Aberrant function and structure of retinal ribbon synapses in the absence of complexin 3 and complexin 4

Ca²⁺-triggered fusion of synaptic vesicles is one of the most tightly regulated membrane fusion reactions. It is executed by SNARE complexes, which in most neurons contain the vesicle protein synaptobrevin 2 (also known as Vamp2), and the plasma membrane proteins SNAP25 and syntaxin 1 (Jahn and Scheller, 2006). SNARE complex function at neuronal synapses is controlled by regulatory proteins, which determine the extreme speed and accuracy of synaptic excitation-secretion coupling (Wojcik and Brose, 2007). Complexins are prominent representatives of this group of SNARE regulators (Brose, 2008).

Complexins bind and stabilize assembled SNARE complexes (Bracher at al., 2002; Chen et al., 2002). Their deletion in mice causes profound deficits in release probability, evoked transmitter release and reduced spontaneous release at neuronal synapses (Reim et al., 2001; Xue et al., 2007; Xue et al., 2008), as well as reduced secretory granule fusion in adrenal chromaffin cells (Cai et al., 2008), which indicates that complexins act as positive regulators at or following the Ca²⁺-triggering step of synaptic vesicle fusion. The notion that complexins act as facilitators of SNARE-mediated secretory vesicle fusion is also supported by the observation that overexpression of Cplx2 in wild-type chromaffin cells increases chromaffin granule secretion (Cai et al., 2008). However, data obtained from in vitro fusion assays and the phenotype of complexin-deficient Drosophila mutants are more compatible with a role of complexins as inhibitory pre-fusion clamps that maintain SNARE complexes in a highly fusogenic state but prevent them from executing fusion until triggered by Ca²⁺ (Giraudo et al., 2006; Schaub et al., 2006; Tang et al., 2006; Huntwork and Littleton, 2007). It is therefore possible that complexins exert both positive and negative regulatory effects on SNARE function, the relative contribution of which may differ between organisms and cell types (Xue et al., 2007; Brose, 2008).

Mice have four complexin genes (Cplx1-Cplx4), three of which (Cplx1-Cplx3) are expressed in brain (Reim et al., 2005). Cplx3 expression in mouse brain is very low and does not contribute significantly to synaptic transmission in the brain regions tested so far (Xue et al., 2008). All four complexin isoforms are expressed in the retina, where Cplx1 and Cplx2 are found in conventional synapses of amacrine cells, whereas Cplx3 and Cplx4 are predominantly expressed in ribbon synapses of photoreceptors and bipolar cells (Reim et al., 2005). All complexins bind SNARE complexes and their overexpression rescues the phenotypic deficits of hippocampal neurons lacking Cplx1 and Cplx2, indicating that the basic functional properties of complexins are similar in conventional synapses (Reim et al., 2005).

Ribbon synapses are a specialized subclass of chemical synapses characterized by a presynaptic plate-like organelle, the ribbon, which is anchored to the plasma membrane in close vicinity to voltage-gated Ca²⁺-channels (tom Dieck et al., 2005). Ribbons are typically surrounded by a large number of synaptic vesicles, which can be morphologically divided into docked vesicles that contact the active zone, tethered vesicles that cluster on the surface of the ribbon but do not touch the plasma membrane, and free vesicles that are distributed in the synaptic cytoplasm. Such large vesicle pools are obviously necessary to support the high rates of synaptic vesicle fusion that enable ribbon synapses to respond precisely to graded changes in membrane potentials (Parsons and Sterling, 2003; Heidelberger et al., 2005; Sterling and Matthews, 2005), but little is known about the molecular composition and the functional properties of the presynaptic fusion machinery that determine the unique characteristics of retinal ribbon synapses. To test whether Cplx3 and Cplx4 contribute to the unique efficacy of transmitter release at retinal ribbon synapses, we studied retina function in mice lacking Cplx3 and/or Cplx4.

Cplx4 single-knockout mice were generated by homologous recombination (Fig. 1A). To confirm the mutation of the Cplx4 gene, Southern blot analyses of offspring resulting from interbreeding of mice heterozygous for the Cplx4 mutation were performed (Fig. 1B). The results demonstrated that the respective genotypes (wild-type +/+, heterozygous +/-, homozygous -/-) appeared in the expected mendelian frequency. Western blot analysis using extracts of adult retinae showed that Cplx4 expression is completely abolished in homozygous mutants whereas protein levels in heterozygous animals were reduced (Fig. 1C). As expected, Cplx4 protein was not detectable in wild-type or mutant brains (Fig. 1C) (Reim et al., 2005). As we described previously, deletion of the Cplx3 gene leads to complete loss of Cplx3 protein in brain (Xue et al., 2008). Western blot analyses using extracts from adult retinae showed that expression of Cplx3 is abolished in homozygous Cplx3 knockouts (Fig. 1D). Double mutants lacking Cplx3 and Cplx4 were generated by interbreeding of parents double heterozygous for the Cplx3 and Cplx4 mutations. Again, standard western blot analyses of retinae from adult animals confirmed that expression of Cplx3 and Cplx4 is abolished in homozygous double deletion mutants (Fig. 1E).

