RIM proteins tether Ca²⁺-channels to presynaptic active zones via a direct PDZ-domain interaction

SUMMARY

At a synapse, fast synchronous neurotransmitter release requires localization of Ca²⁺-channels to presynaptic active zones. How Ca²⁺-channels are recruited to active zones, however, remains unknown. Using unbiased yeast two-hybrid screens, we here identify a direct interaction of the central PDZ-domain of the active-zone protein RIM with the C-termini of presynaptic N- and P/Q-type Ca²⁺-channels, but not L-type Ca²⁺-channels. To test the physiological significance of this interaction, we generated conditional knockout mice lacking all presynaptic RIM isoforms. Deletion of all RIMs ablated most neurotransmitter release by simultaneously impairing the priming of synaptic vesicles and by decreasing the presynaptic localization of Ca²⁺-channels. Strikingly, rescue of the decreased Ca²⁺-channel localization required the RIM PDZ-domain, whereas rescue of vesicle priming required the RIM N-terminus. We propose that RIMs tether N- and P/Q-type Ca²⁺-channels to presynaptic active zones via a direct PDZ-domain mediated interaction, thereby enabling fast, synchronous triggering of neurotransmitter release at a synapse.

INTRODUCTION

At a synapse, action potentials induce Ca²⁺-influx into a presynaptic terminal, which triggers rapid synchronous neurotransmitter release, thereby initiating synaptic transmission (Katz, 1969). Release is mediated by synaptic vesicle exocytosis at the active zone, a specialized region of the presynaptic plasma membrane that docks and primes vesicles for exocytosis (Wojcik and Brose, 2007). Fast synchronous release requires co-localization of Ca²⁺-channels with the release machinery at the active zone (Llinas et al., 1992; Meinrenken et al., 2002). Voltage-gated Ca²⁺-channels consist of a pore-forming α1-subunit and accessory β- and α2δ-subunits (Catterall et al., 2005). Presynaptic neurotransmitter release almost exclusively depends on N- and P/Q-type Ca²⁺-channels containing Ca_V2.1 and Ca_V2.2 α-subunits, respectively; R-type Ca²⁺-channels containing Ca_V2.3 α1-subunits may also contribute, whereas L-type and T-type Ca²⁺-channels containing Ca_V1 and Ca_V3 α1-subunits do not (Castillo et al., 1994; Dietrich et al., 2003; Luebke et al., 1993; Poncer et al., 1997; Regehr and Mintz, 1994; Takahashi and Momiyama, 1993; Wu et al., 1999). However, how N- and P/Q-type Ca²⁺-channels are specifically localized to active zones and coupled to the release machinery is unknown.

Active zones are composed of evolutionarily conserved proteins, including Munc13s, RIMs, RIM-BPs (RIM-binding proteins), ELKS’s, and α-liprins (Wojcik and Brose, 2007). Of these proteins, RIMs are likely the central organizers because they directly or indirectly interact with all other known active zone proteins and with synaptic vesicles (Mittelstaedt et al., 2010). RIM proteins are expressed in three principal isoforms (Kaeser et al., 2008; Wang et al., 1997; Wang et al., 2000): RIM1α and RIM2α that contain all RIM domains (i.e., N-terminal Rab3- and Munc13-binding sequences, central PDZ-domains, and C-terminal C₂A- and C₂B-domains with an intercalated PxxP sequence that binds to RIM-BPs); RIM1β and RIM2β that are identical to RIM1α and RIM2α but lack the N-terminal Rab3-binding sequences; and RIM2γ, RIM3γ, and RIM4γ that are composed only of C₂B-domains, and are not considered here further. Genetic experiments in C. elegans and mice revealed that RIMs are essential for synaptic vesicle docking and priming and for presynaptic plasticity (Castillo et al., 2002; Fourcaudot et al., 2008; Gracheva et al., 2008; Kaeser et al., 2008; Koushika et al., 2001; Schoch et al., 2002; Schoch et al., 2006), but their mechanism of action remains unclear.

Several presynaptic proteins were shown to interact with Ca²⁺-channels. However, none of the reported interactions is selective for N- and P/Q-type Ca²⁺-channels. For example, (i) the RIM C₂A- and C₂B-domains bind to α1-subunits of L- and N-type Ca²⁺-channels (Coppola et al., 2001), (ii) the RIM C₂B-domain interacts with the β4 Ca²⁺-channel subunit (Kiyonaka et al., 2007) that associates with all Ca²⁺-channels subtypes but is not required for neurotransmitter release (Qian and Noebels, 2000), and (iii) the proline-rich sequences of RIMs bind to RIM-BPs (Wang et al., 2000) that in turn bind to L-, N- and P/Q-type Ca²⁺-channels (Hibino et al., 2002). Thus, no plausible hypothesis at present suggests how N- and P/Q-type Ca²⁺-channels are selectively recruited to presynaptic active zones. More importantly, no presynaptic protein has been identified that is essential for recruiting N- and P/Q-type Ca²⁺-channels to presynaptic terminals. Only α-neurexins, which are presynaptic cell-adhesion molecules, were found to be required for presynaptic Ca²⁺-channel function (Missler et al., 2003). α-Neurexins, however, are also essential for the organization of other components of the presynaptic release machinery, and no molecular mechanism is known that links α-neurexins to active zones or Ca²⁺-channels.

Using an unbiased yeast two-hybrid screen, we here identify a direct interaction of P/Q- and N- type Ca²⁺-channels with RIM PDZ-domains. To test whether RIMs act to localize Ca²⁺-channel to active zones, and whether this function requires the PDZ-domains of RIMs, we generated conditional double knockout (KO) mice of all RIM isoforms that contain PDZ-domains. Using these mice, we show by electrophysiological recordings, Ca²⁺-imaging, and quantitative immunostaining of Ca²⁺-channels that RIMs are essential for localizing Ca²⁺-channels to release sites. Moreover, we show that only two RIM sequences are required for localization of Ca²⁺-influx to active zones, their PDZ-domains that bind to Ca²⁺-channels, and their proline-rich sequences that bind to RIM-BPs which in turn bind to Ca²⁺-channels. Thus, we propose that RIMs tether Ca²⁺-channels to active zones via two parallel but essential interactions, direct binding of Ca²⁺-channels to the RIM PDZ-domains that is specific for N- and P/Q-type Ca²⁺-channels, and indirect binding of Ca²⁺-channels to RIMs via RIM-BPs that is shared among different types of Ca²⁺-channels.

