For the deletion of β-Nrxns, we used a previously described triple β-Nrxn floxed mouse strain (βKI; available from JAX Labs as B6;129-Nx1^{TM 2Sud} Nx2^{TM 2Sud} Nx3^{TM 2Sud}/J, RRID:IMSR_JAX:008416) (Anderson et al., 2015). In βKI mice, the 5′-exons specific for β-Nrxns are flanked by two loxP sites, which can be recognized and subsequently excised by Cre recombinase introduced, for example, by lentivirus to create a conditional knockout (cKO). In addition, a constitutive triple β-Nrxn knockout (β-TKO) was generated by crossing B6;129-Nx1^{TM 2Sud} Nx2^{TM 2Sud} Nx3^{TM 2Sud}/J with transgenic mice B6.FVB-Tg (Ella-Cre) C5379Lmgd/J (RRID:IMSR_JAX:003724) expressing Cre recombinase. The βTKO mice were subsequently bred to C57BL/6J (RRID:IMSR_JAX:000664) for several generations to remove Ella-Cre transgene. All animal experiments were carried out at the University of Münster according to government regulations for animal welfare and approved by the Landesamt für Natur, Umwelt und Verbraucherschutz (LANUV, NRW, Germany), license numbers 84-02.05.20.11.209 and 84-02.04.2015.A423.

The lentivirus system expressing active or inactive Cre recombinase fused to enhanced green fluorescent protein (EGFP) used here was derived from the laboratory of Tom Südhof and was used in primary hippocampal or cortical cultures to effectively delete, for example, Mint proteins (Ho et al., 2006), RIM proteins (Kaeser et al., 2011), splice inserts of Nrxns (Aoto et al., 2013, 2015), or entire neurexin genes (Anderson et al., 2015; Trotter et al., 2019; Klatt et al., 2021), as shown by immunoblots for complete deletion of the respective genes. The production of lentiviruses and infection of neuronal cultures with lentiviruses in our lab have been previously described (Klatt et al., 2021). Briefly, the helper plasmids pRSV-REV (RRID:Addgene_12253), pMDLg/gRRE (RRID:Addgene_12251), and pVSVG (RRID:Addgene_8454) were co-transfected with lentiviral expression vectors [FSW-NLS-GFP-Cre or FSW-NLS-GFP-Cre^mut (Y324F)] (de Jong et al., 2018; Klatt et al., 2021) into the human embryonic kidney (HEK) 293T cells using Lipofectamine 2000 (Thermo Fisher Scientific, Waltham, MA, United States, Cat #11668019). About 72 h after transfection, viral supernatant was collected and centrifugated at 500 g for 10 min at 4°C. The supernatant was aliquoted and immediately frozen at -80°C with maximal storage time of 4 weeks. At 5 days in vitro (DIV5), 150 μl of lentiviral supernatant (Cre or Cre^mut) was added dropwise to neuronal cultures for 3 days. Neurons infected either with Cre- or Cre^mut-lentiviruses could be detected by nuclear expression of EGFP driven by a synapsin promoter. Based on the GFP-Cre or GFP-Cre^mut autofluorescence and colabeling with MAP2 as a marker of neuronal cell bodies and dendrites (see section “Immunocytochemistry”), we routinely tested virus preparations for the effectiveness of transduction. We normally determined the efficiency between 92 and 99% and excluded preparations below 95% from this study.

Hippocampal neurons were prepared in HBSS from β-Nrxn floxed mice as described (Brockhaus et al., 2018). Briefly, primary hippocampal neurons derived from timed-pregnant dams at E17.5 were dissociated by 0.25% trypsin and plated onto 18-mm glass coverslips (Menzel-Glaseser) coated with poly-L-lysine (Sigma-Aldrich, St. Louis, MO, United States, Cat #P1524) at the density of 55,000 cells/coverslip in plating medium. After 3–4 h at 37°C neuronal plating medium containing minimum essential medium (MEM), 10% horse serum, 0.6% glucose, and 1 mM sodium pyruvate, coverslips were inverted onto a 70–80% confluent monolayer of mouse astrocytes grown in 12-well plates (Falcon), and incubated in neurobasal medium (plus B27, 0.5 mM glutamine, and 12.5 μM glutamate). After 3 days, media were refreshed with neurobasal medium (GIBCO, ThermoFisher Scientific, Waltham, MA, United States, Cat #21103-049) supplemented with B27, 0.5 mM glutamine, and 5 μM AraC. Cultured neurons were maintained at 37°C in a humidified incubator with an atmosphere of 95% air and 5% CO₂.

