Vesicular glutamate transporters (VGLUTs) 1 and 2 are expressed by neurons generally accepted to release glutamate as a neurotransmitter, whereas VGLUT3 appears in populations usually associated with a different classical transmitter. We now demonstrate VGLUT2 as well as the vesicular GABA transporter (VGAT) in a subset of presynaptic terminals in the dentate gyrus of the rat hippocampal formation. The terminals are distributed in a characteristic band overlapping with the outer part of the granule cell layer and the inner zone of the molecular layer. Within the terminals, which make asymmetric as well as symmetric synapses onto the somatodendritic compartment of the dentate granule cells, the 2 transporters localize to distinct populations of synaptic vesicles. Moreover, the axons forming these terminals originate in the supramammillary nucleus (SuM). Our data reconcile previous apparently conflicting reports on the physiology of the dentate afferents from SuM and demonstrate that both glutamate and GABA may be released from a single nerve terminal.
The notion that a neuron releases the same chemical transmitter from all of its synaptic terminals, has prevailed for many decades (Eccles et al. 1954). However, neuroactive peptides are known to be released along with a classical transmitter, and multiple vesicular neurotransmitter transporters colocalize and individual nerve endings release multiple transmitters (Sulzer and Rayport 2000; Boulland et al. 2004). One report documents simultaneous release of both glycine and γ-aminobutyric acid (GABA) from the same synaptic vesicle (Jonas et al. 1998). Another report shows the corelease of glutamate and acetylcholine (Li et al. 2004). In the former case, both GABA and glycine are inhibitory, in the latter, both glutamate and acetylcholine are excitatory. In contrast, the terminals of hippocampal mossy fibers have been suggested to contain and release glutamate as well as GABA (Gutierrez 2000; Bergersen et al. 2003; Safiulina et al. 2006), and the vesicular glutamate transporter (VGLUT) 3 is present in subsets of neocortical and hippocampal terminals containing the vesicular GABA transporter (VGAT) (Gras et al. 2002; Boulland et al. 2004). However, it is unknown whether these transporters reside on the same set or on different sets of synaptic vesicles.
Despite functional similarity between VGLUT1, VGLUT2, and VGLUT3, they have differential cellular localization and trafficking, suggesting isoform-specific physiological roles (Fremeau et al. 2004; Chaudhry, Boulland, et al. 2008). In the central nervous system, VGLUT1 and VGLUT2 show complementary patterns of expression in subsets of glutamatergic neurons (Fremeau et al. 2001; Varoqui et al. 2002). VGLUT2 is enriched in glutamatergic terminals with high probability of transmitter release and VGLUT2-specific short-term plasticity has been demonstrated (Fremeau et al. 2004). Moreover, VGLUT2 has been detected transiently at high levels in specific brain regions during early postnatal development, and it has been suggested to play a regulatory role in development (Boulland et al. 2004; Nakamura et al. 2005).
In the hippocampus, VGLUT1 is the predominant isoform, but VGLUT2-containing terminals are enriched in bands that partially overlap with the principal cell body layers (in CA2 and in the dentate gyrus). In particular, VGLUT2 immunoreactive terminals are characteristically distributed in a band (in the following designated the “VGLUT2-band,” see Fig. 1A), coinciding with the outer part of the granule cell layer (GCL) and the immediately adjacent part of the inner zone of the molecular layer of the dentate gyrus (Fremeau et al. 2001). We have now characterized the VGLUT2-labeled structures in this band through postnatal development. We find that VGLUT2 colocalizes with VGAT in terminals of axons originating in SuM, which make both asymmetric and symmetric synapses onto the somatodendritic compartment of dentate granule cells. We also show that VGLUT2 and VGAT proteins reside on distinct populations of synaptic vesicles, indicating the potential for differential release.