RESULTS

A screen for synaptic proteins binding to Ca²⁺-channels

We performed yeast-two hybrid screens for proteins that interact with the cytoplasmic C-terminal sequences of P/Q- and N-type Ca²⁺-channels. We chose these Ca²⁺-channel baits because P/Q- and N-type Ca²⁺-channels mediate nearly all presynaptic Ca²⁺-influx, and their C-termini have been implicated in targeting Ca²⁺-channels to the active zone (Catterall et al., 2005). The C-termini of N- and P/Q-type Ca²⁺-channels contain three conserved sequence motifs (Fig. 1A and Fig. S1A): an SH3-domain binding motif (RQLPGTP), a PNGY motif, and a C-terminal sequence motif (DxWC). Among 84 and 134 independent prey clones obtained in P/Q- and N-type Ca²⁺-channel screens, 33 and 16 prey clones, respectively, represented RIM-BPs (Fig. 1B), consistent with earlier studies (Hibino et al., 2002). In addition, 2 and 3 independent prey clones, respectively, contained RIM1 fragments, whose only overlapping sequence encoded the PDZ-domain (Fig. 1C).

Figure 1. Direct interaction of P/Q- and N-type Ca²⁺-channels with RIM PDZ-domains.

A. Structure of the α1-subunits of P/Q- and N-type Ca²⁺-channels. Following the 4 × 6 transmembrane regions (I-IV), P/Q-type and N-type Ca²⁺-channels contain a C-terminal cytoplasmic tail with conserved SH3-domain binding sequences (PxxP), PNGY motifs, and C-terminal sequence motifs (DxWC).

B & C, Summary of the RIM-BP (B) and RIM prey clones (C) isolated in yeast two-hybrid screens with the C-terminal sequences of N- and P/Q-type Ca²⁺-channels.

D-F, Liquid yeast-two hybrid assays with baits containing wild-type C-terminal sequence of the P/Q-type Ca²⁺-channel and the three indicated RIM prey clones (D); baits containing the C-terminal sequence of the N-type Ca²⁺-channel without or with point mutations in the PxxP (PxxP^M) or the PNGY sequence (PNGY^M), or with a deletion of the 4 C-terminal residues (ΔC^term), and the indicated RIM prey clones (E); and baits containing the indicated RIM1 domains (PDZ, PDZ-domain only; C₂A or C₂B, C₂A- or C₂B-domains only; C₂AB, both C₂-domains and the intercalated PxxP motif) and preys consisting of the wild-type C-terminal sequence of the N-type Ca²⁺-channel (F, left bars), or the indicated mutants of this Ca²⁺-channel (F, right bars). For all assays, pLexN served as a control; a.u. = arbitrary units; n.d. = not detectable (means ± SEMs).

G, Analysis of P/Q-type Ca²⁺-channel binding to the RIM1 PDZ-domain by NMR spectroscopy. ¹H-¹⁵N HSQC spectra of the ¹⁵N-labeled RIM1 PDZ-domain (38 μM) were acquired in the absence (black contours) and presence (blue contours) of unlabeled P/Q-peptide (0.1 mM). Selected cross-peak assignments from residues on the periphery of the binding site are indicated; cross-peaks from three lysine residues in the binding pocket that shift upon peptide binding are labeled in bold, underlined typeface (K651, K653 and K694).

H, Model of the RIM1 PDZ-domain (blue ribbon diagram) bound to the six C-terminal residues of the P/Q-type Ca²⁺-channel peptide represented as a stick model with color-coded atoms (carbon, yellow; oxygen, red; nitrogen, blue; sulfur, orange). Strand βB and helix αB, the two structural elements that line the peptide-binding site (Lu et al., 2005), are indicated.

I, Close-up view the surface of the RIM1 PDZ-domain peptide-binding pocket with the bound P/Q-type Ca²⁺-channel peptide (colors are identical to panel H). For additional ¹H-¹⁵N HSQC spectra and affinity measurements by isothermal titration calorimetry, see Figure S1.

The direct interaction of the RIM1 PDZ-domain with P/Q- and N-type Ca²⁺-channels was unexpected, prompting us to quantify it using liquid yeast two-hybrid assays. We found that the RIM1 PDZ-domain strongly and specifically bound to P/Q- and N-type Ca²⁺-channels, whereas no other RIM1 domain tested exhibited Ca²⁺-channel binding activity (Figs. 1D-1F). Mutations in the three conserved motifs of the cytoplasmic sequences of N- and P/Q-type Ca²⁺-channels showed that only the C-terminal sequence motif was essential for binding to RIM1 PDZ-domains, as predicted for a PDZ-domain interaction (Figs. 1E and 1F).

To validate the interaction of RIM PDZ-domains with Ca²⁺-channels by an independent method, we employed NMR spectroscopy. We produced recombinant ¹⁵N-labeled RIM1 PDZ-domain, and acquired ¹H-¹⁵N heteronuclear single quantum coherence (HSQC) spectra in the absence or presence of unlabeled peptide from the C-terminus of the P/Q-type Ca²⁺-channel (residues RHDAYSESEDDWC). Previous studies of the ¹H-¹⁵N HSQC spectrum of the RIM1 PDZ-domain (Lu et al., 2005) allowed us to assign cross-peak shifts induced by the Ca²⁺-channel peptide to specific residues in the RIM PDZ-domain (see supplemental methods). We found that the C-terminus of P/Q-type Ca²⁺-channels directly bound to the ligand-binding pocket of the RIM1 PDZ-domain (Figs. 1G and S1B). Additional ¹H-¹⁵N HSQC experiments revealed analogous binding of the C-terminus of N-type Ca²⁺-channels to the RIM1 PDZ-domain, and of Ca²⁺-channels to the RIM2 PDZ-domain (Figs. S1C and S1D).

An atomic model suggests that the Ca²⁺-channel sequence fits well into the PDZ-domain binding pocket (Figs. 1H and 1I). The binding envisioned in this model agrees with the shifts we observed in the HSQC spectra in cross-peaks from residues such as K651, K653 and K694, which contribute to the binding pocket of the RIM1 PDZ-domain, and move upon P/Q-type Ca²⁺-channel peptide binding (Fig. 1G). Furthermore, we confirmed these interactions by isothermal calorimetry with the RIM1 PDZ-domain (Figs. S1E and S1F), yielding dissociation constants similar to those of other PDZ-domain interactions (P/Q-type Ca²⁺-channel, 10.3 ± 0.6 μM (N = 0.93); N-type Ca²⁺-channel, 23.4 ± 1.7 μM (N =0.93); Wiedemann et al., 2004). Thus, RIM PDZ-domains stoichiometrically interact with the C-terminal sequences of N- and P/Q-type Ca²⁺-channels, which are not found in L- and T-type Ca²⁺-channels.