Mice lacking Cplx3 or Cplx4 and double-knockout animals showed no obvious phenotypic changes and bred normally. To assess changes in protein composition of mutant synapses, we compared the levels of several synaptic proteins (Cplx1-Cplx4, syntaxin 1, syntaxin 2, syntaxin 3, SNAP25, synaptobrevin 2, Bassoon, Ribeye, synaptophysin, synaptotagmin 1, VGLUT1, VIAAT, RIM, Munc13-1-Munc13-3) in mutant and wild-type retinae by western blotting (Fig. 2). Except for the ribbon-specific structural protein Ribeye, for which we observed lower levels in the Cplx3/4 double-knockout mutants, the amounts of tested proteins were comparable between the three different types of homozygous knockouts and their respective wild-type controls.

To study the altered Ribeye expression levels in more detail and to test whether the deletion of Cplx3 and Cplx4 causes subtle alterations of Cplx1 and Cplx2 expression that are not detectable by standard western blotting, we analyzed the amounts of Cplx1 and Cplx2 in retina extracts of Cplx3/4 double mutants by quantitative western blot analysis. Because Cplx3 and Cplx4 are ribbon-synapse-specific SNARE-regulating proteins, we also examined whether their lack influences the expression of key retinal SNARE complex components (syntaxin 1, syntaxin 3, SNAP25, synaptobrevin 2) or structural proteins such as the active zone cytomatrix protein Bassoon. The results of these quantitative analyses, expressed as a percentage of wild-type levels and summarized in Table 1, showed that only the Ribeye expression levels in Cplx3/4 double mutants were altered significantly (62.5±9.1% of wild-type levels, n=3, P<0.05). Ribeye expression levels are not reduced in Cplx3 and Cplx4 single-knockout mice (Cplx3: 99.9±13.4%, n=4, P=0.997; Cplx4: 103.5±21.8%, n=3, P=0.897).

To assess the functional consequences of Cplx3 or/and Cplx4 deletion, we performed ERG recordings (Figs 3, 4). In order to study photoreceptor function and synaptic transmission in the outer retina, we first analyzed the amplitudes and implicit times of a- and b-waves, which principally originate in photoreceptor and depolarizing bipolar cells, respectively (Nusinowitz and Heckenlively, 2006). We found that depletion of Cplx3 alone (Fig. 3A-C) prolonged the implicit time for the b-wave in Cplx3 knockout mice, especially at higher light intensities where cones become active (Fig. 3C); however, a- and b-wave amplitudes remained unchanged (Fig. 3B). In Cplx4 single-knockout mice (Fig. 3D-F), a-wave amplitudes (Fig. 3E) and b-wave implicit times (Fig. 3F) were unchanged as compared with controls. Only a minor reduction of the b-wave amplitudes was detected at the lowest light intensities, where only rods are active (Fig. 3E), as was a slight increase of a-wave implicit times at medium intensities, where the a-wave becomes visible (Fig. 3F). These results corroborate previous results from immunocytochemical studies, which showed that Cplx3 is mainly present in cone terminals and only weakly expressed in rod terminals whereas Cplx4 is mainly present in rod terminals in the OPL (Reim et al., 2005).

In contrast to the single Cplx3 and Cplx4 knockout mice, deletion of both complexins (Fig. 3G-I) led to a significant reduction of the b-wave amplitudes at all light intensities. Moreover, a-wave amplitudes were slightly increased at medium intensities in Cplx3/4 double-knockout mice (Fig. 3H). In addition, b-wave implicit times were prolonged at all intensities, as were a-wave implicit times at medium intensities (Fig. 3I). Because the strong reduction in b-wave amplitude in the Cplx3/4 double-knockout mice cannot be explained by the results from either of the single knockouts (Cplx3 or Cplx4) or by an additive combination of both (Cplx3 and Cplx4), which is at least qualitatively the case for the a- and b-wave implicit times (compare Fig. 3C,F,I), a cooperative effect of Cplx3 and Cplx4 deletion has to be postulated to cause the observed changes in the b-wave response amplitude (Fig. 3A,D,G vs 3B,E,H).