Fixation of cultured neurons at DIV 18 was performed with 4% formaldehyde and 4% sucrose in phosphate buffered saline (PBS) (pH 7.4) for 15 min at RT. After several washing steps, fixed cells were incubated in a blocking solution containing 5% next-generation sequencing (NGS) and 0.3% Triton X-100 in PBS for 30 min to block non-specific interactions. Primary or secondary antibodies were diluted in antibody solution (5% NGS in PBS). Coverslips were incubated with primary antibodies for 2.5 h, followed by washing steps. Incubation with secondary antibodies was done for 1 h. Finally, coverslips were mounted using Deko fluorescence mounting medium and visualized by fluorescence microscope with 63× oil immersion objective and maximal resolution. Fluorescent intensity measurement was performed using the ImageJ software (NIH Image, RRID:SCR_003073).

Samples from cultured neurons and cerebellar tissue were examined with a transmission electron microscope (Libra 120, Zeiss, Jena, Germany) at 80 kV, and imaged with a 2,048 × 2,048 CCD camera (Tröndle, Moorenweis, Germany). The number of DCVs was counted in axonal areas (225 μm²) of cultured hippocampal neurons (n = 3 virus infection experiments) and 350 μm²/total number of synapses in the cerebellar cortex.

Coverslips with cultured hippocampal neurons (DIV 17–18) were placed in the recording chamber of an inverted microscope (Observer.A1, Zeiss, Jena, Germany) with a bath solution (32°C) containing (in mM): 145 NaCl, 3 KCl, 2 CaCl₂, 1 MgCl₂, 11 glucose, 10 HEPES, and pH 7.4 adjusted with NaOH. For recordings of pharmacologically isolated miniature excitatory postsynaptic currents (mEPSCs), 10 μM bicuculline and 1 μM tetrodotoxin were added. Whole-cell patch clamp measurements from neurons with a pyramidal cell-like shape were done with a 2.5–3 MΩ pipette containing (in mM): 140 K-gluconate, 1.5 MgCl₂, 1 CaCl₂, 10 HEPES, 4 Na-ATP, 0.5 Na-GTP, 10 EGTA, and pH 7.3 with KOH.

Continuous voltage clamp recordings at −70 mV holding potential with 10 kHz sampling rate, filtered at 3.3 kHz, were obtained with an EPC10 amplifier and the Pulse software (HEKA, Reutlingen, Germany). About 100 consecutive mEPSCs per neuron were identified and evaluated for inter-event interval (IEI), amplitude, rise-time (10–90%), and half-width with TaroTools in IgorPro (Wavemetrics). TaroTools is a set of procedures for IgorPro to analyze neurophysiological data, and was written by Taro Ishikawa, Jikei University School of Medicine, Tokyo, Japan. The tool integrates into IgorPro as an add-in and can be downloaded upon request as four separate files from https://sites.google.com/site/tarotoolsregister/registration.

Statistical analysis of each experiment was performed with the GraphPad Prism software (GraphPad Prism, RRID:SCR_002798). Data are shown as mean ± SEM. To assess the statistical significance between the two groups, we used a two-tailed unpaired Student’s t-test. Significance differences are indicated in detail in the corresponding figure legends.

To delete β-Nrxns in primary hippocampal neurons, we transduced cultures from triple βNrxn-floxed mice (βKI) at DIV 5 with lentivirus expressing active Cre recombinase (Cre). Conditional deletion of floxed β-specific exons by Cre-expressing lentivirus allows removal of β-Nrxns in comparison to inactive Cre recombinase (Cre^mut) or β-KI neurons as controls. We adopted and characterized an extensively used lentivirus system (see section “Lentivirus Production and Infection”) previously and demonstrated the effectiveness of recombination events and complete deletion of β-Nrxns by immunoblots (Klatt et al., 2021). Consequently, here the distribution of DCVs was studied at DIV18, using the same culture conditions and procedures as in our earlier report (Klatt et al., 2021).

