The human cognition-enhancing CORD7 mutation increases active zone number and synaptic release

Abstract

Humans carrying the CORD7 (cone-rod dystrophy 7) mutation possess increased verbal IQ and working memory. This autosomal dominant syndrome is caused by the single-amino acid R844H exchange (human numbering) located in the 3₁₀ helix of the C₂A domain of RIMS1/RIM1 (Rab3-interacting molecule 1). RIM is an evolutionarily conserved multi-domain protein and essential component of presynaptic active zones, which is centrally involved in fast, Ca²⁺-triggered neurotransmitter release. How the CORD7 mutation affects synaptic function has remained unclear thus far.

Here, we established Drosophila melanogaster as a disease model for clarifying the effects of the CORD7 mutation on RIM function and synaptic vesicle release. To this end, using protein expression and X-ray crystallography, we solved the molecular structure of the Drosophila C₂A domain at 1.92 Å resolution and by comparison to its mammalian homologue ascertained that the location of the CORD7 mutation is structurally conserved in fly RIM. Further, CRISPR/Cas9-assisted genomic engineering was employed for the generation of rim alleles encoding the R915H CORD7 exchange or R915E, R916E substitutions (fly numbering) to effect local charge reversal at the 3₁₀ helix. Through electrophysiological characterization by two-electrode voltage clamp and focal recordings we determined that the CORD7 mutation exerts a semi-dominant rather than a dominant effect on synaptic transmission resulting in faster, more efficient synaptic release and increased size of the readily releasable pool but decreased sensitivity for the fast calcium chelator BAPTA. In addition, the rim CORD7 allele increased the number of presynaptic active zones but left their nanoscopic organization unperturbed as revealed by super-resolution microscopy of the presynaptic scaffold protein Bruchpilot/ELKS/CAST.

We conclude that the CORD7 mutation leads to tighter release coupling, an increased readily releasable pool size and more release sites thereby promoting more efficient synaptic transmitter release. These results strongly suggest that similar mechanisms may underlie the CORD7 disease phenotype in patients and that enhanced synaptic transmission may contribute to their increased cognitive abilities.

Introduction

Presynaptic active zones (AZs) constitute fundamental building blocks of neuronal connections, and contribute to neuronal computation and memory formation.^1–4 They harbour a well-defined set of proteins, which mediate highly regulated neurotransmitter release occurring with sub-millisecond accuracy.² Defects in the orchestrated function of presynaptic proteins typically have devastating consequences. Correspondingly, several neurological diseases have been associated with malfunction of presynaptic proteins.⁵ A striking exception of the observation that a protein defect leads unambiguously to functional defects is the CORD7 syndrome. Affected individuals suffer from an autosomal dominant cone-rod dystrophy, ultimately leading to blindness.⁶ However, they also show largely enhanced cognitive abilities characterized by increased verbal IQ and working memory.⁷

The CORD7 syndrome is caused by a single-amino acid mutation within the 3₁₀-helix of the C₂A domain of RIM (Rab3-interacting molecule).^2,8 RIM is a multi-domain AZ protein,^9,10 encoded by the RIMS1 locus (human gene name as opposed to the Drosophila locus name rim), whose interaction with other core AZ proteins such as RIM binding protein (RBP), Munc13, α-Liprin, and ELKS/CAST, as well as with Ca²⁺ channels² is thought to figure centrally in neurotransmitter release by providing a structural lattice that organizes release sites. Co-expression with voltage-gated calcium channels in HEK293 cells revealed that the CORD7 exchange in RIM1 increased the activation and suppressed the inactivation component of the channel currents.¹¹ However, the impact of the CORD7 mutation on synaptic function has remained largely unclear.

At presynaptic release sites, distance between Ca²⁺ channels, vesicular release sensors and the size of the readily releasable pool (RRP) of vesicles are key determinants of the fidelity of neurotransmitter release.^12,13 Tight nanodomain coupling (within tens of nanometres)^12,14,15 promotes fast and efficient release. It can be distinguished from loose coupling,^16–19 but how these differences are molecularly implemented and controlled is still subject of intense investigations.^20–22 Given the central role of RIM at AZs²³ we here studied how RIM, its C₂A domain, and the CORD7 mutation affect synaptic release at the neuromuscular junction (NMJ) of Drosophila.

Materials and methods

Fly stocks

Flies and larvae were raised at 25°C on standard cornmeal and molasses medium.

Fly stocks generated in this study

Internal stock IDs given in brackets after the genotype:

• w¹¹¹⁸; +; rim^ΔC2A attP DsRed⁺/TM3, Sb; (RIM98)

• w¹¹¹⁸; +; rim^ΔC2A attP DsRed⁻/TM3, Sb; (LAT473)

• w¹¹¹⁸; +; rim^ΔC2A attP{C₂A-WT(pLM5) mW⁻}/TM6b, Tb; (LAT542/AM28)

• w¹¹¹⁸; +; rim^ΔC2A attP{C₂A-R915H(pLM8) mW⁻}/TM6b, Tb; (LAT545/AM29)

• w¹¹¹⁸; +; rim^ΔC2A attP{C₂A-R915E,R916E(pLM9) mW⁻}/TM6b, Tb; (LAT549/AM32)

• w¹¹¹⁸; P{UAS-RIM^wt w⁺}^attP40/CyoGFP; +; (RIM52)

• w¹¹¹⁸; P{UAS-RIM^R>Hw⁺}^attP40/CyoGFP; +; (RIM53)

Other fly stocks

Reference or stock ID given after the genotype:

• w¹¹¹⁸; +; rim^Ex73; (rim^KO allele)²⁴

• w*; ok6-GAL4 mW⁺; +;^25,26

• y¹ M{GFP[E.3xP3] = vas-Cas9.RFP-}^ZH-2A w¹¹¹⁸;;; (BDSC#55821)

• M{vas-int.Dm}^ZH-2A;;;²⁷

Transgene construction

Transgenes for genomic targeting

All primer sequences used in this study are provided in Table 1.

pU6-gRNA

CRISPR/Cas9 cutting sites 5′ and 3′ of the genomic region of the rim/CG33547 gene encoding the C₂A domain were identified by CRISPR Optimal Target Finder.²⁸ The genomic sequence of all CRISPR/Cas9 cleavage sites were confirmed by DNA sequencing of PCR fragments encompassing the suggested sites prior to cloning. Target-specific sequences for gRNAs (gRNA#1: ATAGCGCATGGAATAAC/CTGTGG; gRNA#2: AGTCTAATGTATACACG/TGTGGG; sequences are in 5′–3′ order; the PAM site is underlined, a forward slash indicates the Cas9 cutting site) were synthesized as 5′-phosphorylated oligonucleotides (gRNA#1: mh_46F/mh_26R; gRNA#2: mh_47F/mh_28R), annealed to produce matching overhangs, and ligated into the BbsI sites of the pU6-BbsI-chiRNA vector pTL621²⁹ generating vectors pMH4 (gRNA#1) and pMH5 (gRNA#2).