Since Cplx3 and Cplx4 are also abundant in the inner retina, we additionally analyzed the oscillatory potentials (OPs), which are most probably generated by bipolar cell transmission to amacrine cells, which in turn interact with bipolar, amacrine and ganglion cells. OPs occur mainly at higher light intensities, which activates the cones (Nusinowitz and Heckenlively, 2006; Weymouth and Vingrys, 2008). Fig. 4A-C demonstrates the parameters we used to analyze the OPs, which are superimposed upon the ERG b-wave. The responses (Fig. 4A) were bandpass filtered (Fig. 4B), followed by fast Fourier transformation (FFT) and estimation of peak power and peak frequency (Fig. 4C). Depletion of Cplx3 resulted mainly in a reduction of peak power and peak frequency of the OPs (Fig. 4D,E). This becomes manifest as a large reduction of OP amplitudes and a slight reduction of OP frequency in the original recordings (Fig. 3A, Fig. 4A).

In contrast to oscillations in Cplx3 mutant retinae, Cplx4 depletion (Fig. 4F,G) slightly but significantly increased the peak frequency in the FFT analysis (Fig. 4F). Peak power, however, remained unchanged (Fig. 4G). In the original recordings, the effect of increased OP frequency was barely visible. The unchanged OP amplitudes, however, clearly distinguished the Cplx4-knockout mice from the Cplx3-deletion mutants (compare Fig. 3D and 3A or Fig. 4A).

In the Cplx3/4 double mutants we found a significantly reduced OP peak power, which we had observed in the Cplx3 single-knockout mice before. In the case of the OP peak frequency the opposite effects in Cplx3 and Cplx4 single knockouts cancelled each other out in the double-knockout mice, which resulted in an unchanged peak frequency of the small, remaining OPs (Fig. 4H). These results again indicate an additive action of Cplx3 and Cplx4 deletion because an effect in one of the single knockout mice (Cplx3 or Cplx4) resulted in a qualitatively similar effect in the Cplx3/4 double-knockout mice.

Our results show that the loss of Cplx3 exerts effects in both plexiform layers since the increase in b-wave implicit times in Cplx3-knockout mice (Fig. 3C) must be mainly attributed to changes in transmission from photoreceptors to depolarizing (light on) bipolar cells, whereas the change in OPs most probably results from inner retinal processing with a major contribution from amacrine cells (Weymouth and Vingrys, 2008). This is in accordance with results from immunocytochemical studies, which showed that Cplx3 is predominantly present in photoreceptor terminals in the OPL and glycinergic amacrine cells and rod bipolar cell terminals in the inner plexiform layer (IPL) (Reim et al., 2005). By contrast, Cplx4 depletion had only minor effects in both synaptic layers, with the increase in OP frequency being attributable to changed transmission from cone bipolar cells (Reim et al., 2005).

Based on the ERG data described above it would be predicted that the mutant mice have vision deficits but are not completely blind. We therefore examined Cplx3/4 double-knockout and control mice in behavioural tests that assessed vision (Fig. 5; see Materials and Methods for details). In the cued platform test in the Morris water maze, Cplx3/4 double-knockout mice (n=11) required more time to reach the visible platform than wild-type controls (n=11; Fig. 5A). Repeated measures two-way ANOVA revealed a significant effect of the genotype (F(1,20)=7.40, P=0.0132), a significant effect of time (F(1,20)=14.54, P=0.0011), and no effect of interaction (F(1,20)=0.10, P=0.7609). By contrast, in the visual cliff test, Cplx3/4 double-knockout mice showed a similar performance as wild-type controls (Fig. 5B). In the rotarod test of motor function, Cplx3/4 mutants also performed normally (Fig. 5C). The fact that double mutant mice show longer escape latencies in the cued platform test, but a normal learning curve in this task and normal motor coordination, indicates that vision of Cplx3/4 double-knockout mice is perturbed. Apparently, this perturbation is rather mild because double mutant mice performed as well as wild-type littermates in the visual cliff test, a behavioural assay that mainly relies on contrast perception and is thus less demanding on visual acuity than the cued platform test.

In order to detect alterations in retinal structure due to Cplx3 or/and Cplx4 deletion mutations, we examined the retinal anatomy and neuronal morphology of adult wild-type and Cplx3 and Cplx4 single- and double-knockout mice. Vertical cryostat sections of retinae of wild-type and knockout mice were stained with antibodies that label various types of retinal neurons (Fig. 6), i.e. horizontal cells (anti-calbindin; Fig. 6B), amacrine cells (anti-calretinin; Fig. 6B), cone photoreceptors (anti-peanut agglutinin; Fig. 6C), and rod and cone bipolar cells (anti-CaB5; Fig. 6C). The overall structure of the retina of single- and double-knockout mice with respect to thickness and lamination was comparable to the wild-type retina (Fig. 6A), and the various retinal neurons displayed normal morphology with respect to their stratification patterns in the two synaptic layers of the retina, the OPL and IPL.

In addition to our western blot analyses we also tested whether deletion of Cplx3 or/and Cplx4 led to changes in localization of the remaining complexin isoforms (Fig. 7) and the SNARE components syntaxin 1, syntaxin 3, SNAP25 and synaptobrevin 2 (Fig. 8). Immunocytochemical analyses of vertical cryostat sections of retinae of knockout mice revealed that the synaptic distribution of the indicated proteins in all three types of homozygous mutants was comparable with wild-type controls.