We first monitored the normal distribution of DCVs in control neurons (Cre^mut) by immunolabeling ChrgA which is a common matrix protein of DCVs (Bharat et al., 2017; Dominguez et al., 2018). Colabeling of ChrgA with antibodies against NFH as a specific intermediate filament component of mature axons reliably revealed ChrgA-positive clusters in the identified axons (Figures 1A₁,A₂), resulting in a partly overlapping pattern (Figure 1A₃). Conversely, when ChrgA (arrows, Figure 1B₁) was co-stained in control neurons for the dendritic marker MAP2 (arrowheads, Figure 1B₁), it did not overlap with ChrgA labeling. We also observed that only a subset of axons was ChrgA-positive, which is in accordance with a study showing that DCVs occur in only 18% of axonal boutons in hippocampal slices (Sorra et al., 2006). Successful infection of the labeled neurons by lentivirus with Cre^mut recombinase could be controlled by a green-fluorescent nucleus due to the expression of GFP as a fusion protein with Cre^mut (Figures 1B₃,B₄). We then performed the same set of labelings with cultures infected by active Cre to analyze β-Nrxn-deficient (cKO) neurons (Figures 1C₁–C₄). We compared the effectiveness of GFP-Cre and GFP-Cre^mut expression rates in cultured neurons colabeled by MAP2 antibodies as shown in Figures 1B₂,B₃,C₂,C₃, and determined indistinguishable coexpression rates (mean ± SEM for Cre^mut control: 96.5% ± 0.85, Cre cKO: 96.2% ± 0.78, n = 3 cultures per genotype; p = 0.9193, two-tailed unpaired t-test). Importantly and similar to control, we found that ChrgA labeling was not overlapping with MAP2-positive dendrites in cKO neurons (arrows and arrowheads, Figures 1C₁,C₂), indicating that the deletion of β-Nrxns does not lead to the re-distribution of ChrgA-positive DCVs. Robust nuclear labeling of GFP-Cre recombinase (Figures 1C₃,C₄) again confirmed that ChrgA/MAP2-positive neurons were successfully infected by Cre lentivirus. However, the identified axons of cKO neurons contained less abundant ChrgA staining if compared to controls (Figures 1D₁,D₂). To validate this hypothesis, we measured the intensity of ChrgA fluorescence within a defined axonal window (yellow box, Figures 1D₁,D₂). The results revealed a 20% reduction of ChrgA fluorescence intensity in axons of β-Nrxn cKO neurons compared to control neurons [Figure 1E; data normalized to Cre^mut control: 1.0 ± 0.05 arbitrary units (AU), Cre cKO: 0.82 ± 0.02, p = 0.0014]. These data show that the presence of β-Nrxns is required for normal levels of ChrgA in axons of hippocampal neurons, a surprising finding because β-Nrxns have not been linked to DCVs earlier.

The reduction of ChrgA fluorescence in axons of Cre-infected cultures (Figure 1) raised the question if this was caused by diminished incorporation of the matrix protein into DCVs or a bona fide reduction of the number of DCVs in β-Nrxn-deficient neurons. To distinguish between the possibilities, we analyzed the distribution of ultrastructurally identified DCVs in the same hippocampal cultures (Figure 2). While axons of these cultured primary hippocampal neurons are extremely thin, they regularly form presynaptic terminals as en passant boutons that line up at close distances of about 5 μm (Staras and Branco, 2010; Staras et al., 2010). We, therefore, focused our electron microscopic analysis on the presynaptic boutons in which typical DCVs could easily be recognized that had a diameter of about 80 nm (Cre^mut control: 83.9 ± 2.26 nm, Cre cKO: 79.9 ± 1.9, n = 90 DCVs from three cultures per genotype; n.s., two-tailed unpaired t-test), in line with earlier studies (Sorra et al., 2006). In control neurons (Cre^mut), DCVs were frequently located in typical boutons with presynaptic vesicle clusters (arrows, Figures 2A₁–A₄), clearly discernable synaptic cleft and postsynaptic density. DCV-containing boutons were interspersed with those without any visible DCV (asterisks, Figures 2 A₁,A₄). While the general distribution pattern was similar in Cre-infected cKO neurons, the number of DCVs in presynaptic boutons of β-Nrxn cKO seemed lower, very rarely exceeding a single DCV per synapse (arrows, Figures 2B₁,B₂,B₄). Moreover, we also encountered more boutons that were devoid of any DCV in cKO cultures (asterisks, Figures 2B₂–B₄). Quantification of the area density of DCVs confirmed this observation because there was a 50% reduction in cKO (Cre) neurons compared to Cre^mut controls (Figure 2C; Cre^mut: 29.1 ± 2.3 DCVs/225 μm², Cre: 14.7 ± 1.4, p < 0.0001). This reduction of the area density of DCVs seemed to be based on a shift from boutons that contain one or two DCVs to synapses with no DCV in β-Nrxn-deficient neurons. In support, counting the respective numbers of presynaptic boutons containing 0, 1, or 2 DCVs revealed a 40% higher number of terminals with zero DCVs in β-Nrxn-deficient neurons (Figure 2D; Cre^mut: 59.3 ± 3.84, Cre: 83.3 ± 3.18, p < 0.0001) whereas the number of those containing 1 or 2 DCVs was clearly diminished by about 50 or 80%, respectively (Figure 2D; for DCV = 1 Cre^mut: 24.3 ± 2.19, Cre: 11.3 ± 2.33, p = 0.0062; for DCV = 2 Cre^mut: 12.3 ± 1.33, Cre: 2.3 ± 0.67, p = 0.015). Interestingly, the shift from boutons that contain DCVs to terminals with no DCV was not reflected by changes in the area size of these boutons (Figure 2E). Area size was indistinguishable in β-Nrxn-deficient Cre and Cre^mut control neurons for DCV = 0 (Figure 2E; Cre^mut: 0.2 ± 0.02, Cre: 0.23 ± 0.02, p = 0.3141), for DCV = 1 (Figure 2E; Cre^mut: 0.18 ± 0.01, Cre = 0.2 ± 0.02, p = 0.7094), and for DCV = 2 (Figure 2E; Cre^mut: 0.21 ± 0.02, Cre = 0.22 ± 0.02, p = 0.8229). Thus, our electron microscopic results demonstrated that the diminished ChrgA fluorescence described in Figure 1 is due to an actual reduction of the number of presynaptic DCVs in the absence of β-Nrxns, leading to a higher number of boutons without any DCV. Our results could imply that the capacity of releasing neuromodulators is strictly controlled in hippocampal neurons, and that shifting of subpopulations of boutons from 1 or 2 DCVs to 0 DCV or vice versa might be used to modulate the function of the neuronal network.

