RIM-binding protein couples synaptic vesicle recruitment to release sites

Chemical synapses are the fundamental building blocks of neuronal communication, allowing for a fast and directional exchange of chemical signals between a neurotransmitter releasing presynaptic and receiving postsynaptic target cells. To couple synaptic vesicle (SV) release to electrical stimulation by action potentials, Ca²⁺ ions entering the cell through voltage-gated Ca²⁺ channels activate the Ca²⁺ sensor synaptotagmin that is anchored on SVs to trigger fusion events (Jahn and Fasshauer, 2012; Stanley, 2016). The presynaptic site of SV fusion (“active zone” [AZ]) is covered by an electron-dense scaffold (“cytomatrix”) formed by a set of large conserved multidomain proteins (Südhof, 2012; Van Vactor and Sigrist, 2017). How individual AZ scaffold proteins intersect mechanistically with the SV cycle still remains largely enigmatic, though such knowledge was of importance to properly model synapse function in healthy and diseased circuits.

On the level of the individual AZs, recent data from both cultivated rodent neurons and Drosophila in vivo neurons suggest that the AZ cytomatrix provides stable SV release sites or “fusion slots” via the clustering of the critical release factor (m)Unc13. The number of such release sites seems to be determined independently of the mechanisms controlling release probability (Böhme et al., 2016; Reddy-Alla et al., 2017; Sakamoto et al., 2018). How SV release sites might be coupled and integrated with additional processes organized by the AZ scaffold, such as the recruitment of SVs under high demand periods of resupply (e.g., high action potential frequencies), remains a largely open question.

The physically extended ELKS family protein (glutamic acid [E], leucine [L], lysine [K],and serine [S]-rich protein) Bruchpilot (BRP) not only operates as the fundamental building block of the Drosophila AZ scaffold (Wagh et al., 2006) but was also shown to promote SV recruitment depending on a binding motif at its extreme, AZ membrane-distal C terminus (Hallermann et al., 2010b). Adaptor proteins physically and functionally connecting such extended scaffold proteins, which sample the cytoplasm, and the SV release sites at the AZ membrane might well be relevant in this regard. Rab3-interacting molecule (RIM)-binding proteins (RIM-BPs) are an evolutionarily conserved family of extended AZ scaffold proteins, obviously critical for SV release at Drosophila and mammalian central synapses (Acuna et al., 2015, 2016; Grauel et al., 2016; Hibino et al., 2002; Liu et al., 2011). RIM-BP family proteins with the C-terminal SH3-II/III domains bind to Ca²⁺ channels (Hibino et al., 2002; Liu et al., 2011), release factor Unc13A (Böhme et al., 2016), and RIM, another critical AZ scaffold protein. At Drosophila neuromuscular junction (NMJ) synapses, rim-bp null alleles provoke a most severe functional phenotype (Liu et al., 2011), stronger than for rim and rim-bp single mutants at most mammalian synapses (Andrews-Zwilling et al., 2006; Betz et al., 2001; Brockmann et al., 2019). Here we exploit the severity of the synaptic Drosophila rim-bp phenotype for a detailed structure-function analysis at Drosophila larval neuromuscular (NMJ) synapses. We find that elimination of the individual RIM-BP SH3 domains affected transport to AZs severely; however, the low levels of properly locating RIM-BP-ΔSH3 variants were still able to rescue the “nanoscopic” defects of rim-bp null mutants AZ scaffolds. Our crystallographic analysis of the RIM-BP fibronectin III–like (FN-III) domain cluster describes an extended “hinge” in the protein center, which likely plays conformational roles. Finally, the N-terminal region (NTR) of RIM-BP was required to properly organize the overall nanoscopic architecture of the AZ scaffold. While SV docking and numbers of (m)Unc13 clusters remained at normal levels, the NTR was specifically needed for efficient SV recruitment. Thus, we suggest that RIM-BP family proteins evolved as adaptors physically and functionally connecting SV release sites with BRP/ELKS-dependent SV recruitment processes.

RIM-BP family proteins are highly conserved across animal phyla (Mittelstaedt and Schoch, 2007). The reason for their fixed overall domain organization remains unclear but might well reflect an overall functional principle encompassing the entire protein. The single Drosophila RIM-BP protein is critical for SV release, but equally needed for AZ structural organization at NMJ glutamatergic synapses, with rim-bp null situations displaying a highly disordered core BRP-scaffold when inspected with nanoscopy (Liu et al., 2011). We sought to reconcile the RIM-BP domain architecture with the several discrete functionalities assigned to the protein.

We started by asking which interactions might be relevant for RIM-BP’s role in the nanoscopic patterning of the AZ scaffold. All RIM-BP family members display three SH3 domains, whose specific sequence identity (Fig. S1 A) and order have been conserved through evolution (Mittelstaedt and Schoch, 2007). While the first SH3-I domain is located in the N-terminal half, a pair of SH3 domains (SH3-II and SH3-III) is located in the C-terminal half of RIM-BPs (Fig. 1 A). SH3 domains mediate protein–protein interactions (PPIs) by binding to proline-rich motifs. RIM-BPs bind via their SH3-II/III domains to voltage-gated Ca²⁺ channel α1 subunits (Hibino et al., 2002; Wang et al., 2000, 2002), with sequence and position of interaction motifs and domains highly similar between Drosophila and mammals (Siebert et al., 2015).

First, we probed discrete RIM-BP domains for interaction with other AZ core proteins (BRP, RIM-family), the AZ-residing release factors (i.e., Unc13A), and the intracellular Ca²⁺ channel intracellular C terminus. Thus, we used a systematic, robot-based yeast two-hybrid (Y2H) matrix screening protocol (Worseck et al., 2012). Here, all interactions were tested in two bait–prey and two prey–bait configurations, comprising a total number of 2,424 pairwise tests (Stelzl et al., 2005; Venkatesan et al., 2008; Hegele et al., 2012; see Materials and methods).

In our Y2H analysis, the RIM-BP SH3-II/III domains as expected bound robustly to ligands containing K/RPPXP motifs: voltage-gated Ca²⁺ channel α1 subunit Cacophony (Kittel et al., 2006; Liu et al., 2011), Unc13A (Böhme et al., 2016), and kinesin adaptor Aplip1 (Siebert et al., 2015; Fig. 1 B). In contrast, we did not observe any interactions of the first RIM-BP SH3 domain to these or other PXXP ligands (data not shown). Taken together, the C-terminal SH3 domain tandem of Drosophila RIM-BP seems to specifically bundle interactions to AZ release and transport factors.

Still, direct interactions of the RIM-BP SH3-II/III domains with BRP, the only representative of the ELKS family in Drosophila and core AZ scaffold organizer (Kittel et al., 2006; Wagh et al., 2006), could not be detected. Sequence homology between Drosophila RIM-BP and mammalian RIM-BP1 also extends across the entire NTR but is particularly pronounced in regions of predicted high α-helical character (Fig. S1 B). We recombinantly expressed two small variants of the NTR, NTR^8-151, and NTR^8-254. Equilibrium circular dichroism (CD) spectra revealed indeed a predominantly α-helical folding for both constructs (Fig. S1 C). Additionally, CD-based thermal melting/reannealing suggested the presence of a folded domain that could refold after thermal denaturation (Fig. S1 D). Numerous ER/K motifs present in the NTR may stabilize α helices (Swanson and Sivaramakrishnan, 2014) and may lead to the formation of coiled-coils (Lupas and Bassler, 2017). Notably, in the Y2H assay, the NTR (residues 1–600) displayed robust interactions with a highly conserved region of BRP (residues 100–650; Fig. 1 B) but not with any other of the known RIM-BP–interacting proteins. The region of BRP that exhibits Y2H interactions with the RIM-BP NTR is contained in both major BRP isoforms (BRP-190 and BRP-170) and previously was shown to be crucial for overall AZ integrity (Fouquet et al., 2009; Matkovic et al., 2013).

As Y2H by itself is insufficient to verify interactions, we first intended to use in vitro interaction testing between recombinantly expressed proteins to probe this interaction using standard pull-down assays. Unfortunately, while the relevant RIM-BP region (residues 1–600) could per se be expressed, a major fraction of the material despite efforts remained largely insoluble, precluding stringent in vitro binding analysis of this region. Consistent with an interaction between RIM-BP and BRP protein, however, we could efficiently coprecipitate RIM-BP with BRP and vice versa from Drosophila head lysates (Fig. S2 A).

To use an unbiased approach when exploring the direct binding of RIM-BP to other AZ scaffold proteins from in vivo material, we entailed chemical cross-linking in combination with mass spectrometry (for a schematic representation, see Fig. 1 C; Leitner et al., 2016; Liu and Heck, 2015). More specifically, we applied the disuccinimidyl sulfoxide cross-linker on a synaptosome preparation from Drosophila brains (Fig. S2, B and C), which we previously developed (Depner et al., 2014). Through extensive proteomics analysis, thousands of cross-links across the synaptic proteome were retrieved (data not shown). These cross-links represent residue-to-residue connections that are within a maximum distance restraint defined by the cross-linker in 3D space. Notably, for RIM-BP, these analyses in total provided five cross-links, which were all positioned within the RIM-BP NTR and connected to a specific stretch of BRP (residues 300–600; Fig. 1 D), consistent with the Y2H analysis data (Fig. 1 B). Thus, our cross-linking proteomics analysis provides independent evidence from in vivo–derived material that the NTR of RIM-BP and a region covering residues 300–600 of BRP mediate contact between these two key AZ scaffold organizers. The fact that only this scaffold–scaffold interaction, but not e.g., interactions of the conserved SH3 domains, could be retrieved by cross-linking points toward a stable “stoichiometric” nature of the RIM-BP::BRP interaction.

These cross-links represent residue-to-residue connections that are within a maximum distance restraint defined by the cross-linker in 3D space. Notably, for RIM-BP, these analyses in total provided five cross-links, which were all positioned within the RIM-BP NTR and connected to a specific stretch of BRP (residues 300–600; Fig. 1 D),

The structural organization and functional role of the highly conserved array of FN-III domains (Fig. S1 D) in the RIM-BP central part (Fig. 1 A) remained enigmatic so far, despite the recent finding of big potassium channels (BK-channels) binding to this motif (Sclip et al., 2018). We determined the crystal structure of a corresponding RIM-BP fragment (residues 745–1042; RIM-BP^745-1042; Fig. 2 A) at 2.45 Å resolution to compare the organization of the FN-III region in RIM-BP to other FN-III arrays and precisely delineate domain borders for subsequent genetic manipulation of the protein in vivo.

An asymmetric unit of the crystals contained two RIM-BP^745-1042 chains, in which residues 745–838, 843–945, and 946–1,042 formed the three FN-III domains (Fig. 2 B, Fig. S2 E, and Table S1). FN-III regions in several proteins have been shown to homodimerize (Carr et al., 2001). However, bioinformatics analysis (Krissinel and Henrick, 2007) suggested that RIM-BP^745-1042 crystal contacts are not biologically relevant. CoMulti-angle light scattering showed consistently that RIM-BP^745-1042 is monomeric in solution (MW_theoretical: 32,466 D, MW_n: 32,610 ± 107 D [MW, molecular weight]).