Cplx3 and Cplx4 are the only complexins present at photoreceptor ribbon synapses (Reim et al., 2005). To analyze photoreceptor ribbon synaptic structure in adult Cplx3 and Cplx4 single- and double-knockout mice, we labelled three major presynaptic proteins of the photoreceptor ribbon complex, i.e. Bassoon, Ribeye and Piccolo (Schmitz et al., 2000; tom Dieck et al., 2005). We found no phenotypic changes in photoreceptor ribbon synapses of Cplx3 or Cplx4 single-knockout mice (not shown). In the Cplx3/4 double-knockout mice, however, the OPL appeared disorganized compared with that of the wild type. Moreover, many photoreceptor ribbon complexes were disintegrated as seen in high magnification views of single ribbon synaptic sites, which were triple labelled for Bassoon, Ribeye and Piccolo (Fig. 9A,B).

To further investigate this finding, we analyzed the photoreceptor ribbon synapses of adult Cplx3/4 double-knockout and wild-type mice qualitatively and quantitatively using electron microscopy. We prepared single ultrathin sections and randomly photographed a few hundred photoreceptor terminals from double-knockout and wild-type retinae. Representative examples of photoreceptor ribbon synapses of Cplx3/4 double-knockout mice are shown in Fig. 9C-E, and quantitative data are summarized in Fig. 9F. In the adult Cplx3/4 double-knockout retina, 40±13% (±s.d.; n=343) of photoreceptor terminals contained normal-looking, presynaptically anchored ribbon synapses, as compared with 59±8% (n=290) in the wild-type retina. An example is shown in Fig. 9C. The photoreceptor ribbon is anchored at the presynaptic site, and the postsynaptic elements, horizontal cell processes and rod bipolar cell dendrites, invaginate into the photoreceptor terminal and form a triadic configuration adjacent to the ribbon site (Fig. 9C). Occasionally, we found presynaptically anchored ribbons that were club-shaped rather than ribbon-shaped (Fig. 9D), but most strikingly, in the adult Cplx3/4 double-knockout retina we found a number of photoreceptor terminals (24±6%, n=343) that contained spherical free-floating ribbons (Fig. 9E,F). They may represent breakdown products of anchored ribbons, because in the wild-type retina hardly any spherical ribbons occurred in the photoreceptor terminals (1±0%, n=290; Fig. 9F). The number of photoreceptor terminals without presynaptic ribbons was comparable between the Cplx3/4 double-knockout retina (36±9%, n=343) and the wild-type control (40±8%, n=290).

We showed previously that Cplx3 is present in brain and retina, whereas Cplx4 is solely expressed in retina (Reim et al., 2005). Because Cplx3 and Cplx4 are localized at ribbon synapses of photoreceptors and bipolar cells and because they possess structural features that set them apart from Cplx1 and Cplx2, we postulated that these two members of the complexin family might contribute to the unique release efficacy of retinal sensory neurons (Reim et al., 2005). The present study demonstrates, that Cplx3 and Cplx4 are indeed necessary for efficient transmitter release and, ultimately, for the maintenance of synaptic structure in the retina. Their loss causes vision deficits resulting from changes in signal transmission in both synaptic layers of the retina and the breakdown of photoreceptor ribbon synaptic sites.

ERG recordings revealed striking differences in photoreceptor synaptic transmission between Cplx3/4 double-knockout mice and wild-type controls. The b-wave of the ERG, a measure of synaptic transmission in the light-on channel of the outer retina, was significantly reduced in its amplitude and altered in its timing over the whole stimulation intensity range (Fig. 3H,I). This indicates that the continuous adjustment and fine-tuning of transmitter release at the photoreceptor ribbon synapse, which is necessary to faithfully reflect the changes in membrane potential to changing light intensities, is defective. Based on these findings, we conclude that Cplx3 and Cplx4 are important positive regulators of transmitter release at photoreceptor ribbon synapses. Interestingly, a strong reduction of b-wave amplitude is only seen if Cplx3 and Cplx4 are simultaneously deleted, which indicates a cooperative mode of action of Cplx3 and Cplx4. In contrast to changes in the b-wave amplitude, an increase in b-wave implicit time was observed also in Cplx3 single knockouts, mainly at the intensity range where cones are activated, but not in Cplx4-knockout mice (Fig. 3C,F). In view of our previous immunocytochemical observations, according to which Cplx4 is located mainly in rod terminals (Reim et al., 2005), this finding was very surprising as it shows that Cplx4 deficiency alone is not sufficient to alter transmission in rod photoreceptor terminals. It is possible that this lack of an effect in Cplx4 knockouts is due to the weak Cplx3 expression in rod photoreceptor terminals (Reim et al., 2005), which may suffice to compensate the Cplx4 deficiency. It remains to be determined, however, why the b-wave implicit time is even more prolonged, especially at low light intensities, if Cplx4 is deleted in addition to Cplx3.