The presence of DCVs within or near presynaptic boutons is an important aspect because they represent the specific release sites for DCVs in hippocampal neurons where they are mostly found at the periphery of the synaptic vesicle clusters (Persoon et al., 2018). As the deletion of β-Nrxns has an effect on synaptic vesicle release (Anderson et al., 2015; Klatt et al., 2021), we tested if the alignment of ChrgA-positive DCVs with the active zone component Bassoon was altered in cKO neurons. Colabeling of ChrgA and Bassoon revealed prominent presence of ChrgA only in a subpopulation of axons (arrows, Figures 3A₁,B₁) as also seen above in ChrgA labelings (Figures 1B₁,C₁). In contrast, Bassoon immunostaining showed a ubiquitous punctate pattern (Figures 3A₂,B₂), as expected from this canonical presynaptic marker protein (Gundelfinger et al., 2015; Biederer et al., 2017). Importantly, the general patterns of ChrgA and Bassoon distribution did not differ between controls (Cre^mut) and β-Nrxn-deficient (Cre) neurons. Furthermore, the overlay of ChrgA and Bassoon immunofluorescence images at lower and higher resolution (merged, Figures 3A₃,A₄,B₃,B₄) revealed that DCVs and active zone components are mostly juxtaposed with very little or no overlap (arrows, Figures 3A₄,B₄). These findings are in accordance with the earlier data that reported an average distance of 0.34 μm of DCVs from the active zone (Persoon et al., 2018). In addition, extrasynaptic staining of ChrgA outside Bassoon-positive boutons indicated that DCVs are trafficking along an axon (Figures 3A₃,A₄,B₃,B₄) both in controls (Cre^mut) and β-Nrxn-deficient (Cre) neurons. They are supposed to be transported bidirectionally and captured sporadically in en passent synapses (Wong et al., 2012).

To brace against the possibility that the lack of more severe changes of the presynaptic organization was due to an inefficient cKO, we measured miniature excitory postsynaptic currents (mEPSCs). We performed voltage clamp recordings in the same primary hippocampal cultures shown before (Figures 1–3A,B), and analyzed 50–100 consecutive mEPSCs in each neuron (Figures 3C,E). We found about 80% longer IEIs in β-Nrxn-deficient (Cre) compared to Cre^mut control neurons (Cre^mut: 46.8 ± 2.1 ms, n = 575/7; Cre: 83.7 ± 6.7 ms, n = 647 events/7 cells; p < 0.001). In contrast, the amplitude and kinetics of the mEPSCs were not affected by the lack of β-Nrxns (Figures 3D,F) because no differences were seen for mean amplitudes (Cre^mut: 45.7 ± 1.1 pA, Cre: 46.9 ± 1.2 pA; p = 0.4731), rise-time (Cre^mut: 0.49 ± 0.01 ms, Cre: 0.48 ± 0.01 ms; p = 0.57), and half-width (Cre^mut: 1.05 ± 0.03, Cre: 1.11 ± 0.03; p = 0.16). Together, the difference in mEPSC frequency without changes in maximal amplitudes and kinetics indicated that our lentivirus-mediated deletion in primary hippocampal cultures was effective, and the degree of deficiency was comparable to the results found in an earlier investigation of other neuronal subpopulations lacking β-Nrxns (Anderson et al., 2015).

