|Home | About | Journals | Submit | Contact Us | Français|
Adult neurogenesis occurs in the hippocampus and the olfactory bulb of the mammalian CNS. Recent studies have demonstrated that newborn granule cells of the adult hippocampus are postsynaptic targets of excitatory and inhibitory neurons, but evidence of synapse formation by the axons of these cells is still lacking. By combining retroviral expression of green fluorescent protein in adult-born neurons of the mouse dentate gyrus with immuno-electron microscopy, we found output synapses that were formed by labeled terminals on appropriate target cells in the CA3 area and the hilus. Furthermore, retroviral expression of channelrhodopsin-2 allowed us to light-stimulate newborn granule cells and identify postsynaptic target neurons by whole-cell recordings in acute slices. Our structural and functional evidence indicates that axons of adult-born granule cells establish synapses with hilar interneurons, mossy cells and CA3 pyramidal cells and release glutamate as their main neurotransmitter.
Increasing evidence supports the hypothesis that neurogenesis is a physiologically important event in the adult hippocampus, but the precise role of newly generated cells in the hippocampal network remains unknown1-5. Understanding how newborn granule cells may contribute to information processing in the adult hippocampus requires a detailed analysis of their connectivity and function. Previous studies have clearly demonstrated that newborn neurons in the adult hippocampus receive morphologically mature axo-somatic, axo-dendritic and axo-spinous synapses and that those inputs arising from the entorhinal cortex and local inhibitory interneurons are fully functional6-13. To influence information processing in the hippocampal network, new granule cells must also contact the appropriate neuronal targets, but this capability has not yet been demonstrated because of technical challenges.
In this study, we searched for structural and functional evidence for output synapses of adult-born granule cells. Mossy fibers, the axons of granule cells, are known to contact inhibitory interneurons of the granule cell layer (GCL), hilus and CA3 area as well as excitatory mossy cells in the hilus and CA3 pyramidal cells14. We used retroviral expression of the green fluorescent protein (GFP) to identify newly generated neurons15 and carried out immuno-electron microscopy of labeled cells. We found axo-dendritic and axo-spinous synapses that were formed by labeled terminals with hilar interneurons and CA3 neurons. Those contacts started to form by the third week of neuronal development, even before newborn cells had reached complete maturation. To determine whether these synapses are functional, we expressed the light-gated cation channel channelrhodopsin-2 (ChR2) in the progeny of dividing neural precursors of the adult dentate gyrus. Postsynaptic responses evoked by light stimulation of newborn granule cells were then recorded in acute hippocampal slices. We found that axons of newborn neurons released glutamate onto a variety of postsynaptic neuronal targets in the dentate gyrus.
Injections of the retroviral vector CAG-GFP into the dentate gyrus of 6-to 7-week-old mice resulted in the labeling of newly generated granule cells with GFP6,15. We observed axons and axon terminals in the hilus and the CA3 areas using confocal microscopy. We found that the mossy fiber boutons in CA3 were substantially larger at 28 and 75 days post viral injection (dpi) than those at 17 dpi (Fig. 1). At later time points, the mossy fiber boutons had long extensions, as previously described16. The size of the GFP-labeled mossy fiber boutons at 75 dpi was similar to that of the mature neurons born during early postnatal development ((P = 0.18) Fig. 1b,c,d,f). Notably, the size of the mossy fiber boutons in the hilus did not change significantly during the maturation of newborn neurons (P = 0.65). Furthermore, the average size of the mossy fiber boutons in CA3 was significantly larger than that in the hilus (P < 0.001, Student’s t test), except for new neurons at 17 dpi.
We then used immuno-electron microscopy for GFP and serial-sectioning to analyze boutons that were formed by labeled axons. A total of 44 axonal boutons were sectioned, all of which formed synapses on either CA3 or hilar neurons (Fig. 2). We found synapses on thorny excrescences, dendritic spines or dendrites, which were unambiguously identified by the presence of at least four presynaptic vesicles within 100 nm of the active zone, a well-defined synaptic cleft, and postsynaptic density. We observed large mossy terminals synapsing with thorny excrescences (Fig. 2a,b) as well as boutons synapsing with spiny dendrites in the CA3 area (Fig. 2c) and with thorny excrescences in the hilus (Fig. 2). In addition, the thin GFP-positive axons (~0.1 to 0.2 μm in diameter) also formed en passant synapses with thin dendritic shafts at each of the time points in both CA3 and the hilus (Fig. 2d,e). In our serial sections, these thin dendrites were aspiny, suggesting that they arose from GABAergic interneurons17. These aspiny dendrites were postsynaptic to both the GFP-positive axons and other thin axons that lacked GFP (Fig. 2d,e). Furthermore, there were no intervening glial processes between the thin GFP-positive axons and the unlabeled axons. Further evidence of newborn granule cells forming synapses with GABAergic interneurons is provided below. Thin GFP-positive axons also synapsed onto the spines of thin dendrites, probably arising from spiny interneurons located in the stratum lucidum of CA3 (ref. 18).