rim^ΔC2A HDR plasmid

A 1.1 kb fragment encoding the 3′ homology arm was amplified from w¹¹¹⁸ genomic DNA using primers mh_65F/mh_66R, cut with SapI and inserted into de-phosphorylated SapI-digested pHD-DsRed-attP (pTL620) generating plasmid pMH15. pMH15 was digested with AarI and genomic 0.2 kb PCR product amplified with primers mh_77F/mh_79R, also digested with AarI, were ligated generating plasmid pMH25. Finally, pMH25 was digested with NheI/SpeI and ligated with a 1.1 kb NheI/SpeI-PCR product amplified off genomic DNA with primers mh_82F/mh_83R producing the final HDR vector pMH14.

attB-rim-C₂A^rescue plasmid

A 2.5 kb fragment of genomic DNA from strain w¹¹¹⁸ was amplified using primers lm_30F/lm_31R, digested and ligated to a 6.1 kb Notl/Ascl-fragment of pGE-attB-GMR (pTL370) generating rim-C₂A^rescue vector pLM5.

attB-rim-C₂A^R915H plasmid

A 8.5 kb Hpal/Xhol fragment was released from pLM5, and a 0.2 kb Hpal/Xhol fragment from pTL823 (insert synthesized by Genscript). Both fragments were ligated generating rim-C₂A^R915H vector pLM8.

attB-rim-C₂A^R915E,R916E plasmid

A 8.5 kb Hpal/Xhol fragment was released from pLM5, and a 0.2 kb Hpal/Xhol fragment from pTL825 (insert synthesized by Genscript). Both fragments were ligated generating rim-C₂A^R915E,R916E vector pLM9.

UAS transgenes

Plasmids for Drosophila transformation were engineered through the Drosophila Gateway system as follows. The Drosophila wild type full-length rim cDNA (GenBank ID: KF534710.1)²⁴ through a NotI/XbaI digest and inserted into the pENTR1Adual plasmid generating entry vector pJG42 (RIM^wt). Additional entry vectors for single and double mutated RIM variants were produced by exchanging a DNA fragment of pJG42 encoding the CORD7 site region for matching rim fragments containing the mutations. These were produced by subcloning the wild-type 0.5 kb HindIII/NcoI-fragment of pJG42 into pMCS5 (yielding pTL305) to reduce the plasmid size and allow for an outward PCR of the entire vector. The rim^R>H mutation was introduced via the reverse outward-PCR primer. The PCR-amplicon was cut at an EagI site and self-ligated to recircularize the plasmid. From the resulting clone pTL310 (RIM^R>H) the 0.5 kb HindIII/NcoI fragment was used to replace the wild-type region in pJG42 and generate entry vector pTL431. All entry vector inserts were fully sequenced to ensure no errors were introduced through the cloning procedures. To create the final transformation vectors, pJG42 and pTL431 were recombined with the pTW-attB destination vector²⁷ through an LR reaction according to the manufacturer’s instructions (Invitrogen), and correct recombinant clones were identified by suitable restriction digests (5xUAS-RIM^wt = pTL434; 5xUAS-RIM^R>H = pTL435). All UAS-RIM transgenes were inserted into the attP40 landing site carried on the second chromosome³⁰ using ΦC31 integrase-based transformation³¹ by BestGene Inc.

Transgenes for protein expression

A PCR-amplified DNA fragment encoding the RIM C₂A domain of Drosophila (encoding residues 836–962 of RIM) was subcloned via primer-delivered BamHI/NotI sites in a pGEX-6P-2 vector (GE Healthcare) providing a N-terminal GST-tag (pTL423).

CRISPR targeting

All transgenesis steps were performed at Bestgene Inc. For CRISPR/Cas9 mediated engineering gRNA#1 (pMH4) and gRNA#2 (pMH5), and the matching rim^ΔC2A-HDR plasmid (pMH14) were injected into w¹¹¹⁸ flies carrying a germline-expressing vas-Cas9 source as described before²⁸ producing the rim^ΔC2A, DsRed⁺ allele. Correct gene targeting was confirmed by subsequent sequencing of PCR fragments covering breakpoints between genomic/transgenic DNA amplified off genomic DNA of respective adult transgenic flies using primers tl_843F/tl_687R (covering the 5′ breakpoint), tl_679F/tl_840R (covering the 3′ breakpoint), and across the deleted fragment using primers tl_849F/tl_850R. The 3xP3-DsRed transformation marker was removed from rim^ΔC2A, DsRed⁺ flies by expressing a germline Cre source, and confirmed by PCR genotyping. Subsequent insertion of the different attB-rim-C₂A transgenes into rim^ΔC2A, DsRed⁻ flies (LAT473) by ΦC31-mediated transgenesis was performed by Bestgene Inc.

Protein production

Recombinant protein expression was performed in Escherichia coli strain BL21(DE3) in Luria-Bertani medium. Protein expression was induced at 18°C by addition of 0.3 mM Isopropyl β-D-1-thiogalactopyranoside and allowed to proceed for 16 h. Cell pellets were resuspended in 50 mM Tris-HCl pH 8.0, 250 mM NaCl, 5% v/v glycerol and 3 mM β-mercaptoethanol (β-ME) buffer, and lysed by lysozyme treatment and sonication. Lysates were incubated with Glutathione-sepharose beads (GE Healthcare), and the beads were washed with Buffer A (50 mM Tris-HCl pH 8.0, 250 mM NaCl, 3 mM β-ME). Proteins were eluted from the beads with Buffer A supplemented with 12 mM reduced glutathione, followed by dialysis in Buffer A and removal of the GST-tag by 3C-protease cleavage. Uncleaved proteins and GST were removed by reverse affinity purification with Glutathione-sepharose beads, followed by size exclusion chromatography into a 30 mM sodium citrate pH 6.0, 150 mM NaCl and 1 mM dithiothreitol buffer.

