and 0.3 Na₂GTP, pH 7.2, to which either 0.1 or 5 mM EGTA was added for pre- or postsynaptic recordings, respectively. Pre- and postsynaptic series resistances (R_S) were ∼10–25 or 3–10 MΩ, respectively, and compensated during the experiment such that the remaining R_S did not exceed 10 and 3 MΩ in pre- and postsynaptic recordings, respectively. Postsynaptic currents were corrected offline for the remaining R_S error (Meyer et al. 2001). For the measurement of release evoked by presynaptic Ca²⁺ “tail” currents (Fig. 3), 0.3-ms ramps from −70 to +60 mV, followed by a variable duration at +60 mV and by a 0.3-ms ramp back to −70 mV, were applied. Presynaptic Ca²⁺ currents were corrected for leak and capacitative currents with a P/5 protocol. Presynaptic membrane capacitance (C_m) was estimated by cancelling the first exponential decay component of the capacitative transients in response to 10-mV hyperpolarizing voltage-clamp steps. This procedure estimates the resting C_m value of a first membrane compartment represented by the giant terminal and proximal axon, whereas more distal axonal compartments will not influence this estimate (Wölfel and Schneggenburger 2003). Ca²⁺ current amplitudes measured at 0 mV were divided by the C_m estimate to calculate Ca²⁺ current density (Fig. 2, B1–B3).

Reverse transcription and quantitative PCR.

The expression level of different RIM isoforms was determined by quantitative (q)PCR with isoform-specific probes. For this, acute brainstem slices were made as described above and cochlear nucleus tissue including the ventral cochlear nucleus (VCN) and the dorsal cochlear nucleus was dissected with fine forceps in half frozen RNase-free phosphate-buffered solution. The tissue from each mouse was pooled and RNA was extracted with an RNeasy Micro Kit (Qiagen). In total, n = 3 C57Bl6 mice were used at P14. A recent gene expression study in the auditory brainstem suggests that the expression of RIM1 and RIM2 should not significantly change between P9 and P11 (the age range used for the recordings; see above) and P14 (Körber et al. 2014). The samples from each mouse were processed separately; each sample was processed in triplicate for qPCR. One-hundred nanograms of RNA from each animal were used to make complementary DNA (cDNA), using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). The reverse transcription (RT) reaction was incubated at 25°C for 10 min, at 37°C for 120 min, and at 85°C for 5 s. Real-time PCR was performed separately for each gene in triplicate with TaqMan Universal PCR Master Mix (Applied Biosystems). RT products and PCR reagents were distributed to a 384-well-plate by a Hamilton liquid handling system (Hamilton). Quantitative real-time PCR was done by a 7900HT Fast Real-Time PCR System (Applied Biosystems). The cycle threshold (CT) was defined at a level of 0.2 fluorescence unit.

As mRNA probes, we used the Taqman assays (Applied Biosystems, Rotkreutz, Switzerland) targeting exon-exon boundaries containing the isoform-specific first exons (see Fig. 1, A and B, blue brackets; Kaeser et al. 2008b; Wang and Südhof 2003). In addition, the total amount of RIM1 or RIM2 transcripts was determined by using assays targeting exon-exon boundaries that are shared by all isoforms (see Fig. 1, A and B, red brackets). The following assays are available from Applied Biosystems: β-actin (Rn00667869_m1), RIM1α (Mm00694774_m1), RIM2α (Mm00453603_m1), and RIM2_total (Mm00453627_m1). The assays for RIM1_total, RIM2β, and RIM 2γ were custom designed. We also attempted to detect RIM1β mRNA (Kaeser et al. 2008b) with a custom-designed probe, which, however, failed to detect any signal (not shown). The efficiency of each qPCR reaction was validated using a dilution series of brainstem cDNA, and the mRNA concentrations relative to β-actin were corrected for qPCR efficiency of each probe. All assays showed similar amplification efficiencies (0.88–0.93).

Data analysis and statistics.

