|Home | About | Journals | Submit | Contact Us | Français|
Oswald Steward, University of California-Irvine, Cellular Neurobiology and Neuronal Plasticity
Geometry of the dendritic tree and synaptic organization of afferent inputs are essential factors in determining how synaptic input is integrated by neurons. This information remains elusive for one of the first brainstem neurons involved in processing of the primary auditory signal from the ear, the bushy cells (BCs) of the ventral cochlear nucleus (VCN). Here, we labeled the BC dendritic trees with retrograde tracing techniques to analyze their geometry and synaptic organization after immunofluorescence for excitatory and inhibitory synaptic markers, electron microscopy, morphometry, double tract-tracing methods, and 3-D reconstructions. Our study revealed that BC dendrites provide space for a large number of compartmentalized excitatory and inhibitory synaptic interactions. The dendritic inputs on BCs are of cochlear and non-cochlear origin, and their proportion and distribution are dependent on the branching pattern and orientation of the dendritic tree in the VCN. Three-dimensional reconstructions showed that BC dendrites branch and cluster with those of other BCs in the core of the VCN. Within the cluster, incoming synaptic inputs establish divergent multiple-contact synapses (dyads and triads) between BCs. Furthermore, neuron-neuron connections including puncta adherentia, sarcoplasmic junctions and gap junctions are common between BCs, which suggests that these neurons are electrically coupled. Together, our study demonstrates the existence of a BC network in the rat VCN. This network may establish the neuroanatomical basis for acoustic information processing by individual BCs, as well as for enhanced synchronization of the output signal of the VCN.
Dendrites integrate excitatory and inhibitory information received by neurons, providing the main cell-surface sites for synaptic inputs (Johnston et al., 1996; Häusser, 2001). Advances in electrophysiology and computational modeling demonstrate that the shape of the dendritic tree and its orientation, length and branching pattern are crucial factors for determining how signals coming from individual synapses are integrated (Segev and London, 2000; Gulledge et al., 2005). Synaptic inputs have qualitatively different effects on the output of neurons depending on their location on the dendritic tree (Crook et al., 1998; Magee and Cook, 2000; Spruston, 2000), and synapse location may play a role in mechanisms of cellular working memory (Nielsen, 2003; Morita, 2008). Further, dendritic geometry has important influences on the spread of actively propagating electrical signals (Häusser, 2001), maintenance of synchrony in neuronal networks (Traub et al., 2001; Gansert et al., 2007; Goldberg et al., 2007), and reception of input from divergent synapses between two neurons (Zhang et al., 2003). For all these reasons, the importance of understanding the geometry of the dendritic tree and the spatial distribution of its inputs has become widely accepted (Agmon-Snir et al., 1998; Morita, 2008).
In spite of the growing evidence of the relevance of dendrites in normal and abnormal synaptic processing and brain wiring, very little is known of the role of dendrites in auditory function. One of the first brainstem neurons involved in processing primary auditory signals from the ear, the bushy cells (BCs) of the ventral cochlear nucleus (VCN) exhibit an unusual dendritic morphology but the role of the dendrites in the normal function of these cells remains elusive. Two types of bushy cells, spherical and globular bushy cells, have been identified (Osen 1969; Brawer et al., 1974). Both types have dendritic trees consisting of one or two primary dendrites that branch repeatedly to produce a complex tufted dendritic arborization (Cant and Morest, 1979a; Tolbert et al., 1982; Rouiller and Ryugo, 1984). Clarification of the synaptic organization and geometry of BC dendrites should increase our understanding of the role of BCs in normal and abnormal auditory function.
Both types of BCs encode features of the acoustic waveform and convey precise temporal information to upper auditory structures (Friauf and Ostwald, 1988; Smith et al., 1991, 1993; Cant and Benson, 2003). Large synaptic complexes formed by the endbulbs of Held on spherical bushy cell somata guarantee the transmission of a high-fidelity copy of auditory fiber activity (Pfeiffer, 1966; Brawer and Morest, 1975; Ryugo and Sento, 1991). However, compared to the auditory nerve, spherical bushy cells are more highly synchronized to the acoustic stimulus (Joris and Smith 2008). The biological substrate for the enhanced synchronization is unclear because only one or a few endbulbs terminate on each spherical BC (Cant and Morest, 1979b; Sento and Ryugo, 1989; Ryugo and Sento, 1991; Joris and Smith, 2008).
This study focuses on describing the synaptic organization of BC dendrites, including the proportion, location and neurotransmitter content of different types of synaptic inputs along the dendritic tree. We report that excitatory and inhibitory inputs from cochlear and non-cochlear origins make divergent multiple-synaptic contacts between neighboring BCs. We also describe an intricate dendritic network that establishes junctional connections between BCs to form neuronal clusters. We conclude that the BCs in the rat VCN form an integrated network that may underlie previously described physiological response properties.
A total of 11 adult Sprague-Dawley rats (P40 days) were used in compliance with NIH guidelines concerning the care and use of animals in biomedical research. The handling and care of the rats prior to and during the experimental procedures were approved and supervised by the University of Connecticut Institutional Animal Care and Use Committee (IACUC). All experiments conformed to local and international guidelines on the ethical use of animals and all efforts were made to minimize the number of animals used. For the surgical procedures, including the transcardial perfusion of fixatives and injection of neuronal tracers, the animals were deeply anesthetized with a mixture of ketamine (70 mg/kg body weight) and xylazine (4.0mg/kg body weight) administered intraperitoneally.