Crystallization and structure determination

Crystals of the Drosophila melanogaster RIM C₂A domain were obtained using the sitting drop vapor diffusion technique at 20°C. A Mosquito robot (TTP LabTech) was used to set up 200 nl-size drops with 1:1 and 1.3:0.7 ratios of protein to mother liquor. Rectangular crystals developed in 2 days when mixing the protein at 4.8 mg/ml concentration with a 0.2 M tri-lithium citrate, 20% w/v polyethylene glycol 3350 solution. Crystals and were cryo-protected with 20% v/v glycerol, flash-cooled in liquid nitrogen and diffracted to 1.93 Å at the Diamond Light Source beamline I04. The space group was determined as P212121 with one molecule per asymmetric unit. Crystallographic data were integrated by XDS and scaled in XSCALE.³² The structure was solved by molecular replacement in the programme MrBump³³ using the rat RIM2 C₂A domain (PDB 2BWQ) as search model. Iterative model building with COOT³⁴ and refinement in PHENIX.refine³⁵ using TLS restraints yielded the final model. All figures were prepared in PyMol (PyMOL Molecular Graphics System, Schrödinger LLC). The electrostatics surface potential was calculated with APBS.³⁶ The structure was deposited under PDB ID 4TS6.

Two-electrode voltage clamp recordings

Two-electrode voltage clamp (TEVC) recordings (Axo Clamp 2B amplifier, Axon instruments) were performed in extracellular haemolymph-like solution (HL-3)³⁷ that contained (in mM): NaCl 70, KCl 5, MgCl₂ 20, NaHCO₃ 10, trehalose 5, sucrose 115, HEPES 5, CaCl₂ 1.0, pH adjusted to 7.2. Recordings were obtained from NMJs at longitudinal abdominal muscles 6/7 in segments A2 and A3 at 20–22 ± 1°C room temperature using intracellular electrodes with resistances 10–20 MΩ (filled with 3 M KCl). Holding voltage was −60 mV for evoked excitatory postsynaptic currents (eEPSCs) and −80 mV for miniature EPSCs (mEPSCs). Only muscle cells 6 with an initial membrane potential of ≥−50 mV and ≥4 MΩ input resistance were analysed. Synaptic responses were generated by pulses of 0.3 ms length and 5–15 V amplitude, applied via a suction electrode (filled with extracellular solution) and filtered by an 10 kHz low-pass filter. We applied a paired-pulse protocol with 0.2 Hz frequency and 30 ms interpulse intervals, averaged 10 EPSCs per muscle cell if not noted oherwise and analysed the data with Clampfit (Molecular Devices). To estimate RRP size, a train of 50 pulses was applied at 20 Hz. Linear fits to the last 20 pulses were applied to cumulatively plotted eEPSCs and back extrapolated.^38,39

Focal recordings

EPSC recordings were performed in extracellular haemolymph-like solution (HL-3)³⁷ that contained (in mM): NaCl 70, KCl 5, MgCl₂ 20, NaHCO₃ 10, trehalose 5, sucrose 115, HEPES 5, pH adjusted to 7.2 with 1 N NaOH. CaCl₂ concentration was 1.0 mM. We only recorded from NMJs at abdominal muscles 6/7 in segments A2 and A3. Room temperature was constantly 20 ± 2°C and temperature of the bath solution was maintained at 20 ± 1°C with the help of a constant displacement pump. The focal electrode (resistances 450–550 kΩ, filled with HL-3) was positioned under optical control on proximal boutons (i.e. nearby the access site of the nerve to the muscle) under optical control. Synaptic currents were low-pass filtered with 20 kHz and stored with Patchmaster software using an EPC10 double patch clamp amplifier (HEKA electronics). EPSCs were evoked by stimulating the respective motor neuron via a suction electrode. We applied a paired pulse protocol at 0.2 Hz with 30 ms interpulse intervals, each pulse consisting of 1 ms length and 7 V amplitude. About 60 traces per recording site were averaged and analysed with Igor Pro 6.05 (Wavemetrics). Rise time was measured from 10% to 90% of the peak eEPSC amplitude and tau decay was fitted from the 50% eEPSC amplitude. Measurement of the synaptic delay was performed as illustrated in the inset of Fig. 3D. BAPTA-AM (Invitrogen) was dissolved in DMSO, which contained 10% Pluronic (F-127, Invitrogen), to create a 10 mM stock solution. Larvae were dissected and incubated for 20 min in 10 µM BAPTA-AM diluted in HL-3 or in Pluronic in DMSO diluted in HL-3 for controls. After incubation, CNS was removed and focal recordings were performed as described above. For fluctuation analysis, focal electrodes with resistances of 1.4–1.8 MΩ were used to record eEPSCs elicited by the paired pulse stimulation protocol described above at bath temperatures of 20 ± 1°C with the help of a constant displacement pump. Approximately 20 traces per recording site and concentration were analysed in consecutive extracellular Ca²⁺ concentrations (0.5, 1.0, 1.5, 3.0, 5.0, 0.5 mM) using Igor Pro 6.37 (Wavemetrics). eEPSC variances and the variance of the variances were calculated according to Neher and colleagues^40–42 as described previously.^15,38 Variance-mean amplitude plots were fitted using the formula Var(I) = I/N + qI, with I meaning the mean eEPSC amplitude, q quantal size and N the number of release-ready vesicles. The release probability P_r for each concentration was calculated as P_r = I/(Nq).

Ca²⁺ imaging

Ca²⁺ imaging was performed in principle as described previously.^43,44 Male wandering third-instar larvae were dissected in ice-cold, Ca²⁺ free HL-3 and incubated in 5 mM Oregon-Green BAPTA-1 488 (Invitrogen) and 1 mM Alexa Fluor568 (Invitrogen) in HL-3 on ice for 10 min. After incubation, the preparation was washed with HL-3 for 10 min. Synapses were imaged with a Scientifica 2P system. Excitation light (800 nm) from a Tsunami laser was focused onto the NMJ using a 60× dipping objective (Olympus, NA 1.0) and emitted light was detected by 2 PIMS-PMT-20 photomultiplier tubes. Single action-potential evoked spatially averaged Ca²⁺ transients were imaged from proximal type Ib boutons of NMJs on muscles 6/7 in abdominal segments A2 and A3 at 1.0 mM [Ca²⁺]_Ex. One to two NMJs per preparation and one to six boutons per NMJ were included into the analysis. Line scans (8–12 per bouton) were made at a frequency of 333 Hz. ΔF/F describes the fluorescence change following an action potential and was quantified as [F(t) − F_baseline)/(F_baseline − F_background].⁴³ Measurements with F_baseline < 210 analog/digital (A/D)-values were excluded from analysis.

