Fast Synaptic Subcortical Control of Hippocampal Circuits

See allHide authors and affiliations

Science  16 Oct 2009:
Vol. 326, Issue 5951, pp. 449-453
DOI: 10.1126/science.1178307

Subcortical Network Regulation

Subcortical neuromodulatory centers dominate the motivational and emotional state–dependent control of cortical functions. Control of cortical circuits has been thought to involve a slow, diffuse neuromodulation that affects the excitability of large numbers of neurons relatively indiscriminately. Varga et al. (p. 449) describe a form of subcortical control of cortical information processing whereby strong, spatiotemporally precise excitatory input from midbrain serotonergic neurons produces a robust activation of hippocampal interneurons. This effect is mediated by a synaptic release of both serotonin and glutamate and impacts network activity patterns.


Cortical information processing is under state-dependent control of subcortical neuromodulatory systems. Although this modulatory effect is thought to be mediated mainly by slow nonsynaptic metabotropic receptors, other mechanisms, such as direct synaptic transmission, are possible. Yet, it is currently unknown if any such form of subcortical control exists. Here, we present direct evidence of a strong, spatiotemporally precise excitatory input from an ascending neuromodulatory center. Selective stimulation of serotonergic median raphe neurons produced a rapid activation of hippocampal interneurons. At the network level, this subcortical drive was manifested as a pattern of effective disynaptic GABAergic inhibition that spread throughout the circuit. This form of subcortical network regulation should be incorporated into current concepts of normal and pathological cortical function.

Subcortical monoaminergic systems are thought to modulate target cortical networks on a slow time scale of hundreds of milliseconds to seconds corresponding to the duration of metabotropic receptor signaling (1). Among these ascending systems, the serotonergic raphe-hippocampal (RH) pathway that primarily originates within the midbrain median raphe nucleus (MnR) is a key modulator of hippocampal mnemonic functions (2). Contrary to the slow modulatory effect commonly associated with ascending systems, electrical stimulation of the RH pathway produces a rapid and robust modulation of hippocampal electroencephalographic activity (35). Anatomical evidence shows that MnR projections form some classical synapses onto GABAergic interneurons (INs) in the hippocampus (6), potentially providing a substrate for a fast neuromodulation of the hippocampal circuit. Recent reports of the presence of glutamate in the serotonergic system (79) raise the possibility that this system may use a coordinated action of serotonin [5-hydroxytryptamine (5-HT)] and glutamate to rapidly activate elements of the hippocampal network.

To investigate this form of ascending neuromodulation, we first performed a series of in vitro experiments. Because RH axons are intermingled with intra- and extrahippocampal axons of various origins, we virally targeted a channelrhodopsin2–enhanced green fluorescent fusion protein construct (ChR2-eGFP) to the MnR (fig. S2, a to e) and subsequently used ChR2-assisted photostimulation in acute rat hippocampal slices to selectively activate the labeled RH input (10, 11). eGFP-positive (eGFP+) fibers formed a dense plexus mostly immunoreactive for 5-HT (85 ± 4% 5-HT in eGFP+ RH fibers and 37 ± 3% eGFP+ in 5-HT+ RH fibers; n = 3 animals, fig. S2f) (11) in strata oriens/alveus and at the border of strata radiatum and lacunosum-moleculare (sr/sl-m) of CA1 and CA3 regions, matching previously documented laminar distribution of RH fibers (6, 12). We therefore focused on INs located in these areas to perform whole-cell voltage recordings. Two-photon imaging revealed that eGFP+ RH axons often formed what appeared to be climbing-type contacts at multiple close appositions with IN dendrites (Figs. 1A and 2A). To functionally characterize these putative contacts, we photostimulated the ChR2-eGFP+ axons using brief light exposures. Single light pulses (1 ms) evoked large-amplitude excitatory postsynaptic potentials (EPSPs) in most recorded INs (amplitude: 6.47 ± 1.33 mV, n = 22 out of 27 INs; Fig. 1, B and C). In many cases, this input was sufficient to trigger action potential output from the activated cell (n = 7 out of 22; Fig. 2, A and B). Light-evoked EPSPs had a short latency (2.67 ± 0.17 ms, n = 22), small temporal jitter (SD of latency: 0.42 ± 0.07 ms, n = 22), a fast rise time, and a mono-exponential decay (20 to 80 rise time: 3.15 ± 0.31 ms; decay tau: 105.71 ± 10.79 ms, n = 22; Fig. 1, B and C). Repetitive photostimulation reliably evoked trains of EPSPs in INs (Fig. 1, D and E). Single-pulse photostimulation did not evoke detectable responses in CA3 pyramidal cells (CA3PC, n = 9 cells) and produced either no response (25 out of 29 cells) or only a slow, small hyperpolarization in CA1 pyramidal cells (CA1PC, peak hyperpolarization; single pulse: 0.31 ± 0.18 mV, n = 4; train: 0.91 ± 0.13 mV, n = 21; Fig. 1D).