An altered timing of signal transmission from photoreceptors to bipolar cells in the OPL may also lead to a modified temporal structure of synaptic processing in the IPL, thus causing alterations in the OPs. Therefore, it is difficult to distinguish in the ERG recordings between indirect effects of altered ON-bipolar cell timing on OPs and direct effects generated by altered transmission from bipolar cells to amacrine cells or between amacrine cells. Cplx3-knockout mice show a clear reduction in peak frequency and peak power (Fig. 4D,E), whereas OPs of Cplx4-knockout mice are characterized by an increased peak frequency and an unchanged peak power (Fig. 4F,G). Both results would be compatible, however, with the fact that Cplx3 is contained in rod bipolar cell terminals and glycinergic amacrine cells, thus supporting the notion of a direct effect of Cplx3 depletion in the IPL via changed transmitter release from rod bipolar cells and/or changed transmitter release from amacrine cells. In addition, the small but significant increase in peak frequency in the Cplx4-knockout mice most probably resulted from alterations in transmission at the cone bipolar cell terminals since the b-wave implicit times were completely unchanged and Cplx4 is found in cone bipolar terminals (Reim et al., 2005). In contrast, the peak frequency of Cplx3/4 double-knockout mice is unchanged and the peak power is significantly reduced (Fig. 4H,I). This can be explained by a pure addition of the single-knockout effects in the Cplx3 and Cplx4 mutants. Although these explanations remain somewhat speculative at the moment, it is obvious that the OPs are differentially altered in the tested mouse lines, and the results demonstrate that the synchronization of signal transfer in the IPL is affected by the loss of Cplx3 and/or Cplx4.

Our finding, that the loss of Cplx3 and Cplx4 leads to compromised synaptic transmission of ribbon synapses, corroborates the functional phenotype of Cplx1/2/3 triple-knockout mice, which we analyzed with regard to excitatory and inhibitory synaptic transmission in hippocampal and striatal autaptic cultures (Xue et al., 2008). Synaptic transmission in these triple knockout neurons, which are completely Cplx-deficient, is characterized by severely reduced evoked transmitter release, normal or slightly reduced spontaneous release, and a strong reduction in the probability of fast Ca²⁺-triggered release, which is due to a decrease in the apparent Ca²⁺ sensitivity of the release process (Xue et al., 2008). Taken together, our previous and present data indicate that complexins are positive regulators of synaptic vesicle fusion at all tested mammalian synapses. This notion, which is further supported by the fact that overexpression of Cplx2 in wild-type chromaffin cells increases chromaffin granule secretion (Cai et al., 2008), is in contrast to the phenotype of complexin-deficient Drosophila neuromuscular junctions, in which evoked synaptic transmission is strongly reduced but spontaneous release is massively increased, indicating that complexin has a facilitatory as well as a partially inhibitory, clamp-like function in transmitter release (Huntwork and Littleton, 2007). The reason for this discrepancy is currently unknown, but it is possible that complexins generally have both, intrinsic facilitatory and clamp-like functions, which emerge with different penetrance depending on the organism or experimental protocol used (Brose, 2008; Xue et al., 2008).

In principle, Cplx3 and Cplx4 are able to act in a similar fashion to Cplx1 and Cplx2. Their SNARE-binding domains are evolutionarily highly conserved, they all bind SNARE complexes, and they all rescue the transmitter release deficit of Cplx1/2 double-knockout neurons upon overexpression (Reim et al., 2005). However, Cplx3 and Cplx4 have structural features that could be of significant functional relevance in the highly specialized ribbon synapse. Both contain farnesylation sites at their C-termini that regulate their synaptic targeting and membrane association (Reim et al., 2005), and it is possible that farnesylation-mediated membrane association of Cplx3 and Cplx4 increases the concentration of the two proteins at defined sites of vesicle fusion. Interestingly, vesicles that are tethered to the retinal ribbons appear to be already primed for rapid, Ca²⁺-dependent fusion, regardless of their proximity to the plasma membrane. Based on the assumption that these ribbon-tethered vesicles represent a major fraction of the readily releasable vesicle pool for fast neurotransmitter release (Heidelberger, 1998; Heidelberger et al., 2002; Heidelberger et al., 2005; LoGiudice et al., 2008), it was suggested that synaptic vesicles at ribbon synapses fuse in a compound fashion (Parsons and Sterling, 2003). In this case, Cplx3 and Cplx4 would be expected to be located in close vicinity to ribbons and/or tethered vesicles. Unfortunately, this prediction could not be tested here because the available antibodies that we used for light microscopic immunocytochemistry do not function convincingly in immunoelectron microscopy in our hands.

