Long-Range–Projecting GABAergic Neurons Modulate Inhibition in Hippocampus and Entorhinal Cortex

See allHide authors and affiliations

Science  23 Mar 2012:
Vol. 335, Issue 6075, pp. 1506-1510
DOI: 10.1126/science.1217139


The hippocampus and entorhinal cortex play a pivotal role in spatial learning and memory. The two forebrain regions are highly interconnected via excitatory pathways. Using optogenetic tools, we identified and characterized long-range γ-aminobutyric acid–releasing (GABAergic) neurons that provide a bidirectional hippocampal-entorhinal inhibitory connectivity and preferentially target GABAergic interneurons. Activation of long-range GABAergic axons enhances sub- and suprathreshold rhythmic theta activity of postsynaptic neurons in the target areas.

The excitatory projections connecting the hippocampus and entorhinal cortex (1) account for the functional interdependence of these two brain regions (24). Excitatory neurons in the hippocampus and entorhinal cortex are under control of local γ-aminobutyric acid–releasing (GABAergic) interneurons (5, 6). Some GABAergic neurons also project long distance. For example, long-range–projecting GABAergic cells connect hippocampus with medial septum (79) and other extra-hippocampal brain areas (10, 11), suggesting that interregional GABAergic connectivity might be less rare than was previously assumed (12).

To test for the presence of hippocampal GABAergic neurons projecting to the medial enthorinal cortex (MEC), we injected the retrograde tracer fluorogold (FG) into the MEC of wild-type mice (fig. S1). In addition to the expected labeling of numerous excitatory cells, we found FG+ neurons in stratum oriens and stratum radiatum of CA1 and in the hilus of the dentate gyrus (DG), indicating retrogradely labeled GABAergic cells. We detected FG-labeled cells coexpressing somatostatin (SOM) in stratum oriens of CA1 (23 cells, nine mice) and also in the hilus of the DG (14 cells, nine mice) (Fig. 1, A and B).

Fig. 1

long-range–projecting hippocampal SOM+ neurons target superficial layers in the MEC. (A) Schematic drawing showing the location of retrogradely labeled SOM+ cells in CA1 and DG of the dorsal hippocampus after FG injection into the MEC. (B) Confocal images of a SOM+/FG–labeled neuron in stratum oriens of CA1 (top row) and the hilus of the DG (bottom row). Scale bar, 10 μm. (C) Coexpression of SOM and ChR2-mCherry in CA1 stratum oriens after virus injection into the dorsal hippocampus in a SOMCre mouse. ChR2-mCherry expression was restricted to SOM+ cells (yellow arrows). Scale bar, 30 μm. (D) Digitally encoded (red) mCherry+ axons of long-range–projecting hippocampal SOM+ cells detected in medial septum (MS, left) and contralateral DG (right). Scale bar, 100 μm. (E) Digitally encoded (red) mCherry+ axons of long-range–projecting hippocampal SOM+ cells in the MEC. Projections indicated by numbers are shown as higher magnification. Scale bar, 150 μm. (F) Confocal images of a mCherry and FG double-labeled cell in stratum oriens of the CA1 region after virus injection into dorsal hippocampus and FG injection into the MEC. Scale bar, 10 μm.

To provide direct evidence for the presence of hippocampal SOM+ neurons projecting to the MEC, we injected the adeno-associated viral (AAV) vector AAV DIO ChR2-mCherry (13) into the dorsal hippocampus of SOMCre mice (Fig. 1C and fig. S2), achieving specific expression of the fluorescent fusion protein ChR2-mCherry in SOM+ neurons (Fig. 1C). We detected mCherry-labeled axons of SOM+ hippocampal neurons in the medial septum and the contralateral DG (Fig. 1D) (14, 15). We also detected labeled axons originating from hippocampal SOM+ neurons in the striatum (fig. S3) and MEC (Fig. 1E). In the MEC, long-range–projecting axons crossed orthogonally from layer VI (LVI) to LII, branched in the transition zone between LII and LI, and extended horizontally within LI over a distance of up to several 100 μm (Fig. 1E). We usually observed one to three projections per 50-μm section. Combining virus injection into the dorsal hippocampus with retrograde tracer injection into the MEC, we found individual CA1 neurons that were co-labeled with mCherry and FG (Fig. 1F). The population of hippocampal SOM+ long-range–projecting neurons targeting the MEC appears to be distinct from the one projecting to the medial septum (fig. S4).