To retrogradely label dendrites of BCs, animals received an injection of either dextran fluorescein isothiocyanate (D-FITC; #D-1820; Molecular Probes, Eugene, OR, USA) or biotinylated dextran amine (BDA; #D-1956; Molecular Probes) into the left trapezoid body (TB) and the adjacent medial nucleus of the trapezoid body (MNTB) (Fig. 1A). Injections at this location have been used successfully to label retrogradely spherical and globular BCs in cats (Tolbert et al., 1982). In our study, we targeted the corresponding location in the rat brainstem. The coordinates were calculated using the atlas of the rat brain (Paxinos and Watson, 1998), and the tracers were delivered using the stereotaxic procedure described elsewhere (Gomez-Nieto et al., 2008b). To label BC dendrites for the confocal microscopy study, large injections (0.15 – 0.2 µl) of D-FITC (10% in distilled water) were pressure-delivered with a Hamilton syringe (#710, Hamilton Co., Reno, NV) attached to a Stereotaxic Injector (Stoelting Co., Wood Dale, IL). For the electron microscopy study, small injections (~ 0.05 µl) of BDA (10% in distilled water) were delivered over 5 minutes with the syringe microdrive. After tracer injections, animals were allowed to recover for 4–7 days.
To label the auditory nerve three rats previously injected with D-FICT in the TB-MNTB were perfused through the heart with 0.5% paraformaldehyde (PFA) in 0.1 M phosphate buffer (PB), pH 7.4. The cochlea was removed and crystals of the lipophylic dye DiI (#D-3911, 1.1'-dioctadecyl-3,3,3',3' tetramethylindocarbocyanine perchlorate; Molecular Probes) were placed on the exposed auditory nerve. The brains were stored in 4% PFA for approximately 30 days at room temperature protected from light, and were processed for confocal microscopy as described below.
Five Sprague-Dawley rats with D-FICT injections in the TB-MNTB were used. After the postinjection survival time, animals were perfused through the heart with 4% PFA in 0.1M PB, brains were removed from the skull, and coronal sections were cut with a Vibratome at 40 µm thickness. Single immunofluorescence was performed to determine the distribution of VGLUT1, VGLUT2 and VGAT along dendrites of bushy cells labeled with D-FITC as described elsewhere (Gomez-Nieto et al., 2008b). We used polyclonal antibodies generated in rabbits against Strep-Tag fusion protein containing amino acid residues 456–560 of rat VGLUT1 (#135 302; Synaptic Systems, Göttingen, Germany), Strep-Tag fusion protein containing amino acid residues 510–582 of rat VGLUT2 (#135 402; Synaptic Systems), and the 17 amino acid peptide sequence near the carboxy terminal region of rat VGAT (VHSLEGLIEAYRTNAED; #AB5062P; Chemicon, Temecula, CA; now part of Millipore, Billerca, MA). These antibodies have been tested in preadsorption experiments that blocked efficiently and specifically the corresponding signals (manufacturer’s technical information; see also Zhou et al., 2007). On Western blots of cerebellum and cochlear nucleus the rabbit polyclonal antibodies for VGLUT1 and VGLUT2 recognize a single band migrating at ~ 60 kDa and ~ 65 kDA, respectively (Zhou et al., 2007) and ~ 55 – 60 kDa for the polyclonal rabbit anti-VGAT (manufacturer’s technical information). Washes and dilutions of antisera were done in 0.1 M phosphate buffered saline (PBS), pH 7.4. The sections were incubated in the blocking solution (6% normal goat serum in PBS) for 1 hour, followed by overnight incubation with primary antibodies at 1:1000 dilutions. After removal of primary antisera, sections were reacted for 2 hours with secondary antibody (Cy3 goat anti-rabbit; #111-165-003; Jackson Immunoresearch, West Grove, PA).
In addition, to determine the codistribution of VGLUT1 and VGLUT2 on dendrites of bushy cells, we performed double immunofluorescence for the two isoforms. For these experiments, we used a protein A purified mouse monoclonal antibody against VGLUT2 (#MAB5504; Chemicon, now part of Millipore) in combination with the polyclonal antibody against VGLUT1. The mouse monoclonal antibody MAB5504 was raised against recombinant protein from rat VGLUT2 and it has shown species reactivity to the mouse and rat. The immunogen for this antibody is the whole VGLUT2 protein (information obtained from the manufacture). The antibody has not been epitope mapped by Chemicon, so the epitope for MAB5504 is not known. On Western blots VGLUT2 immunostaining has a predicted molecular eight of 60 kDa (as shown by a comparable product from Abcam Cambridge, MA). As secondary antibodies we used the Cy3 goat anti-mouse (#115-165-166; Jackson Immunoresearch) and Cy5 goat anti-rabbit (#115-175-003; Jackson Immunoresearch). As a control in this set of experiments, double immunolabeling for the monoclonal (MAB5504) and polyclonal (#135 402) antibodies for VGLUT2 generated the same immunolabeling pattern and labeled the same synaptic endings in the cochlear nucleus (Supplemental Fig.1) and cerebellum (data not shown).
In the double tract-tracing experiments (D-FITC and DiI), we also performed a single immunolabeling with polyclonal antibodies for either VGLUT1 or VGLUT2. As secondary antibody we used Cy5 goat anti-rabbit (blue emission) to distinguish the signal from the DiI (red emission) labeled auditory nerve endings. After immunofluorescence sections were mounted on slides and coverslipped with ProLong® Antifade Kit (#P7481, Molecular Probes). In all immunofluorescence experiments, omission of primary antibody resulted in no staining of the preparations. Since VGLUT immunoreactivity is well documented in the cerebellar cortex (Takamori et al., 2001), we used the cerebellum as the positive control (data not shown).
Three rats with BDA injections in the TB-MNTB followed the procedure for electron microscopy. Animals were perfused with fixative containing 4% PFA and 0.5 % glutaraldehyde 0.1M PB. Brains were removed and sectioned (60 µm thick) in the coronal plane with a Vibratome. Brainstem sections were also processed for tracer visualization as described in detail elsewhere (Gomez-Nieto et al., 2008a). After brightfield microscopic examination only sections with BDA-labeled BCs, visualized as a black reaction product, were selected. The sections were postfixed with 1% OsO4, and flat-embedded in EMBed-812 resin (Rubio et al., 2008). The small BDA injection labeled only a few BCs (1–3) that were dispersed within the nucleus. This allowed us to select single BCs for the ultrastructural analysis. The sections containing each single BC were glued on blank resin blocks, and sectioned with a Leica ultramicrotome. Semithin (0.5–1 µm) and serial ultrathin sections (silver-gold interference ~80 nm) were collected on Formvar-coated single slot grids.