Conditional KO mice for presynaptic RIM isoforms

The binding of RIM PDZ-domains to Ca²⁺-channels was unexpected because RIM PDZ-domains interact with the active-zone protein ELKS (a.k.a. Rab6-binding protein, CAST, or ERC; Ohtsuka et al., 2002; Wang et al., 2002). To test whether the RIM PDZ-domain physiologically binds to presynaptic Ca²⁺-channels, we generated conditional double KO mice in which all RIM isoforms containing PDZ-domains (RIM1α, 1β, 2α, and 2β) can be deleted by cre-recombinase (Figs. 2A-2C and S2), thereby enabling us to avoid the lethality of RIM-deficient mice (Kaeser et al., 2008; Schoch et al., 2006). We cultured neurons from newborn conditional double KO mice, and infected these neurons with lentiviruses expressing EGFP-tagged active or inactive cre-recombinases. Rescue experiments were performed by co-expressing various RIM proteins from the same lentiviruses via an IRES sequence (Kaeser et al., 2009).

Figure 2. Conditional deletion of RIM proteins in mice.

A, Structure of the RIM2 gene (a.k.a. Rims2). Exons are shown as black boxes and numbered, positions of exons containing the initiator codons for RIM2α, RIM2β and RIM2γ are labeled 1’, 1” and 1’”, respectively. The first exon that is shared by all RIM2 isoforms (exon 26) was used for gene targeting in the conditional RIM2αβγ KO mice (shaded blue area).

B, RIM2αβγ targeting strategy. The diagram shows (from top to bottom) an expanded map of the RIM2 gene surrounding exon 26; the targeting vector (C = ECFP-tetracysteine tag in exon 26; blue triangles = loxP sites; N = neomycin resistance cassette; green circles = frt recombination sites; DT = diphtheria toxin gene cassette); the knockin allele (KI); the RIM2αβγ^floxed allele (neomycin resistance cassette was removed by flp-recombination); and the KO allele (cre recombination deleted exon 26, creating a non-translated, unstable mRNA).

C, Domain structures of RIM1α, 1β, 2α, 2β, and 2γ that are deleted in the RIM1/RIM2 conditional double KO neurons. Coils surrounding the N-terminal Zn²⁺-finger domain (Zn) signify Rab3-binding sequences.

D, Representative immunoblots of RIM1 and RIM2 proteins in cultured hippocampal neurons from RIM1/RIM2 double conditional KO mice infected with lentiviruses expressing inactive (control) or active cre-recombinase (cDKO). Neurons were infected on DIV3, and analyzed at the indicated times (DIV6-DIV14).

E, Representative images of cDKO and control neurons stained with antibodies to MAP2 (green) and synapsin (red). Scale bar = 5 μm, applies to all images.

F, Quantitations of the size and density of synapses analyzed as shown in E (control, n=18 neurons/3 independent cultures; cDKO neurons, 17/3).

G, Electron micrographs of osmium tetroxide- (top) or phosphotungstic acid-stained (bottom) control and cDKO neurons (scale bars, 200 nm).

H, Quantitations of synaptic ultrastructure in electron micrographs. Docked vesicles are defined as vesicles touching the plasma membrane.

Data in F and H show means ± SEMs. Statistical significance by Student’s t-test : ***, p<0.001.; for additional detailed information, see Fig. S2 and Table S1.

Immunoblotting demonstrated that after 10 days in vitro (DIV10), neurons expressing active cre-recombinase (referred to as cDKO neurons) lack RIM proteins, whereas neurons expressing inactive cre-recombinase (referred to as controls) retain RIM expression (Fig. 2D). Despite lacking RIM proteins, cDKO neurons exhibited an overall normal morphology with unchanged synapse size and density (Fig. 2E, 2F). Electron microscopy revealed that in cDKO neurons, the number of docked synaptic vesicles per active zone was decreased nearly two-fold (Fig. 2G, 2H), consistent with a role for RIM in vesicle docking (Gracheva et al., 2008). All other measured parameters were unchanged, and removal of RIMs did not alter the structure of presynaptic dense projections visualized by phosphotungstic acid staining (Fig. 2G).

RIM deletion severely impairs neurotransmitter release

Electrophysiologically, deletion of RIM proteins caused a 3-4 fold decrease in the frequency of spontaneous ‘minis’, in the amplitude of post-synaptic currents evoked by isolated action potentials, and in the size of the readily releasable pool measured by application of hypertonic sucrose (Figs. 3A-3D, and L.D., P.S.K & T.C.S., unpublished data). Moreover, deletion of RIMs significantly decelerated and desynchronized release, as evidenced by an increase in rise times and in rise-time variability (Figs. 3E-3G), but did not change the relative contributions of P/Q- and N-type Ca²⁺-channels to evoked synaptic responses (Figs. S3A and S3B). Similarly, the RIM deletion massively decreased release induced by short stimulus trains (30 stimuli at 10 Hz; Figs. 3H and 3I), decelerated release as manifested by a relative increase in delayed release (Figs. 3J and 3K), and again strongly desynchronized release (Figs. 3L and 3M). These data suggest that RIMs are not only essential for vesicle priming, a previously identified RIM function that partly accounts for the decrease in release (Koushika et al., 2001; Schoch et al., 2002), but also for the synchronous timing of fast release that is not accounted for by a priming deficit. Since RIMs directly bind to Ca²⁺-channels (Fig. 1), and a loss of Ca²⁺-channels from presynaptic terminals would explain the impaired synchronous timing of release, we hypothesized that RIM binding to N- and P/Q-type Ca²⁺-channels tethers Ca²⁺-channels to the active zone, thereby increasing the efficiency, speed, and synchrony of release.

Figure 3. RIM deletion decreases, decelerates, and desynchronizes neurotransmitter release.

A & B, Excitatory synaptic responses in cultured hippocampal control and cDKO neurons evoked by an action potential (A) or hypertonic sucrose application (B) (left, representative traces; right, summary graphs of amplitudes and charges; A: control, n=8 neurons/3 independent neuronal cultures; cDKO, n=9/3, B: control, n=9/3; cDKO, n=10/3).

C & D, Inhibitory synaptic responses evoked by an action potential (C) or hypertonic sucrose application (D) (C: control, n=21/4; cDKO, n=18/4; D: control, n=11/3; cDKO, n=11/3).

E-G, Analysis of the kinetics of isolated IPSCs (E, representative traces from control and cDKO neurons; F, 20-80% IPSC rise times; and G, rise time variability as expressed by the standard deviation (SD) of the 20-80% rise time [control, n=31/6; cDKO, n=36/6]).