Material and Methods
Sodium dodecyl sulfate (SDS), SuperSignal West Pico Chemiluminescent Substrate and dimethyl pimelimidate-2HCl (DMP) were from Pierce (Rockford, IL). 10% polyacrylamide Criterion precast gels, nitrocellulose sheets (ECL), electrophoresis equipment and molecular mass markers (Precision plus) were from Biorad (Hercules, CA). Gel documentation equipment was from UVP Inc. (Upland, CA). Magnetic beads coated with protein A were from Dynal (Invitrogen) (Carlsbad, CA). Fluoromount G (water base), uranyl acetate, lead citrate, glutaraldehyde (GA), and nickel grids (for electron microscopy [EM]) were from Electron Microscopy Science (Washington, PA). Phaseolus vulgaris leucoagglutinin (Pha-l) was obtained from Vector (Burlingame, CA) and biotinylated dextran amine (BDA, 10 kDa) was from Molecular Probes (Leiden, The Netherlands). Components of Durcupan ACM, sodium azide (NaN3), diaminobenzidine (DAB), human serum albumin (HSA), bovine serum albumin (BSA), sodium ethylenediamine tetraacetate (EDTA), polyethylene glycol (PEG), Trizma base and Tris–HCl were obtained from Sigma Aldrich (Oslo, Norway). Newborn calf serum (NCS) and normal goat serum (NGS) were from Gibco (Paisley, UK). Paraformaldehyde (PFA) was from TAAB (Reading, UK). Ethanolamine, hydrogen peroxide and propylene oxide were obtained from Fluka (Buchs, Switzerland). Streptavidin-biotinylated horseradish peroxidase complex was from Amersham (Buckinghamshire, UK).
Source, concentrations used and other relevant information on the primary antibodies are summarized in Table S1. Characterization of in-house generated antibodies has been described in previous publications (Bellocchio et al. 1998; Chaudhry et al. 1998; Fremeau et al. 2001; Boulland et al. 2004). Note that all antibodies recognize cytoplasmic epitopes allowing vesicle immunoisolation. The specificity of the commercial antibodies, obtained from Chemicon International (Temecula, CA) against VGLUT2, VGLUT1, and GAD65 and from Synaptic Systems against VGAT (Goettingen, Germany), was confirmed on immunoblots of SDS extracts from brain, cerebellum and hippocampus (data not shown). Mouse anti-CCK (cholecystokinin) (Anti-Gastrin/CCKm#9303) was a kind gift from CURE/Digestive Diseases Research Center/RIA Core, NIH Grant # DK41301. Table S2 summarizes the characteristics of the secondary antibodies: their source, recognizing epitope, fluorochromes/gold particles etc conjugated to the secondary antibodies as well as the concentrations applied in different immunoreactions.
Animals, Tissue Preparation, and Synaptic Vesicle Fractionation
Wistar rats, kept at the animal facilities at the Institute of Basic Medical Sciences, University of Oslo (Norway) or at VU University Medical Center (Amsterdam, The Netherlands), were handled according to the European regulations and under veterinary agreement and supervision. For immunocytochemistry, animals were treated as described earlier (Boulland et al. 2002). Briefly, the animals were deeply anaesthetized and fixed by transcardiac perfusion with 4% PFA in 0.1 M sodium phosphate buffer pH 7.4 (NaPi) except for animals injected with a tracer (see below). For postembedding immunogold, animals were perfusion fixed with PFA/picric acid (4/0.02%) in 0.1 M sodium acetate buffer, pH 6.0 (200 mL, 5 min), followed by the same fixatives in 0.1 M sodium carbonate buffer, pH 10.5 (400 mL, 20 min). Brains were dissected out and postfixed overnight. For synaptic vesicle preparations, animals were killed by stunning and decapitation. Purification of synaptic vesicles was done as previously described (Bogen et al. 2006). Briefly, the 2 whole hippocampi (to ensure that we did not miss any VGLUT2-containing terminals) were quickly dissected out, homogenized in 0.32 M sucrose at 4 °C and centrifuged for 10 min at 1000 × g. The supernatant was in turn centrifuged for 30 min at 21 000 × g. The new pellet was resuspended in ice cold water and further centrifuged for 30 min at 21 000 × g. Finally, the supernatant was diluted in HEPES/K+ tartrate to a final concentration of 0.1 M and processed for immunoisolation or for EM.
