Abstract
Synaptic ribbons are large, dynamic structures in the active zone complex of ribbon synapses and important for the physiological properties of these tonically active synapses. RIBEYE is a unique and major protein component of synaptic ribbons. The aim of the present study was to understand how the synaptic ribbon is built and how the construction of the ribbon could contribute to its ultrastructural plasticity. In the present study, we demonstrate that RIBEYE self-associates using different independent approaches (yeast two-hybrid analyses, protein pull downs, synaptic ribbon–RIBEYE interaction assays, coaggregation experiments, transmission electron microscopy and immunogold electron microscopy). The A-domain [RIBEYE(A)] and B-domain [RIBEYE(B)] of RIBEYE contain five distinct sites for RIBEYE–RIBEYE interactions. Three interaction sites are present in the A-domain of RIBEYE and mediate RIBEYE(A)–RIBEYE(A) homodimerization and heterodimerization with the B-domain. The docking site for RIBEYE(A) on RIBEYE(B) is topographically and functionally different from the RIBEYE(B) homodimerization interface and is negatively regulated by nicotinamide adenine dinucleotide. The identified multiple RIBEYE–RIBEYE interactions have the potential to build the synaptic ribbon: heterologously expressed RIBEYE forms large electron-dense aggregates that are in part physically associated with surrounding vesicles and membrane compartments. These structures resemble spherical synaptic ribbons. These ribbon-like structures coassemble with the active zone protein bassoon, an interaction partner of RIBEYE at the active zone of ribbon synapses, emphasizing the physiological relevance of these RIBEYE-containing aggregates. Based on the identified multiple RIBEYE–RIBEYE interactions, we provide a molecular mechanism for the dynamic assembly of synaptic ribbons from individual RIBEYE subunits.
Introduction
Ribbon synapses are specialized chemical synapses, e.g., in the retina and inner ear, capable to maintain rapid exocytosis of synaptic vesicles for prolonged periods of time (for review, see Fuchs, 2005; Heidelberger et al., 2005; Prescott and Zenisek, 2005; Sterling and Matthews, 2005; Nouvian et al., 2006; Singer, 2007). For this purpose, ribbon synapses are equipped with presynaptic specializations, the synaptic ribbons, which are considered to speed vesicle trafficking (for review, see tom Dieck and Brandstätter, 2006; Nouvian et al., 2006; Sterling and Matthews, 2005). Synaptic ribbons are large presynaptic structures associated with the active zone complex of ribbon synapses (for review, see Wagner, 1997). Synaptic ribbons of photoreceptor synapses are plate-like structures in three-dimensional representations that can be >500 nm in length and depth. In EM sections, retinal synaptic ribbons usually appear bar-shaped, and inner ear synaptic ribbons are usually spherical structures (for review, see Nouvian et al., 2006). Also in the retina, the assembly of the bar-shaped ribbon is believed to go through spherical ribbon intermediates, the so called synaptic spheres (for review, see Vollrath and Spiwoks-Becker, 1996; Spiwoks-Becker et al., 2004). The dimensions of synaptic ribbons in the retina can vary and are subject to changes, e.g., in response to different stimuli (lighting conditions/circadian rhythm), probably reflecting structural adaptations to different degrees of synaptic activity (for review, see Vollrath and Spiwoks-Becker, 1996; Wagner, 1997). Regardless of their shape, synaptic ribbons are associated with large amounts of synaptic vesicles and other membrane compartments (for review, see Sterling and Matthews, 2005).
We have previously identified a novel protein,“RIBEYE,” as a unique and specific component of synaptic ribbons (Schmitz et al., 2000). RIBEYE is present in synaptic ribbons of all vertebrate ribbon synapses (Schmitz et al., 2000, 2006; Zenisek et al., 2004; Khimich et al., 2005; tom Dieck et al., 2005; Wan et al., 2005) (for review, see tom Dieck and Brandstätter, 2006). RIBEYE consists of a unique A-domain [RIBEYE(A)] and B-domain [RIBEYE(B)] that is identical to CtBP2 except for the first 20 aa (Schmitz et al., 2000). The B-domain of RIBEYE binds nicotinamide adenine dinucleotide (NAD+ or NADH [NAD(H)]) with high affinity and belongs to a family of d-isomer-specific 2-hydroxy acid dehydrogenases (Schmitz et al., 2000). The structural analysis of these proteins, e.g., of CtBP1, revealed the presence of two globular subdomains, namely an NAD(H)-binding subdomain (NBD) and the substrate-binding subdomain (SBD) (for a review, see Chinnadurai, 2002; Kumar et al., 2002; Nardini et al., 2003).
Previous data indicated that RIBEYE is the major component of synaptic ribbons (Schmitz et al., 2000; Zenisek et al., 2004; Wan et al., 2005). In the present study, we analyzed functional properties of RIBEYE and demonstrate that RIBEYE is capable to interact with itself. RIBEYE–RIBEYE interactions are mediated through three binding sites in the A-domain and two binding sites in the B-domain enabling multiple RIBEYE–RIBEYE interactions. RIBEYE–RIBEYE interactions can generate the three-dimensional scaffold of the synaptic ribbon and provide a molecular mechanism for the ultrastructural plasticity of these presynaptic structures.
Materials and Methods
Plasmids.
Details on all plasmids and antibodies used in this study are posted in the supplemental Methods (available at www.jneurosci.org as supplemental material).
Yeast two-hybrid methods.