H-K, Synaptic responses elicited by 10 Hz stimulus train in cDKO and control neurons (H, representative IPSCs; I-K, summary graphs of the synaptic charge transfer for the first IPSC (I) and for delayed release (J; release starting 100 ms after the last stimulus), and of the ratio of delayed release/first response (K; control, n=20/4; cDKO, n=21/4).

L & M, Analysis of the kinetics of IPSCs during 10 Hz stimulus trains (L, representative traces for the first 10 IPSCs during a 10 Hz stimulus train [top, first response indicated as a thick line, later responses represented as thin lines], and 20-80% rise times for three sample trains [bottom]; M, standard deviation (SD) of the 20-80% rise times during the 10 Hz stimulus train as a measure of synchrony; control, n=7/3; cDKO, n=9/3).

N-P, Time course of the decrease in IPSCs induced by addition of the membrane-permeable Ca²⁺-chelator EGTA-AM (10 μM; N, sample traces; O, summary graphs; P, decay time constants; control, n=8/3; cDKO, n=9/3). Decay time constants τ were calculated by fitting individual experiments to a single exponential function.

All data are means ± SEMs; *, p<0.05, **, p<0.01, ***, p<0.001 as determined by Student’s t-test. Numerical values of electrophysiology results are in Table S2, further analysis of synaptic responses at elicited at 10 Hz in Fig. S3).

RIM deletion alters the Ca²⁺-dependence of release

To explore the hypothesis that RIMs localize Ca²⁺-channels to active zones, we first examined the speed with which addition of a membrane-permeable Ca²⁺-buffer (EGTA-AM) decreases Ca²⁺-triggered release, measured as the IPSC amplitude (see supplemental methods). EGTA-AM caused a significantly faster rate of decline in IPSC amplitude in RIM-deficient cDKO neurons than in control neurons (Figs. 3N-3P). Thus, EGTA-dependent chelation of Ca²⁺ inhibits release more effectively in RIM-deficient than in control neurons, consistent with a longer distance between Ca²⁺-channels and release sites in RIM-deficient synapses.

We next examined the dependence of release in RIM-deficient synapses on the extracellular Ca²⁺-concentration ([Ca²⁺]_ex; Fig. 4). If RIM-deficient presynaptic terminals contain fewer tethered Ca²⁺-channels, presynaptic Ca²⁺-influx should be decreased, and more [Ca²⁺]_exshould be required for equivalent amounts of release – i.e., release should exhibit a higher [Ca²⁺]_ex-dependence without a change in apparent Ca²⁺-cooperativity (see supplemental methods). However, since Ca²⁺-influx through Ca²⁺-channels saturates at high [Ca²⁺]_ex (Church and Stanley, 1996; Schneggenburger et al., 1999), [Ca²⁺]_ex-titrations underestimate the change in Ca²⁺-dependence of release in mutant synapses, and only relative changes are interpretable.

Figure 4. Mutational dissection of RIM KO phenotype.

A, Diagram of RIM rescue proteins expressed in cDKO neurons via an IRES sequence from the same mRNA as cre-recombinase. The single-letter code above the RIM1α diagram identifies the various domains (R, Rab3-binding α-helical region; Z, Zn-finger region, P, PDZ-domain; A, C₂A-domain; S, proline-rich SH3-binding PxxP motif; B, C₂B-domain); H marks the presence of a human influenza hemagglutinin (HA)-tag.

B, Representative traces of IPSCs evoked at the indicated extracellular Ca²⁺-concentrations [Ca²⁺]_ex in control neurons, cDKO neurons without rescue, cDKO neurons with full-length RIM1α rescue, and cDKO neurons with rescue with the RIM-RZ or the RIM-PASB fragments. Each rescue experiment was performed with independent control groups.

C-K, Summary plots of absolute (C, F, and I) and normalized IPSC amplitudes (D, G, and J; normalized to the 10 mM [Ca²⁺]_ex response) evoked at the indicated [Ca²⁺]_ex, and summary graphs of the Ca²⁺-dependence of release (E, H, and K; expressed as the [Ca²⁺]_ex producing a half-maximal IPSC amplitude (EC₅₀), as determined by fitting in individual experiments the [Ca²⁺]_ex-dependence of the IPSC amplitude (Figs. D, G, J) to a Hill function). Control neurons and cDKO neurons were analyzed in comparison with cDKO neurons rescued with RIM1α (C-E), RIM-RZ (F-H), or RIM-PASB (I-K). C, D: n = 8 neurons/3 independent batches of culture in control, 6/3 in cDKO, 9/3 in cDKO + RIM1α; E, F: n = 7/3 in control, 7/3 in cDKO, 7/3 in cDKO + RIM-RZ; G, H: n = 6/3 in control, 5/3 in cDKO, 8/3 in cDKO + RIM-PASB.

L & M, Summary graphs of 20-80% rise times (L) and rise time variability (M) for the indicated rescue experiments at 2 mM [Ca²⁺]_ex (for sample traces, see Fig. S4O, n = see Fig. 4C-4K).

Data shown are means ± SEMs, ***, p<0.001 by one-way ANOVA, detailed statistical analysis for all data points can be found in Table S3. For Ca²⁺-cooperativity and I_max, see Fig. S4.

RIM-deficient cDKO neurons exhibited a large reduction in neurotransmitter release at all [Ca²⁺]_ex (Figs. 4B and 4C), and a major increase in the [Ca²⁺]_ex-dependence of neurotransmitter release (Fig. 4D). Both phenotypes were equally observed in inhibitory and excitatory synapses, and fully rescued by full-length wild-type RIM1α (Figs. 4A-4D and S4J-S4N). Fitting the [Ca²⁺]_ex-response curve of individual experiments to a Hill function showed that the RIM deletion increased the [Ca²⁺]_ex-requirement for release almost 2-fold, without changing the apparent Ca²⁺-cooperativity of release (Figs. 4E and S4H). Note that in the Hill function fits, the experimentally measured amplitudes suggest near-saturation at 10 mM [Ca²⁺]_ex, allowing direct comparison of the fitted parameters (Fig. S4I).