Two types of BCs can be anatomically distinguished in the VCN based on cytological features. These cells are called spherical or globular bushy cells according to the shape of their cell bodies (Harrison and Irving, 1965; Osen 1969; Brawer et al., 1974). In this study, all the cells analyzed in the confocal and electron microscopy studies were labeled with the tracer and were identified as globular or spherical bushy cells after analyzing the shape of the cell body and the geometry of their dendritic tree (Cant and Morest, 1979a, b; Tolbert et al., 1982; Wang et al., 1998; Cant, 1981; Mahendrasingam et al., 2004; Asako et al., 2005; Pocsai et al., 2007). We studied the most anterior sections of the VCN, which were reported in previous studies of cats to contain a homogeneous population of spherical-BCs and more caudal regions containing globular-BCs (reviewed in Cant and Benson, 2003). The dendrites of BCs analyzed were always filled with the tracer and were classified as primary, secondary and distal (including the tufted dendrites) (Ostapoff and Morest, 1991; Ryugo and Sento, 1991).
Sections processed for confocal microscopy were analyzed with a Leica TCS SP2 confocal laser scanning microscope (Leica Microsystems, Mannheim, Germany) coupled to a Leica DM IRE2 inverted microscope and equipped with argon and helium neon lasers with excitation wavelengths of 458, 476, 488, 543, 568 and 633 nm. The fluorochromes FITC, Cy3, or Cy5 were detected sequentially, stack-by-stack, with the acousto-optical tunable filter system and triple dichroic mirror TD488/543/ 633, using laser lines 488, 546 and 633 nm, respectively. The background was controlled and the photomultiplier voltage (800 V) was selected for maximum sensitivity in the linear range. The objectives used were oil immersion ×40 and ×63/numerical aperture 1.30, giving a resolution of ×150 nm in the xy-plane and ×300 nm along the z-axis (pinhole 1 Airy unit), as well as several electronic zoom factors up to ×1.58. To determine the distribution of the immunolabeled terminals on dendrites of BCs, series of 25–50 confocal images were obtained to generate a maximum-intensity z-projection of stacks and an orthogonal projection (=xy, xz, yz planes for z-stacks series). In addition, an intensity profile of the confocal image was defined to analyze the intensity values of the two channels using the MicroSuite 5 software (Olympus). The intensity profile (~8 µm in length) was drawn perpendicular to the bushy cell dendrite bearing the labeled endings and was used to check the apposition of the terminals to the labeled dendrites. Colocalization of the fluorochromes within positive terminals was always verified in the orthogonal view and one confocal image of 1 µm thickness in the z plane.
For morphometric and quantitative analyses, confocal images were captured using the same imaging parameter settings. In the double immunolabeling for VGLUT1 and VGLUT2 experiments, we studied quantitatively the codistribution and number of VGLUT isoforms in the core and granule cell domain (GCD) of the VCN. We selected 8 areas in each region with an average area of 0.034 mm2 and 0.015 mm2, respectively. A total of 5,452 puncta labeled with VGLUT1, VGLUT2 or both were classified according to the size (area in µm2) as 1) large irregular puncta with the size of endbulb or mossy-like terminals (1.5 to 7.5 µm2), and 2) small puncta with the size of en passant boutons (0.5 to 1.5 µm2) (Zhan and Ryugo, 2007; Zhou et al., 2007). All labeled puncta were classified, counted or measured with ImageJ (version 1.39e-Java 1.5.0, NIH; http://rsb.info.nih.gov/ij). To ensure reliable counting, visual inspections and manual corrections were conducted after each automated count. The number of puncta was divided by the chosen area to yield the percentage of puncta per unit area.
To study the distribution of VGLUT1, VGLUT2 and VGAT labeled puncta along the dendrites of BCs, we counted the number of puncta in apposition with each particular dendritic site. We analyzed dendrites of 18 different BCs classified according to their dendritic orientation towards the center or superficial granular regions of the VCN (core and GCD, respectively).
Sections processed for electron microscopy were examined using a transmission electron microscope (TECNAI G2 Spirit Biotwin G2; FEI, Hillsboro, OR) coupled to a digital camera (AMT2K). Digital electron micrographs were taken at a magnification raging from ×2,900 to ×68,000. The synaptic organization on BC dendrites was studied in 2 BDA-labeled bushy cells. To classify synaptic terminals along BC dendrites we performed a morphometric analysis of the synaptic vesicles (Rubio and Juiz, 2004; Rubio et al., 2008). A total of 1325 synaptic vesicles (58 terminals) were measured using ImageJ software. The selection of the synaptic vesicles was random but limited to only those vesicles with clearly visible plasma membrane. Asymmetric and symmetric synapses were identified following criteria described by Gray (1959). Ultrastructural parameters such as electron-dense appearance of the axoplasm and synaptic vesicles, and packing density of synaptic vesicles in the synaptic terminals were considered for the classification. Parameters of the synaptic vesicles included: area, minor and major diameter of the vesicles and circularity (where a ratio of 1 indicates round and smaller ratios indicate a progressively flatter structure) (Rubio and Juiz, 2004; Rubio et al., 2008). The size of the vesicle was represented by its approximate mean diameter, calculated by [(major axis+minor axis)/2]. Each synaptic terminal was also classified according to the targeted dendritic site (proximal, secondary or distal dendrites). To determine whether differences between two samples were statistically significant we used Student’s t-test with P-values threshold < 0.01. Two independent investigators performed blind morphometric and statistical analyses of the sample and the results were similar. The data were then compared to the qualitative classification of the endings. The morphometric and qualitative analyses showed the same distinct eight types of endings (four with round vesicles and four with pleomorphic or flattened vesicles).