FN-III(1) and FN-III(2) interact via a short β-sheet formed between β-strand G of FN-III(1) and the BC loop of FN-III(2) in both copies of the protein (Fig. 2 B). By contrast, FN-III(3) is more loosely appended to FN-III(2). The BC loop of FN-III(3) and the FN-III(2) AB loop interact with the linker region of FN-III(2) and FN-III(3) in one molecule (Fig. 2, C and D). These interactions are lacking in the second molecule, leading to different orientations of the FN-III(3) domains relative to the FN-III(1–2) domains (Fig. 2 D). Analysis of the apparent FN-III(3) movement by the DynDom server (Girdlestone and Hayward, 2016) revealed a rotation of 32.4° and a translation of −1.3 Å of FN-III(3) relative to FN-III(1–2) due to a bending of the FN-III(2)–FN-III(3) linker region (Fig. 2 E).

Most of the published arrays of FN-III regions adopt an arrangement with no or few direct contacts between the FN-III domains. The tight contacts of the FN-III(1–2) module resemble the arrangement of the D4 and D5 FN-III domains in the extracellular domain of gp130, which allows this region of gp130 to signal without major conformational changes upon ligand binding (Xu et al., 2010). Therefore, while the flexible apposition of FN-III(2) and FN-III(3) indicates that this region represents a hinge-like element within RIM-BP, the FN-III(1–2) region might provide a rigid spacer and/or a preformed binding site for a ligand. However, neither our Y2H analyses (Fig. 1 B) nor pull-down experiments using several FN3 domain constructs and adult fly brain extracts revealed stable binding of the RIM-BP FN-III array to other AZ proteins (data not shown).

Deleting RIM-BP at NMJ synapses provides us with a severe structural and functional phenotype at this highly accessible synapse. Based on this, we started a systematic in vivo structure/function analysis using a series of deletion constructs (see schemes in Fig. 3 A). As extended deletions on genomic level would likely interfere with proper splicing and influence mRNA expression levels in an unpredictable manner, we used inducible cDNA-based constructs. All constructs were equipped with GFP to allow for comparing protein and mRNA expression levels of all constructs via the same sequence tag. We tested for mRNA expression levels in adults (“driving” the constructs with pan-neuronal driver line elav-Gal4) by RT-PCR (against the GFP sequence). No significant expression differences between the to-be-analyzed deletion constructs were detected (Fig. S3 A).

We first tested the rescue activity of all constructs by scoring numbers of surviving adults after expressing the RIM-BP variants in a rim-bp null mutant background by elav-Gal4 in three biological replicates (Fig. 3 B). As expected (Liu et al., 2011), only very few mature rim-bp null mutant animals could be retrieved. In contrast, transgenic expression of the GFP-tagged full-length RIM-BP (RIM-BP-FL^GFP) in rim-bp null mutants restored viability in the expected mendelian ratio (Fig. 3 B). Importantly, larvae rescued by RIM-BP-FL^GFP showed NMJ morphologies, AZ numbers, and BRP levels indistinguishable from WT (Fig. 3 D; Fig. 4 B; and Fig. S3, B–D). Neither was evoked SV release statistically different from WT (Fig. S3, E–G). We would like to emphasize that though Gal4-driven expression of RIM-BP-FL^GFP resulted in clearly increased total protein levels as detected in Western blots as compared with WT (Fig. S3 H), no signs of any gain-of-function effects were detected here. This is likely explained by transport and especially synapse localization naturally limiting RIM-BP amounts at AZs.

At the AZs of rim-bp null mutant larva NMJs (rim-bp^STOP1/S2.01), confocal (Fig. 3 C) and stimulated emission depletion microscopy (STED) images (Fig. 4 A) showed a severe misorganization of the BRP scaffold, as expected (Liu et al., 2011). Single-action potential-evoked SV release was drastically reduced (Fig. S4, C–F), while spontaneous release was not altered (Fig. S4, G–I). The RIM-BP-FL^GFP localized consistently into the AZ scaffolds in both the mutant and WT background (Fig. 3 D). Planar STED views pictured RIM-BP-FL^GFP in the AZ center surrounded by the typical ring of the BRP C-terminal epitope labeled by the Nc82 monoclonal antibody, which resides further from the AZ membrane than the RIM-BP C terminus (Fig. 4 B), as expected from the labeling of the endogenous protein (Liu et al., 2011). Importantly, the STED-nanoscopic BRP distribution was also restored by the expression of RIM-BP-FL^GFP in the rim-bp null background.

To evaluate the functional role of the FN-III domains, we expressed GFP-tagged constructs bearing deletions of individual FN-III domains in WT and rim-bp mutant larvae (Fig. 3, A and E; and Fig. 4 C). Re-expressed RIM-BPΔFN-III(2)^GFP and RIM-BPΔFN-III(3)^GFP localized faithfully to the AZ in the rim-bp mutant background (Fig. 3 E), with their nanoscopic topology in STED images relative to the BPR scaffold appearing very similar to control RIM-BP-FL (Fig. 4, C and H). Although both deletion constructs still rescued the rim-bp null mutant lethality (Fig. 3 B), their protein levels at NMJ terminals appeared reduced in the presence of endogenous RIM-BP, possibly indicating that the full-length RIM-BP protein might have advantages over the FN-III(2) and (3) deletion variants in binding into the scaffold. However, in the rim-bp mutant background, both constructs reached the synaptic terminal in similar amounts as the RIM-BP-FL^GFP construct (Fig. 3 H).

In contrast to these rather weak deficits, deletion of the FN-III(1) domain provoked a nearly total absence of the protein from the synaptic terminal in the rim-bp mutant background (Fig. 3, E and H). Consistently, RIM-BPΔFN-III(1)^GFP did not rescue viability (Fig. 3 B), and BRP scaffolds appeared as deformed as in rim-bp null mutants (Fig. 4, C and H; compare with rim-bp mutant Fig. 4 A). Moreover, neurotransmission of this genotype (RBPΔFN-III[1]^GFP in rim-bp null background) appeared identical to the rim-bp null mutant (data not shown). As we were also unable to detect RBPΔFN-III(1)^GFP in motoneuron somata and axons (data not shown), we have to conclude that deletion of the first FN-III domain renders the protein unstable in the cell body. Surprisingly, however, in the presence of endogenous RIM-BP, RBPΔFN-III(1)^GFP effectively reached the synaptic terminals and integrated into the BRP scaffold (Fig. 3 E). We indeed detected a RIM-BP band corresponding to RBPΔFN-III(1)^GFP in Western blot analysis of larval brains, only in the presence of endogenous RIM-BP and not in the rim-bp null mutant background (Fig. S3 H). The most plausible explanation for this is that the RIM-BP protein containing a deletion of the first FN-III domain is rendered unstable, potentially reflecting inappropriate folding, but stabilizes when given the chance to bind the endogenous, properly folded WT RIM-BP.

These results taken together show that the first FN-III domain in RIM-BP is probably critical to properly fold and/or stabilize the protein. Future analyses will have to elucidate the underlying mechanisms of this stability control.

The results above suggested that RIM-BP scaffold integration might be mediated by regions other than the FN-III array. We therefore turned our attention to the SH3 domains. Deletion of the first SH3 domain (RIM-BP ΔSH3-I^GFP) did not have a major effect on the rescue rate (Fig. 3 B), did not influence the AZ targeting and nanoscopic integration of the protein, and completely rescued the scaffold assembly at NMJ synapses (Fig. 3, F and I; and Fig. 4, D and H). Thus, somewhat surprisingly, our analysis did not uncover any major functional role for the conserved SH3-I domain at NMJ synapses, in correspondence with our inability to detect binding to other AZ structural proteins and release relevant factors (see above).

The RIM-BPΔSH3-II^GFP protein showed slightly reduced rescue rates, while the RIM-BPΔSH3-III^GFP construct lacked rescue activity (Fig. 3 B). While both deletion variants were expressed, they reached the synaptic NMJ terminal only at reduced levels (Fig. 3, F and I; for ΔSH3-II, we measured 46%, for ΔSH3-III 40% of control RIM-BP level), probably due to transport impairments, as ectopic GFP signal could be detected in the neuronal somata of the ventral nerve cord (data not shown). We could, however, detect RIM-BPΔSH3-III^GFP robustly at the NMJ upon an increase in laser strength (Fig. 3 F), and STED microscopy indicated that both RIM-BPΔSH3-II^GFP and RIM-BPΔSH3-III^GFP integrated into the AZ scaffold with apparently normal topology. Importantly, both RIM-BPΔSH3-II^GFP and RIM-BPΔSH3-III^GFP rescued the BRP scaffold nanoarchitecture effectively, even at these reduced levels (Fig. 4, D and H). A double deletion of both SH3 domains (ΔSH3-II+III) led to the nearly complete absence of the protein at the synaptic terminal. In this context, the BRP scaffold appeared as defective as the rim-bp null mutant (Fig. S4, A and B), most probably a direct consequence of a nearly complete lack of RIM-BP in this situation.

Previous analyses using two-electrode voltage clamp recordings had shown that rim-bp null mutants suffer from severely reduced evoked junctional currents (eEJCs) and charge, and show increased paired pulse facilitation, probably reflecting a looser coupling of SVs to Ca²⁺ influx (Liu et al., 2011; Müller et al., 2015). Surprisingly, no significant defect of single AP-evoked release or spontaneous release could be detected for the RIM-BPΔSH3-III rescue situation (Fig. S4, C–I).

As presence of either SH3-II or SH3-III might suffice to organize Ca²⁺ channel α1 subunit binding, we generated a genomic mutant of the Ca²⁺ channel, in which we altered the PXXP motif that is responsible for the binding to RIM-BP (see Materials and methods; Han et al., 2011; Kaeser et al., 2011). We then compared cells expressing genomic constructs of either WT Ca²⁺ channel (cac wt) or a Ca²⁺ channel with mutated PXXP motif (cac APTA) in a cac null background by recording postsynaptic responses to single-action potential stimulation at 1 mM extracellular Ca²⁺. Surprisingly, no significant differences were detected for eEJC response amplitudes (Fig. S5, A–E). However, there was a slight but significant increase in the time to peak of evoked responses after altering the Cac PXXP RIM-BP interaction motif.

Finally, we tested the ability of isolated SH3-III or SH3-II+III fragments to transport and enter into the AZ scaffold. Consequently, we created SH3-III^GFP and SH3-II+III^GFP expression constructs also tagged with a C-terminal GFP. Both constructs were effectively transported to the synaptic terminal but failed to integrate into the AZ scaffold in WT and rim-bp mutant larvae (Fig. S5 F).

In summary, we find that deletion of neither the second nor the third SH3 domain interferes with the ability of RIM-BP to integrate into the AZ scaffold. Moreover, deletion of either SH3 domain does not interfere with the ability of RIM-BP to establish a proper AZ nanoarchitecture. Genetic elimination of the RIM-BP binding motifs of the Ca²⁺ channel α1 subunit intracellular C terminus also did not “phenocopy” the rim-bp null allele. Our data suggest that interactions redundant to the RIM-BP contacts stabilize Ca²⁺ channel: AZ scaffold contacts at NMJ synapses (also see Discussion).