We next examined the connectivity of GFP-positive mossy fibers that synapsed with dendrites of CA3 pyramidal cells (Fig. 3). At 17 dpi, 6 out of 8 GFP-positive mossy fiber boutons had formed synapses with the dendritic shafts of pyramidal cells (Fig. 3a,b), whereas adjacent thorny excrescences and their spines19 were contacted by large, unlabeled mossy fiber boutons. The other two GFP-positive boutons at 17 dpi synapsed onto thorny excrescences that were shared with GFP-negative boutons. At 28 dpi, 10 out of 11 GFP-positive mossy fiber boutons shared both the thorny excrescence and spines with one or two GFP-negative mossy fiber boutons (Figs. 2a and 3c,d), whereas one bouton had synapsed directly with the dendritic shaft of a pyramidal neuron. At 75 dpi, 10 out of 12 GFP-positive mossy fiber boutons were entirely associated with one individual thorny excrescence with spines, and this synaptic complex was enveloped by astrocytic processes (Fig. 3e,f). The other two boutons shared a spine with a GFP-negative terminal.
The number of spines enveloped by each GFP-positive mossy fiber bouton increased significantly from 0.1 at 17 dpi (n = 7) to 2 at 28 dpi (n = 9) and 3.8 at 75 dpi (n = 10) (17 versus 27 dpi, P<0.05; 28 versus 75 dpi, P < 0.05). In comparison, there were 2.9 spines per GFP-negative bouton (n = 11; Fig. 3g). The total cross-sectional area of GFP-positive mossy fiber boutons also substantially increased over this time period, with the greatest increase occurring between 17 and 28 dpi (Fig. 3h). Although some of the presynaptic vesicles were obscured by the immunostaining for GFP, which resulted in an underestimation of their total number, we found a significant increase in presynaptic vesicles from 60 at 17 dpi to 120 at 28 dpi, 179 at 75 dpi and 240 in GFP-negative boutons (17 versus 75 dpi, P < 0.05; 75 dip versus GFP negative, P < 0.005; Fig. 3i). Large dense-core and large clear-core vesicles were also present at all of the time points examined. The number of active zones also increased from 3 at 17 dpi to between 5 and 6 at 75 dpi and in GFP-negative boutons (17 versus 75 dpi, P < 0.05; Fig. 3j). A model for the maturation of axonal boutons is shown in Supplementary Figure 1 online.
We then investigated whether the output synapses formed by newly generated neurons are functional. The most direct approach would involve double patch recordings of presynaptic GFP-positive cells connected to target neurons in acute brain slices. However, such paired recordings are technically challenging, as each cell only contacts a small fraction of all possible targets20,21 and axons may be severed during slice preparation. To overcome this problem, we developed a retroviral vector carrying ChR2 (refs. 22,23) and a fluorescent reporter gene (GFP or monomeric red fluorescent protein 1, mRFP1) with which to transduce the progeny of dividing progenitor cells of the dentate gyrus of adult mice. In these conditions, large populations of adult-born granule cells could be readily activated by blue light in acute hippocampal slices (Supplementary Fig. 2 online). Thus, ChR2 can now be used to evoke neurotransmitter release from the mossy fiber terminals of a large population of adult-born granule cells to find connected target neurons using whole-cell patch-clamp recordings.
To determine whether new neurons establish functional synapses with their targets, we used brain slices (3–4 months after retroviral labeling) containing several dozen fluorescent cells to assess responses to light-driven mossy fiber stimulation in randomly selected putative postsynaptic cells. Cellular targets of mossy fibers are distributed in the hilus, GCL/hilar border and CA3 region, and include GABAergic interneurons, mossy cells and pyramidal neurons21. We carried out whole-cell recordings in randomly selected neurons of these areas and determined their identity by considering electrophysiological and anatomical parameters, including location of the soma, morphology, firing properties and spontaneous synaptic activity (Supplementary Methods, Supplementary Fig. 3 and Supplementary Table 1 online). Briefly, interneurons are located in the hilus and the GCL/hilar border, spike with high frequency and have a pronounced afterhyperpolarization (AHP); mossy cells are located in the hilus, reach lower firing frequencies, lack AHP and show a high frequency of spontaneous synaptic activity; and pyramidal cells are found in the CA3 pyramidal layer and display intermediate spiking and AHP properties.
