Formation and Maintenance of Functional Spines in the Absence of Presynaptic Glutamate Release.

Dendritic spines are the major transmitter reception compartments of glutamatergic synapses in most principal neurons of the mammalian brain and play a key role in the function of nerve cell circuits. The formation of functional spine synapses is thought to be critically dependent on presynaptic glutamatergic signaling. By analyzing CA1 pyramidal neurons in mutant hippocampal slice cultures that are essentially devoid of presynaptic transmitter release, we demonstrate that the formation and maintenance of dendrites and functional spines are independent of synaptic glutamate release.

Introduction

Glutamate is the major excitatory neurotransmitter at synapses in the mammalian brain. Most principal neurons receive glutamatergic presynaptic inputs on spines, short protrusions of neuronal dendrites that act as the major glutamate reception compartments. Accordingly, spines play a key role in neuronal signaling, and the dynamic modulation of their molecular composition, structure, function, and number is at the basis of neuronal circuit changes that mediate, for instance, learning and memory processes (Nishiyama and Yasuda, 2015, Rochefort and Konnerth, 2012). Multiple processes trigger and regulate spine formation, including, for example, developmental regulatory processes such as signaling by growth factors or guidance cues (Bennett and Lagopoulos, 2014, Chen et al., 2008, Klein, 2009, Shen and Cowan, 2010, Stamatakou and Salinas, 2014, Thompson, 2003) and synapse induction and maturation mediated by synaptic cell adhesion proteins and their intracellular interactors (Brose, 2009, Scheiffele, 2003, Südhof, 2008, Tada and Sheng, 2006, Washbourne et al., 2004). In addition, glutamate release itself, along with the consequent activation of glutamate receptors (GluRs) on vicinal postsynaptic dendrites, has been strongly implicated in spine formation and maintenance (Collin et al., 1997, Craig et al., 1994, Galante et al., 2000, Gomperts et al., 2000, Harms and Craig, 2005, Kim and Tsien, 2008, Kwon and Sabatini, 2011, McAllister et al., 1996, McKinney, 2010, Mitra et al., 2011, Richards et al., 2005, Thiagarajan et al., 2005, Turrigiano, 2008). The corresponding data led to the notion that glutamatergic signaling is essential for normal spine formation not only in phases of activity-dependent synaptic refinement later in brain development or during learning and memory processes but also in the initial establishment of neuronal connectivity and its maintenance (Lai and Ip, 2013, Saneyoshi et al., 2010, Yoshihara et al., 2009, Yuste and Bonhoeffer, 2004).

A largely ignored challenge of the notion that glutamatergic signaling plays a key role in spine formation was put forward in the context of studies on mouse mutants lacking Munc18-1 (M18 knockout [M18-KO]) or Munc13-1/Munc13-2 (M13 double knockout [M13 DKO]). In these mutants, spontaneous and action potential-evoked presynaptic transmitter release is abolished, but synapses in mutant cultured neurons and intact brains develop rather normally (Verhage et al., 2000, Varoqueaux et al., 2002). Unfortunately, M18-KOs and M13-DKOs die at birth, and M18-KO neurons degenerate during brain development, so activity-dependent spine formation in situ could not be properly assessed. In a subsequent study, the consequences of the combined genetic elimination of all relevant ionotropic GluRs (iGluRs) in hippocampal CA1 pyramidal neurons (CA1-PNs) were analyzed using conditional KO mice. The complete loss of fast postsynaptic glutamatergic signaling was found to leave dendrite, synapse, and spine development largely unaffected (Lu et al., 2013). It was concluded that the initial development of dendrites, synapses, and spines is independent of postsynaptic glutamatergic signaling and, instead, controlled by cell-intrinsic genetic programs. However, a contribution of glutamatergic signaling to spine formation via metabotropic GluRs (Ronesi and Huber, 2008) or by transient expression of iGluRs after their conditional genetic deletion could not be excluded.

In essence, the available data do not allow full assessment of the role of glutamatergic signaling in spine formation and maintenance. In a seminal review, Yuste and Bonhoeffer (2004) indicated a solution to this problem, suggesting “further analysis—for instance, by culturing tissue from newborn mice—[of] other transmitter-release-deficient mice [i.e. ones that do not show neuronal degeneration like the Munc18-1 mutants] that survive into the period of spinogenesis” and stated that this would “likely … be pertinent to the understanding of the role of activity in spine formation.” We did this by analyzing hippocampal organotypic slice cultures from M13-DKOs in which all presynaptic glutamate and γ-aminobutyric acid (GABA) release is essentially abolished. We show that CA1-PNs in such slices are hardly ever synaptically activated but develop and maintain normal dendrites and normal numbers and types of functional spines that recruit iGluRs.

Thus, the initial formation of dendrites and functional spines and their general maintenance in hippocampal CA1-PNs are independent of presynaptic glutamate release. The underlying spinogenic processes are likely based on a combination of cell-intrinsic genetic programs and the concerted action of guidance cues, trophic factors, and neuronal and synaptic adhesion systems. They appear to generate a basic circuit connectivity upon which activity-dependent spinogenic processes (e.g., Engert and Bonhoeffer, 1999, Kwon and Sabatini, 2011) act to establish fully functional neuronal networks and mediate circuit plasticity.

Results

CA1-PNs in M13-DKOs as a Model to Study Activity-Dependent Dendrite and Spine Synapse Formation and Maintenance

We chose CA1-PNs in organotypic hippocampal slice cultures from M13-DKOs and littermate controls (CTRs) to study the role of transmitter release and synaptic signaling in dendritogenesis and spinogenesis. M13-DKOs show normal brain development at the time of their perinatal death, and cultured M13-DKO CA1-PNs show no changes in neuronal development and synaptogenesis, whereas spontaneous and evoked synaptic glutamate and GABA release is eliminated (Varoqueaux et al., 2002). Further, organotypic hippocampal slice cultures of M13-DKOs show no traces of developmental aberrations (Imig et al., 2014). In contrast, transmitter release in Munc13-2 KOs is essentially normal (Varoqueaux et al., 2002, Breustedt et al., 2010).