As shown above, the NTR of RIM-BP and a region from residues 300–600 from the BRP side seemingly provide contact surfaces between these two key AZ scaffold organizers. To probe the relevance of this scaffold–scaffold contact, we constructed RIM-BP variants, in which residues 24–250 (ΔNTR1) or residues 152–599 (ΔNTR2) were deleted (Fig. 3 A). Both N-terminal deletion variants were expressed at levels comparable to control protein and reached the synaptic terminal effectively in both the rim-bp null mutant and WT background (Fig. 3, G and J; and Fig. S6 A). Thus, RIM-BPΔNTR1^GFP and RIM-BPΔNTR2^GFP are competent for axonal transport, most probably because both constructs still contain the C-terminal SH3-II/III domains. However, despite efficient transport to synapses, RIM-BP-ΔNTR2^GFP was unable to rescue from lethality (Fig. 3 B). Moreover, quantifying NMJs of larvae expressing RIM-BPΔNTR2^GFP in the rim-bp null background showed that overall active zone numbers (BRP cluster counts) and BRP intensities were not significantly reduced, while the size of individual AZs was slightly increased (Fig. S6, B–D). The RIM-BPΔNTR2^GFP levels in STED analyses again appeared similar to control levels (Fig. 4, B and E), while about half of the individual AZ BRP scaffolds appeared as defective as in the null mutant (Fig. 4, A, G, and H).

The RIM-BPΔNTR1^GFP variant showed a similar but less severe phenotype. Also, this NTR mutant still was transported effectively to the NMJ terminals (Fig. 3, G and J). STED microscopy also revealed a range of BRP AZ scaffold deficits, which were again qualitatively similar but quantitatively less pronounced compared with the RIM-BPΔNTR2^GFP rescue phenotype (Fig. 4, E, F, and H). Consistent with its less pronounced nanoscopic phenotype, RIM-BPΔNTR1^GFP rescued from lethality more efficiently than RIM-BPΔNTR2^GFP (Fig. 3 B).

We applied EM to address independently scaffold organization and distribution of SVs upon truncating the NTR (Fig. 5, A and B). Consistent with the STED data, T-bars in EM cross-sections appeared morphologically atypical. Most of them appeared reduced in size and decomposed, reflected in a reduced ultrastructural area, length, and height (Fig. 5 C). Thus, both STED and EM analyses suggest that the NTR plays a nonredundant role in structuring the nanoarchitecture of the AZ scaffold, presumably by binding directly to the master AZ scaffold building block, BRP.

The RIM-BPs, as their name indicates, also bind to RIM family proteins. Two family members, RIM proper (Graf et al., 2012; Müller et al., 2012) and FIFE, an extended Piccolo-RIM family member (Bruckner et al., 2012; Bruckner et al., 2017), function at NMJ synapses in Drosophila. RIM bound via a region containing a K/RPPXP motif in the last part of the protein to the C-terminal SH3-III domain, while also Fife bound the RIM-BP SH3-II//III domains via a K/RLPXXP motif in the center of the protein (Fig. 1 B). Given these conserved interactions, we asked whether deficits of rim-bp mutants might be also easily explained by defects in RIM localization/function.

So far, RIM and FIFE have been analyzed individually at NMJ synapses but not in a double loss of function situation. To explore redundancies between these two members of the RIM family in Drosophila, we subjected both single mutants and the double mutant to electrophysiological analysis (Fig. S7, A and B). As was expected (Graf et al., 2012; Müller et al., 2012), rim single mutants displayed severe release deficits, while release deficits were more moderate in fife mutants (Fig. S7, A and B). Notably, release in a double mutant combination was only as severely affected as in rim single mutants. In rim mutants, UNC13A AZ clusters were not reduced (Fig. S7, C and D), suggesting that RIM is functionally needed also beyond likely redundant roles in release machinery clustering and molecular release site definition.

We further analyzed RIM functionality via STED nanoscopy. Despite efforts, we could not retrieve antibodies to visualize the endogenous protein. Thus, we turned to an on-locus GFP insertion into the rim locus. We found RIM-GFP to be often localized to discrete clusters at the AZ center, consistent with a function in AZ release (Fig. S7, E and F). We then asked whether BRP AZ scaffold assembly would depend on RIM family proteins. The BRP distribution in rim single, fife single, and rim/fife double mutants appeared essentially unaffected (Fig. S7, G–J), at least on the level retrievable with STED microscopy. To test for functional redundancy between RIM-BP and RIM in the nanoscopic AZ patterning, we created a rim, rim-bp double loss-of-function mutant by combining a rim mutant with RNAi against RIM-BP (Fig. S7, K and L). Indeed, we could detect a significant increase in BRP scaffold defects in the double-LOF (loss-of-function) situation for rim and RIM-BP RNAi when compared with the RIM-BP RNAi situation (Fig. S7 L). Thus, RIM seems to operate synergistically with RIM-BP in establishing the AZ nanoscopic architecture.

We continued analyzing the RIM-BPΔNTR2^GFP situation. Despite the overall misorganization of the AZ scaffold, the density of Unc13A-positive release sites in STED images was not reduced but even slightly increased (Fig. 5, E and F). The total numbers of SVs in EM appeared unchanged, and SV docking persisted largely undisturbed (Fig. 5 D). Interestingly, however, we discovered that SVs tethered to the distal parts of the T-bars were significantly reduced, potentially pointing toward deficits in SV recruitment processes (Fig. 5 D).

We then subjected NMJs of larvae expressing RIM-BPΔNTR2^GFP in a rim-bp null mutant background to electrophysiological analysis to test for functional consequences of the RIM-BP N-terminal truncation. Single-action potential-evoked release was slightly but significantly reduced for the RIM-BPΔNTR2^GFP-expressing larvae compared with the control rescue situation (Fig. 6, A–C; same experiment as in Fig. S4, C–F; therefore, same controls). At the same time, no change of short-term facilitation behavior was observed, very different from the rim-bp mutant, which, as expected (Liu et al., 2011), showed dramatic facilitation (Fig. 6 D). We observed a similar but less pronounced trend upon expressing the RIM-BPΔNTR1^GFP construct in the rim-bp null mutant, with the single-action potential-evoked charge being slightly, although not significantly, reduced (Fig. S8, A–C). Short-term facilitation was again unaffected (Fig. S8 D), while the quantal content showed a significant reduction (Fig. S8 E). Based on these data, the RIM-BP NTR obviously plays no major role in the coupling of SVs to release machinery and Ca²⁺ channels.

We then used variance mean analysis of synaptic responses (Clements and Silver, 2000; Liu et al., 2011; Sakaba et al., 2002; Silver, 2003) to further dissect the release function after eliminating the RIM-BP NTR. Thus, we determined the mean amplitude and variance of the eEJCs in a two-electrode voltage clamp mode at Ca²⁺ concentrations between 0.75 and 6 mM (Fig. 6 F) and calculated vesicular release probabilities at the Ca²⁺ concentrations used, applying parabolic fits (Fig. 6 G). The SV release probability was indeed unchanged in the mutants compared with the control over the whole range of concentrations tested (Fig. 6 H), again emphasizing that SV-Ca²⁺ channel coupling is obviously undisturbed. Extrapolation of release site number N (Fig. 6 I) and quantal size (Fig. 6 J) showed slight reductions for the RIM-BPΔNTR2^GFP-expressing situation. Taken together, no signs of defective Ca²⁺ channel/SV coupling could be observed after eliminating the NTR in two assays, consistent with the SH3 domains (but not the NTR) binding to Ca²⁺ channels and the Unc13A release factor. Considered together, the RIM-BP NTR, though critical for proper bundling of the BRP-based AZ scaffold, is obviously largely dispensable for SV/Ca²⁺ channel coupling and molecular release site definition.

The BRP-based T-bar scaffold has been implicated in the activity-dependent recruitment of SVs into the SV releasable pool, which involves the tethering of SVs to the membrane-distal ends of the BRP filaments forming the T-bar “roof” (Hallermann et al., 2010b). As has already been mentioned above, the number of SVs tethered to the distal T-bar was significantly reduced in the absence of the NTR (Fig. 5 D). We therefore asked whether the dynamic recruitment of SVs in episodes of increased release would depend on the RIM-BP NTR2 (Fig. 6, K–M). Accordingly, we subjected the RIM-BPΔNTR2^GFP rescue larvae to trains of 60 Hz and plotted cumulative quantal contents. Back extrapolation to the y axis allowed for the determination of release-ready vesicle pool (RRP) sizes, while the slopes gave refilling rates (Fig. 6 K; extrapolated straight lines are indicated in the graph). No significant change of the RRP size was observed (Fig. 6 L). Instead, the SV refilling rate in the course of enhanced SV release was significantly reduced (Fig. 6 M).

To further explore the latter phenotype, we first asked whether RIM-BP–controlled refilling rates might depend on RIM. However, the corresponding rim mutant phenotype was rather inverse to that of RIM-BPΔNTR2^GFP rescue larvae: the RRP sizes were significantly reduced (Fig. 6, N and O), but refilling rates were unchanged (Fig. 6 P). Elimination of RIM in isolation thus seemingly plays no major role in the recruitment of SVs into release sites, at least not at the NMJ synapse. The RIM-BP NTR, while not playing a major role in release site function and definition, seems to take part in a rate-limiting step of recruiting SVs into their release sites at the base of the AZ scaffold.

Notably, BRP has been implicated in RRP refilling previously, based on a specific mutation (brp^nude) that deletes only the last 17 amino acids of the membrane-distal C terminus of BRP (Hallermann et al., 2010b). Interestingly, refilling rates were reduced similarly in brp^nude as in the RIM-BPΔNTR2^GFP rescue larvae (Fig. 6, Q–S).

Our recent analysis provided evidence for discrete Unc13A clusters operating as SV release sites (Böhme et al., 2016; Reddy-Alla et al., 2017). These Unc13A clusters group physically around the AZ central Ca²⁺ channel cluster with an average distance of ∼70 nm from the AZ and Ca²⁺ channel cluster center at planar STED-imaged AZs (Böhme et al., 2016). The extended BRP protein converges from the membrane-distal C terminus to the membrane-proximal N terminus in a funnel-like manner, with the C termini grouping around the AZ center in planar projections at a distance of ∼100 nm. The BRP N termini, instead, overlay the Ca²⁺ channels in the AZ center directly, consistent with the probable role of BRP in Ca²⁺ channel clustering, redundant to RIM-BP (see Discussion).