Using immmunohistochemistry, electron microscopy (EM), and whole-cell patch-clamp recordings, we subsequently investigated whether long-range–projecting hippocampal SOM+ neurons form inhibitory synapses in the MEC. The mCherry-labeled axons were VGAT+ and VGluT1-negative (Fig. 2A) and established symmetric synapses in LI and LII (Fig. 2B). We tested for functional synapses by laser-stimulating ChR2-mCherry–expressing terminals and recording from putative postsynaptic cells located in the vicinity of the labeled axons (Fig. 2C). Responses could be detected in 60 out of 686 patched cells (n = 41 mice). At +40 mV, the mean amplitude of the postsynaptic currents (PSCs) was 84.15 ± 11.94 pA, and the mean latency was 3.13 ± 0.22 ms (n = 37 cells). PSCs were inhibitory, as indicated by the reversal potential (~–70 mV, n = 25 out of 25 cells), the pharmacological block of the responses by the GABA type A (GABAA) receptor antagonist gabazine (n = 17 out of 17 cells), and the lack of effect using the glutamatergic blockers 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) and (2R)-amino-5-phosphonovaleric acid (AP5) (n = 3 out of 3 cells) (Fig. 2D). All target cells identified through whole-cell recording were located close to or at the transition zone between LI and LII (Fig. 2E and fig. S5).

Fig. 2

Hippocampal SOM+-projecting neurons form inhibitory synapses onto GABAergic neurons in the MEC. (A) VGAT+ varicosities (arrows) of mCherry-labeled long-range projections in layer I/II of the MEC (left) and quantification in the indicated layers of the MEC. (Right) Bar histogram representing average number of VGAT+ varicosities (60 optical sections, 2500 μm2 each, mean ± SEM; five hemispheres of three mice). Scale bar, 5 μm. (B) Electronmicrographs showing serial sections of an immunogold-labeled ChR2-mCherry–expressing axon terminal (at) that forms a symmetric synapse onto a dendrite (d). Scale bar, 0.25 μm. (C) Fluorescence image of mCherry-labeled axons in the MEC (left). Asterisk indicates the location of a target cell shown in the DIC image (right). Presynaptic axon was stimulated with blue laser light, and PSCs were recorded in the target cell. Scale bar, 25 μm. (D) Inhibitory PSCs recorded in a target cell at indicated holding potentials. Responses could be blocked by gabazine but not CNQX and AP5. Scale bars, 20 ms and 50 pA. (E) Schematic drawing of a horizontal section showing the location of target neurons (red dots) that were biocytin-filled for subsequent reconstruction (left). Firing pattern most frequently found in MEC target cells (middle). Scale bars, 200 ms and 20 mV. (Right) A representative reconstructed target cell (dendrites, black; axon, red). Scale bar, 100 μm.

Hippocampal SOM+ long-range–projecting neurons preferentially targeted GABAergic interneurons in the MEC. On the basis of their firing pattern, out of 20 target cells 16 could be classified as interneurons and 4 as stellate cells. The firing pattern of the interneurons was not uniform, indicating that they comprised different subtypes (fig. S6). Biocytin-filling revealed that axons of target cells arborized mainly in LI (n = 6 cells, three mice) (Fig. 2E).

FG injection into the MEC suggested that in addition to SOM+ cells, other hippocampal GABAergic neurons project to the MEC (fig. S1B). AAV DIO ChR2-mCherry injection into the dorsal hippocampus of GADCre mice (16) resulted in a larger number of labeled long-range–projecting axons that were VGAT+ and covered additional target fields in the MEC (such as the presence of branches also in deeper layers) (fig. S7). Nineteen target cells were identified upon laser stimulation, and when tested, on the basis of the firing pattern all target cells were GABAergic (n = 12 out of 12 cells).