We tested light-induced postsynaptic responsiveness in 107 putative target cells that included 76 interneurons, 16 mossy cells, 11 pyramidal cells and 4 unclassified neurons (Fig. 4, an example of a successful recording obtained from a cell identified as a GABAergic interneuron of the GCL/hilar border is shown in Fig. 4a–e). Prolonged light stimulation of the slice elicited a high frequency of inward synaptic currents (Fig. 4c). Repetitive light pulses of short duration delivered at low frequency (10 ms, 0.2 Hz) elicited postsynaptic currents (PSCs) that were time-locked to the stimulus (Fig. 4d,e). For each cell, we averaged tens to hundreds of stimuli presentations to distinguish light-evoked PSCs from uncorrelated spontaneous activity (Fig. 4k,l). PSCs were successfully evoked in 14 target cells (Fig. 4p) that corresponded to GABAergic interneurons (n = 11; two examples in Fig. 4a–e and Fig. 4f–i), one mossy cell (Fig. 4j–l), one CA3 pyramidal cell (Fig. 4m–o) and one unclassified neuron.
Light-evoked synaptic responses should be subject to blockade by the appropriate receptor antagonists. Mossy fiber stimulation should typically evoke glutamatergic responses, although other possibilities have also been proposed, especially in developing neurons or after seizures24,25. Polysynaptic inhibitory and excitatory signals can also be observed if mossy fiber activation recruits local interneurons or mossy or pyramidal cells, respectively26-28. Both types of responses ultimately depend on glutamate release from mossy fibers and should therefore be blocked by the mixed AMPA/NMDA receptor antagonist kynurenic acid (Kyn, 4–10 mM). Light-evoked responses of adult-born neurons showed a pharmacological profile that was consistent with this notion, as Kyn consistently abolished evoked PSCs in a reversible manner (Fig. 5). The metabotropic GluR agonist DCG-IV, which reduces neurotransmitter release at mossy fiber terminals29, also blocked PSCs (example shown in Fig. 5a,b; similar results were obtained in two experiments).
We also tested the effect of the GABAA receptor antagonist bicuculline methiodide (BMI, 20 μM) to identify putative feedforward inhibition. In 3 out of 5 cases, BMI elicited little or no changes in PSC amplitude, whereas subsequent addition of Kyn completely abolished the responses, confirming their glutamatergic nature (Fig. 5c,d,f). In the remaining two cases, BMI blocked a substantial proportion of the PSC, but complete block was also observed in the presence of Kyn, consistent with feedforward GABAergic inhibition that is dependent on glutamate release from mossy fiber terminals (Fig. 5e,f).
It has been recently demonstrated that newborn granule cells of the adult hippocampus receive functional synaptic inputs that can drive them to spike6-13. Our study demonstrates that newborn neurons can also influence the activity of the downstream network, as both structural and functional properties of their synaptic terminals are compatible with a full integration into hippocampal circuitry.
At the structural level, we found that axons arising from adult-generated granule cells made synapses with postsynaptic targets in the CA3 area and the hilus as early as 17 d after cell division and that these synapses continued to mature thereafter. At first, axonal boutons synapsed with dendritic shafts of postsynaptic targets and then contacted spines, or thorny excrescences, a time course that is consistent with data from studies of hippocampal development30,31. In contrast with these studies, however, newly generated neurons had intermediate stages where they synapsed with thorny excrescences that were also contacted by GFP-negative neurons. This observation raises the possibility that the development of mossy fiber synapses from newly generated neurons involves the recruitment of pre-existing postsynaptic targets, where boutons may compete for thorny excrescences in a fashion similar to that observed during the development of their dendritic spines12 (Supplementary Fig. 1). Notably, the time course of mossy terminal maturation appears to be concurrent with dendritic spine maturation on GFP-positive neurons12,13,15, suggesting that the functional input and output of newborn neurons is synchronized and occurs in a discrete time frame. Finally, our structural data strongly suggest that mossy fibers make contacts with interneurons of the hilus and the GCL/hilar border. This observation was confirmed by our functional studies, in which the identity of postsynaptic neurons was unambiguously determined using electrophysiological and anatomical parameters (Supplementary Fig. 3 and Supplementary Table 1).