All photographs shown in the figures were processed with minor modifications in brightness and contrast using Adobe Photoshop ® (version 9.0; Adobe Systems Incorporated, San Jose, CA), and assembled using Canvas 7.0 software.
A total of 120 serial electron micrographs (×4,800 magnification) were used to reconstruct one bushy cell dendrite labeled with BDA. A set of serial sections (13 and 20 images), were selected to reconstruct in detail specific structures. In addition, 3 more series of 8 consecutive electron micrographs each were used to reconstruct the synaptic interactions of another labeled BC dendrite. For better resolution the images were taken at ×11,000, ×23,000 and ×30,000 magnification. Three-dimensional reconstructions were made by alignment of serial electron micrographs of the labeled BC dendrites using Reconstruct software (Fiala, 2005; http://synapse-web.org). In brief, two successive sections were aligned via rotation and translation such that corresponding structures like mitochondria and labeled dendrites in the two sections superimposed. To compensate for distortions introduced by the sectioning, we had to transform the images linearly. Alignment was followed by a contouring of membranes by visually defining markers of different colors. The subsequent linear interpolation between these markers resulted in a polygonal outline membrane. The three-dimensional rendering was generated as VRML files from the stacks of all contoured sections. By filling these stacks with tetrahedra, we calculated volumes and surface areas of the structures of interest. Images and movie documents from the 3-D renderings were obtained using image and video converter software (Cortona; Parallel Graphics, Boston, MA). The structures reconstructed were labeled and unlabeled BC somata and dendrites, synaptic terminals, and membrane specializations. The unlabeled somata and dendrites shared similar utrastructural characteristics as the labeled BC. In addition, we followed previous studies on BC ultrastructure (Cant and Morest, 1979a; Smith and Rhode, 1987; Ostapoff and Morest, 1991; Ryugo and Sento, 1991). These membrane specializations include synaptic contacts, puncta adherentia, sarcoplasmic junctions and gap junctions. To distinguish between membrane specializations, we followed the criteria established elsewhere (Henkart et al., 1976; Cant and Morest, 1979b; Landis and Reese, 1982; Mugnaini, 1985; Brunso-Bechtold and Vinsant, 1988; Rash et al., 1998).
To study the dendritic organization of BCs in the VCN we injected dextran fluorescein (D-FITC) in the area of the trapezoid body (TB) (Fig. 1). We obtained large injections centered in the TB and the adjacent MNTB (Fig 1A). The injection site spread to other nuclei of the superior olivary complex including the medial and lateral superior olives. Following the D-FITC injections we found labeled axons through the TB that led to retrogradely filled cells in the VCN of both sides (Fig. 1B). We focused on the analysis of the caudal and rostral regions of the VCN corresponding to the area of distribution of globular and spherical bushy cells, respectively. All labeled cells included in the confocal microcopy analysis were identified as globular (Fig. 1C) or spherical bushy cells (Fig. 1D) by the morphology of their cell body and dendrites. In a small number of injections and in the most caudal sections, we also observed labeled cells that were identified as multipolar stellate cells (Fig. 1E). This cell type was clearly distinguishable from spherical and globular bushy cells (Fig. 1C–D), and was not included in the analysis. In this study, we found a similar distribution of inputs on dendrites of globular and spherical bushy cells, therefore we referred to both cell types as BCs. Labeled BC somata distributed in the central region (core) of the ventral cochlear nucleus, where the anterior branch of the auditory nerve projects (Ryugo, 2008). One or two retrogradely labeled BC dendrites (with an average length up to 54.5 ± 13.5 µm) were visible on each neuron (Fig. 1C–D). When there were two dendrites, they emerged from opposite poles of the cell body. Some BCs were observed with thick dendritic processes forming a sparse bush (Fig. 1C). In other BCs, the dendrites were thin and formed a dense bush with beaded-enlargements most distally (Fig. 1D). No dendritic spines were observed protruding from the dendrites. The location of BCs within the core was relevant as we observed the targets of the dendrites differed depending on their central or superficial location (see below). In the central region of the core, distal dendrites of BCs branched into a complex nest-like formation, which surrounded nearby BC somata (Fig. 1F–G–H). Up to 5 adjacent BC somata were found surrounded by the dendritic nest that appeared to form a neuronal cluster. In contrast, the superficial BCs oriented their dendrites towards the granular cell domain (GCD; Fig. 1I–J–K). Distal dendrites of these superficial BCs were observed within the GCD, which is preferentially the area of distribution of somatosensory, descending and commissural inputs (Itoh et al., 1987; Zhan and Ryugo, 2007; Zhou et al., 2007).
To determine the presence of synaptic endings along BC dendrites, we combined D-FICT injections with immunofluorescence for excitatory (VGLUT1 and VGLUT2 isoforms) and inhibitory (VGAT) synaptic markers. In general in the rat VCN, VGLUT1 distributed throughout the nucleus including the core (endbulb-like puncta ~5.9 µm2) and the GCD (smaller puncta ~1 µm2) (Supplemental Fig. 2). In contrast, small (~1 µm2) and large (~4 µm2) VGLUT2-puncta concentrated mainly in the GCD, although some labeling was also present in the core (Supplemental Fig. 2). When we performed double immunofluorescence for VGLUT1 and VGLUT2, we found that in general their immunoreactivity and distribution in the VCN did not overlap although we found that 15% of puncta co-labeled for both isoforms (Supplemental Fig. 2). These co-labeled VGLUT1 and VGLUT2 puncta were the largest in size (~7.5 µm2), and were more abundant in the GCD (11%) than in the core (4%) (Supplemental Fig. 2). These results support the general idea that these isoforms might label different glutamatergic synaptic inputs. VGAT labeling was also puncta-like (~1.9 µm2) and was found evenly distributed in the VCN.