The [Ca²⁺]_ex-titration provided us with a facile assay to examine which RIM sequences determine vesicle priming and the [Ca²⁺]_ex-dependence of release. We first tested rescue of the RIM double KO phenotype by the N-terminal RIM-RZ fragment that contains the Rab3-binding (“R”) and Zn²⁺-finger domains (“Z”) of RIM1α, and that had been previously implicated in vesicle priming (Dulubova et al., 2005), and the C-terminal RIM-PASB fragment that contains its PDZ- (“P”), C2A- (“A”), PxxP- (“S” for SH3-domain binding) and C2B-domains (“B”)(Fig. 4A). At 2 mM [Ca²⁺]_ex, the RIM-RZ and RIM-PASB fragments each partially rescued the decrease in release, and alleviated the previously described decrease in Munc13 protein levels in cDKO neurons (Figs. 4F and 4I, S4D, Schoch et al, 2002). However, whereas the RIM-PASB fragment completely reversed the impairment of the [Ca²⁺]_ex-dependence of release in cDKO neurons, the RIM-RZ fragment did not (Figs. 4F-4K). RIM-RZ, conversely, completely rescued the readily-releasable pool, whereas RIM-PASB did not (L.D., P.S.K. & T.C.S., unpublished). Note that different from synaptotagmin mutations (Shin et al., 2009), none of the changes in RIMs changed the apparent Ca²⁺-cooperativity of release (Fig. S4H; see also Table S3).

We next tested whether RIM-RZ or RIM-PASB rescued the speed and synchrony of release. Consistent with the [Ca²⁺]_ex-titrations, the RIM-PASB fragment fully reversed the deceleration and desynchronization of release in cDKO neurons, as assessed at 2 mM [Ca²⁺]_ex, whereas the RIM-RZ fragment did not (Figs. 4L, 4M , and S4O-S4Q). Together, these data indicate that the N-terminal RIM domains function in vesicle docking and priming (Betz et al., 2001; Dulubova et al., 2005; Gracheva et al., 2008; Koushika et al., 2001; Schoch et al., 2002), whereas the C-terminal RIM domains function in the Ca²⁺-dependence and synchrony of release.

RIM deletion impairs presynaptic Ca²⁺-influx

The increased [Ca²⁺]_ex-dependence, decreased synchrony, and lowered speed of release in RIM-deficient neurons supports the hypothesis that RIMs localize Ca²⁺-channels to the active zone. However, these measurements are indirect. Their results could be explained by other hypotheses, for example that RIM directly regulates Ca²⁺-triggering of release by binding to synaptotagmin (Coppola et al., 2001; Schoch et al., 2002). To address this issue with an independent approach, we monitored action-potential induced Ca²⁺-transients by Ca²⁺-imaging in presynaptic boutons and postsynaptic dendrites.

We engineered new active and inactive EGFP-tagged cre-recombinase proteins that exhibit a tight nuclear localization (Figs. S5A-S5C), and expressed them in neurons from conditional KO mice. We then filled single neurons via a patch pipette with Alexa594 and the Ca²⁺-indicator Fluo5F, identified presynaptic axonal boutons and second-order dendrites by imaging Alexa594, elicited isolated action potentials by brief somatic current injections, and monitored the resulting Ca²⁺-transients in boutons and dendrites by imaging Fluo5F in line scans (at 333 Hz, ~100-150 μm from the cell body). To ensure that the observed Ca²⁺-transients were not due to passive depolarizations in response to the somatic current injections, we decreased in control experiments the injected current to a threshold level that only sometimes evoked an action potential. In these experiments, Ca²⁺-transients in boutons and dendrites strictly depended on the induction of action potentials, confirming that we were monitoring action potential-induced Ca²⁺-transients (Figs. S5D and S5E).

In RIM-containing control boutons, isolated action potentials induced a brief, ~100% increase in Fluo5F Ca²⁺-indicator fluorescence, whereas in RIM-deficient cDKO boutons, action potentials induced only a ~50% increase (Figs. 5A-5D). Similar to the impaired [Ca²⁺]_ex-dependence of release (Fig. 4), the decreased Ca²⁺-influx in RIM-deficient cDKO neurons was fully rescued by the C-terminal RIM-PASB fragment (Figs. 5A-5D). Deletion of RIM proteins did not alter dendritic Ca²⁺-transients (Fig. 5C, inset, and Figs. S5F-S5H), suggesting that the RIM deletion did not generally impair Ca²⁺-channel function or Ca²⁺-buffering. Moreover, we detected no change in expression levels of various Ca²⁺-channel subunits in cDKO neurons (Figs. 5E and S5I). Thus, RIM deletions selectively decreased presynaptic Ca²⁺-influx in hippocampal neurons.

Figure 5. RIM deletion decreases presynaptic Ca²⁺-transients.

A, Representative fluorescence images of control neurons, cDKO neurons, and cDKO neurons rescued with the C-terminal RIM-PASB fragment. Neurons were filled via a patch pipette with Fluo5F and Alexa594 (red); nuclear EGFP-fluorescence (produced by the active and inactive cre-recombinase EGFP-fusion proteins; see Figs. S5A-S5C) is shown in green; and coincident Alexa594 and EGFP- or Fluo5F signals are shown in yellow). Insets (bottom right) show areas in dotted rectangles containing a sample axonal bouton (grey lines = positions of the patch pipette; white lines = position of line scans for the Ca²⁺-transients shown in B). Scale bar (bottom left) = 20 μm.

B, Representative action potentials (top); line scans of Ca²⁺-transients in presynaptic boutons induced by these action potentials, and monitored via Fluo5F fluorescence (middle; colored white for better visibility); and quantitations of Ca²⁺-transients (bottom; averaged across the bouton).

C & D, Summary plots of action potential-induced changes in Ca²⁺-indicator fluorescence monitored in presynaptic boutons from control neurons, cDKO neurons, and cDKO neurons rescued with the C-terminal RIM-PASB fragment (C, time course of the Ca²⁺-indicator fluorescence (inset: the same plot for dendrites); D, the cumulative probability of the peak Ca²⁺-indicator fluorescence, expressed as ΔG/G₀). Data in C are means (line) ± SEMs (shaded area); ***, p<0.001 as assessed by two-way ANOVA for peak amplitudes during the first 60 ms after action potential induction (C) or by Kolmogorov-Smirnov test (D); control, n=45 boutons/10 neurons/4 independent cultures; cDKO, n=46/11/4; cDKO + RIM-PASB, n=44/11/4.

E, Immunoblot analysis of Ca²⁺-channel subunit levels in control and cDKO neurons. Blots were probed with antibodies to the indicated Ca²⁺-channel proteins (P/Q-type (Ca_V2.1-A and Ca_V2.1-S) and N-type (Ca_V2.2) α-subunits, and α2/δ and β4 subunits) and control proteins (GDI, GDP-dissociation inhibitor).

For analysis of dendritic Ca²⁺-transients, statistical values, and quantitative assessment of mRNA levels, see Fig. S5 and Table S4.