Retrograde labeling experiments have indicated the presence of long-range–projecting GABAergic cells in the opposite direction, from the entorhinal cortex to the hippocampus (17). When dorsal hippocampi of GAD67EGFP mice were FG-injected, we detected FG/enhanced green fluorescent protein (EGFP) double-labeled cells in LII and LIII of the MEC (10 out of 1147 FG+ cells counted, n = 5 mice) (Fig. 3A). We therefore injected AAV DIO ChR2-mCherry into the MEC of GADCre mice (Fig. 3B) and readily detected mCherry-labeled axons in stratum lacunosum-moleculare/radiatum of the hippocampal CA areas, and stratum moleculare of the DG (Fig. 3, C and D).

Fig. 3

MEC long-range GABAergic neurons form functional synapses onto GABAergic neurons in the hippocampus. (A) Retrogradely labeled GABAergic cell in the MEC after FG injection into the dorsal hippocampus of a GAD67EGFP mouse. Scale bar, 10 μm. (B) mCherry expression in MEC GABAergic cells subsequent to AAV DIO ChR2-mCherry injection into a GADCre mouse. Scale bar, 500 μm. (C) Schematic drawing of a sagittal hippocampal section indicating areas targeted by long-range–projecting MEC GABAergic neurons (red). (D) Fluorescent axons are located in stratum lacunosum-moleculare (lm) of the CA1 and CA3 areas and in stratum moleculare (mo) of the DG. Scale bars, 25 μm (CA1) and 12.5 μM (CA3 and DG). (E) Electronmicrographs showing serial sections of an immunogold-labeled ChR2-mCherry–expressing axon terminal (at) that forms a symmetric synapse onto a dendrite (d). Scale bar, 0.5 μm. (F) Inhibitory PSCs recorded in a target cell in stratum lacunosum-moleculare at indicated holding potentials. Response was blocked by gabazine but not CNQX or AP5. Scale bars, 20 ms and 50 pA. (G) Schematic drawing of a horizontal section of the intermediate hippocampus indicating the location of responsive target cells (red dots). Firing pattern most frequently found in hippocampal target cells (middle). Scale bars, 200 ms and 20 mV. (Right) Corresponding reconstruction in a sagittal section. Scale bar, 100 μm.

We also analyzed the entorhinal-hippocampal connections anatomically and functionally. Fluorescently labeled long-range axons were VGAT+ (fig. S8), and EM analysis of labeled axons in stratum lacunosum-moleculare/radiatum (n = 6 mice) showed that long-range projections formed symmetric synapses (Fig. 3E). Laser stimulation of presynaptic ChR2-expressing long-range projections allowed the identification of 86 responding target cells (out of ~1000 patched neurons from 65 mice) (mean amplitude and latency of PSCs at +40 mV was 101.45 ± 22.75 pA and 4.47 ± 0.50 ms, respectively; n = 19 cells) (Fig. 3F). The GABAergic nature of these synapses was also confirmed by the reversal potential (n = 12 out of 12 cells) and selective gabazine blockage (n = 17 out of 17 cells). All detected target cells were located in dorsal and intermediate hippocampal areas that are known to contain interneurons only—namely, stratum lacunosum-moleculare and deep stratum radiatum in all CA areas and in stratum moleculare of the DG (Fig. 3G). In more than 110 pyramidal and granule cells that were patched, no PSCs could be detected, suggesting a preferential, if not even exclusive, targeting of GABAergic interneurons. The firing pattern (n = 26 cells) (Fig. 3G and fig. S9) and morphology (n = 19 cells) (Fig. 3G) of the target cells confirmed their interneuronal phenotype.

To establish the immunochemical identity of MEC GABAergic projection neurons, we combined FG retrograde labeling and immunohistochemistry, using different interneuron markers. FG/EGFP double-labeling in the MEC was found in a subpopulation of parvalbumin+ (PV) neurons (fig. S10A) and in additional GABAergic neurons of unknown immunochemical identity. Injection of AAV DIO ChR2-mCherry into the MEC of PVCre (18) mice further substantiated the finding that MEC PV+ cells projected to the hippocampus (fig. S10B) and formed functional synapses on hippocampal interneurons (n = 5 electrophysiologically identified target cells in CA1-3 and DG) (fig. S10, C and D).