All data are reported as means ± SD. Statistical significance was calculated with Student's nonpaired two-tailed t-test with unequal variance if not noted otherwise. For the data in Fig. 2, a one-way ANOVA was first applied, which showed that the population means were significantly different for all analyzed parameters (Fig. 2, B1–B4, C1–C5, and D1), except for the number of slowly released vesicles (Fig. 2D2; P = 0.23). Then, a Student's t-test with Bonferroni correction for multiple comparisons was used for a pairwise comparison between the different mutant genotypes (RIM1 cKO, RIM2 cKO, or RIM1/2 double cKO) vs. control mice. In all cases, statistical significance was accepted at the P < 0.05 level. The asterisks in the figures indicate the statistical significance of pairwise comparisons, with P > 0.05 (not significant) indicated by no asterisk and levels of significance of P < 0.05, P < 0.01, and P < 0.001 indicated by one, two, and three asterisks, respectively.

RESULTS

RIM1 and RIM2 are expressed in the cochlear nucleus.

We first analyzed the expression level of the major isoforms of Rim1 and Rim2 genes in the cochlear nucleus by using qPCR approaches and tissue samples from P14 mice (Fig. 1). The VCN contains bushy cells that account for a sizeable fraction of the large neuron types; a subpopulation of the bushy cells, most likely corresponding to globular bushy cells, make calyces of Held (Cant and Benson 2003; Körber et al. 2014; Saul et al. 2008; Xiao et al. 2010). All RIM probes showed CT values 6–10 cycles higher than the abundantly expressed β-actin (Fig. 1C). The RIM2_total probe, designed to detect the RIM2 mRNA common for all isoforms (Fig. 1B, red bracket; see materials and methods), and the probe for RIM2γ gave the lowest CT values (Fig. 1C). After normalization for qPCR efficiency and relative to the expression of β-actin, these two probes and the RIM1_total probe gave the strongest signal, with estimated mRNA content of ∼1.5% relative to β-actin (Fig. 1D).

The probes for RIM1α, RIM2α, and RIM2β gave weaker, but well-detectable, signals (Fig. 1, C and D). Taken together, qPCR indicates that both Rim1 and Rim2 genes are expressed in the cochlear nucleus, with a strong signal for the short RIM2γ isoform. In addition, the two long isoforms RIM1α, and RIM2α are also clearly expressed in VCN neurons (Fig. 1D).

RIM1 and RIM2 redundantly regulate presynaptic Ca²⁺ channel density and RRP.

To study the roles of the individual Rim1 and Rim2 genes in determining presynaptic Ca²⁺ channel density and RRP size at the calyx of Held synapse, we produced conditional single RIM1 or RIM2 KO mice (called RIM1 cKO and RIM2 cKO) and conditional RIM1/2 double KO mice (called RIM1/2 double cKO; see materials and methods). We made paired pre- and postsynaptic recordings at the calyx of Held synapse and applied long presynaptic depolarizations (50 ms to 0 mV) to study both presynaptic Ca²⁺ currents and transmitter release evoked by such pool-depleting stimuli (Fig. 2A). Ca²⁺ currents at 0 mV, close to the peak of current-voltage relationships, had large amplitudes in control mice (1.19 ± 0.32 nA; Fig. 2B1, black data points). In calyces of Held from RIM1 cKO mice, Ca²⁺ currents had a tendency to be larger than in control mice, whereas in RIM2 cKO mice Ca²⁺ currents were somewhat smaller, but neither of these two data sets was significantly different from the control group (P > 0.05; Fig. 2B1). To obtain Ca²⁺ current densities, we normalized the absolute Ca²⁺ current amplitudes by the measured presynaptic membrane capacitance (C_m), which is a proxy of membrane area (Fig. 2B2; see materials and methods). This revealed an unchanged Ca²⁺ current density in RIM1 cKO mice, whereas RIM2 cKO mice had smaller Ca²⁺ current densities compared with control mice (P < 0.05). In RIM1/2 double cKO terminals, there was a more marked, ∼40% reduction of presynaptic Ca²⁺ current density (Fig. 2B3; P < 0.01), in good agreement with our previous observations (Han et al. 2011). Taken together, conditional inactivation of RIM1 alone did not affect presynaptic Ca²⁺ channel density, and conditional deletion of RIM2 led to a partial deficiency in presynaptic Ca²⁺ channel density. Thus the Rim1 and Rim2 genes have redundant functions in determining the Ca²⁺ channel density at active zones of calyx of Held synapses, with a somewhat larger role for Rim2 (see discussion).