Our findings that the RIM-BP SH3-II/III domains bind release effectors physically, among them Unc13A, while the RIM-BP NTR binds a region of BRP, suggest that RIM-BP, due to its conserved architecture, might functionally bridge BRP-associated SV recruitment to the Unc13A-dependent SV release sites. We therefore analyzed the nanoscopic distribution of RIM-BP epitopes at planar imaged AZs. The RIM-BP C termini localized toward the outer BRP C termini (Fig. 7 A, upper panel) and, thus, in the direction of the Unc13A clusters. The RIM-BP NTR epitope, instead, was positioned more toward the AZ center and partially colocalized with the BRP N-terminal epitope (Fig. 7 A, middle and lower panels). Line profiles of single AZs show larger rings for the RIM-BP C-terminal than the RIM-BP N-terminal epitope (Fig. 7 B, upper panel). Quantification of the peak-to-peak distances revealed a significant ring diameter difference (Fig. 7 C). The quantified numbers are depicted on a top view model of the AZ (Fig. 7 D), showing that RIM-BP^Nterm and BRP^Nterm are in close proximity, consistent with both NTRs being in stable physical contact, as indicated by our Y2H and in vivo cross-linking data (Fig. 1). Thus, our nanoscopic analysis suggests a functional architecture in which BRP filaments guide SVs into SV release slots, and RIM-BP might represent a relay factor that bridges between BRP and release sites functionally and topologically.

Our current view of AZ scaffolds emphasizes their role in providing a “smart catalytic surface” integrating sub-functionalities that facilitate and control the SV cycle. Providing SV release sites with a proteinaceous “nano-environment” defining release probability, by precisely defining their spatial distance to voltage-operated Ca²⁺ channels, is probably one fundamental function using these highly conserved, “ancient” protein architectures. The SV release sites probably also have to be coupled to processes retrieving SVs from more membrane-distant positions. Concerning protein architectures harvesting SVs, evolutionary solutions might have been somewhat more flexible, obviously adopted to the synapse-type specific needs in physiologically relevant fusion rates. Identifying proteins that might be involved in structurally and functionally coupling recruitment protein architectures with release sites is an emerging topic. We here provide evidence that RIM-BP with its interactions obviously plays a role in both “nanoscopic locations.” First, through its SH3 domains, it binds RIM, the intercellular C-terminus of the voltage-operated Ca²⁺ channel and also Unc13A, the release factor whose nanoscale positioning we recently identified as critical for SV release definition (Reddy-Alla et al., 2017). Second, we found here that it also binds conserved regions of BRP in the core of the AZ central scaffold. Deleting this interaction surface, though not reducing AZ protein levels, obviously undermines the proper bundling of BRP filaments, at least for a major fraction of AZs. Though it is not easy to prove that individual RIM-BP protein molecules span this distance physically, it is tempting to hypothesize that RIM-BP might literally connect the recruitment of SVs from more membrane distant pools, probably starting at the distal end of the BRP filaments, with their integration into SV release sites (see model, Fig. 7 E).

At mammalian synapses, recent multi-loci genetics have demonstrated most severe deficits when eliminating combinations of AZ scaffold proteins between RIM, ELKS, and RIM-BP family members (Acuna et al., 2016; Wang et al., 2016). Here, quadruple knockouts of RIM1/2, together with RIM-BP1/2 proteins in mice, exhibit a total loss of neurotransmitter release from severe impairments in SV priming and docking, a dramatic loss of AZ scaffold density, with a trans-synaptic effect that impairs the organization of the postsynaptic density (Acuna et al., 2016). These severe synthetic “catastrophic” phenotypes, eliminating ultrastructural specializations and release factor targeting, demonstrate a principal functional redundancy between RIM-BPs, RIMs, and ELKS family proteins. Similar results were also retrieved from work on mice in which RIM1αβ and RIM2αβγ, together with ELKS1α and ELKS2α isoforms, have been completely eliminated. Cultured hippocampal synapses of these mutant mice consequently lose Munc13, Bassoon, Piccolo, and RIM-BP, following a mass disassembly of the AZ scaffold (Wang et al., 2016). Similarly, in this study, we observed that RIM seemingly operates synergistically with RIM-BP in establishing the AZ nanoscopic architecture (Fig. 6, K and L).

Despite the obvious importance of these findings for understanding the collective role of the AZ scaffold, the fact that the domain organization for all specific scaffold proteins has been individually conserved over hundreds of millions of years motivated us to search for ways to demonstrate specific functions of these AZ core scaffold proteins. We here molecularly isolated an additional sub-functionality for the so-far functionally nonconsidered RIM-BP NTR, and provide evidence that its protein architecture might have been conserved for the reason that it operates as an adaptor coupling SV recruitment processes with the membrane-associated SV release sites.

Mechanistic analysis of RIM-BPs so far has largely focused on SH3 domains II and III, which bind to Ca²⁺ channels and RIM in both mammals and Drosophila (Acuna et al., 2015; Hibino et al., 2002; Liu et al., 2011; Wang et al., 2002), and Unc13A in Drosophila (Böhme et al., 2016). Analysis of RIM-BP at Drosophila NMJ synapses was the first to demonstrate a major role of the protein family in neurotransmission (Liu et al., 2011), characterized by a severe reduction in release probability and signs of defective Ca²⁺ channel clustering. We here exploit the glutamatergic NMJ synapses for a stringent genetic analysis with multimodal readouts, also including an analysis of transport and nanoscale integration into the AZ scaffold. Deletion of individual SH3 domains majorly affected RIM-BP transport, probably directly reflecting their high affinity binding to JIP-1 homologue Aplip1, whose deletion also interferes with effective axonal transport of BRP/RIM-BP “packages” (Siebert et al., 2015). Somewhat surprisingly, however, moderately or even strongly reduced levels of RIM-BPΔSH3-II or RIM-BPΔSH3-III, respectively, still restored RIM-BP functionality when expressed in a null background. It obviously might be argued that only combined elimination of SH3-II together with SH3-III might uncover their function collectively. Indeed, Aplip1 binds both SH3-II and III, and AZ levels were decreased even more strongly in the RIM-BPΔSH3-II/III double-deleted variant. However, at least concerning the role of Ca²⁺ channel binding, the AZ scaffold can seemingly compensate for the absence of RIM-BP-SH3 mediated interactions, as Ca²⁺ channels with the highly conserved PXXP RIM-BP binding motif deleted still rertained full activity in a genetic rescue assay (instead of resembling the rim-bp null phenotype). We suggest that redundant interactions stabilize Ca²⁺ channel: AZ scaffold contacts at NMJ synapses, an idea that might be relevant for other synapses as well. Indeed, we showed previously that the N terminus of BRP also binds the Ca²⁺ channel α1 subunit intracellular C terminus (Fouquet et al., 2009). Given the essential character of SH3-III for survival, however, it appears likely that slightly different rules and “vulnerabilities” apply to other synapse types, for example, in the Drosophila brain. In fact, we have found lately that synapses in the Drosophila brain differ substantially in their BRP content, with, for example, interneuron synapses largely lacking BRP (Fulterer et al., 2018). This might result in different efficacy of compensation via BRP and be responsible for the essential character of SH3-III. Put differently, it still appears plausible that the C-terminal close SH3 domains might be of truly essential character dependent on synapse type given the results retrieved at mammalian synapses (Acuna et al., 2015; Grauel et al., 2016; Kaeser et al., 2011).

Finally, the presence of three FN-III domains, despite their invariable character, remained and remains an enigma. FN-III modules are common in extracellular but rarer in intracellular proteins, such as RIM-BP, here found chiefly as the main components of a group of intracellular proteins associated with the contractile apparatus of muscles. The first x-ray structure of the RIM-BPs, which we can present here, depicts a typical FN-III-type organization; however, it is oriented in an extended hinge-like arrangement. It is tempting to speculate that the hinge-like architecture might support RIM-BP conformations or conformational change when connecting release sites with recruitment processes.

Our physiological analysis points toward a discrete function of the RIM-BP N-terminal domain. Analysis of SV recruitment based on the rim-bp null allele is complicated due to the severe release probability deficits dominating the physiological scenario. Nonetheless, a careful analysis, using high Ca²⁺ concentrations to milden the influence of release probability differences, indeed identified recruitment defects (Müller et al., 2015). Remarkably, similar to our NTR-specific deletions, a previous study analyzing the rim-bp null mutant (Müller et al., 2015) for RIM-BP suggested a rate-limiting function for the replenishment of high release probability SVs following vesicle depletion at NMJ synapses. Moreover, in mouse rim-bp2 knockouts, recruitment/replenishment deficits at auditory hair cells were reported (Krinner et al., 2017). In rim-bp null mutants, UNC13A levels are strongly reduced (Böhme et al., 2016), while for the dNTR-2 construct, UNC13A levels and consequently release site number were not reduced (instead rather slightly increased). Thus, our data apparently uncouple the role of RIM-BP for release site organization (via Unc13A clustering) from its role in overall scaffold organization where the NTR plays an important role.

The extended BRP filaments in Drosophila probably operate as “antennae” to harvest SVs from the reserve pool, a process facilitated by the C-terminal amino acids of BRP, made evident by the defects of the brp^nude allele lacking only the last 17 amino acids (Hallermann et al., 2010b). BRP in its first half is highly homologous to ELKS/CAZ-associated structural proteins (CAST) AZ proteins; however, its C-terminal half is specific for insects and obviously evolved here for harvesting SVs. The large vertebrate-specific AZ protein Bassoon was shown to help SV replenishment at the central synapse under conditions of heavy stimulation (Hallermann et al., 2010a). Notably, Bassoon binds to the first SH3 domain of RIM-BP (Davydova et al., 2014), suggestive of convergent evolution where different molecular antennae were used to finally target RIM-BP, the generic “old” adaptor coupling to the membrane close release sites.

In its current form, static STED microscopy as used here only samples average epitope distributions, while the relevant proteins likely are dynamically switched in the course of the SV cycle. As an indication of this, the RIM-BP C-terminal SH3 domains were found to bind both the Ca²⁺ channel C-terminus and Unc13A N terminus (Böhme et al., 2016). As these binding events can hardly be accomplished by a single Unc13A molecule at a single time point, future analysis will have to address the underlying conformational dynamics likely involved here.

Taken together, RIM-BPs might take a generic role of SV replenishment, apart from their obvious role in coclustering of release machinery and Ca²⁺ channels at the AZ membrane. Indeed, the highly conserved domain architecture might have evolved for exactly this reason, to guide SVs into the proper release environment.

The full-length drbp cDNA (according to isoform F) has been obtained via PCR using a forward primer with additional 4 bp (CACC; see Table 1) and then ligated into pENTR-dTOPO (Invitrogen) according to the manufacturer’s instructions. Finally, pENTR-dTOPO-rim-bp FL was recombined with pUAST-attB-rfa-EGFP using the Gateway System (Invitrogen) to yield pUAST-attB-rim-bp FL-EGFP. After confirmation by double-strand sequencing, transgenic flies were generated using the PhiC31 system with defined landing sites in the Drosophila genome.

All additional domain deletions of RIM-BP were generated by using the pENTR-dTOPO-rim-bp FL as a template and overlapping primers in a self sequence– and ligation-independent cloning strategy (SLIC; Li and Elledge, 2012) approach (primer sequences, see Table 1). The following amino acids were deleted: RIM-BP^ΔFN-III(1) 747–826 aa, RIM-BP^ΔFN-III(2) aa 837–938, RIM-BP^ΔFN-III(3) 951–1030 aa, RIM-BP^ΔSH3-I 606-681aa, RIM-BP^ΔSH3-II 1318–1382 aa, RIM-BP^ΔSH3-III 1445-1509aa, RIM-BP^ΔSH3-II+III 1318-1509aa, RIM-BP ^SH3-II+III 1318-1509aa, RIM-BP ^SH3-III 1445-1509aa, for NTRs RIM-BP^ΔNTR1 and RIM-BP^ΔNTR2.