Fig. 1

MnR input excites hippocampal INs in vitro. (A) Two-photon image of an IN filled with Alexa 594 (red) surrounded by a dense network of eGFP+ RH fibers (green) at the border of sr/sl-m of CA1 region in a dorsal hippocampal slice. Dashed boxed region is expanded (lower panel) to show multiple putative contacts (arrowheads) made by an eGFP+ axon on the IN’s dendrite. (B) (Upper panel) ChR2-photostimulation produces a large-amplitude and short-latency EPSP in the IN (red dotted line is exponential fit). (Lower panel) Locations used to measure EPSP kinetics are indicated on an expanded time scale of the rising phase of the EPSP. (C) Summary data of EPSP kinetics. Symbols: individual recordings; bars: mean ± SEM. (D) Image of an IN (red) and a CA1PC (yellow-white) recorded sequentially (left) and averaged responses recorded from the IN (black) and from the PC (gray) to ChR2-photostimulation (right). (E) Summary plot of peak EPSP amplitude during train photostimulation from INs (10 pulses at 20 Hz, n = 10; error bars: SEM).

Fig. 2

RH fibers excite hippocampal INs via 5-HT- and/or VGluT3-containing synapses. (A) Two-photon image of an IN (red) and eGFP+ RH axons (green) at the border of sr/sl-m. (B) ChR2 photostimulation evokes short-latency action potential in the IN (red trace, right: expanded time scale) at resting membrane potential (Vm). At a more hyperpolarized Vm, underlying large EPSPs are observed (individual traces: gray; average: black). (C) The IN in (A) processed for light microscopy (image stack of 100 μm). Black frame marks the dendritic segment analyzed for contacts with RH fibers. (D) Triple immunofluorescent confocal image of the dendritic segment. Red: biocytin (cell) and VGluT3 (boutons); green: eGFP (RH axons); blue: 5-HT. Colocalization or overlaps: yellow: red + green; light blue: blue + green; white: triple labeling. White boxes demarcate the putative contacts examined by triple immunofluorescence on higher magnification (E and F) and by correlated EM (G to J). The small arrow to the left of the upper white frame points to an eGFP−/5-HT+ fiber. (E) High-magnification single-plane confocal images of the possible contact between the dendritic segment and the 5-HT+/VGluT3+ bouton shown in the upper white frame in (D). Separate and overlayed images of the three channels are presented. Arrow: bouton examined by EM (G and I); arrowheads: surrounding eGFP+ elements used as landmarks to identify the area on the low-magnification EM [small arrowheads in (G) and (H)]. (F) Putative contact between the dendritic segment and the 5-HT− bouton designated by the lower frame in (D). (G) Low-power EM image of the contact depicted in (E). (H) Low-power EM image of the contact in (F). (I) High-power EM image [scale bar of (G) is 1 μm in (I)] of the contact in (G) showing the parallel apposition of the dendritic and bouton membranes surrounding a possible synaptic cleft (small white arrow). The 3,3-diaminobenzidine tetrahydrochloride (DAB) precipitate masks synaptic vesicles. (J) The same as in (I) for the 5-HT−/VGluT3+ bouton shown in (H) [scale bar of (H) is 1 μm in (J)]. B: RH boutons; D: IN dendrite.