In retrogradely labeled BC dendrites, we observed that both excitatory (VGLUT1 and VGLUT2) and inhibitory (VGAT) synaptic markers distributed along the dendritic profile (Fig. 2). Analysis of the intensity values on each dendritic segment confirmed that VGLUT1 (Fig. 2A–B–C), VGLUT2 (Fig. 2E–F) and VGAT (Fig. 2G) labeled endings were in apposition with BC dendrites (e.g. Fig. 2D). Our data show that dendrites of BCs receive a large number of excitatory and inhibitory endings, preferentially in the distal tufted segments (Fig. 2I). Primary dendrites receive more inhibitory endings than distal dendrites, which are contacted by more excitatory endings (Fig. 2I). We divided the BCs in two groups depending on whether they were located more centrally or superficially in the core region. The superficial BCs were located within 100 µm of the internal boundary of the GCD. In general, the dendrites in these two groups receive the same percent of excitatory endings (VGLUT1 and VGLUT2 combined) and inhibitory endings (VGAT) (Fig. 2H). However, when the immunolabeling for VGLUT1 and VGLUT2 was analyzed separately we found that their distribution differed depending on the orientation of the dendritic tree (Fig. 2H). BC dendrites entirely distributed in the core presented more VGLUT1 (25%) than VGLUT2- puncta (5%). On the other hand BC dendrites oriented near and within the GCD presented more VGLUT2-puncta (20%) than VGLUT1 (14%) (Fig. 2H). The differences in the patterns of VGLUT1 and VGLUT2 labeling shown for the dendrites in the core versus the dendrites in the GCD were significant (P< 0.05). In both groups VGLUT isoforms were concentrated more on distal dendrites than on secondary or primary dendrites (Fig. 2I).
To confirm that excitatory and inhibitory endings synapse along the dendritic tree of BCs, we injected biotin dextran amine (BDA) in the TB-MNTB as described above, and analyzed the labeled dendrites at the electron microscopic level (Fig. 3, Fig. 4). These injections were smaller than D-FICT injections, and as result only a few (1–3) BCs were labeled within the entire VCN, which allowed us to select and trim one single cell for the ultrastructural analysis (Fig. 3A–B). The data presented here correspond to retrogradely labeled spherical-BCs, although similar results were found on globular-BCs. The electron dense reaction product for BDA filled cell bodies and the entire dendritic tree of spherical-BCs (Fig. 3, Fig. 4). Proximal and distal dendrites received synaptic endings that were ultrastructurally identified as excitatory or inhibitory (Nakajima and Reese, 1983). The excitatory endings made asymmetric synaptic contacts and contained round synaptic vesicles, whereas the inhibitory endings made symmetric synaptic contacts and contained flattened to pleomorphic synaptic vesicles (Fig. 3, Fig. 4). Quantitatively we distinguished four main subtypes of endings in each group that differed in size of the ending and of the synaptic vesicles. A morphometric analysis of the synaptic vesicles showed that these types were distinct (Table 1; Student’s t-test, P < 0.01).
The four types of excitatory endings were classified according to the size of the round synaptic vesicles as giant (GR), large (LR), medium (MR) and small (SR) (Table 1; Fig. 3, Fig. 5A). Only the MR had ultrastructural characteristics similar to those of the endbulb of Held on the cell body. Nevertheless, to verify that the ultrastructure of the MR endings was the same as the endbulb, we compared the area of the synaptic vesicles of the four types of endings with round synaptic vesicles on the dendrites with those of the endbulb of Held on the soma (Fig. 5A). Data showed that only the vesicles of the MR subtype were similar to the ones of the auditory nerve. This result suggests that branches of the auditory nerve synapsed on the dendritic tree of spherical-BCs. The other three types of excitatory terminals (GR, LR and SR) were identified as non-cochlear terminals (Fig. 5A) following qualitative observations and quantitative morphometric parameters. Qualitatively the axoplasm of the GR and LR endings was darker than the enbulbs and the synaptic vesicles appeared more compacted, in particular those of the GR endings. The SR endings had a clear axoplasm, but they only were observed making single synaptic contacts. Quantitatively, the GR and LR endings were larger (~3.7 µm2) than the auditory nerve (~2.8 µm2), whereas the SR terminals were the smallest of the four types (~1 µm2). We studied the distribution and proportion of these endings in different regions of the dendritic tree (Fig. 5B). Consistent with the analysis of the vesicular transporters immunofluorescence, the data showed that the percentage of excitatory endings was larger on the distal dendrites (59%) than on secondary (35%) or primary dendrites (43%).
The four types of inhibitory endings were also classified according to the size and shape of the synaptic vesicles as large, medium and small pleomorphic (LP, MP and SP), and flat (F) (Table 1; Fig. 4). Endings of all four kinds were found on primary and secondary dendritic segments, whereas LP and F terminals tended to concentrate on distal dendrites (Fig. 5B). Putative sources of these endings are intrinsic interneurons in the VCN, glycinergic interneurons in the dorsal cochlear nucleus and descending inputs from upper auditory nuclei (Wu and Oertel, 1986; Schofield, 1994; Ostapoff et al., 1997).
These data indicate that proximal and distal dendrites of spherical-BCs can integrate excitatory and inhibitory inputs from different sources probably including the auditory nerve.
The analysis at the electron microscopic level suggested that branches of the auditory nerve formed synapses along the dendritic tree of BCs, including the tufted distal dendrites (Fig. 3, Fig. 5, Fig. 6A–B). We confirmed this interpretation by DiI injections into the auditory nerve (Fig. 6C–F). Within the VCN core we observed thin DiI axonal collaterals (~1.1 µm in diameter) branching from thick auditory nerve fibers (~2.2 µm in diameter) (Fig. 6D–G). The thinner branches were short (~27 µm in length) and gave rise to small endings which terminated on proximal and distal dendrites of BCs (Fig. 6E, H–K, Fig. 7A–E). The analyses of the confocal orthogonal view verified that these endings were in close apposition to the tufted distal dendrites (Fig. 6K, Fig. 7F). We verified that auditory nerve endings on dendrites were VGLUT1 positive by using a triple labeled method consisting of double tract-tracing experiments using D-FITC for BCs and DiI for the auditory nerve, combined with immunofluorescence for VGLUT1 (Fig. 7; Supplemental Fig. 3A–Q). As expected, we also observed large complex endings of the endbulb of Held around the BC cell body that colabeled for DiI and VGLUT1 (Supplemental Fig. 3F–K).