In contrast to SVs, DCVs are continuously generated and filled with cargo at the Golgi apparatus and then trafficked from the soma to distal release sites (Bharat et al., 2017). Accordingly, reduced numbers of DCVs in presynaptic terminals as observed here (Figure 2) could be due to impaired generation. Such a defect was previously reported in Ca²⁺-dependent activator protein for secretion (CAPS1) KO mice and mechanistically associated with alterations in the trans-Golgi network (tGN) (Sadakata et al., 2013). To explore if the phenotype in β-Nrxn-deficient neurons is due to similar changes of the tGN, we compared in β-Nrxn cKO and control cultures the distribution of syntaxin 6 (Stx6) as a tGN marker protein and GM130 as a cis-Golgi protein, using the same markers as the earlier study (Sadakata et al., 2013). We observed robust Golgi staining for both markers with a characteristically clustered fluorescence pattern in the soma of Cre^mut-infected control and Cre-infected β-Nrxn cKO neurons with no apparent differences (Figures 4A₁,B₁,C₁,D₁). Green-fluorescent nuclei confirmed proper infection of neurons with lentivirus particles (Figures 4A₂,B₂,C₂,D₂; and merged images, Figures 4A₃,B₃,C₃,D₃). These data indicate that the diminished ChrgA fluorescence and the reduced number of DCVs in β-Nrxn-deficient neurons are likely not due to alterations in the morphology of the Golgi apparatus.

Our finding of the requirement of β-Nrxns for normal numbers of DCVs in presynaptic boutons was unexpected and striking because these molecules was not previously been linked to DCVs. In contrast, the role of β-Nrxns in synaptic vesicle release and their synaptogenic activity involving transsynaptic interactions has been intensely investigated (Reissner et al., 2013; Sudhof, 2017; Gomez et al., 2021). While cultured neurons have become a widely used tool to monitor DCV distribution and function in KO models (Sadakata et al., 2013; Dominguez et al., 2018; Persoon et al., 2018, 2019), we aimed to validate our DCV phenotype from primary neurons in a more native environment of brain tissue. For this in vivo analysis, we selected the highly structured cortex of the cerebellum because it allows the analysis of defined synapse populations (Xiao and Scheiffele, 2018) and it contains high expression of Nrxn-2β and Nrxn-3β in addition to Nrxn-1β, which dominates in the hippocampus (Anderson et al., 2015; Fuccillo et al., 2015; Schreiner et al., 2015). Specifically, we analyzed the DCV distribution in the molecular layer of the cerebellum from adult constitutive β-Nrxnβ-TKO mice in comparison to floxed β-Nrxn knock-in (β-KI) controlsβ-KI. Constitutive β-TKO mice were generated from floxed β-KI animals by breeding with a Cre recombinase deleter strain [B6.FVB-Tg (EIIa-cre) C5379Lmgd/J] to achieve germline deletion of all β-Nrxns prior to this study (Klatt et al., 2021).

In the mature cerebellum, the molecular layer predominantly contains two types of excitatory synapses, the majority originating from granule cells via parallel fibers and a minority from climbing fibers of the inferior olive neurons (Napper and Harvey, 1988; Xu-Friedman et al., 2001). We studied the ultrastructure of boutons terminating on postsynaptic spines of the proximal part of dendrites arising from β-KI (Figures 5A₁,A₂) and β-TKO (Figures 5B₁,B₂) Purkinje cells and observed a similar population of boutons with clustered SVs, postsynaptic density (arrowheads), frequent mitochondria, and engulfing astrocytic processes. In β-KI control boutons, usually a single DCV per presynaptic terminal was found at the periphery of SVs (arrows, Figures 5A₁,A₂). However, in β-Nrxn-deficient β-TKO cerebella, most boutons seemed to be devoid of DCVs (“empty synapses,” Figures 5B₁,B₂), corresponding to the shift from boutons with 1 or 2 DCVs to those without any DCV observed in cultured hippocampal neurons (Figure 2). Quantification confirmed this impression and revealed a 50% reduction of DCV numbers in presynapses (Figure 5C; β-KI control: 0.25 ± 0.03 DCVs/presynapse; β-TKO: 0.1 ± 0.02, p = 0.0002). As the area density of excitatory synapses, i.e., the number of synapses per area, was unchanged (Figure 5D; β-KI control: 77.8 ± 2.2 synapses/225 μm²; β-TKO: 83.6 ± 3.0, p = 0.17), our results indicate a specific reduction of the number of DCVs in the intact cerebellum, consistent with the quantitative data from primary hippocampal neurons (Figure 2D).