Immunofluorescence imaging

For all imaging experiments we used NMJs formed on abdominal wall muscles 6/7 in segments A2 and A3 from male, wandering third instar larvae. Larvae were dissected in ice-cold Ca²⁺-free HL-3, fixed with 4% paraformaldehyde in 0.1 M PBS for 10 min and blocked with PBT (PBS containing 0.05% Triton X-100, Sigma) including 5% natural goat serum (Dianova) for 30 min. Primary antibodies were added for overnight staining at 4°C. After three washing steps with PBS (20 min each), preparations were incubated with secondary antibodies for 2–4 h at room temperature followed by three washing steps. For quantification of AZs per NMJ, filets were mounted in Vectashield (Vector Laboratories) and images acquired using an Apotome System (Zeiss, Axiovert 200 M Zeiss, objective 63×, NA 1.4, oil). Antibodies were used in the following concentrations: mouse mAb α-Brp^Nc82 (1:250), goat α-mouse conjugated Alexa Fluor488 (1:250, Invitrogen) and goat α-horseradish peroxidase conjugated Cy3 (α-HRP, 1:250, Jackson Immuno Research). Data were analysed using ImageJ software (NIH) as described previously.⁴⁵ Images were maximum-projected. If more than one image per NMJ was collected, the duplicate signal that was acquired in the first image was used for quantification. Prior to analysis, signals from axons and neighbouring synapses as well as obvious background were removed manually with the ‘Rectangle’ or ‘Freehand selection’ tools in corresponding Brp and α-HRP images. If axons or neighbouring synapses crossed or touched the NMJ directly, images were excluded from analysis. NMJ area and number of Brp spots per NMJ were quantified in α-HRP and Brp images, respectively, using the ‘Analyze Particles’ function in 8-bit converted images applying an intensity threshold that was fixed for data that were compared statistically. If necessary images were background-subtracted using the Rolling ball method (radius = 50 pixels, α-HRP data in Fig. 4B). The number of boutons per NMJ was counted manually.

dSTORM

Super-resolution imaging of Drosophila larval NMJs was performed as previously described.⁴⁶ In brief, larvae were dissected in HL-3, fixed, blocked and stained as described above. After staining, muscle preparations were incubated in 100 mM mercaptoethylamin (MEA, Sigma-Aldrich) buffer in PBS, pH 7.8–7.9 to allow reversible switching of single fluorophores during data acquisition.⁴⁷ Images were acquired using an inverted microscope (Olympus IX-71, 60×, NA 1.45, oil immersion) equipped with a nosepiece-stage (IX2-NPS, Olympus). 647 nm (F-04306-113, MBP Communications Inc.), 644 nm (iBEAM-SMART-640-S, Toptica) and 488 nm (iBEAM-SMART-488-S, Toptica) lasers were used for excitation of Alexa Fluor647/Cy5 and Alexa Fluor488 2, respectively. Laser beams were passed through a clean-up filter (Brightline HC 642/10, Semrock, and ZET 488/10, Chroma, respectively) and two dichroic mirrors (Laser-MUX BS 514-543 and HC-quadband BP, Semrock) onto the probe. The emitted fluorescence was filtered with a quadband-filter (HC-quadband 446/523/600/677, Semrock) and divided onto two cameras (iXon Ultra DU-897-U, Andor) using a dichroic mirror (HC-BS 640 imaging, Semrock). In addition, fluorescence was filtered using a longpass (Edge Basic 635, Semrock) or bandpass filter (Brightline HC 525/50, Semrock) for red and green channels, respectively. Pixel size for the super-resolved red channel was 126 nm. Single fluorophores were localized and high resolution-images were reconstructed with rapidSTORM^48–51 (www.super-resolution.de). Only fluorescence spots with an A/D count higher than 12 000 photons were analysed (10 nm/pixel sub-pixel binning). Antibodies were used in the following concentrations: mouse mAb α-Brp^Nc82 (1:2000) and goat α-mouse F(ab′)₂ fragments (A10534, Invitrogen) labelled with Cy5-NHS (PA15101, GE Healthcare) at a concentration of 5.2 × 10⁻⁸ M in ok6-GAL4 > UAS-RIM^wt and >UAS-RIM^R>H larvae, and mouse mAb α-Brp (nc82; 1:100, Antibody Registry ID: AB_2314866, Developmental Studies Hybridoma Bank) and Alexa Fluor647 labelled secondary F(ab′)₂ fragments goat α-mouse (1:500, A21237, Thermofisher) in rim^rescue/rim^rescue and rim^R>H/rim^rescue larvae. Presynaptic boutons were visualized with Alexa Fluor488 conjugated goat-α-HRP (1:250, Jackson Immuno Research). All example AZs in Fig. 4 are shown in top view (i.e. optical axis perpendicular to the AZ plane).

Localization data were analysed essentially as described previously⁵² with custom written Python code (https://www.python.org/, language version 3.6). RapidSTORM localization tables were directly loaded and analysed. Regions of interest corresponding to the terminal 6 boutons were masked in the reconstructed, binned images from rapidSTORM using FIJI. For cluster analysis we used the Python implementation of hierarchical density-based spatial clustering of applications with noise (HDBSCAN).^53,54 Brp clusters were extracted with the combination 400 and 100 for ‘minimum cluster size’ and ‘minimum samples’, respectively, in RIM^wt and RIM^R>H measurements, and 100 and 25 for ‘minimum cluster size’ and ‘minimum samples’ in rim^rescue/rim^rescue and rim^R>H/rim^rescue measurements. The adjustments were necessary due to multiple changes within the experimental setup between measurements (i.e. microscope hardware, antibody concentrations, type of secondary antibody and conjugated dye). Cluster areas were quantified with 2D alpha shapes using CGAL (Computational Geometry Algorithms Library; https://www.cgal.org) in Python (α = 800 nm²). Exclusion criteria for Brp clusters were area <0.01 µm² and >0.1 µm², Brp localizations per cluster >8000 and mean localization density per cluster >60 000 localizations per µm².

Statistical analyses

Datasets in Figs 2 and 6C and D were analysed with Prism 9 (GraphPad Software) and tested for normal distribution applying a one-sample Kolmogorov-Smirnov test. To compare normally distributed datasets, a one-way ANOVA followed by a post hoc pairwise comparison with Tukey’s correction was applied. Data that were not distributed normally were compared with a Kruskal-Wallis test followed by pairwise comparison with Dunn’s analysis. Analyses in Figs 3–5 and Supplementary material were performed with Sigma Plot 12 (Systat Software) using a parametric t-test or a non-parametric Mann-Whitney rank sum test depending on data distribution assessed through Kolmogorov-Smirnov testing. For multiple comparisons in Fig. 6A and B, a one-way ANOVA was used. Data are reported as mean ± standard error of the mean (SEM) or median (25th–75th percentile). In box plots, horizontal lines represent median; boxes quartiles; whiskers 10th and 90th percentiles; scatter plots show individual data-points. Whisker plots in Fig. 6D show mean ± SEM.