For the amplification of the first fragment (A), the primer pair Fwd_1 and Rev_1 was used. For the amplification of the second fragment (B), the primer pair Fwd_2 and Rev_2 (see Table 1) was used. For annealing of fragments (A and B), the Invitrogen Elongase Enzyme Mix kit was used with the primer pair combination of Fwd_1 and Rev_2. Domain deletions of RIM-BP were generated by using overlapping primers in a SLIC (Li and Elledge, 2012) approach with RIM-BP^FL as a template. Through Gateway reaction, all RIM-BP deletion constructs were shuttled into GAL4/UAS vector (pUAST) containing a C-terminal EGFP tag. After sequencing, the constructs were injected into an attP site containing fly strain (y1 w1118; PBac{y+-attP-3B}VK00002; Bloomington Drosophila Stock Center line 9723) using the services of BestGene Inc.

RIM-BP-RNAi constructs were cloned into the pWalium20 vector following the Harvard Medical School TRIP protocol and injected into the Bloomington line 9744-VK27 and balanced over CyO (BestGene). 71-nucleotide primers were synthesized using isoform-specific primers: RBP2-Fw: 5′-CTA​GCA​GTG​GGC​ACC​GAC​AAT​CAG​CCA​CCT​AGT​TAT​ATT​CAA​GCA​TAG​GTG​GCT​GAT​TGT​CGG​TGC​CCG​CG-3′ and RBP2-Rv: 5′-AAT​TCG​CGG​GCA​CCG​ACA​ATC​AGC​CAC​CTA​TGC​TTG​AAT​ATA​ACT​AGG​TGG​CTG​ATT​GTC​GGT​GCC​CAC​TG-3′.

The mutation of second PXXP motif at position 1690,1693 to APTA was cloned according to the Counter Selection BAC Modification kit obtained from Gene Bridges GmbH.

The rpsL-neomycin (neo) template DNA was used to generate selectable cassettes. The primers consisted of 50-bp homology regions and a sequence for the amplification of the rpsL-neo counter selection cassette. The selectable cassettes were generated in a proofreading PCR with Vent Polymerase (New England Biolabs, Inc.) and the following primer pairs: fw: 5′-GTG​TAA​CGC​ATG​TAC​AGC​ATA​GTT​ACC​CAA​CAC​TGG​CCT​CCC​GAA​GAG​CCG​GCC​TGG​TGA​TGA​TGG​CGG​GAT​CG-3′, and rv: 5′-GTG​TGG​GAC​TGG​CAT​TCA​GCT​TCG​GGA​AAT​TGA​TAT​TGG​TTG​GCT​TGA​TCA​GAA​GAA​CTC​GTC​AAG​AAG​GCG-3′.

APTA point-mutated cac cDNA was used as a template to generate selectable cassettes. The selectable cassettes were generated in a proofreading PCR with Vent Polymerase (New England Biolabs, Inc.) and the following primer pairs: 5′-GTG​TAA​CGC​ATG​TAC​AGC​ATA​G-3′ and 5′-ATG​TGT​GGG​ACT​GGC​ATT​CAG-3′.

All cac constructs were injected into an attP site containing fly strain (PBac {y[+]-attP-3B}VK00033; Bloomington Drosophila Stock Center line 9750) using the services of BestGene Inc.

The full-length drbp cDNA (according to isoform F) domain deletions of RIM-BP were generated by using overlapping primers in a SLIC (Li and Elledge, 2012) approach, the gateway cloning system, or conventional cloning (contact the authors for exact cloning strategies and primers). The following aa were deleted: RIM-BP^ΔFN-III(1) 747–826 aa, RIM-BP^ΔFN-III(2) aa 837–938, RIM-BP^ΔFN-III(3) 948–1030 aa, RIM-BP^ΔSH3-I 606-681aa, RIM-BP^ΔSH3-II 1318–1382 aa, RIM-BP^ΔSH3-III 1445-1509aa, RIM-BP^ΔSH3-II+III 1318-1509aa, RIM-BP ^SH3-II+III 1318-1509aa, RIM-BP ^SH3-III 1445-1509aa, for NTRs RIM-BP^Δ26-250 and RIM-BP^Δ152-599. Through Gateway reaction, all RIM-BP deletion constructs were shuttled into GAL4/UAS vector (pUAST) containing a C-terminal EGFP tag. After sequencing, the constructs were injected into an attP site containing fly strain (y1 w1118; PBac{y+-attP-3B}VK00002; Bloomington Drosophila Stock Center line 9723) using the services of BestGene Inc.

RIM-BP-RNAi constructs were cloned into the pWalium20 vector following the Harvard Medical School TRIP protocol (https://fgr.hms.harvard.edu/files/fly/files/2ndgenprotocol.pdf) and injected into the Bloomington line 9744-VK27 and balanced over CyO (BestGene). 71-nucleotide primers were synthetized using isoform-specific primers: RBP2-Fw: 5′-CTA​GCA​GTG​GGC​ACC​GAC​AAT​CAG​CCA​CCT​AGT​TAT​ATT​CAA​GCA​TAG​GTG​GCT​GAT​TGT​CGG​TGC​CCG​CG-3′ and RBP2-Rv: 5′-AAT​TCG​CGG​GCA​CCG​ACA​ATC​AGC​CAC​CTA​TGC​TTG​AAT​ATA​ACT​AGG​TGG​CTG​ATT​GTC​GGT​GCC​CAC​TG-3′.

The mutation of second PXXP motif at position 1690,1693 to APTA was cloned according to the Counter Selection BAC Modification kit obtained from Gene Bridges GmbH.

The rpsL-neomycin (neo) template DNA was used to generate selectable cassettes. The primers consisted of 50-bp homology regions and a sequence for the amplification of the rpsL-neo counter selection cassette. The selectable cassettes were generated in a proofreading PCR with Vent Polymerase (New England Biolabs, Inc.) and the following primer pairs: fw: 5′-GTG​TAA​CGC​ATG​TAC​AGC​ATA​GTT​ACC​CAA​CAC​TGG​CCT​CCC​GAA​GAG​CCG​GCC​TGG​TGA​TGA​TGG​CGG​GAT​CG-3′, and rv: 5′-GTG​TGG​GAC​TGG​CAT​TCA​GCT​TCG​GGA​AAT​TGA​TAT​TGG​TTG​GCT​TGA​TCA​GAA​GAA​CTC​GTC​AAG​AAG​GCG-3′.

APTA point-mutated cac cDNA was used as a template to generate selectable cassettes. The selectable cassettes were generated in a proofreading PCR with Vent Polymerase (New England Biolabs, Inc.) and the following primer pairs: 5′-GTG​TAA​CGC​ATG​TAC​AGC​ATA​G-3′ and 5′-ATG​TGT​GGG​ACT​GGC​ATT​CAG-3′.

All cac constructs were injected into an attP site containing fly strain (PBac {y[+]-attP-3B}VK00033; Bloomington Drosophila Stock Center line 9750) using the services of BestGene Inc.

Fly strains were reared under standard laboratory conditions at 25°C and 65–70% humidity in incubators. Both male and female larvae were used for analysis in all experiments, except when indicated differently. All pUAST-RIM-BP-GFP constructs were crossed into the WT and into a rim-bp^null (Df2.01 and rim-bp^STOP1) background (Liu et al., 2011) using the UAS-GAL4 expression system (Brand and Perrimon, 1993). The following genotypes were used: WT: +/+ (w1118), UAS-RIM-BP-FL^GFP; UAS-RIM-BPΔFN-III(1)^GFP; UAS-RIM-BPΔFN-III(2)^GFP; UAS-RIM-BPΔFN-III(3)^GFP; UAS-RIM-BPΔSH3-I^GFP; UAS-RIM-BPΔSH3-II^GFP; UAS-RIM-BPΔSH3-III^GFP; UAS-RIM-BPΔSH3-II^GFP+ΔSH3-III^GFP; UAS-RIM-BP-SH3-III^GFP; UAS-RIM-BP-SH3-II+III^GFP; UAS-RIM-BPΔNTR1^GFP; UAS-RIM-BPΔNTR2^GFP; elaV-Gal4 (C155)/X (Bloomington 458) was used for reexpression of all deletion construct (elaV-Gal4/X;; rim-bp^STOP1/TM6B); and ok6-Gal4 (Bloomington 64199) was used for expression of the RIM-BP-RNAi (Sigrist laboratory, see above). We used the MIMIC line from Bloomington (60200, MIMIC line, C-terminal genomic EGFP insertion in the PB isoform) for RIM-EGFP, and we used rim¹⁰³ as rim mutant (Müller et al., 2012)/Df(3R)BSC650 (Bloomington 25740). We used fife^ex1027 as fife mutant (Bruckner et al., 2012)/Df(3L)BSC412 (Bloomington 24916). For brp^nude, see Hallermann et al. (2010b). Deletion constructs were raised at 29°C for IF and STED analysis and at 25°C for viability assay and used protein-rich food (see Liu at al., 2011).

The Y2H analyses were performed as described in Böhme et al. (2016) and are summarized here: DNA fragments encoding various regions of RIM-BP, BRP, and Unc13A were each cloned into two bait and two prey vectors (plasmids pACT4-DM and pGAD426-D3 [encoding Gal4 activation domain fusions at the N terminus]). The bait/prey constructs encode for the following proteins: RIM-BP (Isoform F) 1–600 aa; RIM-BP 610–1042 aa; RIM-BP 745–1024 aa; RIM-BP 1040–1328 aa; RIM-BP 1441–1507 aa; RIM-BP 1441–1844 aa; Cacophony (Isoform A) 1500–1848 aa; Unc13 (Isoform A) 1–606 aa; Aplip1 (Isoform B) 1–250 aa; RIM (Isoform P) 1–227 aa; RIM (Isoform P) 1135–1306 aa; Fife (Isoform D) 474–794 aa; and BRP (Isoform G) 100–650 aa. All cDNA fragments encoding the bait protein fragments were subcloned into two gateway-compatible yeast expression vectors, pBTM116-D9 and pBTMcC24-DM, and transformed into the MATa yeast strain L40ccua. Bait and prey constructs that led to autoactivation (transcription of the reporter genes HIS3, URA3) were removed from the analysis. For mating, 25 ml fresh overnight cultures of the bait strains (−Trp/+His +Ade +Ura +Leu) were distributed into 384-well micro titer plates (40 µl each well) using a pipetting robot (Biomek FX, Beckman Coulter). Each single prey yeast colony grown on media containing −Leu/+His +Ade +Ura + Trp was selected and resuspended into a well of the 40-µl bait strains. The pairwise combinations of bait and prey strains were mixed and transferred immediately onto yeast extract peptone dextrose (YPD) agar plates using a spotting robot (KBiosystems) and incubated for 36 h at 30°C. For detection of protein–protein interactions, diploid yeasts carrying both baits and preys were spotted from YPD agar to synthetic dropout IV media that lacked tryptophan, leucine, histidine, and uracil (−Leu −Trp −Ura −His) agar plates. Interacting bait–prey pairs were identified by growth on selective synthetic dropout IV agar plates (−Leu −Trp −Ura −His) after 5–6 d of incubation at 30°C. Non-autoactivating baits (L40ccU MATa yeast strains) were mated with prey strains at least four times using independently transformed bait and prey yeast colonies (384 array format). Only bait vector/prey vector combinations that showed growth in at least four independent replicas were considered as putative PPIs.