The release of neuropeptides such as BDNF or neuropeptide Y has not only been linked to the modulation of synaptic transmission but also to changes in synaptic plasticity and ultrastructure (McAllister et al., 1999; Poo, 2001; van den Pol, 2012). As most synapses analyzed here likely correspond to parallel fiber terminals (Napper and Harvey, 1988; Xu-Friedman et al., 2001), we tested whether the reduction of DCVs in β-Nrxn-deficient cerebellar granule cells (CGCs) has an effect on their synapse structure by measuring important parameters of their ultrastructure. However, no differences between constitutive β-Nrxn-deficient and β-KI control boutons could be determined for presynaptic area size (Figure 5E; β-KI control: 0.2 ± 0.02 μm², β-TKO: 0.22 ± 0.12, p = 0.3808), number of synaptic vesicle (Figure 5F; β-KI control: 43.7 ± 4.37 vesicles/synapse, β-TKO: 44.0 ± 3.96, p = 0.9530) or active zone length (Figure 5G; β-KI control: 351.2 ± 7.5 nm, β-TKO: 351.2 ± 7.5, p = 0.9530). These results suggest that the reduction of the number of presynaptic DCVs in β-Nrxn-deficient mice did not impair the general differentiation or maintenance of terminals, in line with earlier analyses of this mouse model (Anderson et al., 2015; Klatt et al., 2021).

To finally brace against the possibility that the reduction of DCVs in the β-TKO are due to cell death of entire CGCs or alterations of their Golgi apparatus, required for the generation of DCVs (Bharat et al., 2017), we studied the granule cell layer (GCL) at the light-microscopic level and the integrity of the Golgi apparatus at the ultrastructural level. The cerebella of β-KI control and β-TKO mice were gross anatomically normal, similar to the rest of the overall brain structures, as judged from the comparison of β-KI control and β-TKO mice (Figures 6A₁,A₂). To exclude more subtle changes of the neuronal population relevant to the phenotype of reduced DCVs in parallel fiber boutons, we studied cell numbers in semithin sections of the GCL from β-KI control and β-TKO cerebella (Figures 6B₁,B₂). Our quantification revealed no difference in the area density of CGCs (Figure 6C, data normalized to β-KI control: 1.0 ± 0.03 cells/mm², β-TKO: 1.03 ± 0.04, p = 0.5490). To exclude ultrastructural alterations of the Golgi apparatus, we finally compared critical parameters of Golgi morphology on random cross-sections of CGCs from β-KI controls (Figures 7A₁,A₂) to β-TKO neurons (Figures 7B₁,B₂) by electron microscopy. No major differences of the extent or organization of cis- or trans-Golgi regions or dilatation of cisternae were observed, suggesting an intact morphology of the Golgi apparatus in the absence of β-Nrxns. Specifically, we found an unchanged number of Golgi cisternae (Figure 7C; β-KI control: 5.03 ± 0.18, β-TKO: 4.65 ± 0.16, p = 0.1201) and similar cisternae width (Figure 7D; β-KI control: 36.72 ± 1.15 nm, β-TKO: 39.6 ± 1.71, p = 0.1673). As expected, these dimensions are consistent with the earlier reports that determined, for example, an average width of cisternae of about 38 nm (Emperador-Melero et al., 2018). Moreover, DCVs with a diameter between 68 and 79 nm, i.e., within the determined range for presynaptic DCVs as shown above, could be identified in the vicinity of Golgi apparatus in both β-KI controls (arrows, Figures 7A₁,A₂) and β-TKO neurons (arrows, Figures 7B₁,B₂). However, it was not possible to reliably quantify the number of DCVs after generation at the tGN by electron microscopy because the probability of finding DCVs next to a Golgi apparatus on cross-sections was extremely low, limiting the sample size needed for a thorough investigation. Thus, the analysis of the maturation and trafficking of DCVs in β-Nrxn-deficient neurons will have to be done by entirely different methods. In conclusion, our results from the cerebellar tissue confirm the data from cultured hippocampal neurons, and, together, unequivocally demonstrate that the deletion of β-Nrxns specifically alters the distribution of presynaptic DCVs in neurons.

This study is the first that links a member of the most widely investigated synaptic cell adhesion molecules (β-Nrxn) to DCVs in neurons. We mostly focused our analysis on DCVs in presynaptic boutons of excitatory neurons because (i) β-Nrxns have a known function in the differentiation of these presynaptic terminals and their Ca²⁺-dependent release (Dean et al., 2003; Anderson et al., 2015; Born et al., 2015; Klatt et al., 2021) and (ii) DCVs are preferentially located and released at high rates at synapses compared to extrasynaptic sites or dendrites in hippocampal neurons (van de Bospoort et al., 2012; Tao et al., 2018). Our main conclusion is supported by several lines of evidence that the normal distribution of DCVs in presynaptic boutons depends on β-Nrxns.

First, the numbers of presynaptic DCVs upon the deletion of β-Nrxns are clearly reduced by about 50% from control values. These control values (wild type or Cre^mut) are similar to other studies. The majority, or 60%, of wild-type hippocampal synapses have no/zero DCVs (Dominguez et al., 2018), remarkably similar to the 59% of synapses without DCV determined in our study (Figure 2D). Only a minority of terminals contain one DCV, for example, 25% of synapses as determined by Dominguez et al. (2018), or 24% in our study (Figure 2D); and only a few boutons have two or more DCVs. These numbers are also consistent with additional studies that reported between 45 and 70% of synapses without any DCV (van de Bospoort et al., 2012; Persoon et al., 2018). Surprisingly, a large fraction of wild-type boutons without DCVs might indicate that the capacity of releasing neuromodulators is strictly controlled in hippocampal neurons. Based on this consideration, the phenotype we observed in β-Nrxn KO, i.e., shifting a subpopulation of boutons from 1 to 0 DCV (Figure 2D), might represent a bold change in the function of the neuronal network.