Data availability

The authors confirm that the data supporting the findings of this study are available within the article and its Supplementary material.

Results

RIM and the structure of the C2A domain are evolutionarily conserved

RIM is a multidomain molecule containing five evolutionarily conserved protein motifs (Fig. 1A)⁵⁵ that interact with P/Q- and N-type voltage gated calcium channels (VGCC; through the PDZ and C₂B domains),^56,57 Rab3 (through the Zinc finger domain),⁵⁵ other AZ scaffolding partners including ELKS (PDZ domain),^58,59 RBP (Zinc finger domain and PxxP-motif),⁶⁰ Synaptotagmin and α-Liprin (C₂B domain),¹⁰ and the presynaptic target membrane.⁶¹ This domain architecture is already present in invertebrate organisms including Caenorhabditis elegans and Drosophila melanogaster. RIM C₂A and C₂B domains show considerable homology across different species (Fig. 1C) including over a 3₁₀-helix first observed in structures of the rat homologue that is exposed at the bottom face of the C₂A domain (Fig. 1B). This 3₁₀-helix harbours a missense mutation in a human kindred suffering from cone rod dystrophy type 7 (CORD7), a congenital disease hallmarked by progressive vision loss and increased cognitive abilities.^6,7 The arginine located at the CORD7 site at the N-terminal portion of the 3₁₀-helix is always neighboured by a second arginine forming an RR-motif in all species examined except in the C. elegans homologue UNC-10 (Fig. 1C). Interestingly, the CORD7 site corresponds to the first arginine of the RR-motif in vertebrate RIMs but assumes the second position in Drosophila on the basis of sequence alignment. Both arginines are likely important for RIM function as they define a distinct positively-charged epitope on the bottom domain surface (Fig. 1D and E).

To determine the effect of the CORD7 mutation on synaptic function it was necessary to confirm that the Drosophila RIM C₂A domain also sterically conforms to mammalian homologues. Thus, we recombinantly expressed and crystallized the fly RIM C₂A domain. We obtained a complete diffraction dataset at 1.92 Å resolution (Fig. 1B and Supplementary Table 1), which was used to solve the structure of the fly RIM C₂A domain aided by molecular replacement using the C₂A domain of the rat RIM2 homologue as a model (RIMS2, PDBID: 2BWQ).⁸ The final Drosophila RIM C₂A model, refined to R/R_free of 18.4/21.8%, accounts for residues 836–962 of the protein (Fig. 1B; PDB ID 4TS6). The structure adopts a β-sandwich fold with a small 3₁₀-helix in the loop connecting strands β5-β6, and is highly similar to other C₂ domains having a Cα root-mean-square deviation (RMSD) of just 0.83 Å compared to the rat RIMS2 C₂A variant.⁸ Consistent with the sequence alignments, the side chain of the second arginine (R916) in the fly C₂A domain occupies the same position in space as the CORD7 arginine in rat RIMS2 (R805). Interestingly, the side chain of the fly R915 also projects to the same position in space as the rat R806, the C-terminal neighbour of the CORD7 site. Thus, the surface properties of the 3₁₀-helix are highly similar between fly and rat (and presumably human) RIM despite differences in the alignment of the arginine doublet. We suggest that the RR-motif serves the same role in fly and man, as both arginines contribute to the same epitope of the 3₁₀-helix at the bottom face of the RIM C₂A domain.

The RIM CORD7 mutation is a semi-dominant enhancer of synaptic release

To address the functional relevance of the CORD7 mutation in the context of the C₂A domain we employed a CRISPR/Cas9-mediated genomic engineering approach utilizing homology directed repair for genome editing. Through a pair of gRNA probes we removed a 2.5 kb fragment of the rim locus that contains the coding region for the C₂A domain (Fig. 2A) and replaced it with an attP landing site generating a rim^ΔC2A allele (Fig. 2B). Subsequently, this fly strain was used to receive attB-flanked transgenes into the rim gene through ΦC31 expression (Fig. 2B).^28,62 The transgenes carried the previously removed rim gene fragment with either the wild-type C₂A domain sequence (rim^rescue), a C₂A domain in which the CORD7 mutation was inserted at position R915 (R915H; rim^R>H), or in which the arginine doublet was exchanged for glutamate residues (R915E,R916E; rim^RR>EE). The surface of the 3₁₀-helix of the wild-type RIM C₂A domain appears positively charged (Fig. 1E, C₂A^WT). We anticipated that insertion of the CORD7 mutation will slightly alter this electrostatic profile (Fig. 1D) with the surface of the 3₁₀-helix assuming a neutral charge at the site of the CORD7 mutation (Fig. 1E, C₂A^R>H). In contrast, insertion of negatively charged glutamate residues would result in a pronounced inversion of charge (Fig. 1E, C₂A^RR>EE) compared to the wild-type situation. All alleles generated through the genomic engineering approach were confirmed through PCR genotyping and sequencing (Fig. 2C). Overt developmental or behavioural abnormalities in animals carrying the different engineered rim alleles were not observed.

We first tested the functional consequences of the alleles by TEVC recordings of postsynaptic currents in response to low-frequency nerve stimulation in 1.0 mM extracellular Ca²⁺ at muscle 6/7 neuromuscular junctions of third instar larvae. eEPSCs of rim^rescue animals (34.6 ± 1.5 nA mean ± SEM throughout this article; Supplementary Table 2 contains all descriptive statistics and results of group-by-group statistical comparisons of all TEVC and PPR datasets) compared indistinguishably to rim⁺ control animals (37.3 ± 1.4 nA) that did not undergo the CRISPR/Cas9 genome manipulations demonstrating that our genomic engineering protocol yields precisely constructed rim alleles (Fig. 2D and E). We further confirmed that the rim^Ex73 allele²⁴ behaved as a recessive strong loss-of-function mutation in this assay (rim^KO: 10.7 ± 1.0 nA; rim^rescue/rim^KO: 24.2 ± 2.3 nA). In addition, these results also showed that the wild-type rim^rescue allele is not completely haplo-sufficient for the eEPSC response as rim^rescue/rim^KO heterozygotes displayed significantly lower synaptic responses than rim^rescue homozygotes (Fig. 2D and E). This indicates that the amount of rim gene product is critical for the physiological level of synaptic release.