Electrophoresis and Immunoblotting
SDS-PAGE (polyacrylamide gel electrophoresis) was done at constant voltage (200 V) for 1 h. Gels were electroblotted onto ECL nitrocellulose membranes for 2 h at 50 V. Blots were dried and rinsed once with 0.1 M Tris/HCl, 0.15 M NaCl, 0.05% Tween 20, pH 8 (TBSTw). Blots were blocked for unspecific binding sites with 1% BSA in TBSTw for an hour and incubated overnight, at room temperature (RT), with primary antibodies (Table S1) diluted in the same solution. Blots were rinsed in TBSTw and subsequently incubated with secondary antibodies (Table S2). Another rinsing was followed by incubation with streptavidin–horseradish peroxidase complex (1/500) diluted in the same solution. Blots were rinsed, quickly blotted on absorbent tissue paper, incubated for 2.5 min in chemiluminescence reactants and placed in the chamber of the gel documentation system (see Materials).
Immunoperoxidase Staining and Embedding for Electron Microscopy
Immunoperoxidase staining was done as previously described (Boulland et al. 2002). Briefly, Vibratome sections were treated with 1 M ethanolamine-HCl (EtA) and incubated in 1% of H2O2 NaPi, blocked with 10% NCS in 0.1 M Tris/HCl (pH 7.4), 0.3 M NaCl (TBS) before overnight incubation with primary antibodies diluted in TBS (Table S1). Sections were rinsed, incubated with secondary antibodies diluted in TBS (Table S2) and developed with the streptavidin–peroxidase system using DAB and 0.1% of H2O2. Sections were treated with 1% OsO4 in NaPi, dehydrated in graded ethanols and propylene oxide, and embedded in Durcupan ACM. Ultrathin sections were cut, lightly contrasted with uranyl acetate and lead citrate and observed in a Technai CM10 electron microscope (Fei Company, Hillsboro, OR).
Immunofluorescence staining was done as described previously (Boulland et al. 2004). Briefly, sections were treated with EtA, blocked for 1 h with 10% (v/v) NGS and 3% (w/v) BSA in 0.1 M Tris base, 0.3 M of NaCl 0.5% (w/v) Triton X-100, pH 7.4 (TBSTr) and incubated overnight with primary antibodies (Table S1) diluted in TBSTr containing 3% NGS and 1% BSA. After rinsing, sections were incubated with secondary antibodies coupled to fluorescent dyes (Table S2) and rinsed prior to mounting on slides with Fluoromount G water base. Sections were observed with a Zeiss Axioplan 2 or Axiovert equipped with a Pascal 5 LSM, a Zeiss 510 Meta LSM or a Leica DMRI2 equipped with SP2-AOBS scanner head. To exclude artifacts, Z-stacks were acquired for points showing colocalization. In some cases Z-stacks were deconvoluted with Huygens Essential software (Scientific Volume Imaging, Hilversum, The Netherlands) and a Silicon Graphics processor to be three-dimensional reconstructed with Amira TGS series software (Mercury, Chelmsford, MA). Control sections incubated without primary antibodies showed faint, diffuse staining.
Magnetic beads coated with protein A were rinsed in NaPi and incubated overnight with antibodies (Table S1) diluted in NaPi pH 8, at RT. Beads were rinsed 3 times in 0.2 M triethanolamine, pH 8.2, incubated 30 min with 20 mM DMP followed by 15 min incubation with 50 mM Tris–HCl pH 7.4 and rinsed 3 times with NaPi pH 8. Beads were blocked for unspecific binding sites with 1% BSA in NaPi pH 8 for 1 h at RT and exposed overnight to undiluted crude synaptic vesicles (LS1) fraction at 4 °C overnight. The supernatant fraction was collected while the beads were rinsed 3 times in NaPi pH 8. Finally, SDS and SDS-sample buffer were added to both fractions for SDS-PAGE.