We next analyzed the amount and kinetics of transmitter release in response to pool-depleting long depolarizations (Fig. 2A, bottom). EPSCs were measured in the presence of CTZ (100 μM) and γ-DGG (2 mM) to suppress the desensitization and saturation of postsynaptic AMPA receptors (Neher and Sakaba 2001). Presynaptic depolarizations caused large EPSCs in RIM1 and RIM2 cKO mice, indistinguishable from EPSCs measured in control mice (Fig. 2A, middle, and C1). Similarly, the peak release rates and the number of vesicles released in the fast component were unchanged (Fig. 2, C2 and C3). In RIM1/2 double cKO mice, on the other hand, EPSCs in response to pool-depleting stimuli were strongly reduced in amplitude (Fig. 2C1). We showed previously that the miniature EPSC amplitudes were unchanged in RIM1/2 double cKO mice (Han et al. 2011); therefore, the reduced EPSC amplitudes reflect less transmitter release. Correspondingly, the peak release rates and the fast component of the RRP were smaller than in control mice (Fig. 2, A, C2, and C3). In addition, the EPSC rise times and the fast release time constants, which both reflect the immediate speed of transmitter release, were slowed in RIM1/2 double cKO mice (Fig. 2, A, C4, and C5). The amount of vesicles released in the slow component was unchanged in single RIM cKO (Fig. 2D2), but the slow release time constant was slowed in RIM1/2 double cKO mice (Fig. 2D1; P > 0.01). However, all of the latter parameters (EPSC rise times, fast and slow release time constants, and the number of fast and slowly released vesicles) were not significantly different between RIM1/2 single cKO and control mice (Fig. 2, C4, C5, D1, and D2). Thus removal of RIM1 or RIM2 alone did not significantly change RRP size nor the release kinetics of RRP vesicles.

RIM1 and RIM2 redundantly determine Ca²⁺ channel-vesicle coupling.

We showed above that Rim1 and Rim2 genes have a redundant role in determining RRP size and vesicle release kinetics (Fig. 2). It is possible, however, that removal of single Rim1/2 genes affect the spatial coupling between readily releasable vesicles and Ca²⁺ channels in more subtle ways. To test this possibility, we investigated Ca²⁺ current-release coupling in RIM1/2 single cKO mice, as well as in RIM1/2 double cKO and control mice. We used brief ramp-like depolarizations with different durations at +60 mV (Fig. 3A, top; see materials and methods), thereby varying the number of Ca²⁺ channels that were open during the repolarization phase (Ca²⁺ “tail” current protocol; Borst and Sakmann 1999; Fedchyshyn and Wang 2005). In recordings of RIM1 and RIM2 single cKO mice, such pulses evoked well-measurable Ca²⁺ tail currents and EPSCs, similar to control mice (Fig. 3A). In double-logarithmic plots of EPSC amplitude vs. presynaptic Ca²⁺ charge (Q_Ca)_a, we found steep relations for both the RIM1 and the RIM2 cKO data sets, which largely overlayed with the data set obtained in control mice (Fig. 3, B1 and B2). For each data set, we analyzed the slope (“Ca²⁺ current cooperativity”) by line fits in the double logarithmic space, and the Q_Ca value for which an EPSC of 2 nA was observed. While the slope was 4.6 ± 0.6 for the control data set (n = 11 paired recordings), it was slightly reduced in RIM1 and RIM2 cKO mice (to 3.9 ± 0.7 and 4.3 ± 0.4, respectively; n = 5 and 4, respectively), but these changes were not statistically significant (Fig. 3C; P > 0.05). Similarly, the Q_Ca value for an EPSC of 2 nA was slightly increased in RIM1 cKO mice, but this comparison did not reach statistical significance (Fig. 3D; P > 0.05). Therefore, based on Ca²⁺ tail current experiments in paired recordings, we conclude that removal of RIM1 or RIM2 alone does not significantly change the coupling efficiency between Ca²⁺ channels and readily releasable vesicles at the calyx synapse.