Intriguingly, when we assessed eEPSCs of rim^R>H junctions we noticed that homozygous animals carrying the CORD7 mutation in RIM displayed largely increased currents (54.9 ± 2.5 nA; Fig. 2D and E) compared to rim^rescue homozygotes. To assess the functional consequences of the rim^R>H mutation for synaptic transmission we reduced the rim^R>H allele copy number by generating transheterozygous rim^R>H/rim^rescue animals reflecting the RIMS1 genotype of CORD7 patients.^6,7 Intriguingly, eEPSC amplitudes of the transheterozygotes (rim^R>H/rim^rescue: 44.8 ± 3.3 nA) settled between the values of the homozygous parents (rim^rescue/rim^rescue and rim^R>H/rim^R>H) (Fig. 2D and E). In contrast, when rim^R>H was placed in trans to the rim null allele eEPSCs of rim^R>H/rim^KO animals (26.5 ± 2.5 nA) were indistinguishable from rim^rescue/rim^KO heterozygotes (24.2 ± 2.3 nA; Fig. 2D and E) showing that CORD7 mutation- carrying rim^R>H allele exerts a semi-dominant effect over the rim^rescue but not rim^KO alleles with regard to eEPSC amplitudes. Collectively, this allelic series uncovered that rim gene dosage and the CORD7 change of the C₂A domain of RIM contribute independently to enhanced synaptic release.

Of note, recordings from rim^RR>EE larvae showed eEPSCs (26.5 ± 2.1 nA mean ± SEM) that were larger than in rim^KO animals but not significantly different from rim^rescue and rim⁺ junctions (Fig. 2D and E). This indicates that, in contrast to the CORD7 mutation, the charge reversal at the 3₁₀-helix of RIM’s C₂A domain did not profoundly affect eEPSC amplitudes.

Paired-pulse ratio (PPR) measurements confirmed that removal of RIM at rim^KO NMJs resulted in facilitation of synaptic release (Fig. 2F and Supplementary Table 2), consistent with previous observations of decreased release probability and strong synaptic facilitation at the NMJ of rim hypomorphic Drosophila mutants.⁴³

To exclude cell non-autonomous effects caused by the germline alteration of rim function in engineered rim^R>H mutants we next sought to genetically isolate larval motor neurons from an otherwise wild-type nervous system by use of the binary GAL4/UAS expression toolkit. We generated UAS transgenes containing either the wild-type rim cDNA sequence (RIM^wt; in capital letters to contrast this set of transgenes from rim alleles) or a cDNA containing the R915H substitution (RIM^R>H) and expressed them in larval motor neurons by the ok6-GAL4 driver. The functional consequences of the RIM^R>H mutation on synaptic function were tested with TEVC recordings of postsynaptic currents in response to low-frequency nerve stimulation in 1.0 mM extracellular Ca²⁺ (Fig. 3A). eEPSCs in RIM^R>H animals were larger than in RIM^wt (RIM^R>H: 59.1 ± 4.5 nA; RIM^wt: 35.1 ± 6.5 nA; Fig. 3B and Supplementary Table 3), while mEPSCs and short-term plasticity remained unchanged (Fig. 3C, Supplementary Fig. 1 and Supplementary Table 3) confirming the semi-dominant effect of the rim^R>H allele on synaptic release.

While TEVC recordings report the input to an entire muscle cell derived from many boutons, focal electrodes sample from a subset of synaptic boutons with superior temporal resolution (Fig. 3D and Supplementary Table 3). Peak current amplitudes are lower in focal recordings than with TEVC, while they permit resolving synaptic delays and eEPSC rise times. Consistent with results from TEVC focal recordings from proximal type Ib boutons in 1.0 mM extracellular Ca²⁺ revealed that the RIM^R>H increases synaptic current amplitudes compared to RIM^wt (RIM^R>H: 3.6 ± 0.6 nA; RIM^wt: 1.6 ± 0.2 nA; Fig. 3E and F and Supplementary Table 3). Interestingly, synaptic delay was decreased in RIM^R>H (1.18 ± 0.06 ms and 1.39 ± 0.06 ms; Supplementary Table 3) and in addition, rise and decay time were accelerated compared to RIM^wt (rise time: 0.75 ± 0.1 ms and 0.97 ± 0.1 ms; tau decay: 3.3 ± 0.2 ms and 4.2 ± 0.2 ms; Supplementary Table 3). However, at lower extracellular Ca²⁺ concentration (0.5 mM) eEPSC amplitudes and kinetics were not changed in RIM^R>H (Fig. 3G; amplitude: 2.27 ± 0.74 nA and 1.85 ± 0.81 nA; delay: 1.4 ± 0.07 ms and 1.51 ± 0.25 ms; rise time: 1.06 ± 0.07 ms and 1.06 ± 0.07 ms; tau decay: 4.98 ± 0.59 ms and 4.83 ± 0.64 ms for RIM^wt and RIM^R>H, respectively). This indicates, that the RIM^R>H mutation exerts a Ca²⁺-dependent effect on evoked synaptic release.

Overexpression of RIM^R>H leaves synapse and active zone morphology unaltered

To assess whether any of these functional characteristics observed when overexpressing RIM^R>H in larval motor neurons are caused by changes of synaptic structure, we performed a morphometric analysis of immunohistochemically processed NMJ preparations. To this end, we focused on the critical AZ scaffold protein Bruchpilot (Brp) and used the highly specific monoclonal antibody Brp^Nc82, which recognizes an epitope in the C-terminal part of Brp.^63–65

We quantified the number of Brp puncta per NMJ in RIM^wt and RIM^R>H using Brp^Nc82 and found it indistinguishable between the two groups (Fig. 4A and B and Supplementary Table 4). In addition, the overall NMJ size measured by the extent of the presynaptic α-HRP area and the number of boutons per NMJ were unchanged. Next, to test the impact of the CORD7 mutation on the mesoscale architecture (scale level bridging the nanometer scale of atomic structure and the micrometer scale of cellular ultrastructure)⁶⁶ of individual AZs we performed localization microscopy using dSTORM (direct stochastic optical reconstruction microscopy)^48,50,67 applying a previously approved HDBSCAN analysis.⁵² Since Brp is abundantly present at the presynapse and immunolabels by Brp^Nc82 cover the spatial extent of individual AZs,⁴⁶ we equate ‘Brp area’ with ‘AZ area’ (Fig. 4C and D). Analysis of AZs in type Ib boutons of the NMJ showed no difference in size or Brp composition of individual AZs between RIM^wt and RIM^R>H (Fig. 4D and Supplementary Table 4). We conclude that AZ number and synapse morphology as well as the Brp nanoarchitecture are not altered in RIM^R>H-overexpressing motor neurons.