Electron Microscopy of Vesicles
Synaptic vesicles from hippocampal LS1 fractionation were diluted with 0.1 M K+ tartrate/EDTA (1/3–1/10–1/30). Thirty microliters of sample was applied onto formvar and carbon coated nickel grids (see Materials). For a more efficient adsorption of vesicles onto formvar, the sample applied onto the grids was allowed to evaporate until only a film of solution was formed at the surface of the grid. Grids were immediately placed, overnight or no less than 4 h, in 4% PFA. Grids were rinsed 3 times in NaPi, treated with EtA for 10 min, rinsed 3 times in NaPi, blocked for 1 h with 10% NCS in TBS and incubated overnight with primary antibodies (Table S1) diluted with 1% NCS, 0.01% NaN3 in TBS. After rinsing with 1% NCS in TBS, grids were incubated with secondary antibodies coupled to gold particles of different sizes (Table S2) in TBS containing 1% NCS and PEG (5 mg/10 mL) for 1–2 h. Grids were rinsed and postfixed with 4% PFA and 2.5% GA in NaPi for at least 10 min, rinsed in pure water, postfixed again with 1% OsO4 in NaPi for 10 min and rinsed in pure water. For negative contrasting, grids were rinsed with 3 drops of 1% uranyl acetate in water and immediately dried on absorbent paper as previously described (Ohi et al. 2004). Altogether, 2617 vesicles were investigated. A large number of vesicles from different grids and experiments were investigated to reduce the chances of biased vesicle sampling. Furthermore, there was enough space around the vesicles to allow free access for the antibodies to bind to their specific antigens. The “halo artifact” found around the gold particles in Figure 4A is due to negative staining procedure (for details see Ohi et al. 2004).
Freeze substitution and embedding in Lowicryl was done as previously described (Chaudhry et al. 1995). Ultrathin sections were treated with borohydride/glycine in TBSTr, rinsed and blocked with 2% HSA in TBSTr for 1 h and incubated overnight with primary antibodies (Table S1) in the same solution. Subsequently, sections were rinsed and incubated no less than 1 h 30 min with secondary antibodies (Table S2). After another round of rinsing, sections were lightly contrasted with uranyl acetate and lead citrate, rinsed with water, dried, and observed.
Stereotaxy and Tracing
Animals were deeply anesthetized by injection (i.p./i.m.) of a 4:3 mixture of ketamine (Aescoket, Boxtel, The Netherlands) and xylazine (Rompun, Bayer, Mijdrecht, The Netherlands) to a total dose of 1 mL/kg bodyweight and mounted in a stereotaxic frame (Konings, Arnhem, The Netherlands). The bregma was used as reference and stereotaxic coordinates were calculated according to a rat brain atlas (Paxinos and Watson 1998). A solution of 5% Pha-l in TBS pH 7.4 or 2.5% BDA in TBS pH 7.4 was used for iontophoresis (positive pulsed 6 A and 7.5 A DC current for BDA and Pha-l respectively with 7 s on/off for 10 min) delivery in the brain at bregma −3.9 (mm), lateral −1.6, lateral angle 10°, depth between −8.0 and −8.4. A week following the delivery, animals were deeply anaesthetized and perfusion fixed, as described above, with 4% PFA and 0.1% GA in 0.125 M NaPi, pH 7.4. Brains were dissected out, postfixed for 1 h at 4 °C and sectioned into 200-μm-thick sections on a vibrating microtome (Leica VT 1000 S, Leica Microsystems GmbH, Germany). Sections were cryoprotected before storage at −20 °C, using a solution of 20% glycerol and 2% dimethyl sulfoxide (DMSO, Acros Organics, NJ) in 0.125 M NaPi, pH 7.4. The 200-μm-thick sections were resectioned at 10 μm with a freezing sliding microtome (American Optical Corp., NY) to be processed for immunofluorescence as described above.
VGLUT2 and VGAT Colocalize in Terminals Making Asymmetric as well as Symmetric Synapses
By immunocytochemical investigation we explored the distribution of VGLUT2 in the hippocampus. In agreement with previous reports (e.g., Fremeau et al. 2001), light microscopy of immunoperoxidase-stained brain sections revealed VGLUT2-labeled puncta in the VGLUT2-band in area dentata (Fig. 1A). EM confirmed VGLUT2 immunoreactivity in a subset of axon terminals making asymmetrical synapses onto dendritic shafts and spines (Fig. 1D,E). In addition, strong VGLUT2 staining was detected in a subset of terminals making symmetric synapses onto the cell bodies of granule cells (Fig. 1C) and their proximal dendrites (Fig. 1B).