For the RIM1/2 double cKO synapses, Ca²⁺ tail currents were reduced, and EPSC amplitudes for maximal tail currents were always <0.5 nA (Fig. 3A; note different scale bars for control and RIM1/2 double cKO mice). The smaller Ca²⁺ tail currents are expected because of the smaller Ca²⁺ current amplitudes in calyx of Held terminals in RIM1/2 double cKO mice (Fig. 2). The small amplitudes of EPSCs did not allow us to measure reliable slope values between EPSC amplitudes and Q_Ca (Fig. 3B1, red data points). The low efficiency of action potential-like depolarizations in inducing phasic release in RIM1/2 double cKO synapses is consistent with a slowed intrinsic release speed and an additional coupling deficit in RIM1/2 double cKO mice (Han et al. 2011).

A small decrease in the cleft glutamate concentration in RIM1/2 double cKO synapses.

In the absence of both RIM1 and RIM2, calyx synapses have a reduced pool size as tested by prolonged presynaptic depolarizations (Fig. 2), and the number of docked vesicles at the active zone was reduced, whereas active zone size was unchanged (Han et al. 2011). Thus RIM1 and RIM2 proteins ensure a high density of docked vesicles at each active zone, and in turn the probability of release at a given active zone should be higher in wild-type mice than in RIM1/2 double cKO synapses. If a significant amount of multivesicular release happens at single active zones in control mice, then the decreased release probability in RIM1/2 double cKO mice might decrease the amount of multivesicular release and, therefore, the cleft glutamate concentration. To test this prediction, we measured the blocking efficiency of a low-affinity competitive antagonist at AMPA receptors, γ-DGG, in RIM1/2 double cKO and control synapses. The amount of block by such rapid-off antagonists depends on the effective glutamate concentration in the synaptic cleft (Clements et al. 1992; Wadiche and Jahr 2001).

Fiber stimulation evoked EPSCs recorded at calyx of Held synapses of control mice were large (7.0 ± 4.8 nA; n = 8 recordings), and 500 μM γ-DGG blocked the EPSCs reversibly to 42 ± 4% of their amplitude under control conditions (Fig. 4, A, C, and D). In RIM1/2 double cKO mice, EPSCs had amplitudes of only 1.3 ± 1.3 nA, much smaller than in control mice (P < 0.05). The rapid-off antagonist γ-DGG reversibly blocked EPSCs in RIM1/2 double cKO mice to 35 ± 8% of their control amplitude (Fig. 4, B–D; n = 9); this block was slightly, but significantly, stronger than in control mice (Fig. 4D; P < 0.05). This indicates that the amount of postsynaptic receptor saturation is lower in RIM1/2 double cKO mice compared with control, suggesting that less multivesicular release occurs in RIM1/2 double cKO synapses than in wild-type mice. The relatively small difference of the blocking efficiency of γ-DGG might indicate that in wild-type calyx synapses the probability of multivesicular release is not high under our recording conditions. Indeed, a recent study showed that the amount of multivesicular release depends on the release probability as set by the extracellular Ca²⁺ concentration and that at 2 mM Ca²⁺, the condition used here, multivesicular release starts to appear (Budisantoso et al. 2013).

DISCUSSION

Using conditional genetic elimination of the single Rim1 or Rim2 gene, or of the combination of both genes, we find that these genes have largely redundant presynaptic functions at the calyx of Held synapse. The mRNA transcripts of various RIM1 and RIM2 isoforms were expressed in the cochlear nucleus of young hearing mice (Fig. 1). Genetic inactivation of the individual Rim1 or Rim2 gene did not lead to noticeable changes in the RRP of vesicles as studied by prolonged presynaptic depolarizations (Fig. 2) nor in the efficiency of brief presynaptic depolarizations to induce release (Fig. 3). Similarly, the presynaptic Ca²⁺ current density was largely unchanged in the single KOs, although RIM2 inactivation reduced the Ca²⁺ channel density to some degree, maybe indicating a more prominent role for RIM2 in anchoring Ca²⁺ channels at the active zone. Overall, we conclude that at a large excitatory hindbrain synapse, the calyx of Held, RIM1 and RIM2 have redundant functions in determining the RRP and in anchoring presynaptic Ca²⁺ channels to the active zone.