We used the galactosidase-4 (Gal4)-based Matchmaker Yeast Two-Hybrid System (Clontech) according to the manufacturer's instructions. The cDNA of the respective bait proteins were cloned in frame with the Gal4-DNA-binding domain of pGBKT7. The cDNA of the indicated prey proteins were cloned in frame with the Gal4-activation domain of pACT2 or pGADT7. The bait and prey plasmids confer tryptophan and leucine prototrophy to the respective auxotrophic yeast strains. Yeast strains Y187 and AH109 were used that contain distinct auxotrophic marker genes: AH109 contained MATa, trp1–901, leu2-3,112, ura3-52, his3-200, gal4Δ, gal80Δ, LYS2::GAL1UAS-GAL1TATA-HIS3, GAL2UAS-GAL2TATA-ADE2, and URA3::MEL1UAS-MEL1TATA-lacZ (James et al., 1996); Y187 contained MATα, ura3-52, his3-200, ade2-101, trp1-901,leu2-3,112, gal4Δ, met, gal80Δ, and URA3::GAL1UAS-GAL1TATA-lacZ (Harper et al., 1993). Bait plasmids were always electroporated into AH109 yeast, whereas all prey plasmids were transformed into Y187. Preparation of electrocompetent yeasts and electroporation of yeasts were done as described previously (Helmuth et al., 2001). For identifying transformants, yeasts were plated on the respective selective plates to identify the resulting convertants to the respective prototrophy (drop out media Clontech/QBiogene). For interaction analyses, AH109 yeasts containing the respective bait plasmid were mated with Y187 yeasts containing the respective prey plasmid. Mating was performed for 5 h at 30°C in 1 ml of YPD medium (yeast extract, peptone, and dextrose) with heavy vortexing. For assessing mating efficiency, half of the mated sample was streaked on −LW plates [containing synthetic complete yeast medium without leucine (L) and without tryptophan (W)], and the other half was plated on −ALWH selective plate [containing synthetic complete yeast medium without adenine (A), L, W, and histidine (H)] with 10 mm 3-amino-1,2,4-triazole added. For the matings, pSE1111 and pSE1112 (Bai and Elledge, 1996) as well as the empty bait and prey vectors were used as negative controls. Expression of β-galactosidase (β-gal) marker gene expression were qualitatively analyzed by filter assays and quantitatively with liquid assays as described previously (Wang et al., 1997; Stahl et al., 1999).
Expression of RIBEYE(A)-domain.
RIBEYE(A)-glutathione S-transferase (GST) is difficult to express in conventional prokaryotic expression systems because of its high contents of proline, serine, glycine, and arginine residues (Schmitz et al., 2000) (our unpublished observations). We identified two expression systems to express full-length RIBEYE(A)-GST fusion protein. One source were LPAAT (lysophosphatidic acid acyltransferase)-deficient JC201 bacteria (Coleman, 1990), which express full-length RIBEYE(A)-GST fusion protein although part of it is processed to smaller fragments (see Figs. 1B, 4B, 6A,B, 10B; supplemental Fig. 1B, available at www.jneurosci.org as supplemental material). The second source were methylotropic yeast Pichia pastoris (Cereghino and Cregg, 2000). Electroporation of JC201 and expression and purification of RIBEYE(A)-GST fusion protein was performed according to standard procedures (Schmitz et al., 2000). A certain degree of proteolytic processing of RIBEYE(A) is present in both of these systems. The proteolytic processing cannot be prevented even under optimized fermenting conditions using the BioFlo 110 fermenter (New Brunswick Scientific) with constant oxygenation of the medium, pH control, different induction times, and different induction temperatures (data not shown).
Intracellular expression of untagged RIBEYE(A) in Pichia pastoris.
Pichia pastoris yeast strain GS115 (his4) (Invitrogen) was used for heterologous protein expression (Lin-Cereghino et al., 2005). Yeast cultures were grown at 30°C on synthetic minimal medium containing 0.67% yeast nitrogen base (without amino acids supplemented with ammonium sulfate and appropriate amino acid-base; Formedium). RE(A)pPIC3.5K was electroporated into freshly made electrocompetent yeasts GS115 (his4) as described (Lin-Cereghino et al., 2005). For electroporation, 10 μg of purified and SalI-linearized plasmid DNA was used. Electroporation was performed at 1500 V (BTX ECM 399 electroporator; Biogentronix) with 2 mm gapped prechilled cuvette (Peqlab). Recombinant His+ clones were selected on MD (minimal dextrose) plates (0.67% yeast nitrogen base without amino acids, 0.077% CSM-His, 2% dextrose, 0.00004% biotin, 1 m sorbitol, 1.5% agar–agar). Genomic integration of the electroporated construct was confirmed through genomic PCR (5′-AOX1-primer: GACTGGTTCCAATTGACAAGC; 3′-AOX1-primer: GCAAATGGCATTCTGACATCC). For induction of fusion protein, yeasts were first cultured in BMGY medium (1% yeast extract, 2% peptone, 0.67% yeast nitrogen base without amino acids, 0.00004% biotin, 1% glycerol, 0.1 m potassium-phosphate buffer, pH 6) at 30°C to an optical density at 600 nm (OD600) of 2–6. Induction was achieved in BMMY medium (same as BMGY with 0.5% methanol instead of glycerol) at 30°C, starting the culture at an OD600 of 1. After every 24 h, methanol was replenished to the final volume of 0.5%. After 36 h of induction, the cells were pelleted (1500 rpm, 5 min, 4°C) and processed for extraction of fusion protein. Induced Pichia pastoris yeasts were mechanically cracked with 0.5 mm glass beads (Biospec/Roth) in breaking buffer (50 mm sodium phosphate, pH 7.4, 100 mm NaCl, 1 mm EDTA, 5% glycerol, 1 mm PMSF). For this purpose, 100 μl cell pellet were resuspended in 890 μl of ice-cold breaking buffer. To this mixture, ∼500 μl of glass beads were added. The cracking was performed at +4°C using high-speed vortex (25× vortexing for 30 s; between vortexing, samples were chilled 30 s on ice). The lysate was centrifuged twice (13,000 rpm, 1 h at 4°C). Subsequently, the supernatant was precleared with 20 μl empty glutathione-agarose beads (Fluka) for 1 h at 4°C on a rotary wheel.
Protein pull-down assays using bacterial fusion protein.