RIM PDZ-domain is required for localizing Ca²⁺-influx

We next tested whether the RIM PDZ-domain mediates the RIM-dependent presynaptic localization of Ca²⁺-influx, using systematic rescue experiments combined with [Ca²⁺]_ex-titrations as an assay.

Deletions of either the PDZ-domain or the PxxP-motif from the RIM-PASB fragment (Figs. 6A, top) blocked rescue of the impaired [Ca²⁺]_ex-dependence in RIM-deficient cDKO neurons, whereas deletion of the C₂A- or the C₂B-domain had no effect (Figs. 6A-6E and S6A-S6C). Moreover, expression of a RIM fragment composed only of the PDZ-domain and PxxP motif (‘RIM-PS’) in cDKO neurons fully rescued the impaired [Ca²⁺]_ex-dependence of release (Figs. 6F-6H and S6D-S6F). In contrast, expression of a RIM fragment composed of its two C₂-domains (‘RIM-AB’) did not rescue the [Ca²⁺]_ex-dependence of release, but partially reversed the decrease in IPSC size in RIM-deficient cDKO neurons, suggesting that the C₂-domains are dispensable for localizing presynaptic Ca²⁺-influx, but enhance the efficacy of release.

Figure 6. RIM PDZ-domain and PxxP-motif confer normal Ca²⁺-dependence to RIM-deficient synapses.

A, Domain structures of rescue proteins.

B, Sample traces of IPSCs in control neurons, cDKO neurons, and cDKO neurons rescued with the indicated proteins.

C-E, Systematic rescue analyses of the Ca²⁺-dependence of release in RIM-deficient cDKO neurons with RIM fragments containing three of the four RIM domains present in the RIM-PASB fragment. Absolute IPSC amplitudes (C), IPSC amplitudes normalized to the response at 10 mM [Ca²⁺]_ex (D), and apparent Ca²⁺-affinities (EC₅₀ values; E) are indicated (control, n=9 cells/3 independent batches of cultures; cDKO, n=9/3: cDKO+RIM-ASB, n=9/3; cDKO+RIM-PSB, n=8/3; cDKO+RIM-PAB, n=8/3; cDKO+RIM-PAS, n=10/3).

F-H, Rescue analyses of the Ca²⁺-dependence of release with RIM fragments containing either only the PDZ-domain and PxxP-motif (RIM-PS), or only the C₂A- and C₂B-domains (RIM-AB) of RIM1 (control, n=8/3; cDKO, n=7/3; cDKO+RIM-PS, n=9/3; cDKO+RIM-AB, n=9/3).

Data shown are means ± SEMs; ***, p<0.001 by one-way ANOVA. Cooperative factor n, and I_max can be found in Fig. S6, all numerical data are in Table S5.

To directly probe whether the RIM PDZ-domain is actually essential for localizing presynaptic Ca²⁺-influx to active zones, we next performed rescue experiments with full-length wild-type RIM1α or RIM1α lacking the PDZ-domain (RIM-ΔPDZ). Whereas the former rescued all RIM cDKO phenotypes (Figs. 4C-4E), the latter was unable to reverse the impairment in [Ca²⁺]_ex-dependence of release in RIM-deficient neurons, although it completely rescued the decrease in evoked IPSC amplitude at high [Ca²⁺]_ex-concentrations (Figs 7A-7E, S7A-S7C). Moreover, full-length but not PDZ-domain deleted RIM1α rescued the deceleration and desynchronization of release in RIM-deficient neurons (Figs. 4L and 4M, 7F-7H). Possibly most importantly, RIM1α lacking the PDZ-domain did not restore normal Ca²⁺-influx into presynaptic nerve terminals in RIM-deficient neurons as measured by Ca²⁺-imaging, whereas full-length RIM1α fully rescued Ca²⁺-influx (Figs. 7I and 7J, S7D and S7E). Thus, the RIM-PDZ domain is critical for localizing Ca²⁺-influx to presynaptic active zones, thereby promoting the normal Ca²⁺-dependence, speed, and precision of neurotransmitter release.

Figure 7. RIM function in localizing presynaptic Ca²⁺-influx requires its PDZ-domain.

A, Domain structures of rescue proteins.

B-E, Sample traces and quantitative analysis of Ca²⁺-dependence of release of IPSCs in control neurons, cDKO neurons, and cDKO neurons rescued with the PDZ-domain deficient RIM-ΔPDZ fragment. Absolute IPSC amplitudes (C), IPSC amplitudes normalized to the response at 10 mM [Ca²⁺]_ex (D), and apparent Ca²⁺-affinities (EC₅₀ values; E) are indicated (control, n=10/3; cDKO, n=9/3: cDKO+RIM-ΔPDZ, n= 10/3).

F-H, Speed and synchrony of neurotransmitter release in control neurons, cDKO neurons and cDKO neurons rescued with RIM-ΔPDZ (control, n=7/3; cDKO, n=10/3: cDKO+RIM-ΔPDZ, n=12/3).

I and J, Sample line scans (I) and summary data (J) of action potential evoked Ca²⁺-transients in presynaptic boutons of control neurons, cDKO neurons and cDKO neurons rescued with RIM1α or RIM-ΔPDZ. Data shown are means (line) ± SEMs (shaded area). (boutons: control, n=40 boutons/6 neurons/5 independent cultures; cDKO, n=57/7/5; cDKO+RIM1α, n=51/7/5, cDKO+RIM-ΔPDZ, n=52/7/5; dendrites: control, n=22/6/5; cDKO, n=22/7/5; cDKO+RIM1α, n=19/7/5, cDKO+RIM-ΔPDZ, n=22/7/5). For cumulative peak amplitudes and statistical values, see Fig. S7 and Table S6).

Statistical analyses: *, p<0.05; **, p<0.01; ***, p<0.001; E, G and H, one-way ANOVA ; J, two-way ANOVA for peak amplitudes during the first 60 ms after action potential induction.

The RIM PDZ-domain localizes P/Q-type Ca²⁺-channels to presynaptic termianls

The decrease in presynaptic Ca²⁺-influx in RIM-deficient terminals could be due to a loss of presynaptic Ca²⁺-channels, or to a decrease in their activity. To address this question, we measured by quantitative immunofluorescence presynaptic levels of P/Q-type Ca²⁺-channels, which mediate >80% of the synaptic responses we measure (Fig S3A-S3C). Strikingly, the RIM deletion reduced presynaptic P/Q-type Ca²⁺-channel levels ~40%, but had no effect on the active-zone protein bassoon (Figs. 8A, 8B, S8A and S8B).