Restricted expression of ChR2 served to specifically stimulate large groups of adult-born granule cells at mature stages and detect neurotransmitter release onto target cells as postsynaptic responses that were time-locked to the stimulus (Fig. 4). Together with the morphological profile described above, these observations demonstrate that contacts formed by newborn granule cells onto target neurons are functional glutamatergic synapses (Fig. 5).
Light-evoked PSCs may reflect a direct contact between a mossy fiber terminal and a recorded cell (monosynaptic connection; see scheme in Supplementary Fig. 4 online) or an indirect pathway involving the activation of intermediary neurons (polysynaptic connection; Supplementary Fig. 4). Detailed analysis of postsynaptic responses indicated that 7 out of 14 recordings probably correspond to monosynaptic responses, whereas two were mediated by GABAergic interneurons (see Methods; the remaining five neurons could not be reliably classified). The observation of these polysynaptic responses indicates that newborn granule cells can trigger spiking of postsynaptic neurons.
Our present findings demonstrate that axons of newborn cells establish synaptic connections with mossy cells, interneurons of the hilus and GCL/hilar border, and pyramidal cells, and release glutamate as the main neurotransmitter. New neurons are therefore fully functional units of the hippocampal circuit, being able to effectively integrate input information and transmit it to postsynaptic targets. From a global point of view, these properties are similar to those of all of the other granule cells of the adult dentate gyrus, similar to what has been reported for afferent connections10,12,32. However, it is conceivable that subtle differences might be present in the identity of presynaptic neurons and postsynaptic targets, or in the degree of functional plasticity of synapses formed onto7,11,33 or by these cells.
Female C57Bl/6 mice, 5–7 weeks old, were injected with murine leukemia–based retroviruses as described previously10,15 and were housed with unlimited access to running wheels starting from 3–7 d before virus injection. At the indicated times post–viral injection, depending on the experimental design, mice were anesthetized with a mixture of ketamine and xylazine (100 mg per kg of body weight, 10 mg per kg) and perfused transcardially with 0.9% saline (wt/vol), followed by 4% paraformaldehyde (PFA, wt/vol). The brain samples were postfixed with 4% paraformaldehyde for an additional 2 d. We prepared 50-μm coronal sections with a vibratome. Mice destined for electrophysiological studies received a vitamin A–containing multivitamin supplement to ensure the availability of retinal precursors needed for ChR2 function. The animal protocols were approved by the Salk Institutional Animal Care and Use Committee.
A replication-deficient retroviral vector based on the Moloney murine leukemia virus was used to express either GFP alone15 or ChR2. Two ChR2 constructs were used: CAG–ChR2-GFP, expressing a ChR2-GFP fusion protein23 driven by a CAG promoter, and PGK–ChR2-IRES-RFP, expressing a C-terminally truncated version of ChR2 (amino acids 1–315) and mRFP1 (ref. 34) driven by the phosphoglycerate kinase (PGK) promoter. Retroviral particles were assembled using three separate plasmids containing the capsid (CMV-vsvg), viral proteins (CMV-gag/pol) and transgene (CAG-GFP, CAG-ChR2 or PGK–ChR2-IRES-RFP). Plasmids were transfected onto 293T cells using Lipofectamine 2000 (Invitrogen). Virus-containing supernatant was harvested 48 h after transfection and concentrated by two rounds of ultracentrifugation.
Images of GFP-labeled mossy fiber axons in the CA3 region were acquired at 0.75-μm intervals with the Bio-Rad R2100 confocal system with a plane apochromatic 40× oil lens (numerical aperture, 1.3; Nikon) and a digital zoom of 6. The Bio-Rad image files were subjected to five iterations of deconvolution with the AutoDeblur program (AutoQuant). Maximum projections of z series were created with the Confocal Assistance Program in BMP format and the files were then imported into IGL Trace (http://synapse-web.org/tools/index.stm). Mossy fiber boutons that fit the following criteria were selected for quantification: (i) the diameter of the bouton was more than threefold greater than the diameter of the mossy fiber, (ii) the bouton was connected to the mossy fiber on at least one end and (iii) the bouton was relatively isolated from other boutons for accuracy of tracing. These criteria were arbitrary but nevertheless established an unbiased sampling. The contour of the boutons was traced manually and the enclosed area was calculated by the IGL Trace program. A total of 20–21 boutons were analyzed for each time point in CA3 and a total of 20–66 boutons were analyzed for each time point in the hilus.