For pull-down experiments using pairs of GST- and maltose-binding protein (MBP)-tagged fusion proteins, the GST-tagged fusion proteins were usually kept immobilized on glutathione beads whereas MBP fusion proteins were used as solublized prey proteins if not denoted otherwise. Bait and prey proteins were used in equimolar amounts along with the respective control proteins. Protein concentrations were determined using the Bradford method (Bradford, 1976). For pull-down experiments, fusion protein eluates were precleared with 10 μl of empty glutathione Sepharose beads (per 1 ml of eluate) for 1 h at 4°C. Binding was performed in PBS that contained 0.5% Triton X-100 at 4°C for 12 h on a rotary wheel (500 μl incubation volume) if not denoted otherwise. Pellets were washed five times by adding an excess of PBS/Triton X-100 and subsequent spinning (13,000 rpm, 1 min, 4°C). Pellets were boiled in SDS-sample buffer and subsequently subjected to Western blot analyses with the indicated antibodies.
Miscellaneous methods.
For the preparation of synaptic ribbons, synaptic ribbons were purified as described previously (Schmitz et al., 1996, 2000). Standard protein techniques were performed as described previously (Schmitz et al., 2000). For reprobing of Western blots, nitrocellulose sheets were treated with prewarmed (90°C) stripping buffer (1% SDS, 10 mm β-mercaptoethanol in PBS) and incubated at room temperature for 1 h. Immunofluorescence microscopy was performed as described previously (Schmitz et al., 2000, 2006) using a Zeiss Axiovert 200M microscope (Carl Zeiss) equipped for conventional epifluorescence microscopy with the respective filter sets for enhanced green fluorescent protein (EGFP) and monomeric red fluorescent protein (mRFP) and equipped with an Apotome (Zeiss) to make optical sections. Transfection of COS cells was done as described previously with the DEAE-dextran method (Schmitz et al., 2000). R28 cells were transfected by lipofection using perfectin (Peqlab) according to the manufacturer's instructions. Transfected cells were usually analyzed by fluorescence microscopy 48 h after transfection if not denoted otherwise. Conventional transmission, immunogold electron microscopy, and quantification of radioactive NAD+-binding to RIBEYE fusion proteins were performed as described previously (Schmitz et al., 1996, 2000). Thrombin cleavage of GST-tagged fusion protein was performed mostly as described previously (Chadli et al., 2000).
Results
Homodimerization of RIBEYE(A)
We used the yeast two-hybrid (YTH) system to determine whether the A-domain of RIBEYE can homodimerize. In YTH, we observed a strong self-interaction between the A-domains of RIBEYE as judged by growth on −ALWH-selective plates and expression of the β-galactosidase marker gene activity compared with the respective control matings (Fig. 1A, mating 1; supplemental Fig. 1A, available at www.jneurosci.org as supplemental material). RIBEYE(A) also interacted with full-length RIBEYE [RIBEYE(AB)] (Fig. 1A, mating 6). The RIBEYE-expressing yeasts were not autoactivating in YTH, as demonstrated by the lack of growth on −ALWH plates and analysis of β-galactosidase expression of the respective control matings (Fig. 1A, matings 2–5, 7, 8). Quantitative β-galactosidase activities determined in liquid assays are shown in supplemental Figure 1A (available at www.jneurosci.org as supplemental material). The homodimerization of RIBEYE(A) observed in the YTH system was also confirmed at the protein level using two different pull-down assays (Fig. 1B; supplemental Fig. 1B, available at www.jneurosci.org as supplemental material). Immobilized RIBEYE(A)-MBP fusion protein (but not immobilized MBP alone) bound soluble RIBEYE(A)-GST fusion protein (but not GST alone) (Fig. 1B). Similarly, immobilized RIBEYE(A)-GST specifically bound RIBEYE(A) from crude protein extracts of RIBEYE(A)-transgenic Pichia pastoris (supplemental Fig. 1B, available at www.jneurosci.org as supplemental material). GST control protein alone did not bind RIBEYE(A) from the Pichia pastoris extract. Thus, both YTH and protein pull-down data independently demonstrated that RIBEYE(A) interacts with RIBEYE(A).
Mapping of RIBEYE(A)–RIBEYE(A) interaction
In the rat, RIBEYE(A) consists of the N-terminal 563 aa. To map the interaction sites important for RIBEYE(A)–RIBEYE(A)-interaction, we generated C- and N-terminal deletion constructs of RIBEYE(A) and tested them for their capability to interact with full-length RIBEYE(A) in YTH (Fig. 2A,B; supplemental Fig. 2, available at www.jneurosci.org as supplemental material). Most of the C-terminal region of the A-domain could be removed without abolishing the interaction with RIBEYE(A). RIBEYE(A)1–105 was the shortest N-terminal construct that could interact with full-length RIBEYE(A)-domain (Fig. 2B, prey 7). Therefore, the first 105 N-terminal amino acids contain a binding site for RIBEYE(A), which is subsequently denoted as the “A1” interaction site. We also analyzed N-terminal deletions of RIBEYE(A) for their interaction with full-length RIBEYE(A) (Fig. 2B; supplemental Fig. 2, available at www.jneurosci.org as supplemental material). Interestingly, N-terminal deletion constructs of RIBEYE that did not contain the previously identified RIBEYE(A1)-binding site also interacted with RIBEYE(A) (Fig. 2B, preys 9, 10; supplemental Fig. 2, available at www.jneurosci.org as supplemental material), pointing to a second homodimerization site in the C-terminal region of RIBEYE(A). We identified RIBEYE(A)438–563 as the smallest C-terminal portion of RIBEYE(A) that interacts with RIBEYE(A) (Fig. 2B; supplemental Fig. 2, available at www.jneurosci.org as supplemental material). This C-terminal RIBEYE(A) interaction site is denoted as “A2” in the following text. We further tested whether the midregion of RIBEYE(A), which does neither contain the N-terminal A1 interaction site nor the C-terminal A2 interaction site for its capability to interact with RIBEYE(A). This region in the midportion of RIBEYE(A), denoted as “A3,” also interacted with RIBEYE(A) (Fig. 2B, prey 11). Thus, the A-domain of RIBEYE has three independent sites which are able to interact with full-length RIBEYE(A) (summarized in Fig. 2C). Supplemental Figure 2 (available at www.jneurosci.org as supplemental material) demonstrates that all of the tested RIBEYE constructs were not autoactivating as judged by the absence of growth on −ALWH and lack of expression of β-galactosidase activity.