The overall structure of the Cplx3/4 double-knockout retina is comparable to the wild-type situation, and the various retinal neurons display normal morphology in both synaptic layers. At the ultrastructural level, however, Cplx3/4 double-knockout retinae contain photoreceptor terminals with spherically shaped free-floating ribbons, which are rarely observed in wild-type retina. Such free-floating ribbons are usually found as ribbon precursors in the developing retina (Regus-Leidig et al., 2009), and in much larger numbers than in the Cplx3/4 double-knockout retinae described here, in mice with a mutation in the gene encoding Bassoon, which is thought to be involved in anchoring ribbons at release sites (Dick et al., 2003). Given that Complexins are SNARE regulatory proteins and not involved in active zone formation (Reim et al., 2001), it is unlikely that Cplx3 and Cplx4 are necessary for regulation of proper ribbon biosynthesis, development or anchoring. Accordingly, we did not observe such spherical free-floating ribbons in the retinae of young Cplx3/4 double-knockout mice (age 4 weeks; data not shown). Rather, the ribbon breakdown products seen in Cplx3/4 double-knockout retinae may be a secondary consequence of the altered synaptic activity in the Cplx3/4 double-knockout retina - a notion that is corroborated by the significant decrease of Ribeye in the Cplx3/4 double-knockout mice.

In summary, the present study demonstrates that Cplx3 and Cplx4 are important regulators of transmitter release at retinal ribbon synapses. Their loss leads to aberrant adjustment and fine-tuning of transmitter release at the photoreceptor ribbon synapse, alterations in transmission at bipolar cell terminals, changes in the temporal structure of synaptic processing in the IPL of the retina, and perturbed vision. Our ERG data do not provide us with the necessary information to determine what exactly Cplx3 and Cplx4 contribute to the unique release properties of retinal ribbon synapses, and in which step of the release process they are involved. In an ideal scenario, corresponding experiments would be performed that provided patch-clamp measurements of membrane capacitance changes at individual ribbon synapse terminals during transmitter release. Such patch-clamp experiments are routinely performed with goldfish and frog bipolar cell terminals, but they are extremely difficult with terminals of mouse retinal neurons such as rod bipolar cells, because of two main experimental limitations. First, mouse bipolar cells have rather long and thin axons compared with goldfish and frog preparations, which confounds measurements of axon terminal capacitance changes at the level of cell bodies because of space-clamp problems, at least in our pilot experiments (Wolf Jockusch and K.R., unpublished observations) (see also Zhou et al., 2006). Second, axon terminals of mouse bipolar cells are very small and more ramified than those of other preparations, which so far has precluded direct patch-clamp measurements from terminals. Essentially, a significant improvement of the currently available technology is required in order to determine the exact role of complexins at retinal ribbon synapses.

Cplx3-knockout mice were described previously (Xue et al., 2008). Cplx4-knockout mice were generated by homologous recombination in mouse embryonic stem cells following established procedures (Thomas and Capecchi, 1987). To generate the targeting vector for deletion of Cplx4, a genomic mouse Cplx4 clone (pBlueMG-CPXIVλ20.2) containing two exons was used. In the targeting vector, the first coding exon of the corresponding gene (representing bp 6-188 of Cplx4 cDNA, GenBank accession number AY264291) was replaced by a neomycin resistance cassette (Fig. 1A). Following electroporation and selection, recombinant stem cell clones were analyzed by Southern blotting after digestion of DNA with BamHI. For hybridization, a probe located 5′ of the targeting vector short arm was used. Two positive clones were injected into mouse blastocyts to obtain highly chimeric mice that transmitted the mutation through the germline. Chimeric males were bred with C57Bl6 females. Germline transmission of the mutation was tested by Southern blot analysis (genomic DNA) and immunoblotting (retina extracts). Subsequently, genotyping was performed by PCR. For analyses of Cplx3 and Cplx4 single-deletion mutants, wild-type and homozygous mutant littermates were obtained by interbreeding of heterozygous parents, and in all experiments homozygous mutants were compared with appropriate wild-type control littermates. To establish a Cplx3/4 double-knockout line Cplx3 and Cplx4 single-knockout mice were crossed. For routine experimentation, homozygous Cplx3/4 double-knockout mice from this line were compared with wild-type mice derived from heterozygous Cplx3 knockouts. All animal experiments were performed in compliance with the guidelines for the welfare of experimental animals issued by the State Governments of Lower Saxony and Bavaria (comparable to NIH guidelines).