The CORD7 mutation tightens release coupling

RIM is required for normal levels of presynaptic Ca²⁺ channels²⁴ and, in addition, the CORD7 mutation was shown to modulate VGCC function.¹¹ Therefore, we asked whether the RIM^R>H mutation affects Ca²⁺ signals at the Drosophila NMJ. To test this, we loaded presynaptic terminals with the Ca²⁺ indicator OGB-1 and imaged spatially-averaged Ca²⁺ transients across type Ib boutons in response to single action potential stimulation. We did not observe significant differences in peak amplitudes or decay kinetics of presynaptic Ca²⁺ transients between RIM^wt and RIM^R>H (Fig. 5A and B and Supplementary Table 5). Thus, the RIM^R>H mutation does not affect presynaptic spatially-averaged Ca²⁺ transients, implying unaltered Ca²⁺ influx.

Our data show a striking enhancement of evoked synaptic transmission by the RIM^R>H mutation, which requires a yet unnoted mechanistic basis of RIM function since the number of AZs was unchanged at RIM^R>H NMJs (Fig. 4B). Tighter coupling causes the release process to be less susceptible to Ca²⁺ chelators, and through nanodomain coupling²⁰ the slow Ca²⁺ buffer EGTA has little effect at Drosophila NMJs.⁶⁴ Thus, we loaded presynaptic boutons with the membrane permeable BAPTA-derivative BAPTA-AM (10 µM for 20 min), which results in an intracellular BAPTA concentration in the range of 200–300 µM.⁶⁸ We observed a substantial reduction of the measured eEPSC amplitudes in RIM^wt (dropping to 24%), whereas the RIM^R>H mutation rendered release less sensitive to BAPTA (dropping to 42%, P = 0.006; Fig. 5C–E). These results indicate that the RIM^R>H mutation tightens the coupling distance of synaptic vesicles to the calcium source at the AZ.

Allelic rim^R>H expression increases the number of presynaptic boutons, active zones and RRP size

So far our analysis revealed striking effects of the RIM CORD7 mutation on synaptic function, which could not be linked to structural synapse alterations in this overexpression situation. However, allelic expression of rim^R>H also increased eEPSC amplitudes in TEVC recordings. We wondered whether synaptic morphology was influenced by the CORD7 mutation when expressed under endogenous cis-regulatory control. To address this question, we performed immunohistochemical analyses of NMJs 6/7 in rim^rescue/rim^rescue, rim^rescue/rim^R>H and rim^R>H/rim^R>H animals using Brp^Nc82 again as a marker for presynaptic AZs and α-HRP to measure the extent of the presynaptic plasma membrane (Fig. 6). Surprisingly, we found an increased number of AZs and of presynaptic bouton numbers per NMJ as well as enlarged NMJ size due to the rim^R>H mutation in a dose-dependent manner (Fig. 6A and Supplementary Table 6). Employing dSTORM and HDBSCAN algorithms we then analysed the Brp nanoarchitecture in the three genotypes. We found no effect on AZ area, the number of Brp localizations per AZ and Brp localization density in heterozygous rim^rescue/rim^R>H and homozygous rim^R>H/rim^R>H variants of the CORD7 mutation (Fig. 6B and Supplementary Table 6). These imaging results suggest that allelic expression of rim^R>H changes the overall NMJ morphology without influencing the Brp organization at the mesoscale level.

Finally, we estimated the size of the RRP of synaptic vesicles in animals carrying the rim^R>H allele. Through fluctuation analysis of focal recordings we found that q, N and p_r were unchanged (Supplementary Fig. 2). However, linear fits of average cumulative eEPSC amplitudes in response to 20 Hz trains were back-extrapolated to allow estimations for the RRP (Fig. 6C and D). Applying this approach we determined that the RRP size was increased at rim^R>H synapses (Fig. 6E and Supplementary Table 7).

Discussion

Protein interactions at presynaptic AZs determine fundamental aspects of neuronal communication. Typically, mutations in proteins are deleterious without ambiguity. A significant exception, however, is a specific mutation in the RIM protein that leads to the CORD7 syndrome.⁶ The affected individuals suffer from retinal dystrophy leading to blindness but also show enhanced cognitive capabilities characterized by increases in verbal IQ and working memory.⁷ As revealed by large-scale exome sequencing efforts the abundance of the CORD7 mutation in the general human population was found with a minor allele frequency of 0.017% without further indication on putative phenotypes of affected individuals.⁶⁹ A thorough analysis of the effects of the CORD7 mutation on neuronal structure or synaptic communication that may ultimately cause the remarkable neurological phenotypes of index family members has thus been warranted.

RIM is a central element of AZs and involved in the configuration of AZ shape, composition and function through multiple protein interactions.^{10,11,23,70–72} These roles are at least partially mediated through RIM’s domains, which are structurally and biochemically well defined. The clarification of each domain’s contribution to synaptic function is complicated through the genetic redundancy of four different RIMS loci in addition to the splicing-dependent heterogeneity of their gene products in mammalian model species.⁷³

Here we studied the function of the sole rim homologue CG33547 of Drosophila melanogaster.^24,43 We first employed X-ray crystallography of the fly RIM C₂A domain to demonstrate that the overall domain layout is strikingly similar and has remained highly conserved throughout evolution. We found that the arginine doublet in the 3₁₀-helix, which is affected in CORD7 patients, is also present in the Drosophila C₂A domain and sterically matches the mammalian domain layout.⁸ We then analysed the effect of the CORD7 mutation at the Drosophila NMJ by genomic engineering of alleles and construction of transgenes with a CORD7-equivalent mutation, and by mimicking the RIMS1 genotype of CORD7 patients.