The A1, A2, and A3 interaction modules in the A-domain of RIBEYE allow multiple RIBEYE–RIBEYE interactions
Next, we tested whether the identified RIBEYE(A1), RIBEYE(A2), and RIBEYE(A3) interaction modules in the A-domain of RIBEYE could interact with each other. We tested all possible interaction combinations between RIBEYE(A1), RIBEYE(A2) and RIBEYE(A3) in the YTH system and found that multiple interactions could take place between them. RIBEYE(A1) interacts with RIBEYE(A1), RIBEYE(A2), and RIBEYE(A3) (Fig. 2D,F,G). Similarly, RIBEYE(A2) interacted with RIBEYE(A2) and RIBEYE(A1) but not RIBEYE(A3) (Fig. 2E–G). RIBEYE(A3) interacted with RIBEYE(A1) and RIBEYE(A3) but not RIBEYE(A2) (Fig. 2G). All of these interactions between RIBEYE(A) subdomains characterized in the YTH system were confirmed by protein pull-down analyses using the respective fusion proteins (Fig. 3A–D,F).
Homodimerization of RIBEYE(B)
With the YTH system, we confirmed homodimerization of the B-domain of RIBEYE (supplemental Fig. 3A,B, available at www.jneurosci.org as supplemental material). The homodimerization of RIBEYE(B) is not very surprising: CtBP2, which is identical to RIBEYE(B) (except for the first 20 aa) has been shown previously to homodimerize (Thio et al., 2004). CtBP1 also homodimerizes (Sewalt et al., 1999; Balasubramanian et al., 2003), and the structure of the CtBP1 dimer (tCtBP1) has been resolved (Kumar et al., 2002; Nardini et al., 2003). NAD(H) was found to stimulate the homodimerization of both CtBP1 and CtBP2 (Balasubramanian et al., 2003; Thio et al., 2004).
RIBEYE(B) also interacted with RIBEYE full-length protein indicating that the A-domain of RIBEYE does not prevent homodimerization of RIBEYE(B)-domains (supplemental Fig. 3A,B, available at www.jneurosci.org as supplemental material). The homodimerization of RIBEYE(B) is dependent on amino acids 689–716, which form the αB-loop-αC motif [homodimerization loop (HDL)] of RIBEYE(B) as judged by homology modeling (supplemental Fig. 3D, available at www.jneurosci.org as supplemental material). The αB-loop-αC motif in CtBP1 is important for homodimerization of CtBP1 (Nardini et al., 2003). In agreement with this prediction, homodimerization of RIBEYE(B) is completely abolished if the HDL is deleted (supplemental Fig. 3C, matings 1, 2, available at www.jneurosci.org as supplemental material). RIBEYE(B)ΔHDL (supplemental Fig. 3, available at www.jneurosci.org as supplemental material). The RIBEYE(B) homodimerization interface is denoted as “B1” in the following text.
Heterodimerization of RIBEYE(B) and RIBEYE(A)
We used the YTH system to test whether RIBEYE(B) can also interact with RIBEYE(A). RIBEYE(B) showed a robust interaction with RIBEYE(A) in the YTH system as judged by growth on −ALWH selective plates and β-galactosidase marker gene expression (Fig. 4, mating 1; supplemental Figs. 4, 5, mating 1, available at www.jneurosci.org as supplemental material). This interaction between RIBEYE(A) and RIBEYE(B) was verified at the protein level using protein pull-down analyses (Fig. 4B). RIBEYE(A)-GST fusion protein, but not GST alone, specifically interacted with RIBEYE(B)-MBP fusion protein (but not with MBP alone) as judged by protein pull-down analyses (Fig. 4B). We used the YTH system to map the respective interaction sites for RIBEYE(B)–RIBEYE(A) interaction. Mapping analyses revealed that the A2 interaction site in the C-terminal portion of the RIBEYE(A) is the binding site for RIBEYE(B) (Fig. 5; supplemental Fig. 4B, available at www.jneurosci.org as supplemental material). On RIBEYE(B), the NBD is responsible for the interaction with RIBEYE(A) (Fig. 5C; supplemental Fig. 4C, available at www.jneurosci.org as supplemental material). These mapping data obtained by YTH analyses were confirmed by protein–protein pull-down analyses that showed interaction between RIBEYE(A2) and RIBEYE(B) (Fig. 3E). We used deletion and point mutants of RIBEYE(B) to further analyze the binding site of RIBEYE(A) on the NBD of RIBEYE(B) in detail. First, we tested whether the RIBEYE(B)ΔHDL deletion mutant that is no longer able to homodimerize with RIBEYE(B) (supplemental Fig. 3C,D, available at www.jneurosci.org as supplemental material) is still able to interact with RIBEYE(A). Indeed, RIBEYE(B)ΔHDL is able to interact with RIBEYE(A) as well as with full-length RIBEYE [RIBEYE(AB)] (Fig. 5C; supplemental Figs. 4C, 7, available at www.jneurosci.org as supplemental material). Next, we tested different point mutants of the NBD of RIBEYE(B) for their interaction with RIBEYE(A) and RIBEYE(B). We analyzed RIBEYE(B) point mutants RIBEYE(B)G730A, D758N, I796A, E844Q, F848W, and K854Q, which are located at the outer face of the NBD (Fig. 5D). All of these point mutations did not prevent homodimerization with RIBEYE(B) (supplemental Figs. 5C, 6, available at www.jneurosci.org as supplemental material). Furthermore, RIBEYE(B)D758N, I796A, E844Q, F848W, and K854Q bound NADH as judged by NADH-dependent energy transfer from tryptophan W867 to bound NADH [performed as described by Fjeld et al. (2003)] (data not shown), demonstrating the proper folding of these point mutants. Although these point mutants did not prevent homodimerization of RIBEYE(B), all of these point mutations [except for RIBEYE(B)K854Q] completely abolished interaction with RIBEYE(A) (supplemental Fig. 5A–C, available at www.jneurosci.org as supplemental material). This shows that the two binding interfaces on RIBEYE(B) available for interaction with RIBEYE(A) and RIBEYE(B) are distinct from each other, albeit spatially closely related. The binding site for RIBEYE(A) covers a large portion of the NBD (Fig. 5D; supplemental Fig. 5, available at www.jneurosci.org as supplemental material). Interestingly, RIBEYE(B)G730, which is an essential component of the conserved NAD(H)-binding motif of RIBEYE (Schmitz et al., 2000), appears to be part of the interaction interface for RIBEYE(A): the point mutant RIBEYE(B)G730A (Fig. 5D) that does not bind NAD(H) (supplemental Fig. 5D, available at www.jneurosci.org as supplemental material) can no longer interact with RIBEYE(A) (supplemental Fig. 5C, available at www.jneurosci.org as supplemental material). We interpret the latter result to mean that the binding sites for NAD(H) and for RIBEYE(A) are overlapping to a certain extent (see below and Discussion). The docking site on RIBEYE(B) for RIBEYE(A) is denoted as “B2” in the following text.