As previously shown in the guinea pig (Zhou et al., 2007) we confirmed that VGLUT1 was present in endings of the auditory nerve. Previous studies in mouse and guinea pig reported that VGLUT2 preferentially labeled non-auditory excitatory inputs, some being of somatosensory origin (Nakamura et al., 2005; Zhou et al., 2007). The fact that we observed VGLUT1 and VGLUT2 labeling on BC dendrites, together with the identification at the electron microscopic level of three types of endings that did not share ultrastructural characteristics of the auditory nerve, led us to ask whether VGLUT2 puncta on dendrites had a non-cochlear origin. We combined double immunolabeling for VGLUT1 and VGLUT2 with D-FITC labeled BC dendrites. VGLUT1 and VGLUT2 puncta were found on dendrites of BCs (Fig. 8) with a similar distribution as with the single immunolabeling. The orthogonal view (Fig. 8B) and 1-µm confocal images (Fig. 8C) confirmed that dendritic VGLUT1 and VGLUT2 puncta did not colocalize. This non-overlapping pattern was also evident on BCs somata where VGLUT1 and VGLUT2 puncta distributed in opposite poles of the cell body (Fig. 8D). As described above we demonstrated that auditory nerve endings colabeled with VGLUT1 on the cell body and along the BC dendritic tree (Fig. 7A–F; Supplemental Fig. 3A–Q). When we performed the same analysis with VGLUT2, we did not observe co-localization with the DiI-labeled endbulbs on the BC somata, nor on the small auditory nerve endings on BC dendrites (Fig. 7G–H; Supplemental Fig. 3R–T). Some bipolar BCs that were more superficially located within the core showed a differential pattern of labeling for VGLUT1 and VGLUT2 (Supplemental Fig. 4). One dendritic tree of these neurons is located within the core (as is the cell body) and appears fully decorated with VGLUT1 puncta. The other dendrite is oriented toward the GCD and the distal tufted dendrites that lie near or within this layer are decorated with puncta immunolableled for VGLUT2.
In the core of the VCN, distal dendrites of BCs branch into a complex tuft which surrounds 4 to 5 adjacent BC somata forming a cluster (Fig. 1F–G–H, Fig. 2E, Fig. 6F, Fig. 9A–B). By immunoflorescence for VGLUT1, DiI injections, and electron microscopy, we showed that the endbulb was usually found in apposition to the cell body of one BC and the distal dendrites of another BC (Fig. 9A–E). To better understand the relationship between an endbulb, cell bodies and distal dendrites of BCs we performed a 3D-reconstruction of one BDA-labeled BC. The BDA-labeled dendrites were clearly identified by the characteristic electron dense material. In addition, these distal dendrites were distinguishable from other dendritic profiles because they had distinct organelles including clumps of large mitochondria, abundant membranes of smooth endoplasmic reticulum, multivesicular bodies and clathrin coated pits (Fig. 6J–K, Fig. 9D–E, Fig. 10A, Fig. 11F–H, Supplemental Fig. 5). In total we analyzed a dendritic surface area of 312 µm2 representing a volume of 110 µm3. The 3D-reconstruction revealed that the distal dendritic profile near the cell body of another BC was very complex, and that distal dendritic branches followed the same orientation and direction of the endbulb along the adjacent BC cell body (Fig. 9F, see movie 1 in supplementary data). Of a total of 45.1 µm2 auditory nerve surface, 54% was found in apposition with the adjacent cell body and 23% was in apposition with the labeled BC dendrite. We counted 16 synaptic specializations of the auditory nerve where it contacted the cell body surface, and 5 on the dendrite. Therefore within the cluster, our data show that BCs receive divergent synaptic inputs in which one single excitatory ending contact two or three different BCs (so-called synaptic dyads and triads, respectively) (Fig. 9E, Fig. 10A). Synaptic connections were classified as “dyads”, when one auditory nerve ending was observed synapsing on the cell body of one BC and the distal dendrites of another BC (Fig. 9D–E, Supplemental Fig. 5A–F). The synaptic contacts were named “triads” when a single auditory nerve ending synapsed on three different BCs through the cell body of one BC and the distal dendritic branches of two others (Fig. 10A–D). Occasionally, we also observed that the auditory nerve terminals made divergent dyadic contacts between dendrites of different BCs (Fig. 6J–K). Therefore, the auditory nerve can spread information between BCs by axosoma-dendritic and axodendro-dendritic synaptic dyads and triads. Inhibitory terminals with pleomorphic or flattened vesicles also formed dyads between the cell body of one BC and the distal dendrite of another (Fig. 10E–F, Supplemental Fig. 5A–F). Together, these results indicate the existence of a mechanism to spread excitatory (auditory nerve) and inhibitory information within the BC cluster and suggests a role of distal dendrites in the processing of auditory information.
So far, we presented data indicating that BC dendrites may play a key role in VCN function by receiving cochlear, and non-cochlear inputs. In addition, after filling BCs with neuronal tracers, we observed that their dendrites form an elaborate physical network within the nucleus. Confocal microscopic analysis of D-FITC labeled neurons suggests a close apposition between distal dendrites and/or somata of BCs (Fig. 1F–G–H, Fig. 2E, Fig. 6F, Fig. 9A–B). 3D-reconstruction of serial sections and electron microscopy analysis confirms the existence of dendro-somatic and soma-somatic junctions within the BCs cluster (Fig. 11, Fig. 13, Supplemental Fig. 5, see movie 2 in supplementary data).