Role of RIM proteins for anchoring Ca²⁺ channels at the active zone.

Experiments in cultured hippocampal neurons and at the calyx of Held, a synapse uniquely accessible to direct presynaptic recordings, have shown that RIM proteins determine a high density of presynaptic voltage-gated Ca²⁺ channels (Han et al. 2011; Kaeser et al. 2011). These observations, together with biochemical experiments, have shown that RIM proteins anchor Ca²⁺ channels at the active zone, probably via a direct interaction of the central PDZ domain with voltage-gated Ca²⁺ channel α-subunits (Kaeser et al. 2011). At the calyx synapse the overall Ca²⁺ current density was significantly reduced in RIM1/2 double cKO calyces (∼40%), but the voltage dependence of Ca²⁺ current activation was unchanged, and a large part of the Ca²⁺ currents in RIM1/2 double cKO mice was mediated by ω-agatoxin sensitive P/Q-type channels, similarly as in wild-type mice (Han et al. 2011). Together with the previous results, this suggests that RIM proteins contribute to anchoring Ca²⁺ channels at the presynaptic active zone. Earlier studies have shown an interaction between the C2B domain of RIM1α with the accessory Ca²⁺ channel β4-subunit, and this interaction was also found for the C2B domains of the short RIM2, RIM3, and RIM4 γ-isoforms. This interaction was found to modulate the inactivation of Ca²⁺ currents in heterologous expression systems (Kiyonaka et al. 2007; Uriu et al. 2010). However, this function of RIM proteins is independent from the Ca²⁺ channel anchoring function of the long RIM1 and RIM2 isoforms and cannot account for the presynaptic deficits observed in RIM1/2 KO synapses (Kaeser et al. 2012).

When we inactivated either the Rim1 or the Rim2 gene alone, we found an unchanged presynaptic Ca²⁺ current density in the RIM1 cKO mice and a somewhat smaller presynaptic Ca²⁺ current in RIM2 cKO mice (Fig. 2). Thus it seems that at the calyx of Held, the RIM2 protein is more relevant than RIM1 for anchoring Ca²⁺ channels at the active zone. However, the reduced Ca²⁺ channel density in RIM2 cKO mice did not go along with significant changes of maximal EPSC amplitudes, pool size, or the release efficiency of Ca²⁺ tail currents (Figs. 2 and 3). Whole cell recordings of a nerve terminal cannot distinguish between Ca²⁺ currents arising from the active zone or from extrasynaptic membrane areas. Therefore, if genetic deletion of RIM2 had resulted in a stronger reduction of extrasynaptic Ca²⁺ channels compared with synaptic ones, then such a reduction would be functionally less relevant. Overall, therefore, the present work suggests that the Rim1 and Rim2 genes can replace each other in anchoring Ca²⁺ channels at the presynaptic active zone of the calyx synapse.

These findings are different from the situation at cerebellar parallel fiber synapses where constitutive inactivation of RIM1α alone led to a reduced release probability and reduced presynaptic bouton Ca²⁺ influx (Kintscher et al. 2013). A somewhat predominant role for the RIM2 isoforms in determining the Ca²⁺ channel density at the calyx synapse might also explain the slightly larger Ca²⁺ currents observed in the RIM1 cKO mice, if it is assumed that early tissue-specific inactivation of RIM1 would lead to some compensatory upregulation of RIM2. In wild-type mice, RIM2 mRNA transcripts were well detectable in the cochlear nucleus of P14 mice (Fig. 1). The Ca²⁺ channel anchoring function of RIM1 and RIM2 is mediated by the PxxP motif and PDZ domain (Kaeser et al. 2011) (Fig. 1). Thus, at the calyx of Held synapse, all isoforms that contain at least the PDZ domain and the PxxP motif (RIM1α and RIM1β; and RIM2α and RIM2β; see Fig. 1) could function redundantly in anchoring Ca²⁺ channels at the presynaptic active zone.