Second, the phenotype of reduced numbers of presynaptic DCVs upon the deletion of β-Nrxns is specific as it can be found in two different neuronal populations that normally contain high levels of β-Nrxn, excitatory hippocampal (Figures 1, 2) and cerebellar (Figure 5) neurons, and their synaptic boutons. Neurons from these two different brain regions were chosen as Nrxn-1β is the prominent isoform in the hippocampus with lower levels of Nrxn-2β and Nrxn-3β, whereas in the cerebellum, there were higher levels of Nrxn-2β and Nrxn-3β in addition to Nrxn-1β (Anderson et al., 2015; Fuccillo et al., 2015; Schreiner et al., 2015). The phenotype also appears specific because the deletion of β-Nrxns neither affected the distribution of a presynaptic marker protein, Bassoon (Figure 3), nor the number of synapses (Figure 5D) or ultrastructural parameters such as presynaptic bouton size (Figure 5E), number of SVs (Figure 5F), and the length of active zone (Figure 5G), confirming earlier investigations of β-Nrxn-deficient neurons (Anderson et al., 2015; Klatt et al., 2021).

Third, the phenotype is robust because it can be observed by independent experimental methods. Reduced levels of ChrgA, an intravesicular matrix protein of DCVs (Bharat et al., 2017; Dominguez et al., 2018), were shown by immunofluorescence intensity measurements (Figure 1), and the reduced numbers of ultrastructurally identified presynaptic DCVs were demonstrated using electron microscopy (Figures 2, 5). In fact, the 50% reduction seen here upon the deletion of β-Nrxns is higher than the 30% reduction of DCVs reported in an investigation of CAPS1 KO, which established a new function for CAPS1 (Sadakata et al., 2013). ChrgA and its closely related isoform Chromogranin B, in turn, have been widely used as marker proteins of DCVs in brain neurons (Machado et al., 2010; Bartolomucci et al., 2011). Interestingly, the deletion of these molecules did not compromise the ability of DCVs to fuse with the plasma membrane upon bursts of action potentials (Dominguez et al., 2018). Moreover, the molecules are not essential for DCV distribution to synapses because the deletion of both ChrgA and B in mice revealed a normal number of synaptic DCVs (Dominguez et al., 2018). These results indicate that the reduction of DCVs reported in our study is a direct consequence of deleting β-Nrxns and not a secondary effect of downregulating the expression or loading of DCVs with ChrgA.

Fourth, the phenotype is reliable because it was present in similar magnitude in cultured primary neurons conditionally deleted for β-Nrxns by Cre recombinase-expressing lentivirus (Figure 2) and in the brain tissue from constitutive β-TKO β-TKO (Figure 5). Thus, DCV numbers in presynaptic boutons drop by 50% when β-Nrxns are missing, as shown by quantitative electron microscopy. This is a surprising finding because previous studies have solely identified a role of β-Nrxns in the release of small SVs in cKO neurons, including reduced presynaptic Ca²⁺ influx and glutamate release, diminished spontaneous mini frequencies, and impaired endocannabinoid signaling (Anderson et al., 2015; Klatt et al., 2021). The clear effect of β-Nrxns on DCV distribution has thus far escaped attention but is a remarkable finding due to the low amount of β-Nrxn protein in the brain, estimated for Nrxn1β at ≈0.9 fmol/μg compared to N-cadherin (≈20 fmol/μg) or PSD-95 (≈54 fmol/μg) (Schreiner et al., 2015). The low abundance translates into very low copy numbers at individual terminals, estimated at 7–16 molecules per synapse for all β-Nrxn isoforms combined. Together with our recent observation that a large extrasynaptic axonal pool of these elusive molecules exists (Klatt et al., 2021), the numbers imply that very few copies of a β-Nrxn variant are present at individual synapses at a given time point. However, β-Nrxns are also highly mobile molecules and only transiently confined at boutons (Klatt et al., 2021), opening the possibility that the dynamic behavior of synaptic and axonal β-Nrxns is involved in the regulation of DCVs.