To confirm the loss of Cplx3 and Cplx4 in corresponding single and double-knockout mice, retina and brain extracts were analyzed by SDS-PAGE and western blotting using polyclonal rabbit antibodies raised against Cplx3 and Cplx4 [1:1000 (Reim et al., 2005)]. Syntaxin 3 was used as a loading control for the analysis of Cplx3/4 double-knockout retinae. It was detected with a polyclonal sheep antibody [1:1000 (Morgans et al., 1996)]. For analysis of synaptic proteins, the following antibodies from Synaptic Systems (Göttingen, Germany) were used: monoclonal mouse antibodies to SNAP25 (1:10⁶), syntaxin 1 (1:20,000), synaptobrevin 2 (1:7500), Synaptophysin (1:10,000), and polyclonal rabbit antibodies to Cplx1/2 (1:2500), Cplx3 (1:1000), Cplx4 (1:1000), syntaxin 2 (1:500), syntaxin 3 (1:1000), and synaptotagmin 1 (1:2500). Moreover, we used a polyclonal guinea pig antibody to VGLUT1 (1:5000; Millipore, Eschborn, Germany), a polyclonal rabbit antibody to VIAAT (1:500; Millipore), a polyclonal sheep antibody to syntaxin 3 (1:1000) (Morgans et al., 1996), a monoclonal mouse antibody to RIM (1:500; BD Transduction Laboratories, San Jose, CA), and polyclonal rabbit antibodies to Munc13-1, Munc13-2, and Munc13-3 (Varoqueaux et al., 2005). For quantitative analysis of Ribeye and Bassoon, proteins were separated on 3-8% Tris-acetate gels (Invitrogen, Karlsruhe, Germany) and blotted according to the manufacturer's instructions. For detection we used a monoclonal mouse antibody against CtBP2/RIBEYE (1:4000, BD Transduction Laboratories, San Jose, CA) as well as a polyclonal antibody against Bassoon (1:100, Synaptic Systems). By using fluorescently labelled secondary antibodies signal intensities were estimated on an Odyssey Infrared Imaging System (LI-COR Biosciences, Bad Homburg, Germany). Expression levels were normalized to Actin (1:4000, Sigma, Hamburg, Germany) or CtBP2, which served as loading controls. All western blots were done twice and statistical significance was tested using Student's t-test.

Scotopic ERG responses of Cplx3-, Cplx4- and Cplx3/4-knockout mice at the age of 10-13 weeks were compared to the indicated controls. Animals were dark-adapted overnight before recordings began. Mice were anesthetized by intraperitoneal injections of xylazine (50 mg/kg) and ketamine (20 mg/kg), and the pupils were dilated with 1% atropine sulfate. Surgery and subsequent handling were done under dim red darkroom light. A gold coil electrode was placed on the corneal surface, which was moistened with a thin layer of Methocel, and a platinum needle reference electrode was inserted subcutaneously into the skin covering the skull. Another platinum needle grounding electrode was inserted into the tail. Mice were placed onto a temperature-regulated platform (Roland Consult, Brandenburg, Germany) with the head fixed. Electrical potentials were recorded and bandpass filtered (1-1000 Hz) using a ML132 Bioamplifier (AD Instruments, Hastings, UK) connected to a PowerLab system (AD Instruments) for digitizing and storage. A calibrated Ganzfeld ERG setup was used for light stimulation (Q450; Roland Consult). Ten responses were averaged at each intensity. A-wave response amplitudes were measured relative to baseline, which was determined by the mean voltage within a 30 msecond period before the light flash. The b-wave amplitude was determined from the most negative a-wave trough to the b-wave peak (Fig. 4A). For analysis of oscillatory potentials (OPs) ERG traces were first bandpass filtered (30-300 Hz; Fig. 4B). From this filtered response, a time slice from the maximum of the a-wave to 200 mseconds (grey area in Fig. 4B) was chosen for fast Fourier transformation (FFT). Settings for the power spectrum were: 1024 for FFT size; Welch with 50% overlap. In the FFT power spectrum, peak power and frequency where peak power occurred were measured and plotted against light intensity (Fig. 4C). Analysis was performed with CHART v5 (AD Instruments). Statistical analysis was done with JMP 5.0 (SPSS), and results were plotted using Deltagraph v5 (SPSS, Chicago, IL).

Visual acuity was assessed in the cued platform test. For this purpose, Cplx3/4 double-knockout and wild-type mice were trained to find a submerged escape platform by means of a prominent visual cue. The test was performed in a large circular tank (diameter 1.2 m, depth 0.4 m). The tank was filled with opaque water (25±1°C, depth 0.3 m) and the escape platform (diameter 10 cm) was submerged 1 cm below the surface. The swimming patterns were monitored by a computer and the video-tracking system Viewer 2 (Biobserve, Bonn, Germany). Escape latency, swim speed, path length and trajectory of swimming were recorded for each mouse. Mice were trained for 2 days to swim to a clearly visible platform that was marked with a 15 cm high black flag and placed pseudo-randomly in different locations across trials. Every day, mice went through four trials with an inter-trial interval of 5 minutes. The mice were placed into the pool facing the side wall randomly at one of four start locations and allowed to swim until they found the platform, or for a maximum of 90 seconds. Any mouse that failed to find the platform within 90 seconds was guided to the platform. The animal then remained on the platform for 20 seconds before being removed from the pool.