Redundancy of RIM1 and RIM2 function depends on synapse identity.

We found that the RRP, studied by prolonged presynaptic depolarizations, was not significantly affected in the single RIM1 or RIM2 cKO mice, whereas the RIM1/2 double cKO synapses showed a significantly reduced RRP and slowed release kinetics (Fig. 2; Han et al. 2011). Removal of all RIM1/2 isoforms at the calyx of Held also led to a strong reduction of the docked vesicle number, which paralleled the reduced pool size (Han et al. 2011). Therefore, a large component of the reduction in RRP likely reflects a reduced number of docked vesicles in the double cKO mice. Here, we studied the RRP functionally, and we therefore cannot distinguish between possible docking and priming roles of RIM proteins. Our data show that in RIM1 and RIM2 single cKO synapses, the RRP was not affected, suggesting that neither docking nor priming is strongly affected by inactivating the individual Rim1/2 genes. Previous work showed that the priming function of RIM1 depends on the interaction of the NH₂-terminal Zn-finger domain with Munc13-1 (Betz et al. 2001; Deng et al. 2011; Dulubova et al. 2005); the latter domain is present in RIM1α, RIM1β, and RIM2α (Fig. 1). This suggests that in the conditional RIM1 KO mice in which expression of both the RIM1α and RIM1β isoforms is abrogated (Kaeser et al. 2008b), RIM2α can replace the docking/priming function of the RIM1 isoforms.

Our results, together with recent reports, suggest that the redundancy of RIM isoforms depends on synapse identity. Whereas at the calyx synapse (this study) and in cultured hippocampal synapses (Kaeser et al. 2012) a predominant role of the RIM1 isoforms in anchoring Ca²⁺ channels at active zones was not apparent, a reduction of axonal Ca²⁺ transients was found in cerebellar synapses of conventional RIM1α KO mice (Kintscher et al. 2013). In agreement with the absence of a dominant role for RIM1α at the calyx, we found earlier that constitutive RIM1α KO mice do not have an obvious release phenotype in fiber stimulation experiments at the calyx of Held (F. Felmy and R. Schneggenburger, unpublished observations). This again suggests at most a redundant role of RIM1α with either RIM1β or the RIM2 isoforms at the calyx of Held.

Taken together, experiments with single Rim 1/2 gene inactivation at the large calyx of Held synapse show largely overlapping roles of the two major Rim genes in vesicle priming/docking. The presynaptic Ca²⁺ channel anchoring function also depended on both genes, with a slight predominance of the Rim2 gene. Compared with recent work (Kaeser et al. 2012; Kintscher et al. 2013), it is seen that the redundancy of Rim gene function likely depends on the expression regulation of RIM proteins in individual neuron and synapse types. It is possible that the fast releasing calyx synapse makes use of most of the RIM1 and RIM2 isoforms to maximize fast and fail-safe synaptic transmission (Taschenberger et al. 2002). On the other hand, at some synapses with presynaptic forms of long-term plasticity, fewer RIM isoforms might be present and maybe predominantly those that are permissive for long-term regulation of release (Castillo et al. 2002; Kaeser et al. 2008a). Future studies could disentangle the role of individual RIM protein isoforms in long-term regulation of transmitter release, and the mechanisms that control the degree of coexpression of distinct RIM isoforms in specific neurons and at specific synaptic connections.

GRANTS

The work was supported by Swiss National Science Foundation Grant 3100A0–114069 (to R. Schneggenburger).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: Y.H., P.K., T.C.S., and R.S. conception and design of research; Y.H. and N.B. performed experiments; Y.H. and N.B. analyzed data; Y.H., N.B., P.K., and R.S. interpreted results of experiments; Y.H., N.B., and R.S. prepared figures; Y.H. and R.S. drafted manuscript; Y.H., P.K., and R.S. edited and revised manuscript; Y.H., N.B., P.K., T.C.S., and R.S. approved final version of manuscript.