Analyzing serial ultrathin sections of retrogradely labeled BCs after 3D-reconstruction, we observed that the distal dendrite generated small filopodia-like processes that intermingled within the endbulb to directly contact a unlabeled soma of a putative BC (Fig. 11A–B–C–D–E). We calculated that the total contact surface area of these filopodia on the soma plasma membrane was 10.22 µm2. These dendro-somatic junctions also occurred between larger distal dendrites and the cell body (Fig. 11F–G–H, Supplemental Fig. 5A–F). Ultrastructurally, these dendro-somatic junctions were similar to puncta adherentia (Mugnaini, 1985; Rash et al., 1998) or sarcoplasmic junctions (Henkart et al., 1976; Cant and Morest, 1979b; Landis and Reese, 1982). These junctions were usually observed in two serial ultrathin sections and had an average area of 0.065 ± 0.035 µm2. Near these membrane specializations we observed membranes of the endoplasmic reticulum, and clathrin-coated pits symmetrically distributed at both sides of the junction (Fig. 11, Supplemental Fig. 5B–E). In the dendrite, large mitochondria were also present close to the junction (Fig. 11), indicating high-energy demands associated with these membrane specializations. As discussed above, these dendrites received synaptic contacts from excitatory and inhibitory endings synapsing on the cell body (synaptic dyads and triads, Fig. 9, Fig. 10, Supplemental Fig. 5). Therefore, we have identified a specialized microcircuit formed by presynaptic and postsynaptic neuronal profiles.
3D-reconstruction of serial section electron micrographs allowed us to visualize direct and indirect relations of BDA-labeled BC dendrites. As described above, one labeled BC contacted another unlabeled soma of a putative BC through dendro-somatic junctions. Through the analysis of the serial images we also observed that the unlabeled BC formed soma-somatic junctions, identified as puncta adherentia and gap junctions, with a neighbor BC (Fig. 12, Fig. 13, Supplemental Fig. 5G–H–I). Puncta adherentia were similar to the dendro-somatic junctions and had an average area of 0.05 ± 0.03 µm2. Gap junctions were identified as sections of two adjacent plasma membranes separated by a uniform gap of roughly 3–4 nm in width (Fig. 12E–F–G, Supplemental Fig. 5H–I). These membrane specializations were seen as patches of varying sizes with an average area of 0.05 ± 0.01 µm2. The total surface area of the two BC somata was of 360 µm2 representing a volume of 1,265 µm3 (Fig. 12A, Fig. 13A–C), and the surface contact area occupied by the soma-somatic junctions was of 11.5 µm2. The specialized junctions represented 8% of this soma-somatic contact area. Of the total soma-somatic junctions analyzed (n=22), 31% were gap junctions and 69% puncta adherentia that distributed along the membrane at an interval of 0.2 µm (Fig. 12B–C, Fig. 13A–C). Areas of close membrane appositions were common between distal dendritic processes of two BCs (Supplemental Fig. 6), however we did not observe any of the specialized membrane junctions described above. The presence of these specialized membrane junctions which mediate cell adhesion and electrotonic coupling between cells, together with the divergent synaptic connections, define a BC network in the VCN (Fig. 14; see movie 1 and 2 in supplementary data).
We showed that BC dendrites receive excitatory and inhibitory synaptic endings from cochlear and non-cochlear origins. In the core of the VCN, we reported that the nest-like tufted BC dendrites radiate away from the cell body to branch in the vicinity of 4-to-5 nearby BC somata, forming neuronal clusters as suggested in the cat (Ryugo and Sento, 1991, Ostapoff and Morest, 1991). Within the neuronal cluster, our study further revealed that BCs are linked by direct neuron-neuron connections and receive divergent multiple-contact synapses (dyads and triads) from afferent inputs. Together, they constitute a network that might be important for adjusting the level of baseline activity in single BCs (Fig. 14). This network may establish a neuroanatomical basis for the unsolved puzzle of the enhanced synchronization in spherical-BCs (Joris and Smith, 2008).
Temporal coding in the auditory nerve is transformed in the cochlear nucleus, and BCs can modify the auditory nerve information rather than being a simple relay. A recent report demonstrates that the sound-evoked spike activity of BCs is the result of the integration of acoustically driven excitatory and inhibitory inputs (Kopp-Scheinpflug et al., 2002). This was anatomically correlated with the fact that the BC soma receives excitatory and inhibitory inputs from a variety of sources (Saint Marie et al., 1986; Wu and Oertel, 1986; Schofield, 1994; Mahendrasingam et al., 2004; Spirou et al., 2005). Our study further demonstrated that BC dendrites provide space for a large number of excitatory and inhibitory synaptic interactions. Since the synaptic strength becomes larger as one move along the dendrite, away from the soma (Spruston, 2000; Magee and Cook, 2000; London and Segev, 2001) the location and proportion of synaptic inputs along the dendritic tree can help in understanding the active role of dendrites in synaptic integration as shown in hippocampal pyramidal cells (Häusser, 2001). We showed that distal tufted dendrites of BCs are a major target of the incoming dendritic afferents, including the auditory nerve inputs. Therefore, BC dendrites and in particular distal dendrites provide a site for specific synaptic configuration, which can effectively influence the neuronal output. Moreover, the fact that the auditory nerve innervates BCs dendrites, might explain the small excitatory peaks observed in paired BCs-auditory nerve electrophysiological recording (Young and Sachs, 2008). These authors showed that the presence of excitatory peaks was much more frequent than expected based on the limited number of endbulbs or modified endbulbs on BCs (Ryugo and Sento, 1991).