The visual cliff test was used to assess more profound visual deficiencies, e.g. in contrast perception (Crawley, 1999). The test apparatus consisted of an open-topped box (70×35 cm length × width, 30 cm high). The walls of the box were made from white Perspex, and the base was made from clear Perspex. The box was positioned on the edge of a laboratory bench so that half of the base was placed on the bench (`ground' side), and the other half over the edge of the bench, suspended 1 m above the floor (`air' side). Mice were placed in the middle of the base at the edge of the cliff, and their activity was monitored for 5 minutes by a computer and the video-tracking system Viewer 2 (Biobserve). The percentage of time each mouse spent on the `ground' and the `air' side of the box was calculated.

Motor function, balance and coordination was assessed in the rotarod test (Crawley, 1999). This test involves a rotating drum (Ugo Basile, Comerio, Varese, Italy), which was accelerated from 4 to 40 r.p.m. in the course of 5 minutes. Mice were placed individually on the drum and the latency of falling off the drum was recorded using a stop-watch.

Preparation of retinal tissue and antibody incubation for light microscopic immunocytochemistry were done as described previously (Dick et al., 2003; tom Dieck et al., 2005). Briefly, the eyes were opened and retinae were immersion fixed in the eyecup for 30 minutes in 4% paraformaldehyde (PFA) in phosphate buffer (PB; 0.1 M, pH 7.4). The retinae were frozen in freezing medium (Reichert-Jung, Bensheim, Germany) as sandwiches of wild-type and Cplx3-knockout material, and of Cplx4-knockout and Cplx3/4 double-knockout material. Vertical sections (16 μm thick) of both sandwiches were cut and collected on the same slides to ensure equal immunocytochemical treatment. Primary antibody incubation was carried out overnight at room temperature, secondary antibody incubation for 1 hour. For analysis, labelled sections were examined with a Zeiss Axio Imager Z1 equipped with an ApoTome (Zeiss, Oberkochen, Germany). Images were taken with a ×20 (0.8, Apochromat) or a ×100 (1.3 oil immersion, Plan-Neofluar) objective as stacks of multiple optical sections, and projections were calculated with AxioVision 4.6.3 software (Zeiss). The images were adjusted for contrast and brightness using Adobe Photoshop CS (Adobe, San Jose, CA). To compare Cplx1, Cplx2, Cplx3, Cplx4 and SNARE protein expression in wild-type, Cplx3 and Cplx4 single and Cplx3/4 double-knockout retinae, images were taken with the same exposure time. The following antibodies and reagents were used for immunocytochemistry: monoclonal mouse anti-CtBP2/RIBEYE (1:10,000; BD Biosciences, Heidelberg, Germany), mouse anti-Calbindin (1:1000; Swant, Bellinzona, Switzerland), mouse anti-SNAP25 (1:10,000), mouse anti-synaptobrevin-2 (1:10,000), mouse anti-syntaxin-1 (1:5000), mouse anti-Cplx3 (1:1000), mouse anti-Cplx4 (1:50,000, all from Synaptic Systems), polyclonal rabbit anti-Bassoon sap7f (1:16,000) (tom Dieck et al., 1998), rabbit anti-CaB5 [1:1000; a kind gift from Francoise Haeseleer, University of Washington, Seattle, WA (Haeseleer et al., 2000)], rabbit anti-Calretinin (1:1000; Swant), rabbit anti-Cplx1/2 (1:10,000), rabbit anti-syntaxin-3 (1:2500, both from Synaptic Systems), guinea pig anti-Piccolo-44a [1:800 (Dick et al., 2001)], and FITC-conjugated peanut agglutinin (1:500; Vector Laboratories, Burlingame, CA).

For conventional electron microscopy and good tissue preservation, the fixation was in 4% PFA and 2.5% glutaraldehyde for 2 hours at room temperature, followed by an incubation in 2% osmium tetroxide for 1.5 hours. For analysis, ultrathin sections were examined and photographed with a Zeiss EM10 electron microscope (Zeiss) and a Gatan BioScan digital camera (Gatan, Munich, Germany) in combination with the Digital Micrograph 3.1 software (Gatan, Pleasanton, CA). Images were adjusted for contrast and brightness using Adobe Photoshop CS. To quantify the morphological phenotype of photoreceptor ribbon synapses in the Cplx3/4 double-knockout retina, we examined three wild-type and three homozygous Cplx3/4 double-knockout mice of the same age (1 year). For each animal, we photographed 25 random images of the outer plexiform layer (OPL) at a magnification of ×16,000 using single ultrathin sections (a single image contained about four photoreceptor terminals). The number of the examined terminals was counted, and the terminals were subdivided into three categories: (1) terminals with anchored ribbons, (2) terminals with spherical ribbons, and (3) terminals without ribbons.