Arginine doublets in C₂ domains of AZ proteins, the site at which the CORD7 mutation alters RIMS1’s protein structure, have been implicated in different functions and interactions. The arginine doublet of the Synaptotagmin-1 (Syt1) C₂B domain, which is located on the same side of the domain compared to the position of the CORD7 mutation, was previously shown to allow for lipid interaction.⁷⁴ This doublet is functionally highly relevant as RR-elimination culminates in release failure and is functionally equivalent to the impairment of calcium binding at the top face of the C₂B domain.⁷⁵ In this context the arginine doublet of the C₂B domain appears pertinent for positional priming and site clearance.⁷⁶ We found a profound increase of synaptic current amplitudes, and accelerated rise and decay times caused by the CORD7 mutation implying that the rim^R>H mutation affects synaptic release at AZs in Drosophila too, and thus by analogy also in CORD7 patients. In addition, we found NMJ size and the number of presynaptic AZs per NMJ increased by allelic expression of the CORD7 mutation which might be a contributing factor to release potentiation. Interestingly, also release coupling and changes in RRP size are influenced by the mutation.

The molecular mechanism by which the rim^R>H mutation affects synaptic function remains speculative. It was shown that RIM C₂A and C₂B domains interact with the α- and β-subunits of VGCC, SNARE components and synaptotagmins;^10,56,57 one of these interactions may be mediated by the 3₁₀-helix side of the C₂A domain. At the biophysical level rim^R>H lessens the positive electrostatic charge found at this site of the wild-type protein and would weaken charge-based interaction forces. However, a weaker electrostatic charge effect of C₂A^R>H on this putative target binding may be counteracted by higher hydrophobicity of the histidine amino acid compared to arginine, leading to a net strengthening of this interaction. In contrast, the rim^RR>EE substitution not only inverts the charge of this protein site, but it also offers no compensatory hydrophobic effect causing the weakening or disruption of the putative RIM complex. Thus, the rim^R>H and rim^RR>EE mutations can have opposing molecular effects leading to the observed differences in synaptic response.

Intriguingly, RIM1-mediated modulation of VGCC was altered by the CORD7 mutation in vitro displaying impaired function of L-type Ca_V1.4, the predominant VGCC that controls release from ribbon synapses in the retina, upon co-expression with RIMS1^R>H, while calcium conductance through P/Q-type Ca_V2.1 channels, which partake in central synaptic release, was augmented.^11,77 These findings offer an explanation for the opposing impact of the CORD7 mutation on the visual and central systems. Our presynaptic calcium indicator measurements argue that the rim^R>H mutation does not impact presynaptic Ca²⁺ signals in vivo. Instead, enhanced synaptic release in the presence of the CORD7 mutation is likely caused by a reduced coupling distance of synaptic vesicles to the calcium source at the AZ as shown by the reduced effect of a calcium chelator on eEPSC amplitudes in the presence of RIM^R>H. This is in line with findings showing that RIM C₂A and C₂B domains interact with the α and β-subunits of VGCC, SNARE components and Synaptotagmins.^{11,56,78–80} The RIM-C₂B domain additionally crosstalks with α-Liprin.¹⁰ Rescue experiments of RIM-deficient synapses, however, demonstrate that RIM’s capacity to recruit calcium channels to the AZ does not rely on the presence of either C₂ domain, while, in contrast, both C₂ domains facilitate synaptic release without modulating presynaptic calcium channel influx in vivo.^23,71 In addition, our investigations revealed an increase of the RRP size at Drosophila NMJs. This finding might offer an additional explanation for enhanced synaptic performance in mutants expressing the rim^R>H CORD7 allele since RRP enlargement—representing a ‘pool engram’—was suggested to be suitable for short-term storage of information.⁸¹ Remarkably, this effect was calcium- and dosage-dependent as measurements in the motor neuron-specific overexpression and the allelic expression revealed. One can speculate that the heterogeneity of evoked synaptic release along the NMJ could contribute to these effects too, thus, employing transgenically expressed GCaMP Ca²⁺ sensors imaging to precisely dissect evoked neurotransmission at different release spots may be a fitting approach to better understand this phenomenon.^82,83 To gain further insight into organization of the synaptic vesicle pool, 3D EM tomographic analyses of vesicle distributions within the presynaptic terminal are attractive.⁸⁴ Further investigations are necessary to delineate RIMS1-VGCC interactions and their steric relationship, specifically at retinal and telencephalic mammalian synapses.

A possible cause of the observed release abnormalities at rim^R>H synapses is the interference of the mutation with direct molecular interactions, which occur between the 3₁₀-helix of the C₂A domain and yet unknown partners. Alternatively, the CORD7 mutation may impact protein folding and/or stability of the C₂A domain and, thus, the entire RIM molecule. This may result in altered levels of the RIM pool or subcellular localization of RIM at the presynapse and affect neurotransmission. However, the observed gene dosage sensitive gain-of-function effects of rim^R>H in direct comparison with loss-of-function alleles (rim^RR>EE or rim^KO) argue against a reduction or complete depletion of RIM^R>H at the synapse. Unfortunately, no α-RIM antibody or transgenically tagged RIM tool is currently available to faithfully assess the amount of RIM protein and its potential change under the influence of the CORD7 mutation.

In sum, the use of the Drosophila NMJ as a model system allowed us to uncover the impact of the human CORD7 syndrome mutation of the RIM C₂A domain on speed and efficacy of synaptic release. Future investigations have to detail how the striking cognitive enhancement of CORD7 patients and the functional augmentation of synaptic transmission at Drosophila synapses are accomplished through this discrete change in the RIM C₂A domain.

Acknowledgements

We acknowledge the Diamond Light Source for provision of synchrotron radiation facilities and the Helmchen lab for providing access to the Scientifica 2P system. The authors thank P. Jonas for critically reading the manuscript, J. Shiroma for help with protein purification, F. Köhler, M. Oppmann, C. Häusslein and K. Heise for technical assistance, and E. Graf and N. Ehmann for materials and discussions.

Funding

This work was supported by grants from the German Research Foundation (DFG) to M.He., M.S., R.J.K., N.S. and T.L. (SFB-TR166/projects A4, B4, B6, C3; SFB1047/project A5; FOR2149/projects P1 [265903901] and P3 [265996823]; FOR3004 SYNABS/project P1; KI 1460/5-1; LA 2861/7-1), a SNSF Assistant Professor grant (PP00P3-15) and an European Research Council Starting Grant (‘SynDegrade’, 679881) to M.M., the Biophotonics Initiative of the Bundesministerium for Bildung und Forschung to M.S. (13N11019 and 13N12507) and the IZKF Würzburg to M.M.P. (Z-3/69) and M.He. (N229). I.V. thanks the Wellcome Trust (088497/Z/09/Z) and the Biotechnology and Biological Sciences Research Council UK (BB/J008265/1) for support.

Competing interests

The authors report no competing interests.

Supplementary material

Supplementary material is available at Brain online.

References

Author notes