To assess functional synaptic defects in M13-DKOs, we performed whole-cell voltage-clamp recordings of CA1-PNs at weeks in vitro (WIV) 1, 2, and 3 and filled them with biocytin for subsequent morphometric analysis. CTR CA1-PNs showed a steady increase in the frequency of spontaneous postsynaptic currents (sPSCs) with time in vitro (Figures 1A and 1B; Table S1), whereas M13-DKO CA1-PNs showed extremely low sPSC frequencies at all times tested and only a small increase over time in vitro (Figures 1A–1C; Table S1). CTR sPSC amplitudes increased slightly with time in vitro (Figure 1C; Table S1), likely because the increasing frequency caused more overlapping sPSCs, whereas sPSC amplitudes in M13-DKO cells were similar to CTR levels early in culture and did not increase over time (Figure 1C; Table S1). Corresponding observations were made in a different set of CA1-PNs for miniature excitatory PSCs (mEPSCs) and miniature inhibitory PSCs (mIPSCs) (Figures 1D–1I; Table S1).

Figure 1

Functional Synaptic Inputs to M13-DKO and CTR Neurons

(A) Example traces of sPSCs in M13-DKO (red) and CTR neurons (black).

(B and C) sPSC frequencies (B) and amplitudes (C) in M13-DKO (red) and CTR neurons (black).

(D) Examples traces of mEPSCs in M13-DKO (red) and CTR neurons (black).

(E and F) mEPSC frequencies (E) and amplitudes (F) in M13-DKO (red) and CTR neurons (black).

(G) Examples traces of mIPSCs in M13-DKO (red) and CTR neurons (black).

(H and I) mIPSC frequencies (H) and amplitudes (I) in M13-DKO (red) and CTR neurons (black).

1/2/3 W, WIV1/2/3. Data are shown as mean ± SEM. n Values of cells are given in/above bars. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001 (Student’s t test). See also Table S1.

Our data show that M13-DKO CA1-PNs receive hardly any functional glutamatergic or GABAergic synaptic inputs at WIV1–3.

Dendrite Formation and Maintenance in M13-DKO CA1-PNs

To assess the differentiation of CA1-PNs in organotypic slices from M13-DKOs, we used a Sholl analysis (Sholl, 1953) to determine the overall morphology of neurons that had been analyzed electrophysiologically, filled with biocytin, and stained with Alexa Fluor 555-labeled streptavidin.

We found the overall morphology of M13-DKO and CTR CA1-PNs to be remarkably similar (Figures 2A and 2B), with peak dendrite complexities of apical and basal dendrites at radial distances from the soma of 50–100 μm and 50 μm, respectively, irrespective of the developmental stage and genotype (Figures 2C–2E). The dendrite complexities of M13-DKO and CTR CA1-PNs were identical at WIV3 (Figure 2E). At earlier stages, subtle and statistically mostly insignificant changes were observed in M13-DKO CA1-PNs, including slightly more complex basal dendrites at WIV1, possibly “in search” of release-active inputs, and slightly more complex but shorter apical dendrites at WIV2 (Figures 2C and 2D).

Figure 2

Dendrite Development in M13-DKO and CTR Neurons

(A and B) Confocal fluorescence image of representative M13-DKO (A) and CTR neurons (B) at WIV3, after filling with biocytin and subsequent staining with Alexa Fluor 555-labeled streptavidin. The concentric circles in (A) indicate spheres at 50 μm distance intervals relative to the soma (dashed line, apical dendrite; dotted line, basal dendrite). The white boxes indicate the dendrite subsections shown in Figure 3. Scale bars, 50 μm.

(C–E) Sholl analysis of dendrite complexity in M13-DKO and CTR neurons at WIV1 (C), WIV2 (D), and WIV3 (E). Linear Sholl plots for the number of intersections of dendrites (mean ± SEM) with spheres at 50-μm distance intervals relative to the soma are shown. Basal and apical dendrites were analyzed separately, and the data for basal dendrite intersections are plotted for negative Sholl radius values, with the soma center at 0.

Data are shown as mean ± SEM. n Values of cells are given in the insets. ∗p < 0.05 (Student’s t test).

Our data show that the development and maintenance of the dendritic tree of CA1-PNs are not affected by the almost complete absence of functional glutamatergic and GABAergic inputs.

Spine Synapse Formation and Maintenance in M13-DKO CA1-PNs

To identify, classify, and count dendritic spines in labeled CA1-PNs (Figures 2A and 2B), we analyzed images morphometrically using NeuronStudio. The fluorescence of several entire CA1-PNs was recorded at high resolution (voxel size, 0.05 × 0.05 × 0.126 μm), and images were filtered with a 3D Gaussian filter (σ = 0.7 pixels). We then rendered entire dendritic trees and assessed spines along the dendrites of each entire neuron based on standard parameters for the distinction of thin, stubby, and mushroom-type spines (Rodriguez et al., 2008). In addition, we detected long, thin extrusions with a length of >1 μm, which are typically not recognized by NeuronStudio, by visual inspection and classified them as filopodia. Visual inspection was also used to detect false “spine calls.” This systematic approach was chosen to account for possible changes in spine distribution along dendrites.

The overall distributions of values for spine length, volume, head diameter, and neck diameter were not different between M13-DKO and CTR CA1-PNs. Further, inspection of representative dendrite sections (i.e., first-order branches from main dendrites) revealed a remarkable similarity of spine types, density, and distribution between M13-DKO and CTR CA1-PNs (Figures 3A–3F). Corresponding quantitative data also revealed a striking similarity between M13-DKO and CTR CA1-PNs regarding the numbers and types of spines and their dendritic distribution, with peak spine densities along apical and basal dendrites at 100 μm distance from the soma (Figures 3G–3R). Only very subtle and statistically mostly insignificant differences were observed in M13-DKO CA1-PNs, including slightly lower numbers of filopodia and slightly more spines at WIV1, slightly less thin and stubby spines in basal dendrites at WIV3, slightly more mushroom-type spines at WIV3, and slightly more filopodia in basal dendrites at WIV3.