Mice were transcardially perfused with 4% PFA in 0.1 M phosphate buffer, pH 7.4, and maintained at 4 °C. We removed the brain 15 h after the perfusion was stopped and postfixed it for 72 h in 4% PFA. We then cut 50-μm horizontal vibratome sections and observed them under an epifluorescence microscope. One brain, which had an insufficient amount of labeled cells, was discarded. The sections of the remaining brains were then cryoprotected in 2% glycerol and 20% DMSO (vol/vol) in 0.1 M phosphate buffer for 20 min and freeze-thawed four times in liquid nitrogen. After a treatment in 0.3% hydrogen peroxyde (vol/vol, five times for 5 min) and three 10-min washes in 0.1 M phosphate buffer + 0.5% bovine serum albumin (vol/vol, BSA-C, Aurion), slices were incubated overnight in the primary antibody (rabbit antibody to GFP, 1:500, Chemicon) in phosphate buffer + 0.1% BSA-C for 40 h, at 4 °C on a shaker. After washing in phosphate buffer + 0.1% BSA-C, the sections were incubated for 5 h at 20 ± 5 °C in biotinylated secondary antibody (goat antibody to rabbit (F)ab fragment, 1:200, Jackson Laboratories). To reveal this labeling, we incubated slices for 2 h in avidin biotin peroxidase complex (ABC Elite, Vector Laboratories), followed by 3,3′-diaminobenzidine tetrachloride (Vector Laboratories Kit) for 10–20 min. The sections were then postfixed overnight in 2.5% glutaraldehyde (wt/vol), washed in phosphate buffer 0.1 M, postfixed in osmium tetroxide for 1 h, dehydrated in ascending concentrations of ethanol and then acetone, and embedded in Epoxy resin.
Serial sections at a depth of more than 5 μm were cut at 40-nm thickness and collected on single-slot grids, then contrasted by incubating for 35 min in 5% uranyl acetate solution (wt/vol), followed by 25 min in a Reynolds solution. Serial images of the labeled structures were then collected with a digital camera (MegaView III, SIS) mounted on a JEOL 100 CXII transmission electron microscope at a 19,000× magnification with a filament voltage of 80 kV. Serial images were then aligned and three-dimensional structures were rendered using Reconstruct 1.0 (provided by J. Fiala, Boston University).
Experiments were carried out in 43 slices from 34 mice, 14.9 ± 2.9 weeks (mean ± s.d., range: 8.6–20.6) after injection of the retrovirus. Light-evoked PSCs were detected in 14 neurons in 11 slices from 9 mice, 16.0 ± 1.9 (range: 13.4–20.4) weeks after the injection. Mice were anesthetized and decapitated. Brains were removed and placed into a chilled solution containing 110 mM choline chloride, 2.5 mM KCl, 2.0 mM NaH2PO4, 25.0 mM NaHCO3, 0.5 mM CaCl2, 7 mM MgCl2, 20 mM dextrose, 1.3 mM sodium ascorbate, 0.6 mM sodium pyruvate and 4 mM Kyn. Right hemisphere slices (400 μm thick) were cut in a vibratome and transferred to a chamber containing artificial cerebrospinal fluid (ACSF: 125.0 mM NaCl, 2.5 mM KCl, 2.0 mM NaH2PO4, 25.0 mM NaHCO3, 2 mM CaCl2, 1.3 mM MgCl2, 1.3 mM sodium ascorbate, 3.1 mM sodium pyruvate and 10 mM dextrose). Slices were bubbled with 95% O2 / 5% CO2 (315 mOsm) and maintained at 30 °C. Recordings were carried out in ACSF at room temperature (25 ± 2 °C) using micro-electrodes (3–5 MΩ) pulled from borosilicate glass (KG-33, Garner Glass). In most experiments, ACSF was supplemented with all-trans retinal (0.1–1 μM, Sigma-Aldrich) to favor ChR2 functionality and 200 μM BaCl2 to increase granule cell excitability by blocking inward rectifier potassium channels35 (D.A.L. and A.F.S., unpublished observations), thus increasing the number of granule cells activated by light stimuli. The internal solution contained 120 mM potassium gluconate, 20 mM KCl, 5 mM NaCl, 4 mM MgCl2, 0.1 mM EGTA, 10 mM HEPES, 4 mM Tris-ATP, 0.3 mM Tris-GTP, 10 mM phosphocreatine, Alexa Fluor 488 or 594 (5 μg ml−1, Invitrogen), pH 7.3 and 290 mOsm.