For synaptosomes preparation, see below. The synaptosome pellet was resuspended in homogenization buffer (320 mM sucrose, 5 mM Hepes, 1 mM PMSF, and a protease inhibitor tablet, pH 7.4). Freshly prepared disuccinimidyl sulfoxide cross-linker (http://www.thermofisher.com/order/catalog/product/A33545) was added to a final concentration of 1 mM and incubated for 1 h at room temperature. The cross-linking reaction was quenched by the addition of 50 mM Tris, pH 8.0, for 30 min. The cross-linked synaptosome pellet was denatured in 8 M urea, reduced with 5 mM DTT at 56°C for 30 min and alkylated with 40 mM chloroacetamide for 30 min in the dark. Proteins are sequentially digested with Lys-C (4 h, 37°C, 1:75 [wt/wt] enzyme:substrate ratio) and trypsin (overnight, 37°C, 1:100 [wt/wt] enzyme:substrate ratio). The resulting peptide mixture was desalted using Sep-Pak C18 cartridges (Waters), dried under vacuum, reconstituted in 10% (vol/vol) formic acid, and fractionated by strong cation exchange chromatography. Strong cation exchange fractions were analyzed by liquid chromatography/mass spectrometry (LC/MS) using an UltiMate 3000 UPLC system (Thermo Fisher Scientific), equipped with an in-house packed C18 column for reversed phase separation (column material: Poroshell 120 EC-C18, 2.7 µm; Agilent Technologies) and coupled online to an Orbitrap Fusion mass spectrometer (Thermo Fisher Scientific). LC/MS acquisition was performed using 3 h gradient and a previously established MS2-MS3 method (https://www.ncbi.nlm.nih.gov/pubmed/28524877). In brief, sequential collision-induced dissociation (CID) MS2, and electron transfer dissociation MS2 acquisitions were applied to each MS1 precursor, followed by CID MS3 acquisitions if a unique mass difference (Δ = 31.9721 D) was observed in the CID MS2 spectrum. Data analysis was accomplished using XlinkX software (https://www.ncbi.nlm.nih.gov/pubmed/28524877), and results were reported at 1% false discovery rate (FDR).

Dissections were performed following standard protocols (Andlauer et al., 2014; Andlauer and Sigrist, 2012) and are described here: Third instar larvae were opened in HL3 by opening dorsally along the midline and removing the innards. Filets were fixated with 4% paraformaldehyde in PBS (pH 7.2) for 10 min for all antibodies or ice-cold MeOH for the anti-UNC13A antibody. After fixation, the filets were washed with PBS plus 0.05% Triton X-100 and blocked for 60 min in 5% normal goat serum. The larvae were incubated with primary antibodies at 4°C overnight and subsequently washed in a PBS plus 0.05% Triton X-100 solution for 2 h at room temperature for immunostaining. Larvae were then incubated for 3 h with secondary antibodies at room temperature. Washing procedures were repeated. Immunocytochemistry was equal for both conventional confocal and STED microscopy. Larvae were finally mounted either in Vectashield (Vector Laboratories) or Mowiol (Sigma-Aldrich) for STED.

The following primary antibodies were used: aBRP^C-term (DSHB: Nc82), mouse, for IF 1:250, Developmental Studies Hybridoma Bank; aBRP^N-term (Ex -2), rabbit, for IF 1:500 from our laboratory (see below); aBRP (last200), rabbit, for Western blot (WB) 1:50,000 (Siebert et al., 2015); aRIM-BP^C-term, (9171) rabbit, for IF 1:500, for WB 1:5,000 (Liu et al., 2011); aRIM-BP^N-term, (9172) rabbit, for IF 1:500 from our laboratory (see below); aUNC13A, guinea pig, for IF 1:500 (Böhme et al., 2016), aGFP mouse (A-11120, Invitrogen), aGFP rabbit (cA-11122, Invitrogen) for IF 1:500, aTubulin mouse, for WB 1:100,000 (62204, Thermo Fisher Scientific), HRP-Cy5 conjugated antibody (123–605-021, Jackson ImmunoResearch), for IF 1:250. All secondary antibodies were used 1:500. The following secondary antibodies were used: Alexa 488 anti-mouse (A-11001, Invitrogen) and anti-rabbit (A-11008, Invitrogen), 1:500; Cy3 anti-mouse (ab97035, abcam) and Cy3 anti-rabbit (ab6939, abcam), 1:500; Cy5 anti-mouse (ab6563, abcam) and anti-rabbit (ab97077, abcam), 1:500; Alexa 594 anti-mouse (A32742, Invitrogen) and anti-rabbit (A32754, Invitrogen) 1:500 used for STED; Atto647N anti-mouse (50185, Sigma Aldrich) and anti-rabbit (40839, Sigma Aldrich) and Al647 anti-guinea pig (A-21450, Invitrogen) 1:500 used for STED.

The BRP N-terminal (Ex -2) antibody was raised in rabbit against the peptide sequence RVRRLQELPTVDR, which is only present in the brp exon-2 encoded amino acid sequence (BRP isoform G; aa 19–31). The antiserum was affinity-purified with the peptide.

The RIM-BP^N-term (9172) antibody was raised in rabbit against the following peptide sequence: SANVEEENRRPEKAAAAASKKQKHKQQKSRPRGSHSMPYESMHHHQSAAAAVAAGTTPNGMLDALSLQLRDAEMRRTEIERAHQETLAQIRNLSGSARPDAEAVENLQSRARELEKKVALENVRCEELQIELTSALKAKQASRSACSGMGSVSSGGGATIPTSASSSTVTWAPTISHQDQGSEIDIIMAKIEQDNRVLAELEQPRTSASASMSALPPSSMLSTVNSEFRTISKSELEEELNRYKRAV. A 6×His-tagged fusion protein was used for immunization of rabbit. The amino acid sequence of positions 8–151 aa was used as antigen. The primers used for amplification of the DNA fragment are RBP-N-term FW 5′-AGG​GCG​CCA​TGG​CCA​GCG​CAA​ACG​TG-3′ and RBP-N-term REV 5′-CGG​ATC​CGG​TAC​CTT​AAA​CTG​CTC​TTT​T-3′. The fragment was cloned topET-30 His-tag vector, and Escherichia coli was used as an expression host for expression and purification of the target protein. Following the injection of the animals, the antibody (AB)-containing serum (obtained from Selbaq) was affinity-purified versus the same protein as used for immunization.

Conventional confocal and STED images were acquired with Leica DMI 6000 (SP8) and TCS SP8 gSTED 3× microscopes (Leica Microsystems), respectively. For confocal scans, a HC PL APO CS2 63×/1.40-NA oil objective (Leica Microsystems) was used, and for STED a HC PL APO CS2 100×/1.40-NA oil objective (Leica Microsystems). Images were acquired at ∼20°C, and fluorochromes used are indicated in the antibody section. For detection, HYD (high sensitive) 400–800 nm spectral descanned for green and red channels and PMT 400–800 nm spectral descanned for far red channels were used for confocal scans. For STED, HyD Sp GaAsP were used. The NMJ z-stacks had a step size of 0.2–0.3 µm between single optical slices. All images were acquired using the LAS X software (Leica Microsystems). For all confocal and STED image analysis, the software ImageJ 1.52n was used. For all statistical analysis, the software GraphPad Prism, version 5.01, was used. For STED microscopy, Huygens Deconvolution software was used, applying a theoretical point spread function automatically computed based on pulsed- or continuous-wave STED optimized function and the specific microscope parameters. Default deconvolution settings were applied. Image analysis followed the standard protocol as described in Andlauer and Sigrist (2012) and is described in detail for the different analyses as follows.

For whole NMJ confocal GFP mean pixel intensities for RIM-BP-FL^GFP and RIM-BP deletions^GFP, first the original confocal stacks were converted to maximal projections and the background subtracted. Second, the HRP signal was used as a template for a mask, restricting the quantified area to the shape of the NMJ. This mask was applied to the GFP channel and mean pixel intensity measured. Data distribution was tested following the D’Agostino and Pearson omnibus normality test, all data had a normal distribution, and an unpaired, two-sided t test comparing RIM-BP-FL^GFP to all other genetic conditions was applied. N represents a single NMJ. Analyzed are five animals/two NMJs per animal.

For quantification of single synapses (BRP number/NMJ, BRP mean intensity or BRP area) by confocal imaging, the signal of an HRP channel was used as a template for a mask, restricting the quantified area to the shape of the NMJ. The original confocal stacks were converted to maximal projections. After background subtraction, a HRP mask was created by applying a threshold to remove spurious low-intensity pixels. The segmentation of single spots was performed semiautomatically via the “Find Maxima” routine and by hand with the pencil tool and a line thickness of 1 pixel. The processed picture was then transformed into a binary mask using the same lower threshold value as in the first step. This binary mask was then projected onto the original unmodified image using the “min” operation from the ImageJ image calculator. The individual spots of the resulting images were counted with the help of the “analyze particle” function with a lower threshold set to 5. Single active zone per NMJ (AZ/NMJ) were then analyzed regarding mean pixel intensity, area, and number. AZ/NMJ area was obtained by normalizing the total number of AZs analyzed to the NMJ area measured via HRP. HRP mean pixel intensity was also measured (via the HRP mask) and if not significantly different between the genotypes, BRP mean pixel intensity values were normalized to the HRP mean pixel intensity. Analyzed were two NMJs/animal of three to six animals; n represents a single NMJ. For comparison of WT with RIM-BP-FL^GFP, the D’Agostino and Pearson omnibus normality test could not be performed due to too small n, but data distribution was assumed to be normal. An unpaired, two-sided t test was applied. For all other analyses of this type, data distribution was normal according to D’Agostino and Pearson omnibus normality test, and an unpaired, two-sided t test was applied.

To evaluate the nanoarchitecture of the singe active zone, the BRP^C-term ring was analyzed by STED microscopy. Closed rings were counted as normal rings, open or elongated rings as deformed. Lateral rings at the edge of the bouton as well as AZ marked by only two-clustered BRP “rings” were excluded from analysis. N represents the mean of all AZ (20–40) from 1–3 boutons (one image). Two or three animals with two or three NMJs/animal and one to three boutons/NMJ were analyzed to reach ∼100 AZ/genetic condition. The D’Agostino and Pearson omnibus normality test could not be performed due to too small n, but data distribution was assumed to be normal. A one-way ANOVA was applied, using Tukey post-test.

To determine the number of UNC13A sites at a single AZ by STED microscopy, individual 100–300 AZs in a planar orientation were manually chosen in three animals with two or three NMJs/animal and two or three boutons/NMJ. n represents a single AZ (exception for the double LOF for RIM-BP-RNAi and rim mutant; as the genotype is highly lethal, we only had one larvae to analyze with three NMJs and one to three boutons/NMJ). UNC13A sites were identified via the Find Maxima function using a noise tolerance of 175 within a radius of 400 nm from the center of the BRP^Nc82 ring. Data distribution is normal following the D’Agostino and Pearson omnibus normality test. An unpaired, two-sided t test was applied.