NADH and NAD+ inhibit RIBEYE(A)–RIBEYE(B) interaction
Because RIBEYE(A) docks to a broad interface of the NAD(H)-binding subdomain of RIBEYE(B), we analyzed whether this interaction is dependent on NAD(H). To analyze this question, we applied the pull-down assay described in Materials and Methods. We used RIBEYE(A)-GST as immobilized bait and eluted RIBEYE(B)-MBP as soluble prey protein and checked for interaction of these proteins in the presence of increasing concentrations of NADH/NAD+ (Fig. 6). Increasing concentrations of NADH/NAD+ strongly inhibited RIBEYE(A)–RIBEYE(B) interaction. Both NAD+ as well as NADH strongly inhibited RIBEYE(A)–RIBEYE(B) interaction already at low physiological concentrations. The tested concentrations of NAD(H) are within the cellular concentration range of NAD(H) known from other studies (Zhang et al., 2002; Fjeld et al., 2003). The NAD(H) concentrations did not have any influence on the control pull downs. At all NADH/NAD+ concentrations used, there was no unspecific binding of RIBEYE(B)-MBP to GST alone (supplemental Fig. 8, available at www.jneurosci.org as supplemental material). The identified interactions between the different subdomains of RIBEYE and their regulation via NAD(H) are summarized in Figure 11B.
RIBEYE coaggregates with other RIBEYE molecules in transfected R28 and COS cells
To determine whether the identified RIBEYE–RIBEYE interactions can also occur within the cellular context, we performed cell transfections with the indicated RIBEYE expression constructs that were differentially tagged either with EGFP or with mRFP. For transfection, we used COS7 cells and the R28 retinal progenitor cell line (Seigel, 1996; Seigel et al., 2004). R28 cells express retinal and neuronal marker proteins (e.g., opsins, β-2 arrestin, recoverin, neurotransmitter receptors, and various presynaptic and postsynaptic proteins) in addition to stem cell/precursor cell markers (e.g., nestin) (Seigel et al., 2004). If transfected alone, both RIBEYE(A) as well as RIBEYE(AB) displayed a discrete, spot-like distribution, whereas RIBEYE(B) is diffusely distributed (Fig. 7; supplemental Figs. 9–11, available at www.jneurosci.org as supplemental material), as also described previously (Schmitz et al., 2000). If RIBEYE(A)-EGFP was cotransfected with RIBEYE(A)-mRFP both coaggregated to the same protein clusters as judged by the large extend of colocalization of the EGFP and mRFP signals (Fig. 7C, arrows; supplemental Figs. 10C, 11A, available at www.jneurosci.org as supplemental material). Identical results were obtained when full-length RIBEYE(AB)-EGFP was cotransfected with RIBEYE(A)-mRFP (Fig. 7B, arrows). If RIBEYE(B)-EGFP was cotransfected with RIBEYE(B)-mRFP, both signals remained diffusely distributed (Fig. 7E). Interestingly, whenever RIBEYE(B)-mRFP was cotransfected with RIBEYE(A)-EGFP, RIBEYE(B)-mRFP redistributed from a diffuse distribution [as typical for single transfected RIBEYE(B)], to a patchy, spot-like distribution that is typical for RIBEYE(A) (Fig. 7D, arrow). Part of RIBEYE(B) remained diffusely distributed (Fig. 7D, arrowhead; supplemental Fig. 10D,E, available at www.jneurosci.org as supplemental material) probably because of the NAD(H) sensitivity of the RIBEYE(B)–RIBEYE(A) interaction (see above). NAD(H) is ubiquitously present in the cytoplasm and expected to partly dissociate RIBEYE(A)–RIBEYE(B) complexes. Interestingly, in cells double-transfected with full-length RIBEYE(AB) and RIBEYE(B), RIBEYE(B) virtually completely redistributed from the diffuse distribution to the spot-like distribution typical for RIBEYE(AB) and perfectly colocalized with RIBEYE(AB) (Fig. 7A, arrows; supplemental Figs. 9, 10B, 11B, available at www.jneurosci.org as supplemental material). From these latter experiments, we conclude that both homotypic domain interactions [RIBEYE(B)–RIBEYE(B) interactions] as well as heterotypic domain interactions [RIBEYE(A)–RIBEYE(B) interactions] support the interaction between RIBEYE(AB) and RIBEYE(B). As judged by the nearly complete colocalization of RIBEYE(AB) and RIBEYE(B) compared with cells double-transfected with RIBEYE(A) and RIBEYE(B), we assume that a combination of homotypic and heterotypic domain interactions is probably stronger than a single type of homotypic interactions. Qualitatively identical results were obtained for R28 cells (Fig. 7; supplemental Fig. 9, available at www.jneurosci.org as supplemental material) and COS cells (supplemental Figs. 10, 11, available at www.jneurosci.org as supplemental material). Also in COS cells, RIBEYE(A) coaggregated with RIBEYE(AB) (supplemental Fig. 10A, available at www.jneurosci.org as supplemental material) and RIBEYE(A) (supplemental Figs. 10C, 11A, available at www.jneurosci.org as supplemental material). Similarly, RIBEYE(B) translocated from a completely diffuse distribution to a spot-like distribution if cotransfected with RIBEYE(A) or RIBEYE(AB) [in 98 and 99%, respectively, of double-transfected cells in 100 randomly picked double-transfected cells vs 3% spot-like distribution in cells transfected with RIBEYE(B) only].