Figure 3

Spine Number and Distribution in M13-DKO and CTR Neurons

(A and B) Representative dendrite segments (first-order branches from main dendrites; see white boxes in Figures 2A and 2B) in an M13-DKO (A) and a CTR (B) neuron at WIV3, after filling with biocytin and staining with Alexa Fluor 555-labeled streptavidin. Identified spines were classified as thin (upright rectangles), stubby (squares), and mushroom-type (triangles). Filopodia, which are rarely seen at this developmental stage, were not found in the dendrite sections shown. White boxes indicate the regions shown at higher magnification in (C) and (D) and (E) and (F), respectively. Scale bars, 5 μm.

(C and E) Angular projections of spines in three different angles within ± 60°, in the regions indicated by the white boxes in (A) and (B), respectively. Scale bars, 2 μm.

(D and F) Schematic drawings for (C) and (E), respectively.

(G–R) Quantitative analysis of the numbers and distributions along dendrites of dendritic protrusions (i.e., thin, stubby, and mushroom-type spines and filopodia) in M13-DKO and CTR neurons at WIV1, WIV2, and WIV3. Dendritic spines were counted for each entire neuron, separating data for basal and apical dendrites (Figures 3A and 3B). Each data point represents the mean ± SEM for a 50 μm bin along the dendrite, measured from the soma center. Positive values on the abscissa represent data from apical dendrites, and negative values on the abscissa represent data from basal dendrites. Shown are stubby spines (G–I), mushroom type spines (J–L), thin spines (M–O), and filopodia (P–R) at WIV1, WIV2, and WIV3.

Data are shown as mean ± SEM. n Values of cells are given in the insets. ∗p < 0.05 (Student’s t test). See also Figure S1.

In a control experiment, we morphologically assessed the presynaptic innervation of spines in M13-DKO and CTR CA1-PNs at WIV1–2. Biocytin-filled CA1-PNs labeled with Alexa Fluor 555 streptavidin were co-stained for vesicular glutamate transporter 1 (VGluT1) as a maker of glutamatergic presynaptic boutons, and the apparent overlap of streptavidin and VGluT1 staining in representative dendrite subsections within first-order branches from main dendrites was analyzed as an indicator of close proximity. The fractions of innervated spines, as assessed morphologically, were very similar between M13-DKO (84% ± 1%, n = 4 cells with 267 spines analyzed) and CTR CA1-PNs (83% ± 3%, n = 8 cells with 318 spines analyzed) (Figure S1). Filopodia did not overlap with presynaptic staining, and the degree of presynaptic innervation, as assessed morphologically, did not depend on spine type in either M13-DKO or CTR CA1-PNs.

Our data show that, in CA1-PNs, the numbers, densities, and types of spines and their morphologically assessed presynaptic innervation do not depend on functional presynaptic glutamatergic inputs.

Functionality of Spines in M13-DKO CA1 Pyramidal Cells

We used glutamate uncaging and whole-cell patch-clamp recording (Kwon and Sabatini, 2011) to test directly whether M13-DKO CA1-PNs in organotypic culture are able to generate spines in an activity-dependent manner and whether the spines formed by M13-DKO CA1-PNs contain a functional postsynaptic receptor apparatus.

Repetitive high-frequency glutamate uncaging (HFU; 60 times at 10 Hz, 2-ms duration) induced the formation of new spines in td-Tomato-expressing M13-DKO and CTR CA1-PNs at WIV1 with a success rate of ∼0.60 (Figures 4A and 4B). As expected (Kwon and Sabatini, 2011), with time in culture, the success rate of spine induction by HFU decreased in parallel in M13-DKO and CTR cells to ∼0.22 and ∼0.05 at WIV2 and WIV3, respectively (Figures 4A and 4B), and the efficiency of spine induction showed a similar dependency on the spatial distance from the HFU stimulus in both genotypes (Figure 4C).

Figure 4

Activity-Dependent Spinogenesis and Spine Responsiveness in M13-DKO and CTR Neurons

(A) Images of dendrites from M13-DKO (top) and CTR (bottom) CA1-PNs expressing tdTomato and exposed to HFU (yellow crosses, 60 pulses at 10 Hz) at WIV1. Yellow arrowheads indicate de novo spine formation following HFU. Scale bars, 1 μm.

(B) Success rate of de novo spine formation by HFU at WIV1, WIV2, or WIV3 in M13-DKO (red bars; WIV1, n = 19 trials, 8 cells; WIV2, n = 20 trials, 9 cells; WIV3, n = 21 trials, 10 cells) and CTR CA1-PNs (black bars; WIV1, n = 18 trials, 8 cells; WIV2, n = 25 trials, 10 cells; WIV3, n = 22 trials, 9 cells).

(C) Summary plots of the distance-dependent success rate of de novo spine formation from M13-DKO (red) and CTR CA1-PNs (black).

(D) Two-photon image of a dendritic segment from a CA1-PN recorded in whole-cell voltage-clamp mode in organotypic slice culture. A target spine was exposed to glutamate uncaging test pulses (green cross, eight to ten trials at 0.1 Hz). Scale bars, 10 μm (left) and 1 μm (right).

(E) uEPSCs evoked by glutamate uncaging from M13-DKO (red traces) and CTR CA1-PNs (black traces) measured in whole-cell voltage-clamp mode at −65 mV for AMPAR-uEPSCs (filled circles) and +40 mV for NMDAR-uEPSCs (open circles). The green dotted line (70 ms after the uncaging pulse) indicates measuring points for NMDAR-uEPSC amplitudes.

(F and G) The amplitude of AMPAR-uEPSCs (F) and NMDAR-uEPSCs (G) plotted against relative spine size from M13-DKO (red circles, n = 76 spines, 21 cells) and CTR CA1-PNs (black circles, n = 66 spines, 22 cells).