We did not obtain any evidence that ChrgA immunofluorescence clusters were differently distributed in β-Nrxn-deficient primary hippocampal neurons compared to controls (Figures 1, 3). As the pattern of fluorescence was identical in β-Nrxn-deficient (Cre) vs. Control (Cre^mut) neurons (Figures 1D₁,D₂), an ectopic accumulation seems unlikely. In support, we also did not see the accumulation of DCVs in axons outside synapses by EM (Figure 2). In fact, DCVs are very sparsely distributed in “inter-bouton parts of axons” both in control and β-Nrxn-deficient neurons as seen in our EM samples (Figure 2) and reflected by the lower ChrgA fluorescence intensity (Figure 1). Thus, the deletion of β-Nrxns and the concomitant reduction of spontaneous and evoked synaptic release as shown here (Figures 3C–F) and elsewhere (Anderson et al., 2015; Klatt et al., 2021) leads to a reduction of presynaptic DCVs without ectopic accumulation. This is a different scenario, for example, compared to chronic inactivity by the application of TTX, which caused the accumulation of presynaptic DCVs in a recent study of cultured hippocampal neurons by cryo-EM tomography (Tao et al., 2018).

We also did not observe an accumulation of DCVs around the Golgi apparatus in β-TKO neurons where these organelles are loaded with cargo and budded from tGN cisternae (Figure 7). The unchanged localization of Golgi markers (Figure 4), identical cell density (Figure 6), and intact morphology of the Golgi apparatus (Figure 7) rather suggest that biogenesis of DCVs in general does not depend on β-Nrxns. For quantification of putative changes of the Golgi ultrastructure, we measured the number of Golgi cisternae and diameter of cisternae width (Figures 7C,D), using the same parameters that led to the discovery of a trafficking phenotype of DCVs in Vti1-deficient neurons (Emperador-Melero et al., 2018). The undisturbed Golgi organization accompanying the phenotype of reduced numbers of presynaptic DCVs in the absence of β-Nrxns is striking because the only other study reporting a comparable defect of DCV numbers actually depended on altered Golgi structure and impaired trafficking (Sadakata et al., 2013). In this study, the analysis of cKO mice of the CAPS1 revealed a 30% reduced immunoreactivity for the DCV marker secretogranin II, another member of the granin protein family (Bartolomucci et al., 2011), and a 50 or 70% reduction of the number of presynaptic DCVs in the cerebellum or hippocampus, respectively (Sadakata et al., 2013). While these data are similar to our findings here, CAPS1-deficient mice additionally showed altered expression of the tGN marker protein syntaxin6, dilated tGN cisternae, and reduced numbers of SVs (Sadakata et al., 2013), all of which were absent in our β-Nrxn-deficient neurons. Moreover, other studies of CAPS1-deficient mammalian chromaffin cells and neurons have pointed to an essential role for survival and concluded that it functions strongly in the priming or fusion of DCVs (Liu et al., 2010; Farina et al., 2015; Crummy et al., 2019). In any case, a comparison of these partially overlapping, partially distinct phenotypes indicates that CAPS1 primarily plays a role in the generation, trafficking, and fusion of DCVs, whereas β-Nrxns appear to have a more distinct, regulatory role.

It has been shown that DCVs are highly mobile organelles that undergo long-range translocation and can be captured at the level of individual boutons (Wong et al., 2012), a process that involves phosphorylation of synaptotagmin-4 at presynapses (Bharat et al., 2017) and liprin-α at postsynaptic spines (Stucchi et al., 2018). In fact, the capture of DCVs at active terminals is an important process that might complement the statistically significant presynaptic enrichment of these organelles as shown by Poisson statistics based on randomness (Robinson et al., 2016). Relevant to a presynaptic phenotype, it was demonstrated that the capture of DCVs in hippocampal neurons requires destabilization of DCV/Syt4/KIF1a complexes and increased neuronal activity (Bharat et al., 2017). β-Nrxns themselves depend on microtubule KIF1a-dependent trafficking (Neupert et al., 2015), and synaptic release is reduced in β-Nrxn-deficient neurons as shown here (Figures 3C–F) and earlier (Anderson et al., 2015; Klatt et al., 2021). Consequently, the inability to redirect DCVs from the circular trafficking pool to active boutons (Wong et al., 2012; Bharat et al., 2017) could explain the observations described in our study. Alternatively, the reduced number of DCVs could imply that less active zone material is delivered to active synapses (Tao et al., 2018), reducing the release probability and the readily releasable pool of vesicles, possibly reflected by the reduced mEPSC frequency or elevated IEI (Figure 3E). However, independent of the mechanism that causes the reduction of presynaptic DCVs in β-Nrxn-deficient neurons, it can be predicted that shifting a population of synapses from 1 to 0 DCV (Figure 2D) means a reduction of the neuromodulatory prowess, which in turn might represent a bold change in the function of the neuronal network. Thus, future studies will have to dissect the functional role of β-Nrxns for neuropeptide or neurohormone release.