To directly test whether EPSPs on INs are mediated by synapses formed by RH afferents, we accomplished correlated light and electron microscopic analysis of six appositions between eGFP+ and neurochemically identified boutons and recorded INs. Five of six boutons were glutamatergic, as indicated by positive immunoreactivity for vesicular glutamate transporter type 3 (VGluT3), and one was 5-HT/VGluT3 double immunoreactive (n = 2 INs, two connections tested for 5-HT/VGluT3 and four tested only for VGluT3). All examined appositions were synaptic contacts between the identified boutons and dendritic profiles of INs (Fig. 2, C to J, and fig. S3, a to d).

What type of neurotransmitters mediate fast transmission at RH-IN synapses? First, we tested the contribution of ionotropic 5-HT3 receptors (5-HT3Rs) that are expressed by a subset of hippocampal INs, which are innervated by MnR afferents (6, 13). The selective 5-HT3R antagonist MDL72222 (100 μM) significantly reduced the amplitude of the EPSPs (to 74 ± 5% of control, P < 0.05, n = 16; Fig. 3, A to D) (11). A prominent 5-HT3R antagonist-resistant component was still present and was abolished by subsequent coapplication of AMPA-type (NBQX, 5 μM) and N-methyl-d-aspartate (NMDA)-type (d,l-AP5, 50 μM) ionotropic glutamate receptor (iGluR) antagonists [to 6 ± 1%, P < 0.001, n = 16; the slow hyperpolarization on PCs was sensitive to 5-HT1AR antagonist (S)-WAY 100135, 100 μM; fig. S3e].

Fig. 3

Monosynaptic excitation of INs is mediated by 5-HT3 and AMPA/NMDA receptors. (A) Upper left: Image of an IN (red) and eGFP+ RH fibers (green). Lower left: Expanded image of the dashed boxed region in upper panel; arrowheads indicate the positions of multiple close appositions between a proximal dendritic segment of the IN and eGFP+ axons. Traces to the right of the image are averaged Vm responses evoked by single-pulse and train photostimulation in control (black), MDL72222 (100 μM, green), and MDL72222+NBQX+AP5 (gray). Subsequent coapplication of iGluR antagonists (NBQX 5 μM, d,l-AP5 50 μM) abolishes photostimulation-evoked responses. (B and C) Grouped data (B) and summary graph [(C); error bars: SEM; *P < 0.05; ***P < 0.001] of EPSPs evoked by single-pulse photostimulation in INs. Lines connect individual recordings in control (open circles), MDL72222 (green circles), and MDL72222+NBQX+AP5 (gray circles). (D) Summary plot of peak EPSP amplitude during train photostimulation (n = 10; error bars: SEM). (E to L) EM images of immunoperoxidase-immunogold stainings. Images show GluR1 receptor subunit labeling in the synaptic contacts (arrows) that are established either by VGluT3+ boutons [(E) to (H), DAB-positive dark precipitation] or by PHAL-labeled [(I) to (L), DAB-positive] RH terminals at the border of sr/sl-m. (E to J) Two or three sections of the same synapse with multiple labelings for GluR1 are shown. Scale bar on (E) applies for (F) to (L).

Two lines of evidence further support the dual-component serotonergic/glutamatergic nature of fast synaptic transmission in RH-IN synapses. VGluT3+ terminals at the border of sr/sl-m in the hippocampus CA1 area originate from raphe nuclei (8). Using electron microscopy (EM) and combined immunogold-immunoperoxidase staining, we found that most (19 out of 25) synapses of these above-mentioned terminals and many (10 out of 16) synapses of RH terminals labeled by the anterograde tracer Phaseolus vulgaris leucoagglutinin (PHAL) are GluR1 subunit positive as well (Fig. 3, E to L). Moreover, in several cases, two-photon uncaging of MNI-glutamate (14) at putative synaptic contact sites made by eGFP+ axons evoked fast rising and decaying EPSPs in INs (fig. S3, f to i).