(H) Left: mean AMPAR-uEPSC amplitudes plotted against the mean estimated size of small, medium, medium-large, and large spines for M13-DKO (red circles) and CTR CA1-PNs (black circles). Right: summary graph of AMPAR-uEPSC amplitudes from M13-DKO (red bars; small spines, n = 22; medium spines, n = 17; medium-large spines, n = 20; large spines, n = 17; 21 cells) and CTR CA1-PNs (black bars; small spines, n = 27; medium spines, n = 20; medium-large spines, n = 12; large spines, n = 7; 22 cells).

(I) Summary graph of NMDAR-uEPSC amplitudes from M13-DKO (red bar; n = 76 spines, 22 cells) and CTR CA1-PNs (black bar; n = 66 spines, 21 cells).

Where relevant, data are shown as mean ± SEM. ∗p < 0.05, Student’s t test. See also Table S2.

We next performed whole-cell voltage-clamp recordings to assess the functional properties of individual spines. Simultaneous two-photon imaging of green fluorescence (Alexa Fluor 488) was used to visualize spines and estimate spine sizes. In both M13-DKO and CTR CA1-PNs, glutamate uncaging near visualized spines effectively elicited uncaging-evoked α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor- and NMDA receptor-mediated EPSCs (AMPAR-uEPSCs, NMDAR-uEPSCs), which were measured by holding cells at −65 mV and +40 mV, respectively (Figures 4D and 4E). In both genotypes, the amplitude of AMPAR-uEPSCs, but not of NMDAR-uEPSCs, correlated strongly with spine size, as expected (Sobczyk et al., 2005; Figures 4F and 4G). We finally examined the functional synaptic properties of individual spines of different sizes in M13-DKO and CTR CA1-PNs. Spines of all sizes exhibited a significant decrease in the amplitudes of AMPAR-uEPSCs in M13-DKO CA1-PNs compared with CTR cells (Figure 4H; Table S2). In addition, the amplitudes of NMDAR-uEPSCs were significantly reduced in M13-DKO CA1-PNs (Figure 4I; Table S2).

Our data show that the protein machinery required for activity-dependent spinogenesis is intact in M13-DKO CA1-PNs and that spines formed in the absence of presynaptic glutamate release recruit AMPA and NMDA receptors and are functional.

Discussion

To study the dependence of dendritogenesis and spinogenesis on presynaptic transmitter release, we analyzed CA1-PNs in organotypic slices of M13-DKOs because the loss of Munc13 proteins from cultured CA1-PNs, which express only Munc13-1 and Munc13-2, causes a complete loss of spontaneous and evoked transmitter release without changes in neuronal morphology or signs of degeneration (Varoqueaux et al., 2002), and electron microscopic analyses of M13-DKO CA1-PNs in organotypic slices revealed the presence of spine synapses with only very subtle morphological changes in presynapses (Imig et al., 2014).

Our data show that M13-DKO CA1-PNs receive extremely few synaptic glutamatergic or GABAergic signals (Figure 1). It cannot be excluded that the remaining PSCs originate from very few presynaptic termini containing the third Munc13 isoform, Munc13-3, although Munc13-3 is not detectable in CA1 synapses with currently available tools (Netrakanti et al., 2015), or from Munc13-independent synaptic release, which can occur on very rare occasions (Varoqueaux et al., 2002). Alternatively, the glutamatergic signals arriving at M13-DKO CA1-PNs might originate from constitutive neuronal or even glial secretion and not from typical synaptic active zone inputs on spines, although this seems unlikely in view of the typical “synaptic-type” time course of the corresponding EPSCs. Assuming that any remaining active synaptic inputs to M13-DKO CA1-PNs do not change much during development, all of these notions are compatible with our finding that mEPSC frequencies in M13-DKO CA1-PNs remain in the range of 0.07–0.09 Hz independent of cell maturity (Figure 1), although the total number of spines in CTR and M13-DKO CA1-PNs, most of which are presynaptically innervated (Figure S1), increases relative to the spine number at WIV1 by ∼2-fold at WIV2 and by ∼3-fold at WIV3 (Figures 3G–3R). Irrespective of their nature, the frequency of the remaining functional inputs to M13-DKO CA1-PNs is too low to induce activity-dependent spinogenesis. For instance, 40 uncaging pulses of 0.5-ms length at 0.5 Hz and an Mg2+-free extracellular medium are required to induce a single spine with a success rate of 0.14 in cortical slices from 8- to 12-day-old mice (Kwon and Sabatini, 2011). This frequency is already ∼3-fold higher than the sPSC frequency we observed in entire neurons at WIV1, which have ∼500–900 spines (Figures 3G–3R). More strikingly, in medium including Mg2+, as was used in our experiments, electrical stimulation with two 100-pulse trains at 100 Hz is required for new spine growth, whereas stimulation at 10 Hz is insufficient (Kwon and Sabatini, 2011).

We thus conclude that hippocampal organotypic slices of M13-DKOs are an excellent model to study the role of presynaptic glutamate release in the formation and maintenance of dendrites and spines. The chronic and pervasive nature of the block of synaptic transmitter release in this model is a key advantage compared with strategies involving acute or transient interference using toxins, whose effects likely involve homeostatic alterations.

Our data further show that the formation and maintenance of functional spines in CA1-PNs are essentially independent of synaptic glutamatergic input signaling. These findings corroborate an earlier study showing that the elimination of iGluRs from CA1-PNs has also little effect on spinogenesis (Lu et al., 2013) and are in nice agreement with a parallel and independent study in which the effects of a loss of glutamate release in excitatory forebrain neurons because of the conditional expression of tetanus toxin were analyzed (Sando et al., 2017 [this issue of Neuron]). It thus appears that a developmental program independent of glutamatergic signaling “lays out a carpet” of functional spine synapses along the dendrites of CA1-PNs and, thereby, generates the basic circuit connectivity. Subsequently, activity-dependent processes (e.g., Engert and Bonhoeffer, 1999, Kwon and Sabatini, 2011) can act, by changing spine morphology, function, turnover, or number, to establish a fully functional circuit and to mediate circuit plasticity. The subtle perturbations in M13-DKO CA1-PNs regarding dendrite morphology (Figure 2) and spine number and distribution (Figures 3G–3R) may be due to such activity-dependent processes. Moreover, M13-DKO spines show an ∼30% reduction in uEPSCs when stimulated by glutamate uncaging (Figures 4H and 4I), whereas mEPSCs, which may be triggered synaptically, have similar amplitudes in M13-DKO and CTR CA1-PNs (Figure 1F). These findings are compatible with the notion that AMPA and NMDA receptor recruitment to spines is partially but substantially (∼30%) dependent on presynaptic glutamatergic signaling.