Figure 8. RIM-dependent Ca²⁺-channel tethering linked to synaptic vesicle docking and priming.

A and B, Immunofluorescent stainings (A) and quantitative immuno-localization analyses (B) of P/Q-type Ca²⁺-channels (top panel in A) and presynaptic bassoon (bottom panel in A) in control and RIM-deficient cDKO neurons, and in cDKO neurons rescued with RIM1α or RIM-ΔPDZ (n=3 cultures per condition, *, p<0.05; **, p<0.01 by Student’s t-test compared to control, a second, independent experiment is found in Fig. S8 and Table S7).

C, Model of the presynaptic release machinery. The drawing illustrates the structures of major active zone proteins (RIMs, Munc13s, and RIM-BPs), P/Q- or N-type Ca²⁺-channels, a partially assembled SNARE-complex (composed of synaptobrevin/VAMP on synaptic vesicles and SNAP-25 and syntaxin-1 on the plasma membrane), Munc18-1, complexin, and key synaptic vesicle proteins (Rab3 and synaptotagmin-1 [Syt1]). Domain identification is provided on the top right. We propose that RIMs determine the specific localization of P/Q- and N-type Ca²⁺-channels at the active zone via a direct Ca²⁺-channel/PDZ-domain interaction, and via indirect binding of Ca²⁺-channels to RIMs via RIM-BPs (Hibino et al., 2002). In addition, RIMs form an N-terminal priming complex with Rab3 and Munc13, in which Munc13 likely acts by binding to SNARE complexes (not depicted due to restrictions of the 2-dimensional presentation). Synaptotagmin-1 on the vesicles serves as the Ca²⁺-sensor for exocytosis. With this architecture, Ca²⁺-channels and Ca²⁺-sensors are in close proximity, accounting for the speed, synchrony and extent of release.

Finally, to test whether the presynaptic localization of P/Q-type Ca²⁺-channels depends on the RIM PDZ-domain, we examined rescue of the Ca²⁺-channel localization deficit in RIM-deficient cDKO neurons with full-length wild-type RIM1α, or RIM1α lacking the PDZ-domain (Figs. 8A and 8B). In agreement with the electrophysiological and Ca²⁺-imaging results described above (Fig. 7), deletion of the PDZ-domain rendered RIM1α unable to rescue the decrease in Ca²⁺-channel levels in presynaptic terminals (Figs. 8A and 8B), suggesting that the RIM PDZ-domain is essential for tethering of Ca²⁺-channels to presynaptic terminals.

DISCUSSION

Based on protein/protein interaction studies (Fig. 1), generation of conditional KO mice (Fig. 2), electrophysiological recordings (Figs. 3, 4, 6, and 7), Ca²⁺-imaging (Figs. 5 and 7), and quantitative immunofluorescence (Fig. 8), we here propose that the PDZ-domains of RIM proteins stoichiometrically interact with N- and P/Q-type Ca²⁺-channels in vitro, that RIM proteins are essential for tethering Ca²⁺-channels to presynaptic terminals, and that the RIM PDZ-domain is required for this function. In addition, RIMs indirectly interact with Ca²⁺-channels via their RIM-BP binding sequence (Hibino et al., 2002), which we show is also essential. Thus, RIMs perform two parallel interactions with Ca²⁺-channels: a direct interaction via their PDZ-domains that is specific for N- and P/Q-type Ca²⁺-channels, and an indirect interaction via RIM-BPs that is not specific for N- and P/Q-type Ca²⁺-channels. Our results suggest a physiologically validated mechanism by which Ca²⁺-influx is localized to the active zone, as required for fast, synchronous triggering of neurotransmitter release (Fig. 8C).

The RIM PDZ-domain/Ca²⁺-channel interaction

We found that the C-terminal sequences of N- and P/Q-type Ca²⁺-channels specifically bind to RIM PDZ-domains. R-type Ca²⁺-channels have a similar C-terminal sequence, and may also interact, whereas L- and T-type Ca²⁺-channels do not (Figs. 1 and S1). The PDZ-domain/Ca²⁺-channel interaction was surprising because the PDZ-domain proteins Mints and CASK were previously shown to bind to Ca²⁺-channels (Maximov et al., 1999), and because the RIM PDZ-domain is known to bind to ELKS proteins (Ohtsuka et al., 2002; Wang et al., 2002). However, it remains unclear whether these previously described interactions are physiologically important; in fact, Mint- and CASK-deficient synapses exhibit multiple abnormalities that do not resemble a Ca²⁺-influx impairment (Atasoy et al., 2007; Ho et al., 2006), whereas ELKS2-deficient synapses display increased release at inhibitory synapses (Kaeser et al., 2009), and ELKS levels are not detectably changed in RIM cDKO neurons (data not shown). Thus, it seems unlikely that the effects we observe here are indirectly mediated via Mints, CASK, or ELKS. It is possible, however, that different PDZ-domain binding reactions compete with each other at the active zone. For example, ELKS-binding to RIM PDZ-domains may inhibit Ca²⁺-channel binding, and thereby attenuate neurotransmitter release; this inhibitory role of ELKS binding could be regulated during plasticity, which might account for the central role of RIM in short- and long-term plasticity (Castillo et al., 2002; Fourcaudot et al., 2008; Kaeser et al., 2008; Schoch et al., 2002).

The RIM PDZ-domain is essential for localizing Ca²⁺-channels to the active zone

Using rescue experiments in RIM-deficient neurons, we found that the RIM PDZ-domain was invariably required in various RIM rescue constructs to reverse the impairment in presynaptic Ca²⁺-influx in RIM-deficient neurons (Figs. 4-7), and for localizing P/Q-type Ca²⁺-channels to presynaptic boutons (Fig. 8). In addition, loss of RIM-BP binding sequences blocked rescue of Ca²⁺-influx (Fig. 6). These experiments suggest that RIMs tether Ca²⁺-channels to the active zone via two parallel interactions: directly by binding to Ca²⁺-channels via their PDZ-domains, and indirectly by binding to RIM-BPs which in turn bind to Ca²⁺-channels (see model in Fig. 8C, and reconstitution of the tripartite complex in Fig. S8C). Our data also account for the specificity of N- and P/Q-type Ca²⁺-channels in release, since the RIM PDZ-domain/Ca²⁺-channel interaction is specific for these Ca²⁺-channel types (Fig. 1).