MnR activation may therefore robustly engage hippocampal INs via a fast serotonergic/glutamatergic excitation. To directly test this prediction, we first juxtacellularly recorded hippocampal neurons during electrical stimulation of the MnR in vivo (n = 60 cells; 8 cells were labeled and recovered, 4 cells were anatomically identified) (11). We found a subset of neurons that MnR stimulation activated with a short latency (8.9 ± 0.58 ms, n = 16; Fig. 4) (11). These fast activated cells showed a robust (firing: 491.2 ± 155.2% of control; success rate: 65.7 ± 5.3%; Fig. 4, figs. S4 to S6, and table S2) and temporally focused activation (duration: 8.05 ± 1.03 ms; jitter: 1.7 ± 0.19 ms). All of these parameters except success rate were significantly different from those of a second subset of neurons that were activated more slowly (latency: 15 to 50 ms; n = 7, P < 0.05; table S2).

Fig. 4

A subset of hippocampal INs is activated by MnR stimulation in vivo with short latency by a partly glutamatergic mechanism. (A and B) Immunofluorescent identification of a fast-activated IN [(A) neurobiotin: green; (B) cholecystokinin: blue]. (C) MnR stimulation–induced activation of this IN is manifested as the concentration of evoked spikes within 10 ms of stimulation (overlaying 20-ms windows of raw data, n = 115). (D) Firing pattern of the IN during the non-theta to theta transition. (E and F) Response of this IN is characterized by raster plots (upper) and peristimulus time histograms (PSTHs) (lower panels, PSTH bin size = 1 ms) showing the MnR stimulation–induced fast activation (E) and the subsequent long-lasting inhibition [(F), bin size = 10 ms]. Note the robust increase in spike count (13 times that of the control, 6-ms latency). (G) Distribution of response latencies of neurons activated by MnR stimulation within 50 ms after stimulus onset (n = 23). The peak of fast responders can be clearly distinguished (n = 17 in red). (H) The fast response is largely attenuated by local application of NBQX: Doubling the duration of NBQX application (1st: 2 min; 2nd: 5 min) almost totally attenuates the fast response. After switching off NBQX ejection, the fast response recovers (1st and 2nd wash out). (I) Change of success rate of fast activation during all NBQX experiments (2 min, n = 6; 5 min, n = 3) after continuous drug iontophoresis. NBQX reduced the success rate in all instances, and its effect was time (dose) dependent.

Evidence of the prominent glutamatergic component of the MnR-stimulation–mediated excitatory drive was also found in vivo. Local iontophoresis of NBQX (n = 6; table S3) (11) significantly suppressed the success rate of the fast activation (2 min: to 53.8 ± 4.5%; P < 0.05, n = 6; 5 min: to 29.7 ± 11%, n = 3; Fig. 4, H and I) in a dose-dependent and reversible way, without altering the latency and nonsignificantly increasing the jitter of the response (table S3) (11). We determined the position of fast responding INs in the hippocampal network by identifying labeled cells immunocytochemically and analyzing their state-dependent activity (table S1 and figs. S4 to S6).

That hippocampal INs are highly interconnected (15) raises the possibility that the activated components of the inhibitory circuit could, through feed-forward inhibition, themselves play a role in producing the elevated spike precision described above. Further analysis of our MnR-stimulation data revealed that the initial increase in IN spiking was often followed by a period of suppressed activity (by 90.6 ± 4%, 10 out of 16 cells; latency: 15 ± 1.71 ms; duration: 144.85 ± 17.32 ms; fig. 4F). Two further observations indicated that this postexcitatory suppression was caused by MnR-triggered inhibition. First, this effect appeared to be sensitive to 5-HT3R antagonists, as ondansetron (1 to 2 mg/kg, intraperitoneal injection) significantly reduced the duration of the inhibitory period (from 148.5 ± 32.85 ms to 83.25 ± 14.13 ms; n = 5; 5 to 15 min after injection, P < 0.05) and increased the jitter of the primary facilitatory response (fig. S9h and table S3).