Based on our data, the mechanisms that control spinogenesis and spine synapse maintenance in the absence of synaptic glutamatergic signaling remain unknown. It is unlikely that ambient glutamate released from glia cells plays a significant role in this context because the genetic elimination of all relevant iGluRs in CA1-PNs leaves dendrite, synapse, and spine development largely unaffected (Lu et al., 2013). An ultimate assessment of the role of glial glutamatergic signaling in spine formation and maintenance will require a targeted elimination of glutamate release from glia cells. This is currently impossible, mainly because the mechanisms of glial glutamate release are still unclear and, hence, controversially discussed, so corresponding genetic interference strategies, which would have to be employed on top of the M13-DKO studied here, cannot be properly designed.

A more likely mechanistic explanation is that the normal spinogenesis and spine synapse maintenance in the absence of synaptic glutamatergic signaling are caused by a combination of cell-autonomous processes that are dictated by genetic programs, as proposed based on the analysis of CA1-PNs lacking iGluRs (Lu et al., 2013), and signaling by guidance cues, trophic factors, and neuronal and synaptic adhesion systems. Indeed, constitutive transport pathways involved in membrane, protein, and organelle trafficking appear to be normal in M13-DKOs, so key presynaptic and postsynaptic proteins, including scaffold proteins, adhesion proteins, and cell surface receptors, are properly targeted (Augustin et al., 1999, Varoqueaux et al., 2002). Further, secretion from large dense-core vesicles, which contain growth factors and neuropeptides, is partially intact in M13-DKOs (van de Bospoort et al., 2012). In view of these characteristics of M13-DKOs, we propose that the genetic programs that control initial dendritogenesis and spinogenesis are complemented by signaling via trophic factors and guidance cues (Bennett and Lagopoulos, 2014, Chen et al., 2008, Klein, 2009, Shen and Cowan, 2010, Stamatakou and Salinas, 2014, Thompson, 2003) and by adhesion systems that initiate synaptogenesis, control synapse maturation, and promote spine formation and maintenance by recruiting the key protein components of synapses (Brose, 2009, Scheiffele, 2003, Südhof, 2008, Tada and Sheng, 2006, Washbourne et al., 2004). That activity-independent synaptogenic processes can be triggered by synaptic adhesion systems is nicely illustrated by the fact that presynaptic and postsynaptic adhesion proteins (e.g., neurexins and neuroligins) can trigger postsynaptic and presynaptic specializations, respectively, in neuronal dendrites and axons, even when they are presented on the surface of non-neuronal cells (Scheiffele et al., 2000, Graf et al., 2004).

STAR★Methods

Key Resources Table

Contact for Reagent and Resource Sharing

Requests for resources, reagents, and further information should be directed to and will be fulfilled by the Lead Contact Nils Brose (brose@em.mpg.de).

Experimental Model and Subject Details

Mouse Lines

Mouse breeding was done with permission of the Niedersächsisches Landesamt für Verbraucherschutz und Lebensmittelsicherheit (LAVES; 33.19.42502-04-15/1817). All animals were kept according to the European Union Directive 63/2010/EU and ETS 123 in individually ventilated cages (IVCs) under specific pathogen-free conditions at 21 ± 1°C and 55% ± 10% relative humidity. The light/dark cycle was 12 h/12 hr. Mice were group-housed in IVCs type I superlong (435 cm2 floor area; TECHNIPLAST). The mouse environment consisted of food and tap water ad libitum, bedding, and nesting material. Cages were changed once a week. The animal health status was controlled daily by animal caretakers as well as by a veterinarian. Systematic health monitoring was carried out quarterly according to FELASA recommendations with either NMRI sentinel mice or animals directly taken from the colony. Health monitoring consisted of serological analyses and microbiological, parasitological, and pathological examinations. The mouse colony used for experiments did not exhibit signs of the pathogens routinely tested for. Mice lacking Munc13-1 (Unc13A) and Munc13-2 (Unc13B) (Augustin et al., 1999, Varoqueaux et al., 2002) were initially generated using 129/ola embyronic stem cells. Since the establishment of the initial M13-DKO line, the line has been bred into the C57BL/6N background more than 10 times. As the remaining 129/ola genetic contribution has not been assessed in the M13-DKO colony, the line must be regarded to be on a mixed C57BL/6N;129/ola background, with a minimal 129/ola contribution. M13-DKO and CTR littermates, obtained by interbreeding Unc13A;Unc13B and Unc13A;Unc13B mice, were used for all experiments. CTRs with genotypes Unc13A;Unc13B, Unc13A;Unc13B, and Unc13A;Unc13B show no obvious changes in brain morphology or cage behavior, and cultured neurons and hippocampal slices of mice with these genotypes show essentially normal transmittter release (Varoqueaux et al., 2002, Breustedt et al., 2010). The gender of animals used for experimentation was not checked because all previous studies on M13-DKOs had indicated that the gender does not affect the M13-DKO phenotype.