In addition, our analyses of the excitatory synaptic markers revealed that BC dendrites receive different proportion of VGLUT isoforms according to their location in the VCN. This is of relevance considering that VGLUT isoforms are associated with different specific pathways in the cochlear nucleus (Zhou et al., 2007). We demonstrated that VGLUT1 inputs on BC dendrites are of a cochlear origin, whereas VGLUT2 inputs are non-cochlear. This VGLUT2 non-cochlear projection to BCs dendrites presumably corresponds to somatosensory projections to the VCN (Zhou et al., 2007). Supporting this somatosensory innervation, we identified on BC dendrites, synaptic endings with similar ultrastructural characteristics as mossy-like endings (Zhan and Ryugo, 2007). These data together with the fact that the majority of dendritic afferents non-colocalized for both VGLUT isoforms suggest that BCs dendrites might integrate multisensory information. The existence of multisensory integration of inputs through BC dendrites opens the possibility of modulation and synaptic plasticity in BCs, as mossy fibers adjust synaptic strength in the cerebellum (Sola et al., 2004). Other possible sources of modulation in BCs are the cholinergic (Oertel and Fujino, 2001) and serotoninergic systems (Thompson and Thompson, 2001). Besides the endbulbs and mossy-like endings on BC dendrites, we described two other types of endings with round vesicles, one of which shared similar ultrastructural parameters as cholinergic endings (Benson et al., 1996; Gomez-Nieto et al., 2008a). In addition to excitation, inhibition on BCs is important for the adequate processing of acoustic signals (Kopp-Scheinpflug et al., 2002), and neurons in the VCN are sensitive to GABA and glycine (Wu and Oertel, 1986; Caspary et al., 1993; Ebert and Ostwald, 1995). BC dendrites appeared highly decorated with VGAT endings, and we identified four types of inhibitory inputs synapsing on BC dendrites containing either flat or pleomorphic synaptic vesicles that could be considered glycinergic or GABAergic, respectively (Altschuler et al., 1986; Rubio and Juiz, 2004). In sum, the interaction and particular distribution of auditory and putative non-auditory inputs along the dendritic tree might be important for controlling the level of baseline activity in single BCs. Further electrophysiological studies are needed to determine how the inputs on dendrites shape and regulate the final BC output.
Compared to the auditory nerve and in response to low-frequency tones at high intensities, a subpopulation of BCs show enhanced phase-locking (representation of temporal information) relative to the auditory nerve. This synchronization enhancement and entrainment might be possible if BCs serve as coincidence detectors, which would indicate that the increase in the firing rate will occur as result of an increase in the correlation among the inputs to the neuron (Joris et al., 1994a, 1994b). Nevertheless, the anatomical substrate for such mechanism is still unresolved (Joris and Smith, 2008). Our study revealed that a large number of auditory nerve endings including endbulb terminals innervate BC dendrites. Particularly relevant is the fact that the endbulb makes dyads and triads in the form of axodendro-somatic synaptic contacts (our results; Ryugo and Sento, 1991; Ostapoff and Morest, 1991). Therefore, a single endbulb discharges on multiple BCs. Divergent projections are also necessary to link brain areas serially and hierarchically, as in higher auditory networks (Winer and Lee, 2007). Our study provides strong evidence of a divergent mechanism in which the incoming inputs might disperse activity among BCs. If inhibition is essential to the output discharge of BCs (Kopp-Scheinpflug et al., 2002), inhibitory inputs might similarly affect multiple neurons. In this study, we reported inhibitory inputs making divergent dyadic contacts within the BC cluster, suggesting a general mechanism to disperse activity through the nucleus. In sum, it is plausible that the divergent multiple-contact synapses of cochlear and non-cochlear inputs on BCs clusters underlie the morphological substrate for the enhanced synchronization of BCs firing in the VCN.
The spherical BC output signal exhibits enhanced synchronization compared to auditory nerve fibers (Joris et al., 1994a, b; Paolin et al., 2001; Joris and Smith, 2008). Such synchrony can occur if neurons communicate with each other using chemical synaptic activity or direct electrical coupling via gap junctions (electrical synapses) (Fukuda and Kosaka, 2000; Saraga et al., 2006). Early studies reported the existence of gap junctions in the VCN as well as in the neural circuitry of the dorsal cochlear nucleus (Sotelo et al., 1976; Wouterlood et al., 1984; Mugnaini 1985). Our study revealed that BCs are connected through three types of membrane junctions in addition to standard synapses: puncta adherentia, sarcoplasmic junctions, and gap junctions. According to our data derived from 3D-reconstructions, these junctions were found in considerable proportions and distributed strategically to connect BCs through dendrites and cell bodies. Puncta adherentia function as adhesion complexes (Peters et al; 1991; Honda et al, 2006) and they are frequently found between auditory structures (Sätzler et al., 2002; Ryugo et al., 2006). In contrast, the specialized sarcoplasmic junctions are involved in coupling excitatory events at the surface membrane with intracellular processes (Henkart et al., 1976; Landis and Reese, 1982). Synaptic dyads and triads on coupled BCs occur in close proximity to sarcolema junctions suggesting a mechanism to synchronize and modulate the surface electrical activity generated by the divergent synaptic connections. The coupling of neurons through gap junctions allows ultra-fast network synchrony (Fukuda and Kosaka, 2000; Saraga et al., 2005). Gap junctions between BC somata were identified ultrastructurally, suggesting that BCs are electrotonically coupled. The existence of coupling between BC somata which make dendro-somatic sarcolemma junctions with other BCs, together with the existence of divergent excitatory and inhibitory synapses on these neuronal clusters, lead us to argue for a BC network within the VCN (Fig. 12). Dendrites of BCs are essential structures of this network, particularly the nest-like distal dendrites which serve as hot spots in the communication between BCs. Therefore, this BC network may establish the neuronal basis for the enhanced synchronization detected in the VCN output (Joris et al., 1994a, b; Paolin et al., 2001; Joris and Smith, 2008).
We thank Nathan Maltezos for excellent assistance in the morphometric analyses of the synaptic vesicles. We gratefully acknowledge Drs. David Ryugo and Andrew Moiseff for comments and critically reading the manuscript. The authors thank NIH/NIDCD RO1 DC006881, and NSF DBI-0420580 for funds to purchase the Tecnai G2 Spirit Biotwin electron microscope.