Photostimulation experiments were used to investigate this network effect in vitro. With a low concentration of chloride ([Cl]in) in the intracellular solutions, monosynaptic EPSPs were often followed by readily observable GABAergic (i.e., gabazine-sensitive, n = 2) inhibitory postsynaptic potentials (IPSPs; fig. S9, a and b) (11). When present, these IPSPs were preferentially blocked by the 5-HT3R antagonist (n = 5 out of 13 INs; fig. S9, b, e, and f). The most likely interpretation of this result (and observed in 7 out 13 INs; fig. S9, c to f) is that the recorded IN is itself innervated by a subset of INs that are strongly excited by RH afferents. In agreement with this, IPSPs were disynaptic in origin, as the evoked responses were abolished by the coapplication of 5-HT3R and iGluR antagonists (amplitude: to 7 ± 2%, P < 0.05, n = 13; fig. S9, e to g).

Finally, we found that MnR activation of the inhibitory circuit is effectively transmitted onto the output elements, i.e., principal cells of the hippocampal circuit. When recording from PCs with low [Cl]in in the intracellular solution in vitro, we observed robust disynaptic GABAergic IPSPs (fig. S10, a to c). This fast inhibition was also observed in PCs in vivo as a form of short-latency cessation of PCs spiking when recorded juxtacellularly (fig. S10d; n = 3) and as a short-latency hyperpolarization when recorded intracellularly (fig. S10e) during MnR stimulation (11).

The present demonstration of a fast synaptic activation of hippocampal interneurons by MnR afferents via glutamate/serotonin cotransmission puts the subcortical control of cortical information processing in a fundamentally new perspective. The current view of a slow and diffuse modulation conveying emotional, motivational, or other state-dependent tuning (1, 16) is now complemented by the ability to carry out target-selective synaptic actions with high temporal and spatial resolution. This additional ability may promote the rapid formation and selection of particular hippocampal local representations or modes of information processing (17, 18), possibly through fast alterations in the relative contribution of the different classes of interneurons to rhythmic population activity (13, 19) (fig. S1). Finally, the present observations lend support to a proposed role of glutamate in disorders that have traditionally been considered to be subcortical in origin, such as depression, anxiety, and certain components of schizophrenia (2022).

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S10

Tables S1 to S3


  • * These authors contributed equally to this work.

  • Present address: Department of Neuroscience, Columbia University, New York, NY 10032, USA.

References and Notes

  1. Supplementary methods and results are available as supporting material on Science Online.
  2. The efficiency in recruiting inhibitory neuron groups of the fast modulation described herein implies that the pathological alteration of this connection would dramatically shift the inhibitory and excitatory balance leading, ultimately, to the impairment of the target network.
  3. We thank A. Lee and B. Kocsis for their comments on the manuscript. We thank L. Acsady and B. K. Andrásfalvy for helping to establish the collaboration between V.V. and A.L. We thank B. Shields, K. Lengyel, and G. Goda for technical assistance. V.V. and G.N. were János Bolyai Research Fellows. This work was supported by grants from the NIH (NIH-MH-54671), Howard Hughes Medical Institute (HHMI55005608), and Hungarian Scientific Research Fund (OTKA K-60927 and NKTH-OTKA CNK 77793). Author contributions V.V., A.L., J.C.M, and T.F.F. conceived the experiments. V.V. conducted the in vivo physiological experiments with the help of A.D. B.V.Z performed the molecular experiments. A.L. performed the in vitro experiments. Z. B. and G.N performed most of the histological experiments. N.H. performed the in vitro experiments for the specificity of the NBQX effect. V.V. and A.L. analyzed most of the physiological data except firing pattern analysis carried out by B.H. A.L., V.V., B.V.Z., J.C.M., and T.F.F. wrote the paper.

Stay Connected to Science

Navigate This Article