Tissue Culture

To determine the correct gestation day of pregnant female mice carrying M13-DKO and CTR embryos, females of the desired phenotype were mated with corresponding males and checked for vaginal plugs. At gestation day 18, as determined based on mating day and/or vaginal plug check, pups were obtained by hysterectomy, and organotypic hippocampal slice cultures were prepared according to published procedures (Imig et al., 2014, Stoppini et al., 1991). Pups were decapitated, and the brains were placed into cutting solution (97 mL HBSS, 2.5 mL Glucose 20%, 1 mL 100 mM kynurenic acid). The hippocampi were removed and transferred with a sterile pasteur pipette onto a tissue chopper stage under a preparation hood. Excess liquid was removed with sterile autoclaved tissues. Slices of 300 μm thickness were cut perpendicular to the longitudinal axis of the hippocampus and transferred in liquid with a plastic pasteur pipette to a plastic dish. Slices were then placed onto small membrane confetti on cell culture inserts. Maximally four hippocampal slices were cultured in each cell culture insert. Slices were maintained in culture medium (22.44 mL ddH2O, 25 mL 2xMEM, 25 mL BME, 1 mL GlutaMAX, 1.56 mL Glucose 40%, 25 mL horse serum). The medium was first changed 24 hr after preparation and then 2-3 times per week. Slices were cultured for 1, 2, or 3 weeks at 37°C/5% CO2.

Method Details

Experimental Design

Datasets were obtained with 6 (uncaging experiments), 4-5 (morphological analyses of dendrites and spines - except for the WIV1 datasets, 2), and 2-3 (electrophysiological analyses of synaptic transmission) independent slice culture preparations. The corresponding n values for the numbers of cells or spines assayed and of the number of glutamate uncaging trials are given in the respective figure legends. Morphological analyses of dendrites and spines were performed with the experimenter blind to the genotype. Experiments involving electrophysiological readouts could not be blinded effectively because the massive phenotypic defect of M13-DKOs is obvious in whole-cell patch-clamp analyses. Inclusion criteria for electrophysiological experiments were a stable series resistance (∼15 MΩ in the synaptic transmission measurements; ∼20-40 MΩ in the uncaging experiments) and a stable steady-state holding current (< 200 pA). Inclusion criteria for the morphological analyses of dendrites and dendritic spines were a pyramidal morphology and complete perfusion/staining of the neurons assessed. Neurons with non-pyramidal morphology, fluorescence staining background around the soma (indicating leakage during filling and/or recovery), or obviously non-stained dendritic areas were excluded form the analysis. In the context of uncaging-induced spine formation, only a new spine that grew during the initial 5 min time-lapse period after the HFU stimulus was considered as HFU-induced de novo spine formation. No more than three spinogenesis trials were performed on the same neuron. If there was uncertainty concerning the status of a new spine because of undulations in the dendrite, swellings in the z axis, or neighboring spine movement, the spine was excluded. For all neurons included in our analysis, overall spine density did not significantly change over the 1 hr imaging session. This ensured that the de novo spinogenesis being examined did not take place due to a general decrease in cell health, as spine/filopodia-like structures can form when cell health is compromised.

Electrophysiology, Cell Perfusion, and Imaging

At WIV1, WIV2, or WIV3, typically two CA1-PNs per CA1 region of a given slice were patch-clamped and simultaneously perfused with biocytin for subsequent staining with Alexa Fluor-555-labeled streptavidin. The internal solution contained 100 mM KCl, 50 mM K-gluconate, 10 mM HEPES, 4 mM ATP-Mg, 0.3 mM GTP-Na, 0.1 mM EGTA, and 0.2% biocytin (pH 7.4, 300 mOsm). The external solution was carbogen-saturated artificial cerebrospinal fluid (ACSF) (120 mM NaCl, 26 mM NaHCO3, 10 mM D-glucose, 2 mM KCl, 2 mM MgCl2, and 2 mM CaCl2, and 1 mM KH2PO4 - 304 mOsm). For measurements of mPSCs, 1 μM TTX was used. To measure mIPSCs and mEPSCs, AMPA and GABAA receptors were blocked with 10 μM of NBQX or bicuculline, respectively. Slices were acclimatized for 30 min in an interface chamber with external solution before patching. CA1-PNs were then whole-cell voltage clamped at −70 mV for at least 15 min to allow full exchange of the pipette solution containing biocytin with the dendritic lumen (Pusch and Neher, 1988), using an EPC-10 amplifier with Patchmaster 2 software (HEKA/Harvard Bioscience). Generally, for CA1-PNs that were later to be used for morphometric analyses, no pharmacological channel or receptor blockers were added. After removing the patch pipette at the end of the recording, slices were kept for at least 15 min in external solution to allow for recovery, immediately followed by fixation (24 hr, 5% formaldehyde in ACSF for the first 15 min then 5% formaldehyde in isotonic PBS at pH 7.4. Slices then were permeabilized for 2 hr with 0.3% Triton X-100 (Roche), and stained with 2.0 μg/ml Alexa-555-labeled streptavidin (Molecular Probes/Thermofisher) and 0.4 μM 4’,6-diamidino-2-phenylindole (DAPI; Sigma-Aldrich). Some slices were stained additionally with antibodies against the excitatory presynapse marker VGluT1 (rabbit antiserum 135302 at 1:500 dilution; Synaptic Systems) and Alexa 488-labeled secondary antibodies (Molecular Probes/Thermofisher), both in the presence of 10% goat serum. Slices were then washed, carefully removed from the underlying membrane insets, and mounted with glass coverslips in aqueous polyethylene glycol mounting media (Aqua Poly Mount 18606; Polysciences). Fluorescence in stained samples was imaged using a confocal microscope (Leica SP5; Leica) with an 8 kHz resonant scanner, a 100 x/1.44 NA oil immersion objective (Leica HCX PL APO CS; Leica), and a white light laser light source (SuperK EXTREME EXW-12, NKT Photonics A/S). Fluorescence was excited at 555 nm, and emission was recorded between 565 nm and 650 nm. Data acquisition was controlled by Leica software (Leica APS-AF) with a stitching extension, allowing us to image an entire CA1-PN in a single stack of images. Images were 3D Gaussian filtered (σx,y 0.7, σz 1.0; ImageJ, National Institutes of Health, Bethesda, MD, USA), using custom-written macros to handle the large datasets. To ensure analysis of only CA1-PNs, DAPI fluorescence was imaged to obtain overview images with a 10 x/0.4 NA long distance objective (Leica HCX PL APO CS; Leica). Only cells with a pyramidal shape and a location in CA1 pyramidal layer were used for further analysis. For segmentation of entire dendritic trees and subsequent spine analysis, we used NeuronStudio (CNIC, Mount Sinai School of Medicine, New York, NY, USA) (Rodriguez et al., 2003, Rodriguez et al., 2006, Rodriguez et al., 2008, Wearne et al., 2005).