Active zone functions of RIMs

Our findings corroborate the notion that RIM proteins are central organizers of active zones by showing that besides their role in vesicle docking and priming (Betz et al., 2001; Gracheva et al., 2008; Kaeser et al., 2008; Koushika et al., 2001; Schoch et al., 2002; Schoch et al., 2006), RIMs are essential for tethering Ca²⁺-channels to active zones (Figs. 3-8). Moreover, RIM proteins perform additional functions, as indicated by the fact that although the RIM C₂-domains had no detectable role in Ca²⁺-influx, they boosted neurotransmitter release, possibly by binding to α-liprins, the β4 Ca²⁺-channel subunit, or other proteins (Kiyonaka et al., 2007; Schoch et al., 2002). Thus, RIMs occupy the center of an interaction network in the molecular anatomy of the active zone, and influence all aspects of neurotransmitter release (Fig. 8C). However, our data also raise new questions. How do RIM proteins function in priming – only via their interaction with Munc13’s (Betz et al., 2001; Schoch et al., 2002), or via other effectors? How do RIM C₂-domains boost release without altering Ca²⁺-influx? Why does deletion of just one RIM isoform, RIM1α, which has only a partial effect on neurotransmitter release due to its redundancy with other RIM isoforms, block multiple forms of presynaptic long-term synaptic plasticity (Castillo et al., 2002; Fourcaudot et al., 2008; Kaeser et al., 2008)? With the availability of double conditional KO mice described here, these questions can now be addressed.

EXPERIMENTAL PROCEDURES

In vitro protein binding assays

Two yeast two-hybrid screens of a rat brain cDNA library using bait vectors encoding the Ca_V2.2 N-type Ca²⁺-channel C-terminus (residues 2163-2339) or Ca_V2.1 P/Q-type Ca²⁺-channel C-terminus (residues 2213-2368) and liquid yeast to hybrid essays were performed as described (Wang et al., 1997). Of the 134/84 isolates with the N-type/P/Q-type bait, 8/16 clones corresponded to RIM-BP1, 8/17 to RIM-BP2, and 3/2 to RIM1. For mapping of the interaction region and the relative strength, yeast strain L40 was co-transformed with the various Ca²⁺-channel or RIM1 bait vectors and the Ca²⁺-channel or RIM1 prey vectors. HSQC spectroscopy was performed with rat RIM1 (residues 596-704, expressed as described (Lu et al., 2005)), and non-labeled P/Q-type Ca²⁺-channel peptides were synthesized. ¹H–¹⁵N HSQC spectra were acquired in a Varian Inova500 spectrometer at 40-200 μM protein.

Generation of double conditional RIM KO mice

The RIM2αβγ targeting vector was constructed from a λ-phage DNA clone isolated from a genomic library, and conditional RIM2αβγ KO mice were generated by homologous recombination in R1 embryonic stem cells. The recombined stem cells were used for blastocyst injections to obtain chimeric mice. After germline transmission of the mutant allele, the newly generated conditional RIM2αβγ KO mice were crossed to conditional RIM1αβ KO mice (Kaeser et al., 2008).

Electrophysiology

Whole-cell patch-clamp recordings were performed in cultured hippocampal neurons at DIV13-15. Synaptic responses were elicited by a local stimulation electrode, and were acquired with a multiclamp 700B amplifier. The extracellular solution contained (in mM) 140 NaCl, 4 KCl, 2 CaCl₂, 2 MgCl₂, 10 HEPES-NaOH pH 7.3, and 10 glucose, with 315 mOsm, and either 50 μM picrotoxin (EPSCs) or 10 μM CNQX and 50 μM D-APV (IPSCs). For all electrophysiological experiments, the experimenter was blind to the condition/genotype of the cultures analyzed.

Ca²⁺-imaging

All experiments were performed in a Zeiss LSM 510 confocal microscope. Cultured hippocampal neurons were examined at DIV14-18 in whole-cell patch-clamp configuration after filling with Fluo5F and Alexa594 dyes for 10 min. Action potentials were induced by short, somatic current injections through the patch pipette (typically 5 ms, 600 pA); Ca²⁺-transients were measured with line scans through presynaptic boutons and second order dendrites at a frequency of 333 Hz, typically 100-150 μm away from the neuronal cell body. Fluorescent signals were quantified as mean region of interest and plotted as G-G₀/G₀ (G = average green emission in a given line; G₀ = average of 20 line scans before action potential induction). The experimenter was blind to the condition/genotype until all recordings and analyses were completed.

Immunofluorescence staining of cultured neurons

Cultured neurons were fixed in 4% paraformaldehyde, permeabilized in 0.1% Triton X-100/3% bovine serum albumin and incubated overnight with anti-Ca_V2.1 rabbit polyclonal antibodies (Alomone labs, 1:100), or anti-bassoon rabbit polyclonal antibodies (Synaptic Systems, 1:250), and anti-synapsin mouse monoclonal antibodies (Synaptic Systems, 1:1000). Alexa-Fluor 546 anti-mouse and Alexa-Fluor 633 anti-rabbit secondary antibodies were used for detection with a confocal microscope. Single sections were acquired with identical settings applied to all samples in an experiment, and were used to quantify levels of P/Q-type Ca²⁺-channels and bassoon in ImageJ (NIH); the data were normalized to synapsin staining, and expressed relative to control cultures. The experimenter was blind to the condition/genotype in all experiments.

Data anlaysis

All data are shown as means ± SEM. Statistical significance was determined by one-way ANOVA (some electrophysiological recordings), two-way ANOVA (Ca²⁺-imaging peak amplitude), Kolmogorov-Smirnov test (cumulative distribution of peak amplitudes in Ca²⁺-imaging), χ-test (mouse survival analysis), or Student’s t-test (all other experiments). 95% confidence intervals for [Ca²⁺]_ex-titration data and fitting parameters were calculated based on the covariance matrix. All numerical and statistical values and the tests used can be found in the Suppl. Tables 1-7.

Miscellaneous

Mixed hippocampal cultures and lentiviruses generated in transfected HEK293T cells expressing EGFP-tagged active or inactive cre recombinases followed by an IRES sequence for expression of rescue constructs were produced as described (Kaeser et al., 2009). SDS/PAGE gels, immunoblotting, and electronmicroscopic analyses were done according to standard methods described in the supplementary materials. All animal experiments were performed according to institutional guidelines. A detailed methods section can be found in the supplemental materials.

RESEARCH HIGHLIGHTS.

• At presynaptic active zones, Ca²⁺-channels bind to PDZ-domains of RIM proteins

• Deletion of RIM proteins decreases presynaptic Ca²⁺-influx and vesicle priming

• RIMs tether Ca²⁺-channels to presynaptic active zones for fast, synchronous release

• RIM PDZ- but not C₂-domains are essential for tethering presynaptic Ca²⁺-channels