Interestingly, if RIBEYE(AB)-EGFP-transfected cells were analyzed already a few hours after transfection, the RIBEYE(AB)-containing aggregates appeared smaller and more numerous than at later time points suggesting that the smaller protein clusters could mature/coalesce to the bigger protein aggregates that are predominant at later time points (supplemental Fig. 9, available at www.jneurosci.org as supplemental material).
In conclusion, the coaggregation and colocalization data in the transfected COS and R28 cells indicate that the interaction sites between RIBEYE(A) and RIBEYE(B), either between the same type of domains (A–A, B–B) or between different domains (A–B), are also available within a cellular context.
Electron microscopy of RIBEYE-containing aggregates in transfected R28 cells
RIBEYE is the major component of synaptic ribbons and RIBEYE forms large protein aggregates in transfected cells (Fig. 7). We analyzed the ultrastructural appearance of the RIBEYE-containing aggregates by electron microscopy to find out whether these structures have similarities with synaptic ribbons (Fig. 8). Using conventional transmission electron microscopy, we observed large electron-dense aggregates in RIBEYE-EGFP-transfected R28 cells (Fig. 8A–J), which were absent in control cells (K). Similar, large electron-dense protein aggregates were also present in RIBEYE(AB)-EGFP-transfected COS cells but not in EGFP-transfected COS cells (data not shown). The large aggregates typically displayed a spherical shape with a diameter between of 200–500 nm. These electron-dense structures were often surrounded by vesicles which in part were physically attached to the electron-dense aggregates via thin electron-dense stalks (Fig. 8A–J, arrowheads). These large spherical structures were strongly positive for RIBEYE by immunogold labeling with antibodies against RIBEYE (Fig. 8L–N) but not reactive with antibodies against tubulin (O) or RIBEYE preimmune serum (P) (control incubations). These spherical structures have similarities to spherical synaptic ribbons of inner hair cells (for review, see Nouvian at al., 2006). Beside the large electron-dense particles we also found smaller aggregates which showed physical contacts between each other and which sometimes appeared to coalesce into larger, electron-dense structures (Fig. 8E,F). These structures were also partly physically linked to surrounding vesicles and show some resemblance to synaptic spheres, intermediate structures in the assembly and disassembly of synaptic ribbons (for review, see Vollrath and Spiwoks-Becker, 1996) (see Discussion).
RIBEYE coaggregates at bassoon-containing sites in retinal R28 progenitor cells
To further address the physiological relevance of the RIBEYE aggregates, we tested whether these structures are related to bassoon, a physiological interaction partner of RIBEYE at the active zone of ribbon synapses (tom Dieck et al., 2005). Bassoon is endogenously expressed in R28 retinal precursor cells as judged by immunocytochemistry (Fig. 9), Western blotting (supplemental Fig. 9D, available at www.jneurosci.org as supplemental material) and reverse transcription-PCR (data not shown). Bassoon is distributed in R28 in a spot-like manner (Fig. 9D, arrowheads). The RIBEYE clusters in RIBEYE(AB)-EGFP-transfected R28 cells primarily formed around this bassoon-containing clusters and colocalized with bassoon (Fig. 9A–C, arrows). The preferential colocalization between RIBEYE and its physiological interaction partner bassoon emphasizes the physiological relevance and ribbon-like partial function of the RIBEYE-containing protein aggregates.
Purified synaptic ribbons recruit externally added RIBEYE(A) and RIBEYE(B)
Next, we tested whether isolated, purified synaptic ribbons can recruit externally added RIBEYE(B)-GST and RIBEYE(A)-GST fusion proteins (Fig. 10). GST alone was used as control protein. Purified synaptic ribbons bound soluble RIBEYE(A)-GST and RIBEYE(B)-GST fusion proteins (Fig. 10A,B). GST control protein did not bind to synaptic ribbons (Fig. 10Aa, lane 6, Ba, lane 6) demonstrating the specificity of binding. Thus, the RIBEYE–RIBEYE interaction sites are accessible on synaptic ribbons and available to recruit externally added, additional RIBEYE proteins. We verified that the binding of RIBEYE(A) to purified synaptic ribbons is independent of the attached GST tag by removing the GST tag by thrombin cleavage (Fig. 10C). Untagged RIBEYE(A) cosedimented with purified synaptic ribbons but not without synaptic ribbons, further confirming the specific binding of RIBEYE(A) to purified synaptic ribbons. To further evaluate the binding of RIBEYE(A) to synaptic ribbons, we also tested whether RIBEYE(A1), RIBEYE(A2), and RIBEYE(A3) (used as purified MBP-tagged fusion proteins) were able to bind to purified synaptic ribbons. RIBEYE(A1)-MBP and RIBEYE(A3)-MBP bound to synaptic ribbons whereas MBP alone did not demonstrating the specificity of the interaction. Interestingly, RIBEYE(A2)-MBP did not bind to purified synaptic ribbons although it efficiently interacted with RIBEYE(A) subunits [RIBEYE(A1), RIBEYE(A2)] in protein pull-down assays (Fig. 3). We interpret these findings that the RIBEYE(A2)-binding sites/options are probably unavailable or blocked by other proteins on purified synaptic ribbons (see Discussion). The recruitment of additional RIBEYE subunits to preexisting ribbons could explain the known dynamic growth and ultrastructural plasticity of synaptic ribbons (see Discussion). Figure 11 depicts a simplified model that schematically shows how synaptic ribbons could be built from individual RIBEYE subunits via the identified RIBEYE–RIBEYE interactions.