To investigate the phenotype of VGLUT2-containing terminals, we performed double and triple immunofluorescence labeling of hippocampal sections with VGLUT2 antibodies and markers for GABAergic synapses. VGLUT2 colocalizes with VGAT in a subset of puncta resembling terminals in the VGLUT2-band (Fig. 2A) and in the pyramidal cell layer of the CA2 region (Fig. 2B). Furthermore, the VGLUT2 and VGAT-containing terminals also stain for GAD65 (Fig. 2C), the crucial enzyme in GABA biosynthesis. Three-dimensional reconstruction of Z-stack images shows colocalization in all planes (inset, Fig. 2C), ruling out the possibility of different proteins in closely apposed terminals falsely appearing to be colocalized with VGLUT2. VGLUT2-mediated, and VGAT-mediated corelease of glutamate and GABA is further supported by triple labeling experiments demonstrating that the postsynaptic targets of VGLUT2+/GAD65+ terminals contain N-methyl D-aspartate (NMDA) receptors as well as GABAA receptors (Fig. 2D,E). Three-dimensional reconstructions also illustrate double stained terminals impinging on postsynaptic elements containing NMDA and GABAA receptors (insets, Fig. 2D,E). We also detect gephyrin and PSD95 postsynaptic to VGLUT2+/VGAT+ terminals (data not shown). Finally, immunogold electron microscopy results indicate corelease of glutamate and GABA as they show pronounced colabeling for VGLUT2 and VGAT in nerve terminals making asymmetric (Fig. 2F) and symmetric (Fig. 2G) synapses onto somatodendritic domains of dentate granule cells.
The Proportion of VGLUT2+/VGAT+ Terminals Increases through Postnatal Development
To explore a potential for corelease of glutamate and GABA in synaptogenesis, we quantified the appearance, in the VGLUT2-band, of terminals singly and doubly labeled for VGLUT2 and VGAT at P7, P14 and in the adult. At P7, the number of doubly stained terminals in the measured area (see Fig. S1), expressed as the percentage compared with the total number of VGLUT2-containing vesicles in the adult, is only 5% (Fig. 3). In contrast, there are 4 times more terminals singly stained for VGLUT2 than doubly stained for VGLUT2 and VGAT (Fig. 3). However, the number of terminals stained only for VGLUT2 increases more slowly than the number of doubly stained terminals. In the adult, the number of terminals containing both VGLUT2 and VGAT in the field sampled surpasses the number of terminals expressing only VGLUT2 (Fig. 3).
VGLUT2 and VGAT Localize to Distinct Subpopulations of Synaptic Vesicles
Immunogold EM analysis of intact brain sections cannot distinguish between the colocalization of VGLUT2 and VGAT on the same vesicles, as opposed to localization on different vesicles, as the resolution of the immunogold method is similar to the diameter of a synaptic vesicle and to the intervesicular distance. We therefore prepared crude synaptic vesicles (LS1) from hippocampus, spread them on EM grids and double labeled them for VGAT and VGLUT2 using secondary antibodies coupled to gold particles of different sizes (a representative micrograph is shown in Fig. 4A). As expected from the immunofluorescence, a higher proportion of the vesicles was labeled by large particles (for VGAT) than by small particles (for VGLUT2). No labeling was detected when the primary antibody was omitted (data not shown). Surprisingly, no double labeled vesicles could be identified among the 2617 vesicles examined, suggesting that VGAT and VGLUT2 are sorted to different sets of vesicles.