Whole-cell recordings (Axopatch 200B, Molecular Devices) were filtered at 2 kHz, digitized (Digidata 1322A, Molecular Devices) and acquired at 20 kHz onto a PC using pClamp 9 (Molecular Devices). In voltage clamp experiments, a holding potential of −80 mV was used for granule cells and −60 or −70 mV for postsynaptic neurons. Action potentials were elicited in current clamp by current steps of 400 ms from 50 to 800 pA in 50-pA increments from a −60-mV baseline membrane potential.
Light stimuli were delivered through the epifluorescence pathway of a Leica DMLFS microscope (100 W Hg lamp, FITC filter cube, 63× water immersion objective, numerical aperture 0.9) and controlled through a UNIBLITZ VMM-D1 shutter (Vincent Associates), commanded from the acquisition software. Slices were briefly scanned for ChR2 expression (GFP or RFP positive) in granule cells and whole-cell patch-clamp recordings were made from neighboring cells in the GCL/hilar border, hilus or pyramidal layer. Light pulses of 10 ms were then delivered at 0.2–4 Hz while we recorded postsynaptic membrane currents in voltage-clamp.
We measured the latency of light-evoked spikes in ChR2-positive neurons from the start of the light pulse to the peak of the action potential. The latency to onset of PSCs was measured from the start of the light pulse to the time when the average PSC reached 10% of the peak amplitude.
We estimated the mono-or polysynaptic nature of PSCs using measurements of latency and kinetics combined with pharmacological treatments. A limitation of our approach is that synaptic delays can be estimated, but not measured precisely, as the latency from the light stimulus to the presynaptic spike varies across granule cells according to their ChR2 expression level (Supplementary Fig. 2). When stimulated at 1 Hz with 10-ms light pulses, ChR2-positive neurons spike with latencies of 10–24 ms (median 19.3, n = 9). The latencies of monosynaptic PSCs result from adding spike latencies and mossy fiber synaptic delays, which were reported to be ~3–6 ms36,37. Some of the light-evoked PSCs (7 out of 14) showed properties that strongly suggest a monosynaptic nature (see examples in Fig. 4a–l): (i) median latency to onset of 17.3 ms (range 14.2–22.0 ms), (ii) fast rise time of average traces (median 1 ms, range 0.7–2.1 ms), also reflecting little jitter in latency36, and (iii) lack of BMI effect, thus ruling out the possibility of polysynaptic GABAergic signals (Fig. 5c,d). In contrast, in two additional recordings, PSCs showed high latencies (25.0 and 25.6 ms), slow rise times (15.5 and 15.3 ms) and sensitivity to both Kyn and BMI, all evidence of polysynaptic pathways (see examples in Fig. 4m–o and Fig. 5e).
Light-evoked PSCs often had two peaks, separated, on average, by ~12 ms. Individual traces showed that failures could be observed in each component independently (Fig. 4d,h). We propose that both peaks correspond to responses to spike doublets induced in presynaptic granule cells by light stimuli (Supplementary Fig. 2). In fact, in granule cells that responded with more than one spike to 10-ms light pulses, the first two spikes were separated, on average, by ~12 ms. Also, the s.d. of the latency of the second spike was about tenfold higher than that of the first one, thus accounting for the broader appearance of the second PSCs in average traces (Fig. 4e,i). Furthermore, second peaks are not polysynaptic GABAergic signals, as they were insensitive to BMI (Fig. 5c,d).
The comparison between hilar and CA3 boutons was done using Student’s t test. All other statistical analyses were carried out with one-way ANOVA followed by Fisher’s PLSD post hoc analysis when a significant difference was found with ANOVA. All data were presented as means ± s.e.m.
Note: Supplementary information is available on the Nature Neuroscience website.
We would like to thank J. Jepsen and N. Nashi for technical assistance, S. Herlitze for providing the ChR2 construct, L. Petreanu for helpful discussions, J. Simon for artwork, J. Fiala for providing software and M.L. Gage for helpful suggestions to improve this manuscript. We also acknowledge the support of the Picower Foundation, Lookout Foundation, McDonnell Foundation, US National Institutes of Health grants NS050217 (to F.H.G.) and NS038331 (to C.E.R.), Agencia Nacional de Promoción Científica, Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), and the Howard Hughes Medical Institute (to A.F.S.). D.A.L. and G.L. were supported by fellowships from CONICET.