Discussion
Synaptic ribbons are large and dynamic macromolecular constructions in the active zone of ribbon synapses. At present, it is not clearly understood how the synaptic ribbon is made and how it functions in the synapse. In the present study, we demonstrated that RIBEYE is a scaffold protein that contains multiple interaction sites for other RIBEYE molecules. Noteworthy, the RIBEYE–RIBEYE interactions involve sites in the A-domain as well as in the B-domain of RIBEYE, i.e., three distinct interaction sites in the A-domain (A1, A2, A3) and two in the B-domain (B1, B2). We have shown that these five interaction sites allow either homotypic domain interactions [interactions between same type of domains: RIBEYE(A)–RIBEYE(A), RIBEYE(B)–RIBEYE(B)] or heterotypic domain interactions [RIBEYE(A)–RIBEYE(B)]. Homotypic domain interactions can be either homotypic or heterotypic concerning the subdomain involved. A homotypic domain interaction, e.g., RIBEYE(A)–RIBEYE(A), can be mediated either by homotypic subdomain interactions, e.g., RIBEYE(A1)–RIBEYE(A1), or by heterotypic subdomain interactions, e.g., RIBEYE(A1)–RIBEYE(A2). The cotransfection experiments demonstrated that RIBEYE proteins interact with each other and coaggregate into the same protein clusters. Given the fact that RIBEYE is the major component of synaptic ribbons (Schmitz et al., 2000; Zenisek et al., 2004; Wan et al., 2005), the multiple protein interactions of RIBEYE provide a molecular mechanism how the scaffold of the synaptic ribbon can be created. RIBEYE–RIBEYE interactions could directly link the individual RIBEYE units to each other. Because RIBEYE is present throughout the entire synaptic ribbon, RIBEYE–RIBEYE interactions could thus generate and stabilize the macromolecular structure of the synaptic ribbon. The proposed modular model of synaptic ribbons could explain how the scaffold of the synaptic ribbon is formed mostly from a single protein component (RIBEYE). In agreement with this hypothesis, the RIBEYE aggregates in transfected R28 cells possess structural and functional similarities with synaptic ribbons. RIBEYE(AB)-transfected R28 cells formed electron-dense large protein aggregates that were partly associated with surrounding vesicles and membrane compartments. The electron-dense aggregates were usually round in shape and resembled spherical synaptic ribbons of inner hair cells (Nouvian et al., 2006). Bar-shaped/plate-shaped ribbons were not observed in the RIBEYE-transfected cells. Thus, the spherical synaptic ribbon appears to be the “basal” type of synaptic ribbon structure that is built from RIBEYE and most likely additional factors are needed to build plate-shaped ribbons from spherical ribbons. The colocalization of RIBEYE with its physiological interaction partner bassoon in R28 cells emphasizes the physiological relevance of the RIBEYE-containing protein aggregates and suggest that the RIBEYE-containing aggregates fulfill partial ribbon-like functions. Because RIBEYE is not the only component of synaptic ribbons (Schmitz et al., 2000; Wan et al., 2005), it cannot be expected that RIBEYE alone makes fully mature ribbons, e.g., with a dense and regular association of synaptic vesicles. Very likely, additional ribbon components are necessary to provide full-ribbon function and structure.
In principle, the multiple interaction sites present on RIBEYE can be important for both intramolecular and intermolecular RIBEYE–RIBEYE interactions. Intermolecular RIBEYE–RIBEYE interactions could provide the three-dimensional scaffold of the synaptic ribbon as discussed above. Intramolecular RIBEYE–RIBEYE interactions could shield the interaction sites from unwanted intermolecular interactions to keep the protein soluble. Such a shielding of binding sites could be particularly important during development and to prevent the assembly of synaptic ribbons at unwanted, unphysiological subcellular sites (e.g., outside of the presynaptic terminal).