To test this further, we used immunoisolation from LS1 hippocampal fractions to enrich VGLUT2-containing or VGAT-containing vesicles. The enriched fractions (B), together with depleted supernatant fractions (S) and total crude vesicular fractions (T) were immunoblotted for VGLUT1, VGLUT2, VGAT, or other vesicle-associated proteins (Fig. 4B). The fraction immunoisolated for VGLUT2 showed strong labeling for VGLUT2, but none for VGAT or VGLUT1 (left panel). The supernatant fraction showed no VGLUT2 staining, but labeled for VGAT and VGLUT1, indicating efficient isolation of all VGLUT2-containing vesicles. Similarly, VGAT labeling of the fraction immunoisolated for VGAT (right panel) was nearly as strong as the labeling of the total fraction, and the supernatant showed depletion, again consistent with efficient immunoisolation. VGLUT1 or VGLUT2 staining was not found in the VGAT-enriched fraction, but was found in the supernatant fraction at the same level as in the total vesicular fraction. All 3 fractions stained for synaptophysin (p38) or VAMP2, consistent with the presence of synaptic vesicles. The immunoisolation results thus agree with the immunogold staining of purified synaptic vesicles, verifying that VGAT and VGLUT2 localize to different sets of synaptic vesicles.
VGLUT2 and VGAT-Containing Terminals are Projections from the SuM
To determine the cellular origin of terminals colocalizing VGLUT2 and VGAT we used multiple staining for VGLUT2, VGAT and markers of specific subsets of GABA interneurons. A subpopulation of basket cells projects to the GCL and expresses VGLUT3 (Boulland et al. 2004; Somogyi et al. 2004). Although VGLUT3 colocalizes extensively with VGAT in the GCL (Fig. 5A), the VGLUT3+ terminals have no detectable VGLUT2 immunoreactivity (Fig. 5B). The VGLUT3−, CCK+ subpopulation of basket cells also project to the cell bodies and proximal dendrites of the granule cells (Boulland et al. 2004; Somogyi et al. 2004). However, hippocampal sections triple stained for CCK, VGLUT2, and VGAT did not show triple colocalization (Fig. 5C). CCK and VGAT colocalized in a subset of terminals (bright blue, double arrowheads; right inset), but terminals containing both VGLUT2 and VGAT (yellow, single arrowheads; left inset) were not stained for CCK. The VGLUT2+ terminals therefore do not belong to the VGLUT3 or CCK expressing populations of interneurons.
In a previous report, VGLUT2-containing terminals in the hippocampus were suggested to originate in the supramammillary nucleus (SuM), based on their characteristic distribution conforming with that of the axonal input from SuM (Fremeau et al. 2001). To test this possibility, anterograde tracers (BDA or Pha-l) were injected stereotaxically in the SuM (Fig. 5D). Additional staining for VGLUT2 followed by confocal microscopy, deconvolution, and three-dimensional reconstruction showed extensive colocalization in varicosities and in terminal-like structures in the VGLUT2-band, whereas intervening axon stretches were devoid of VGLUT2 immunoreactivity (Fig. 5E), consistent with a scarcity of synaptic vesicles in passing axons. Furthermore, the tracer and VGLUT2 colocalized with VGAT (Fig. 5F) as well as with GAD65 (data not shown). Thus, a population of neurons in the SuM projects to the VGLUT2-band and sorts VGLUT2 and VGAT into distinct populations of synaptic vesicles.
Previous studies have reported ubiquitous coexpression of VGLUT3 with vesicular transporters for GABA, monoamines and acetylcholine (Gras et al. 2002; Boulland et al. 2004; Seal and Edwards 2006). However, there is also evidence for VGLUT2 colocalizing with other vesicular transporters within a cell or even within a terminal. Despite the complementary pattern of expression initially reported for VGLUT1 and 2 in subsets of glutamatergic nerve terminals (Fremeau et al. 2001; Varoqui et al. 2002), colocalization of these isoforms has been demonstrated in small populations of nerve terminals (Boulland et al. 2004; De Gois et al. 2005). VGLUT2 is also transiently expressed at high levels in specific brain regions during early postnatal development (Boulland et al. 2004). The switch between VGLUT2 and VGLUT1 indicates additional age- and location-specific roles for VGLUT2-mediated glutamate release (Fremeau et al. 2004; Nakamura et al. 2005). Likewise, rat alpha-motoneurons release acetylcholine at the neuromuscular junction, whereas VGLUT2-containing axon collaterals mediate the excitatory feedback to the inhibitory Renshaw cells in the spinal cord (Nishimaru et al. 2005). Thus, our data are consistent with other observations indicating release of multiple transmitters from a single neuron.