Distance measurements were performed by STED microcopy. A line profile was laid across vertical oriented synapses and the peak to peak distance measured. Data distribution is normal following the D’Agostino and Pearson omnibus normality test. A one-way ANOVA was applied using Tukey post-test. N represents a single AZ from three animals with two or three NMJs/animal and one or two boutons/NMJ; a total of 50–90 AZ/genotype was analyzed.

Three biological replicates were performed and 100–200 offspring per F1 generation analyzed. The genotypes were as follows: as driver line elaV-Gal4/X (homozygous);; rim-bp^STOP1/TM6B crossed either to the RIM-BP-FL^GFP rescue construct/II (homozygous); rim-bp^null (Df2.01)/TM6B or the eight deletion constructs (see Fig. 3 A; ΔRIM-BP^GFP/II (homozygous); rim-bp^null (Df2.01)/TM6B). Applying the Mendelian laws, the theoretically possible outcome of the correct genotype (elaV-Gal4/+; ΔRIM-BP^GFP/+; rim-bp^null (Df2.01)/rim-bp^STOP1) is due to lethality of the balancer chromosomes if homozygous (TM6B/TM6B) 33%. The survival rate was thus normalized to those 33% for all genotypes. Statistical analysis was performed with GraphPad Prism 5 software using an unpaired, two-sided t test comparing RIM-BP-FL^GFP with all other genetic conditions. As n = 3 for each biological replicate (per replicate, 100–200 animals were analyzed), the D’Agostino and Pearson omnibus normality test could not be performed due to too small n, but data distribution was assumed to be normal.

Sequences used in this study were obtained using a nucleotide and protein BLAST search from the publicly available database of the National Center for Biotechnology Information. The phylogenetic tree, multiple global sequence alignments (Blosum65), and editing (indications of deletion constructs, coiled-coil domain and interaction sites identified in cross-linking mass spectrometry) were performed in a Windows platform with the Geneious program version 9.1.8. Dark color within alignment indicates amino acid identity; light color indicates amino acid similarity.

Two-electrode voltage clamp (TEVC) recordings were performed essentially as previously reported (Matkovic et al., 2013). They comprised spontaneous recordings (miniature excitatory junction currents: mEJCs, 90 s), single evoked (evoked excitatory junction currents: eEJCs, 20 repetitions at 0.2 Hz) and high-frequency recordings (paired-pulse 10 ms or 30 ms interstimulus interval, PP10 or PP30, 10 repetitions at 0.2 Hz; 60 pulses at 100 Hz for cumulative quantal content computation) as well as mean variance analysis (4× eEJC protocol at different c[Ca²⁺]). All experiments were performed on third instar larvae raised at 25°C on semi-defined medium (Bloomington recipe). The dissection and recording medium was extracellular haemolymph-like solution 3 (Stewart et al., 1994; composition in mM: 70 NaCl, 5 KCl, 20 MgCl₂, 10 NaHCO₃, 5 trehalose, 115 sucrose, and 5 Hepes, pH adjusted to 7.2). Dissection was performed in ice-cold Ca²⁺-free HL3 medium, while mEJC, eEJC, and high-frequency recordings were performed in 1.5 mM Ca²⁺ HL3 at room temperature. Data for mean-variance analysis were recorded at 0.75, 1.5, 3, and 6 mM Ca²⁺ by starting with a bath volume of 2 ml 0.75 mM Ca²⁺ HL3, consecutively removing 1 ml of the former bath solution and adding 1 ml of 2.25, 4.5, or 9 mM Ca²⁺ HL3, respectively, while mixing carefully with a pipette and giving 1 min of acclimation period before the next measurement. For all physiological recordings, intracellular electrodes with a resistance of 15−25 MΩ (filled with 3M KCl) were placed at muscle 6 of the abdominal segment A2/A3 for all physiological recordings. The data acquired were low-pass filtered at 1 kHz and sampled at 10 kHz. The command potential for mEJC recordings was −80 mV, and −60 mV for all other recordings. Only cells with an initial membrane potential between −50 and −70 mV and input resistances of ∼4 MΩ were used for further analysis.

The eEJC and paired-pulse traces were analyzed for standard parameters (amplitude, rise time, decay, charge flow, paired-pulse [PP]-ratio) by using a semiautomatic custom-written Matlab script (Mathworks, version R2009a). The 100-Hz trains were analyzed for amplitudes by using another semiautomatic custom-written Matlab script that calculates eEJC amplitudes by measuring peak to baseline directly before the onset of the response. The quantal content of each response was calculated by dividing the amplitude by the mean quantal size of the respective genotype. Release-ready vesicles (y intercept) and refilling rate (slope) were determined by back extrapolation of the last 300 ms of cumulative quantal contents. Mean variance analysis was basically performed as described previously (Reddy-Alla et al., 2017). In short, the amplitudes of 20 repetition traces per c[Ca²⁺] were averaged and plotted against mean variance of the amplitudes (SD²) to obtain the mean versus variance plot and parabolic fits. Second-order polynomial fits (SD² = q × Ī − Ī²/N) were performed per cell where q is the quantal size, Ī is the mean current amplitude, and N is the number of release sites. Vesicular release probability was calculated by P_VR = Ī/(N × q) per cell. The parabolas in Fig. 4 represent fits to the mean values of a full dataset per genotype. Stimulation artifacts in eEJC and paired-pulse recordings were removed for clarity. The mEJC recordings were analyzed with pClamp 10 software (Molecular Devices). GraphPad Prism v5.01 (GraphPad Software, Inc.) was used for all fitting procedures.

Data were analyzed using GraphPad Prism v5.01. Data distribution is normal following the D’Agostino and Pearson omnibus normality test. If data had a normal distribution, an unpaired, two-sided t test was applied for comparison of two conditions or a one-way ANOVA using Tukey post-test if more than two conditions were compared. If data did have a normal distribution, the nonparametric, two-tailed Mann–Whitney test was applied for comparison of two conditions or the nonparametric Kruskal–Wallis test with the Dunn’s post-test if more than two conditions were compared. For standard TEVC analysis, n = 1 cell, and one or two cells from four to six animals were analyzed. For variance-mean analysis, n is one animal, and seven to nine animals were analyzed.

Conventional embedding was performed as described previously (Matkovic et al., 2013) and described here: dissected third instar larvae were fixed with PFA (10 min; 4% PFA and 0.5% glutaraldehyde in 0.1 M PBS) and glutaraldehyde (60 min; 2% glutaraldehyde in 0.1 M sodium cacodylate), washed in sodium cacodylate buffer, and post-fixed with 1% osmium tetroxide and 0.8% KFeCn in 0.1 M sodium cacodylate buffer (1 h on ice). After washing with sodium cacodylate buffer and distilled water, the samples were stained with 1% uranyl acetate in distilled water. Samples were dehydrated and infiltrated in epon resin. Subsequently, muscles 6/7 of the abdominal segment A2/3 were cut out. Collected in an embedding mold, the blocks were polymerized and cut into thin serial sections (65–70 nm). These sections were post-fixed and poststained with uranyl acetate/lead citrate. Micrographs were taken with an EM (JEM 1011; JEOL) equipped with a camera (Orius 1200A; Gatan) using the DigitalMicrograph software package (Gatan).

The plasma membrane and the electron-dense T-bar were detected by eye and labeled manually for quantification. The T-bar roof size was measured by a straight line connecting the furthest distance of the upmost T-bar dense material (in relation to the plasma membrane). The T-bar area was obtained by surrounding the dense material and measuring the area of the region of interest created.

Data were analyzed using GraphPad Prism v5.01. Data distribution is normal following the D’Agostino and Pearson omnibus normality test. An unpaired, two-sided t test was applied. All ns represent the number of quantified T-bars/SVs at a T-bar in four to six animals. No statistical methods were used to predetermine sample sizes, but our sample sizes are similar to those generally employed in the field.

Larval CNS protein extraction was performed as follows: 10 CNS were dissected from third instar larvae. The tissues obtained were sheared manually in 5 µl of 2% SDS aqueous solution using a micropistill fitting tightly into a 1.5-ml cup. An amount of 0.5 µl of a 10% Triton X-100 aqueous solution and 5 µl of 2× sample Laemmli buffer was added, and samples were heated at 95°C for 10 min. Each sample was centrifuged for 2 min at 16,000 g to pellet the debris. An amount of 10 µl (equivalent to 10 larval CNS) was subjected to denaturing SDS-PAGE using a 6% Tris/HCl gel.

Proteins were then transferred onto a nitrocellulose membrane blocked with 5% skim milk in 1× PBS supplemented with 0.1% Tween-20 and probed with rabbit anti-RIM-BP (1:5,000; Liu et al., 2011) and mouse anti-tubulin (1:100,000) diluted in 5% milk in 1× PBS, supplemented with 0.1% Tween-20. After washing, secondary anti-rabbit (111–035-045, Dianova) or anti-mouse (115–035-146, Dianova) IgG HRP-conjugated antibodies were used for detection in conjunction with an enhanced chemiluminescence (RPN 2232, GE Healthcare ECL Prime) detection system with Hyperfilm ECL (GE Healthcare).

Pulldown experiments with synaptosomes confirm the PPI between both proteins in vivo. Approximately 6,000 fly heads were collected, and synaptosomes were purified via differential centrifugation (see below and Depner et al., 2014). 20 µg of indicated rabbit antibodies each were coupled to 50 µl Protein A–coated agarose beads. Bead-antibody complexes were incubated with solubilized and precleared synaptosome membrane preparations (LP1) for at least 4 h. After four washing steps with immunoprecipitationbuffer (containing 20 mM Hepes, pH 7.4, 200 mM KCl, 2 mM MgCl₂, and 1% Triton X-100), antibody-antigen complexes were eluted with 60 µl 2× denaturing protein sample buffer each. For Western blot analysis, 1% of input and 10 µl of eluate were loaded. Probing was performed with RbαBRPlast200 (1:50,000, Depner et al., 2014) and RbαRIM-BPC-term (1:5,000, Liu et al., 2011).

The procedure involves decapitation of adult flies (sieving, ∼6,000 heads), pulverization, homogenization (320 mM sucrose, 4 mM Hepes, protease inhibitors [complete, 11873580001, Roche]) and differential centrifugation (from low speed to higher speed: 1,000–15,000 g) of fly heads, which allows subsequent purification of presynaptic and postsynaptic components (Depner et al., 2014).