Sholl Analysis and Spine Analysis

To investigate dendrite length and branching, we used the NeuronStudio segmentation data and a custom-written Excel worksheet template to count the number of branches as a function of distance to the soma (Sholl, 1953). Briefly, the algorithm defines a series of concentric spheres around the center of the soma in radius steps of 50 μm and counts the connections between neighboring vertices that cross these spheres. We used the NeuronStudio segmentation algorithms to localize, classify, and count dendritic spines and to determine their distribution along CA1-PN dendrites (Rodriguez et al., 2003, Rodriguez et al., 2006, Rodriguez et al., 2008, Wearne et al., 2005). Spines were classified into stubby, long/thin, and mushroom-type (Rodriguez et al., 2008), and were counted separately for each spine type as a function of distance along the dendrite relative to the soma. Parameters were kept as suggested (Rodriguez et al., 2008), except for the ‘threshold correction’, which we set to 10%, and the maximum height of spines, which was set to 3 μm. In addition, we used visual inspection of the images to count long and thin extrusions with a length of > 1μm, which were designated as filopodia. To assess the colocalization of spines with presynaptic terminals, a subset of samples containing CA1-PNs filled with biocytin and labeled with Alexa Fluor-555 streptavidin were co-stained for VGluT1, a maker of glutamatergic presynaptic boutons, using a specific primary antibody (rabbit antiserum 135302 at 1:500 dilution; Synaptic Systems) and an Alexa 488-labeled secondary antibody (Molecular Probes/Thermofisher). Colocalization of Alexa Fluor-555 and Alexa Fluor-488 signals was assessed visually in representative subsections of ∼100 μm secondary and tertiary apical dendrites close to the somata. In brief, green fluorescence images representing VGluT1 staining were convolved using a 5 × 5 pixel edge-detection matrix, in which all pixels were −1 except the center pixel, which was +24, followed by Gaussian filtering with σ = 3 pixels. Under these conditions, some voxels of the presynaptic staining should overlap with some voxels of stained spines. Overlap was detected visually in color image stacks. The fraction of spines with overlapping VGluT1 positive boutons was determined visually.

Glutamate Uncaging Experiments

M13-DKO and CTR embryos were recovered by hysterectomy at embryonic day 18 (E18) and decapitated, and the brains were placed in ice-cold Hanks’ balanced salt solution (HBSS). Organotypic hippocampal slice cultures using the embryonic brain were prepared as described (Stoppini et al., 1991). Expression vectors were delivered 2-3 days before imaging using biolistic gene transfer (180 psi) as described (Woods and Zito, 2008), except that 12 μg of tdTomato expression vector was coated onto 6-8 mg of gold particles. CA1-PNs at WIV1-3 at depths of 20-40 μm were imaged using a two-photon microscope (Prairie Technologies) with a pulsed Ti::sapphire laser (MaiTai HP DeepSee, Spectra Physics) tuned to 920 nm (2-2.5 mW at the sample) in a specific recirculating artificial cerebrospinal fluid (uACSF; 127 mM NaCl, 25 mM NaHCO3, 1.25 mM NaH2PO4, 2.5 mM KCl, 25 mM D-glucose, aerated with 95%O2/5%CO2). For each CA1-PN, image stacks (512 × 512 pixels; 0.035 μm/pixel) with 1 μm z-steps were collected from one segment of secondary or tertiary apical and/or basal dendrites at 30-80 μm from the soma. All images shown are maximum projections of 3D image stacks after applying a median filter (2 × 2) to the raw image data. Uncaging of MNI-glutamate was achieved as described (Oh et al., 2013). In brief, high-frequency glutamate uncaging (HFU) stimuli consisted of 60 pulses of 2-ms duration, delivered at 10 Hz by parking the beam at a point ∼0.5 μm from the edge of a dendrite with a pulsed Ti::sapphire laser (MaiTai HP, Spectra-Physics) tuned to 720 nm (14-15 mW at the sample). Imaging-only experiments to induce de novo spinogenesis were carried out at 30°C in an uACSF-based external solution containing 2 mM Ca2+, 0 mM Mg2+, 1 μM TTX, and 2.5 mM MNI-glutamate. No more than three spinogenesis trials were performed from the same CA1-PN. Whole-cell recordings (electrode resistances 5-8 MΩ; series resistances 20-40 MΩ) were performed at 25°C on visually identified CA1-PNs within 40 μm of the slice surface using a MultiClamp 700B amplifier (Molecular Devices). To record uncaging-evoked excitatory postsynaptic currents (uEPSCs), CA1-PNs were patched in voltage-clamp configuration (Vhold of −65 mV and +40 mV for AMPA-receptor-mediated and NMDA receptor-mediated uEPSCs, respectively) using a cesium-based internal solution (135 mM Cs-methanesulfonate, 10 mM HEPES, 10 mM Na2 phosphocreatine, 4 mM MgCl2, 4 mM Na2-ATP, 0.4 mM Na-GTP, 3 mM Na L-ascorbate, 0.2 mM Alexa Fluor 488, ∼300 mOsm, ∼pH 7.25) in uACSF containing 2 mM Ca2+, 1 mM Mg2+, 0.001 mM TTX, and 2.5 mM MNI-glutamate. uEPSC amplitudes from individual spines were quantified as the average (8-10 test pulses of 1-ms duration at 0.1 Hz) from a 2-ms window centered on the maximum current amplitude after uncaging pulse delivery for AMPA currents and from a 10-ms window between 70 and 80 ms after stimulus for NMDA currents. Laser pulses were delivered by parking the beam at a point ∼0.5 μm from the center of the spine head (720 nm; 14-15 mW at the sample). Only a new spine that grew during the initial 5 min time-lapse period after the HFU stimulus was considered as HFU-induced de novo spine formation. If there was uncertainty concerning the status of a new spine because of undulations in the dendrite, swellings in the z axis, or neighboring spine movement, the spine was excluded. Integrated green (Alexa Fluor 488) fluorescence intensities were measured from background-subtracted green fluorescence using the integrated pixel intensity of a boxed region surrounding the spine head. Estimated spine size was calculated by normalizing the fluorescence intensities (as described above) for each individual spine to the mean fluorescence intensities measured from four ROIs on the dendritic shaft (Woods et al., 2011). Relative spine size was determined by dividing the estimated size of an individual spine by the mean estimated size of all spines on the same dendritic segment (Oh et al., 2015). Correlation was examined by Pearson’s correlation.