There are a number of reports supporting the release of glutamate and GABA from the same terminal. A family of postsynaptic proteins, the neuroligins, controls the formation and functional balance between excitatory and inhibitory synapses (Prange et al. 2004; Chih et al. 2005; Dean and Dresbach 2006). Neuroligin 2 accumulates specifically postsynaptic to GABAergic terminals but also allows presynaptic glutamatergic differentiation (Graf et al. 2004). Moreover, glutamate release may itself induce postsynaptic differentiation and assembly of its receptors at non-glutamatergic synapses (Pizzi et al. 2006; Dean and Dresbach 2006). Although observations differ in detail, markers for glutamatergic and GABAergic activity have been demonstrated at the mossy fiber-to-CA3 synapses in the hippocampus (Ottersen and Storm-Mathisen 1984; Sandler and Smith 1991; Bergersen et al. 2003; Gutierrez 2005; Safiulina et al. 2006; Uchigashima et al. 2007). The parent cell bodies of the mossy fibers, the granule cells, which are the targets of the SuM afferents, express GABAA receptors (Sperk et al. 1997) as well as NMDA receptors (Thompson et al. 2002), suggesting responsiveness to glutamate and GABA. As glutamate-mediated activation of NMDA receptors is required for the refinement of GABAergic synapses (Anderson et al. 2004; Dean and Dresbach 2006), the observed VGLUT2 could mediate glutamate release to stabilize VGAT-mediated inhibitory input to the granule cells. Indeed, our data show that a large portion of VGLUT2+/VGAT+ terminals make symmetric synapses onto the cell bodies and dendrites of the dentate granule cells.
The finding that tracer injected into the SuM colocalizes with VGLUT2 and VGAT in the VGLUT2-band is consistent with several reports on hippocampal projections from the SuM (Mizumori et al. 1989; Borhegyi and Leranth 1997) and with the Allen Brain Atlas of in situ hybridization for VGLUT2 in SuM. Consistent with an origin in the SuM, VGLUT2, and VGAT do not colocalize with markers for local interneurons in the VGLUT2-band. Further, our finding that VGLUT2 and VGAT colocalize in the terminals derived from the SuM reconciles previous reports on either inhibitory or excitatory transmission. Potent, long-lasting inhibition of granule cells has been observed after stimulation of the SuM (Segal 1979). However, we also found VGLUT2-containing SuM derived terminals at asymmetric synapses, consistent with previous observations that SuM fibers make asymmetric synapses (Magloczky et al. 2000) and accumulate 3H-D-aspartate (Kiss et al. 2000) suggesting glutamatergic transmission.
Finally, we show that VGLUT2 and VGAT are segregated on different synaptic vesicles. Such differential localization has not been shown before and suggests the potential for differential regulation of glutamate and GABA release. This dual phenotype may be important for homeostatic plasticity (Turrigiano and Nelson 2004) and could be involved in regulating the ratio between the glutamatergic and GABAergic input (Liu 2004) at different physiological and pathological conditions. VGLUT2 and VGAT may be regulated differentially in response to hormones and drugs (Ottem et al. 2004; De Gois et al. 2005; Chaudhry, Edwards, et al. 2008) and VGLUT2 transcription is upregulated by growth hormones (Aihara et al. 2000). A dual phenotype may represent an advantageous reduction in metabolic cost and in error of signaling as differential regulation can be achieved by only one cell (Somogyi 2006).
Hippocampal theta rhythm is associated with successful learning, REM sleep and exploratory locomotor activity (Buzsaki 2002). As theta activity may be conveyed from SuM through its projections to the dentate gyrus (Vertes and Kocsis 1997; Pan and McNaughton 2004), the differential release of an excitatory and an inhibitory transmitter from this projection may contribute to hippocampal information transfer and synaptic plasticity.
Research Council of Norway.
Conflict of Interest: None declared.
- presynaptic terminals
- gamma-aminobutyric acid
- cell nucleus
- dentate gyrus
- hippocampus (brain)
- membrane transport proteins
- nerve endings
- synaptic vesicles
- gaba transporter
- vesicular glutamate transport proteins
- hippocampal formation