Two constructs for the NTR were prepared (NTR^8-151 and NTR^8-254) and cloned into the pETM11 vector with an N-terminal His₆-tag. The protein production of NTR^8-151 and NTR^8-254 is identical. The expression vector was transformed in E. coli BL21 Roesetta2 cells. Overexpression was performed using auto-induction media at 37°C until an OD ∼0.7 was reached and subsequently cooled down to 18°C (Studier, 2005). Cells were harvested by centrifugation (10 min, 6,000 rpm at 4°C). Buffer A was used (50 mM Tris/HCl, pH 7.4, 250 mM NaCl, 1 mM DTT, and 20 mM imidazole) for resuspension of the cell pellets. Cells were lysed by homogenization at 4°C, and the lysate was cleared by centrifugation (1 h, 21,000 rpm at 4°C). A Ni²⁺-NTA (column volumes [cv] ∼1 ml; GE Healthcare) was equilibrated with buffer A. The NTR was loaded on the column and washed with 10 cv of buffer A. The NTR was eluted in a linear gradient to buffer A with an imidazole concentration of 400 mM. Pooled fractions were dialyzed overnight into buffer B (50 mM Tris/HCl, pH 7.4, 100 mM NaCl, and 1 mM DTT). A MonoQ HR 16/10 column (GE Healthcare) was equilibrated with buffer B before protein loading. Subsequently, the column was washed with 5 cv of buffer B. The NTR was eluted in a linear gradient to buffer B but with 1 M NaCl over 10 cv. 2 mg of Tobacco Etch Virus (TEV) protease were added to the pooled fractions and the protein dialyzed overnight into buffer C (10 mM Tris/HCl, pH 7.4, 100 mM NaCl, and 1 mM DTT). Size exclusion chromatography was performed with a HighLoad Superdex S75 26/60 column (GE Healthcare), equilibrated with buffer C. Pooled protein fractions were concentrated with Amicon-Ultra 10,000. Protein concentrations were determined by the absorbance at 280 nm.

The DNA coding for the three FN-III domains comprising aa 745–1042 was cloned into the pGEX-6P vector (GE Healthcare) fused to an N-terminal GST-tag. Expression and cell harvesting were performed as described above. The pellet was resuspended in buffer D (50 mM Tris/HCl, pH 7.5, 500 mM NaCl, 10% [vol/vol] glycerol, and 2 mM DTT). Cells were broken by sonication at 4°C, and the supernatant was cleared by 1 h of centrifugation (20,000 rpm at 4°C). The supernatant was loaded on 5 ml GSH-sepharose beads (GE Healthcare) equilibrated with buffer D. The bound protein sample was washed with 10 cv of buffer A, and a subsequent high-salt wash with buffer A with 1 M NaCl followed by 10 cv of buffer A. The protein was cleaved from the beads overnight by the addition of 2 mg Precission protease in buffer D. Size exclusion chromatography of FN-III(1–3) and a HighLoad Superdex S75 26/60 column (GE Healthcare) were equilibrated with 10 mM Tris/HCl, pH 7.5, 500 mM NaCl, and 10% (vol/vol) glycerol. Pooled protein fractions were concentrated with Amicon-Ultra 10,000 to 6 mg/ml.

Selenomethionine-labeled protein was overexpressed in E. coli BL21pLys Express cells in minimal medium supplemented with amino acids including selenomethionine instead of methionine (Van Duyne et al., 1993). The DTT was added to all buffers to a final concentration of 4 mM for protein purification. Incorporation of selenomethionine and molecular mass of proteins was confirmed by MALDI-TOF mass spectrometry.

The influence on the thermal stability of proteins under different buffer conditions was measured with the Mx3005P qPCR system (Agilent) in a 96-well plate format using a laboratory-made screen. Each well contained 135 µl buffer, 10 µl protein (1–2 µg/µl), and 1 µl of 10x SYPRO Orange dye (Invitrogen). The program consisted of three steps: step 1 was a preincubation for 1 min at 20°C, and steps 2 and 3 were cycles comprising the temperature increase of 1°C within 20 s. The temperature gradient proceeded from 25 to 95°C at 1°C per minute.

The FN-III(1–3) was concentrated to 5 mg/ml for the crystallization experiment. Initial crystals of FN-III were obtained at 291 K in a sitting drop setup. Crystals appeared after 1 d in a sitting drop using a reservoir solution with a volume of 700 µl composed of 16% (wt/vol) polyethylenglycol 3350 and 300–400 mM KH₂PO₄. The crystallization drop was mixed from 1 µl protein and 1 µl reservoir solution. Initial crystals were obtained of a FN-III construct spanning residues 748 to 1039 diffracted to 10 Å. The diffraction quality of the initial crystals could be significantly improved by streak seeding. Fine tuning of the construct to a core region of ranging from 745 to 1042 improved the resolution to 3.6 Å. Differential scanning fluorimetry was employed to optimize the protein buffer. Addition of 10% (vol/vol) glycerol in all buffers increased the stability by 5°C and improved the diffraction quality of the crystals further. The crystals were transferred to a reservoir solution supplemented with 20% (vol/vol) glycerol and subsequently flash-cooled in liquid nitrogen for cryoprotection. Selenomethionine-labeled protein was concentrated to 4 mg/ml and crystallized under identical conditions as described for the native protein.

Synchrotron diffraction data were collected at the beamline 14.2 of the MX Joint Berlin laboratory at BESSY (Berlin, Germany) or beamline P14 of Petra III (Deutsches Elektronen Synchrotron, Hamburg, Germany). X-ray data collection was performed at 100 K. Diffraction data were processed with the XDS package (Kabsch, 2010; Table S1). A randomly generated set of 5% of the reflections from the diffraction dataset was used to calculate the free R-factor and excluded from the refinement. Initial phases of the FN-III were determined by a three-wavelength multiwavelength anomalous diffraction experiment. Two selenomethionine sites per monomer could be identified with the program SHARP (Vonrhein et al., 2007). An initial model was built with the program BUCCANEER (Cowtan, 2008). The structure was first refined by applying a simulated annealing protocol and in later refinement cycles merely by maximum-likelihood restrained refinement implemented in PHENIX (Afonine et al., 2012), followed by iterative model building cycles with COOT (Emsley et al., 2010). Model quality was evaluated with PROCHECK (Laskowski et al., 1996) and MolProbity (Chen et al., 2010). Figures were prepared using PyMOL. Domain movements were analyzed with the DynDom software (Girdlestone and Hayward, 2016). The atomic coordinates and structure factor amplitudes have been deposited in the Protein Data Bank under the accession no. 6Q9M.

Proteins were dialyzed against CD buffer (10 mM K₂HPO₄ and 50 mM NaCl) at 4°C overnight and diluted to a final concentration of 1.2 mg/ml. All spectra were recorded with a Jasco J-810 spectropolarimeter using quartz cuvettes with 0.1 mm path length. Initial CD spectra were collected at wavelengths between 190 and 240 nm at 20°C. The α-helical content was calculated with the CDNN CD spectral deconvolution software (version 2.1, Gerald Böhm). The CD melting profiles were recorded by heating the samples by 2°C/min from 20 to 95°C, following the CD signal at 222 nm.

Total RNA was purified from ∼20 adult fly heads using the NucleoSpin RNA Isolation kit (Macherey-Nagel) according to the manufacturer’s instructions. 1 µg RNA was reverse-transcribed into cDNA using M-MuLV Reverse transcription enzyme (200,000 units/ml, New England Biolabs) and random primers (100 pmol/µl, Invitrogen). Quantitative RT-PCR was performed using a StepOnePlus Real-Time PCR System (Thermo Fisher Scientific). Reactions were performed in triplicates in MicroAm Optical 96-well Reaction Plates (Thermo Fisher Scientific) using Luna Universal qPCR Master Mix (New England Biolabs). Target gene expression was quantified relative to Act5C using the ΔΔCT method including primer efficiency (Pfaffl, 2001). The following primer sets were used (5′-3′, forward and reverse): total RIM-BP forward 5′-CCG​GCC​GCG​GCG​GTC​TGA​GCC-3′ and reverse 5′-GTG​TAG​GGG​AAG​CAG​GGC​GAG-3′, and GFP forward 5′-GGC​GTG​CAG​TGC​TTC​AGC​CG-3′ and reverse 5′-CGT​CCT​TGA​AGA​AGA​TGG​TG-3′. N represents three independent experiments (n = 3). As n = 3, the D’Agostino and Pearson omnibus normality test could not be performed due to too small n, but data distribution was assumed to be normal. A one-way ANOVA was applied using the Tukey post-test.

Fig. S1 shows that RIM-BP and the RIM-BP SH3 domain are conserved across species. Fig. S2 shows RIM-BP:: BRP pulldown experiments, synaptosome preparation, and RIM-BP FN-III domain conservation. Fig. S3 shows validation of the RIM-BP-FL^GFP rescue construct and Western blot analysis of Drosophila larval brains. Fig. S4 shows RIM-BP SH3-III domain analysis. Fig. S5 shows that RIM-BP SH3-II/III binding to Ca²⁺ channels is dispensable for RIM-BP integration into the AZ matrix and SH3-II/III domain localization. Fig. S6 shows that RIM-BP NTR2 is required for AZ scaffold stability and synaptogenesis. Fig. S7 is an analysis of RIM-family protein function in electrophysiology, scaffold assembly, and release site positioning. Fig. S8 shows that RIM-BP NTR1 promotes SV recruitment into the release sites. Table S1 shows diffraction data collection, phasing, and refinement statistics for FN-III(1).

We thank C. Alings for support in the crystallization experiments. We acknowledge access to the CD spectrometer of Prof. B. Koch. We thank T. Schneider for access and support at beamline P14 of PETRA III (Deutsches Elektronen Synchrotron, Hamburg, Germany). We accessed beamlines of the BESSY II (Berliner Elektronenspeicherring-Gesellschaft für Synchrotronstrahlung II) storage ring (Berlin, Germany) via the Joint Berlin MX-Laboratory sponsored by the Helmholtz Zentrum Berlin für Materialien und Energie, the Freie Universität Berlin, the Humboldt-Universität zu Berlin, the Max-Delbrück Centrum, and the Leibniz-Institut für Molekulare Pharmakologie. Finally, we would like to thank Paul Sigrist for his support.

We were supported by the Deutsche Forschungsgemeinschaft (SFB958/A6 and TRR186/A3). We thank C. Weise (SFB958/Z3) for mass spectrometric analysis. A.G. Petzoldt was supported by the SFB958 women support program and the Berlin Institute of Health Gender Equality Fund.

The authors declare no competing financial interests.

Author contributions: A.G. Petzoldt performed all genetic experiments, designed experiments, and performed all confocal and STED imaging, including analysis. T.W.B. Götz conducted all electrophysiological experiments, except for the Cac analysis, which was done by E. Knoche. J.H. Driller, J. Lützkendorf, and U. Stelzl performed the Y2H analysis. J.H. Driller and B. Loll performed the crystal structure analysis. J.H. Driller, J. Lützkendorf, S. Liu, S. Reddy-Alla, S. Mertel, and C. Beuschel cloned constructs. S. Liu performed recombinant expression experiments. J.H. Driller performed WB analysis. V. Ugorets and J. Lützkendorf performed RT-PCR. T. Matkovic-Rachid performed the EM analysis. M. Lehmann contributed to the STED microscopy. N. Ramesh and A.G. Petzoldt created the model. B. Kuropka and C. Freund performed mass spectrometry on the FN-III constructs. F. Liu performed the cross-linking experiment. M.C. Wahl and S.J. Sigrist designed the study and wrote the manuscript.