Quantification and Statistical Analysis

Data are presented as mean ± SEM. Statistical analyses were carried out using standard statistical software packages in Microsoft Excel or GraphPad Prism 5. For comparisons of M13-DKO and WT datasets from littermate samples, statistical differences were assessed by Student’s t test, based on near-normal distribution and similar variance. Success rate of de novo spinogenesis across development was compared by two-sided Fisher’s exact test. Correlation was examined by Pearson’s correlation analysis. Depending on the experimental readout, the values of n represent the numbers of cells assayed (Figures 1, 2, and 3), the number of glutamate uncaging trials (Figure 4B), or the number of spines assayed (Figures 4B and 4G–4I).

Author Contributions

A.S., C.I., J.S.R., and N.B. conceived the project. C.I. and J.S.R. established the long-term slice culture system and neuron injection protocols. B.H.C. designed the protocols for morphological analyses of neurons. A.S. conducted the patch-clamp electrophysiological analyses of spontaneous synaptic activity and designed and performed the quantitative morphological analyses of neurons. B.A. conducted the patch-clamp electrophysiological analyses of spontaneous synaptic activity. W.C.O. performed the glutamate uncaging experiments. H.B.K. designed and supervised the glutamate uncaging experiments. H.K. provided conceptual and methodological input. J.S.R. and N.B. supervised the project. A.S., H.B.K., J.S.R., and N.B. wrote the manuscript with input from all coauthors.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Goat anti-Rabbit IgG Secondary Antibody, Alexa Fluor 488	Thermo Fisher Scientific	Cat. No. A-11008; RRID: AB_143165
Rabbit anti-VGLUT1 antibody	Synaptic Systems	Cat. No. 135 302; RRID: AB_887877
Chemicals, Peptides, and Recombinant Proteins
Alexa Fluor 488 hydrazide	Thermo Fisher Scientific	Cat. No. A10436
Alexa Fluor 555 streptavidin conjugate	Thermo Fisher Scientific	Cat. No. S32355; RRID: AB_2571525
(-)-Bicuculline methochloride	Tocris Bioscience	Cat. No. 0131
Biocytin hydrochloride	Sigma-Aldrich	Cat. No. B1758
DAPI: 4′,6-Diamidine-2′-phenylindole dihydrochloride	Sigma-Aldrich/Roche	Cat. No. 10236276001
MNI-caged-L-glutamate	Tocris Bioscience	Cat. No. 1490
Mounting glue: Aqua-Poly/Mount	Polysciences	Cat. No. 18606
NBQX disodium salt	Tocris Bioscience	Cat. No. 1044
Tetrodotoxin	Tocris Bioscience	Cat. No. 1078
Tetrodotoxin citrate	Tocris Bioscience	Cat. No. 1069
Kynurenic acid	Sigma-Aldrich	Cat. No. K3375
GlutaMAX Supplement	Thermo Fisher Scientific	Cat. No. 35050038
Experimental Models: Organisms/Strains
Mouse: M13-DKOs: Unc13atm1Bros; Unc13btm1Rmnd	Augustin et al., 1999, Varoqueaux et al., 2002	RRID: MGI 2654056
Recombinant DNA
tandem-dimer Tomato (TdTomato) DNA	Kwon et al., 2012	http://dx.doi.org/10.1038/nn.3256
Software and Algorithms
AxoGraph X 1.3.1	John Clements	https://axograph.com; RRID: SCR_014284
Clampfit 10.3	Molecular Devices	https://www.moleculardevices.com/systems/conventional-patch-clamp/pclamp-10-software
GraphPad Prism 5	GraphPad Software	http://www.graphpad.com; RRID: SCR_002798
ImageJ	National Institutes of Health	https://imagej.nih.gov/ij; RRID: SCR_003070
Leica LAS AF	Leica Microsystems	http://www.leica-microsystems.com; RRID: SCR_013673
NeuronStudio segmentation algorithms	CNIC, Mount Sinai School of Medicine, New York, NY, USA	http://research.mssm.edu/cnic/tools-ns.html; RRID: SCR_013798
OriginPro 8.5	OriginLab	http://www.originlab.com
Patchmaster v2X80	HEKA/Harvard Bioscience	http://www.heka.com/products/products_main.html#soft_pm; RRID: SCR_000034
Other
Basal Medium Eagle (BME)	Thermo Fisher Scientific	Cat No. 41010026
Hank’s Balanced Salt Solution, Ca2+, Mg2+ (HBSS)	Thermo Fisher Scientific	Cat No. 24020091
Helios Gene Gun System	Bio-Rad	Cat No. 165243
Horse Serum, heat inactivated	Thermo Fisher Scientific	Cat No. 26050088
McIlwain tissue chopper	Ted Pella	Cat No. 10180-220
Millicell cell culture inserts	Merck Millipore	Cat No. PICM03050
Millipore membrane confetti	Merck Millipore	Cat No. FHLC04700
Minimum Essential Medium (MEM)	Thermo Fisher Scientific	Cat No. 11700077