To directly investigate whether long-range GABAergic cells modulate the activity of targeted cells, we recorded sub- and suprathreshold activity in targeted interneurons in slices of the MEC and the hippocampus during stimulation of long-range–projecting axons. Laser stimulation at 8 Hz for 1 s enhanced rhythmic firing of the postsynaptic neuron, as indicated by the reduction of the action potential number during the first half of each 125-ms-pulse interval in target cells in the MEC (n = 5 cells, P = 0.03) and in the hippocampus (n = 6 cells, P = 0.02) (Fig. 4A). Thus, axonal stimulation resulted in an increase in rhythmicity in theta range, as shown by the spectrogram of individual neurons (Fig. 4B) and the comparison of theta power during and before stimulation (n = 5 and 6 and P = 0.03 and 0.02 for MEC and hippocampus target cells, respectively) (Fig. 4C). There was no significant difference in the overall firing rate before, during, and after stimulation (fig. S11A). Axonal stimulation at 40 Hz frequency did not affect rhythmic firing of target cells (n = 7 and 5 for MEC and hippocampus, respectively) (fig. S11, B and C). Axonal stimulation at 8 Hz also increased subthreshold oscillations at theta frequency (n = 10 and 7 and P = 0.0002 and 0.02 for MEC and hippocampus target cells, respectively) (fig. S11, D to G).

Fig. 4

Activation of GABAergic long-range projections enhances rhythmic activity in the MEC and the hippocampus. MEC target cells in hippocampal AAV DIO ChR2-mCherry–injected SOMCre mice [(A) to (C), left] and hippocampal target cells in MEC AAV DIO ChR2-mCherry–injected GADCre mice [(A) to (C), right] cells were patched and depolarized to suprathreshold potentials. Long-range projections were stimulated at 8 Hz (red ticks). (A) Overlay of 20 unfiltered traces recorded in a target cell, with indicated enlargement of action potential firing during the stimulation period. Scale bars, 40 mV and 500 ms. Histogram below indicates mean number of spikes ± SEM within the first 62.5-ms interval directly after laser stimulation (red bar) and the subsequent 62.5-ms interval (black bar). (B) Spectrogram showing that activation of long-range–projecting axons entrains target cell to fire rhythmically at theta range frequency. (C) Increase in theta power (7 to 9 Hz) in target cells during laser stimulation. (D) Representative unfiltered (top) and filtered (3 to 7 Hz, bottom) trace of DHPG/NBQX–induced CA1 theta oscillations. (Right) Averaged power of 3- to 7-Hz oscillations normalized to power before stimulation for wild-type (black) and GADCre-injected mice (red). Scale bars, 500 ms and 0.1 mV.

Last, because theta oscillations can be induced pharmacologically in acute hippocampal slices (19) we analyzed whether recruitment of long-range GABAergic cells affected network activity. We recorded (S)-3,5-dihydroxyphenylglycine (DHPG)/2,3-dihydroxy-6-nitro-7-sulfamoyl-benzo[f]quinoxaline-2,3-dione (NBQX)–induced theta oscillations in the CA1 region. Hippocampal 8-Hz laser stimulation of MEC-derived axons increased theta power (3 to 7 Hz) significantly in slices obtained from MEC virus-injected GADCre (n = 14 slices, P < 0.05) but not wild-type mice (n = 7 slices, P > 0.1) (Fig. 4D). Stimulation at 40 Hz did not change gamma power (n = 12 slices) (fig. S11H).

Using optogenetic viral tracing, we identified long-range GABAergic neurons connecting the hippocampus and the MEC. Furthermore, we provided functional evidence that long-range GABAergic neurons target local interneurons whose activity they modulate. It has been postulated that long-range–projecting GABAergic neurons might be an ideal substrate to precisely coordinate activity between distant brain regions (20). Long-range GABAergic neurons in the hippocampal-entorhinal formation might well account for the highly synchronized theta activity in the hippocampus and entorhinal cortex (21) and thus contribute to the proposed mechanisms underlying spatial and temporal coding and ultimately spatial memory (22, 23).

Supporting Online Material

Materials and Methods

Figs. S1 to S11


References and Notes

  1. Acknowledgments: We thank K. Deisseroth for the AAVs, A. Vogt for help in generating mice, I. Preugschat-Gumprecht and R. Hinz for their technical assistance, and P. H. Seeburg for helpful discussions. This work was supported by a European Research Council grant (GABAcellsAndMemory grant 250047, to H.M.) and a German Ministry of Education and Research (BMBF) grant 01GQ1003A to H.M. and E.C.F.
View Abstract

Navigate This Article