It is likely that the interaction between different RIBEYE domains and RIBEYE molecules is regulated. In the present study, we found that NAD(H) is an important regulator of RIBEYE interactions. RIBEYE(A)–RIBEYE(B) interactions are efficiently inhibited by low, physiological concentrations of NAD(H). Both NADH and NAD+ are very efficient in disrupting RIBEYE(A)–RIBEYE(B) complexes. Thus, NAD(H) appears to act as a molecular switch that distinguishes between two different types of RIBEYE–RIBEYE interactions: in the presence of NAD(H), RIBEYE(A)–RIBEYE(B) interactions are disassembled (this study), whereas RIBEYE(B)–RIBEYE(B) interactions are favored as judged by the NAD(H)-induced dimerization of CtBP2 (Thio et al., 2004). The binding interface on RIBEYE(B) for RIBEYE(B) interaction is spatially closely related but distinct from the binding interface on RIBEYE(B) for RIBEYE(A). This was shown by the analyses of point and deletion mutants of RIBEYE(B) that affect one type of interaction [RIBEYE(A)–RIBEYE(B) interaction] but not the other [RIBEYE(B)–RIBEYE(B) interaction] (Fig. 5C,D; supplemental Figs. 5, 6, available at www.jneurosci.org as supplemental material). The binding of NAD(H) could induce a conformation of RIBEYE(B) that favors homodimerization of RIBEYE(B) and that is incompatible with the formation of RIBEYE(B)–RIBEYE(A) heterodimers. RIBEYE(B)G730 is an essential part of the NADH-binding motif and the RIBEYE(B)G730A point mutant no longer interacts with RIBEYE(A). Therefore, one possible mechanism for the NADH-induced dissociation of the RIBEYE(A2)–RIBEYE(B) interaction could be that the NAD(H)-binding region of RIBEYE(B) is also part of the binding interface with RIBEYE(A). If NADH binds to RIBEYE it could displace RIBEYE(A) from RIBEYE(B) and stimulate homodimerization of RIBEYE(B). By this way of thinking, NAD(H) would favor RIBEYE complexes that contain a homodimerized B-domain, which is likely important for RIBEYE function. Additionally, RIBEYE(B) displaced from RIBEYE(A2) would make the A2-binding module available for RIBEYE(A)–RIBEYE(A) interactions. By this mechanism, binding of NAD(H) could potentially initiate the assembly of synaptic ribbons. The NADH concentrations used in the present study are well in the range of the known cellular concentrations of NADH (Zhang et al., 2002; Fjeld et al., 2003) and thus very likely capable in regulating RIBEYE–RIBEYE interactions in situ.
The suggested modular assembly of the synaptic ribbon from individual RIBEYE units also provides a molecular explanation for the ultrastructural dynamics of synaptic ribbons by the addition or removal of RIBEYE subunits or rearrangements of RIBEYE–RIBEYE complexes. The ribbon recruitment experiments showed that binding sites for additional RIBEYE subunits are accessible and available on synaptic ribbons at a molecular level. Isolated synaptic ribbons (Schmitz et al., 1996, 2000) are able to bind externally added RIBEYE(B) and also RIBEYE(A). The multiple RIBEYE–RIBEYE interaction sites in the A-domain suggest a predominantly structural role of the A-domain as previously suggested (Schmitz et al., 2000). Probably large portions of RIBEYE(A) are likely “buried” in the core of the synaptic ribbons. Still, part of the A-domain is accessible in isolated synaptic ribbons and therefore partly exposed. In ribbon pull-down experiments (Fig. 10), RIBEYE(A1) and RIBEYE(A3) but not RIBEYE(A2) did bind to purified synaptic ribbons. Because RIBEYE(A2) can bind to both A1 and A2 interaction sites but not to the A3 interaction site, we suggest that A1 and A2 are located in the core of the ribbon, where these sites are not available for interaction with RIBEYE(A2). In contrast, the A3 region appears at least partly exposed on purified synaptic ribbons where it is free to interact with other protein, i.e., externally added RIBEYE(A1) and RIBEYE(A3) (Fig. 10D). Binding of RIBEYE(B) probably occurs via homodimerization of RIBEYE(B)-domains based on homologous findings with CtBP2 (Balasubramanian et al., 2003; Thio et al., 2004). This homodimerization is favored by the presence of NADH. Interestingly, RIBEYE(B) of synaptic ribbons does not bind RIBEYE(A2), although the respective fusion proteins can interact in an NAD(H)-dependent manner. Therefore, the RIBEYE(B)-binding site for RIBEYE(A2) might be blocked or the binding disfavored, e.g., by RIBEYE(B) homodimerization, or inhibited by NAD(H) bound at synaptic ribbons via the NBD of RIBEYE. Clearly, these working hypotheses have to be analyzed by future investigations and testing these assumptions will shed further light on the understanding of the construction and assembly of synaptic ribbons and how they work in the synapse.
In conclusion, our data show that RIBEYE is a scaffold protein with ideal properties to explain the assembly of synaptic ribbons as well as its ultrastructural dynamics via the modular assembly mechanism. The capability to interact with other RIBEYE proteins in multiple ways could explain how a single protein, RIBEYE, builds the scaffold for the entire ribbon (Fig. 11). Our transfection experiments actually show that RIBEYE can form aggregates that resemble spherical synaptic ribbons. The proposed modular assembly of the synaptic ribbon from individual RIBEYE subunits provides a molecular basis for the ultrastructural plasticity of synaptic ribbons (e.g., changes in size and shape of the ribbon). The binding of externally added RIBEYE to purified synaptic ribbons mimics the growth of synaptic ribbons that occurs in situ, e.g., under darkness in the mouse retina (Balkema et al., 2001; Spiwoks-Becker et al., 2004; Hull et al., 2006). Similarly, RIBEYE-aggregates increased in size over time in light microscopy (supplemental Fig. 9, available at www.jneurosci.org as supplemental material) and RIBEYE aggregates appeared to be able to coalesce into larger structures at the ultrastructural level. The regulation of RIBEYE–RIBEYE interactions, e.g., by NAD(H), could contribute to the regulation of structural plasticity of synaptic ribbons.
Footnotes
-
This work was supported by Deutsche Forschungsgemeinschaft Grant SFB530 TP C11 and HOMFOR (F.S.), and by Research to Prevent Blindness and National Institutes of Health Infrastructure Grant R24EY016662 (G.M.S.). We thank Drs. M. J. Schmitt and L. Köblitz (Saarland University) for the donation of plasmids and yeast strains; Dr. Susanne tom Dieck (Max Planck Institute for Brain Research Frankfurt) for the gift of anti-rat RIBEYE(A) antibody; Dr. H. T. McMahon (Medical Research Council, Cambridge, UK) for the donation of JC201 bacteria; and Dr. Jutta Schmitz-Kraemer for critically reading this manuscript.
- Correspondence should be addressed to Dr. Frank Schmitz, Department of Neuroanatomy, Institute for Anatomy and Cell Biology, Saarland University, Medical School Homburg/Saar, Kirrbergerstrasse, Building 61, 66421 Homburg/Saar, Germany. frank.schmitz{at}uniklinik-saarland.de