Abstract
A key mode of neuronal communication between distant brain regions is through excitatory synaptic transmission mediated by long-range glutamatergic projections emitted from principal neurons. The long-range glutamatergic projection normally forms numerous en passant excitatory synapses onto both principal neurons and interneurons along its path. Under physiological conditions, the monosynaptic excitatory drive onto postsynaptic principal neurons outweighs disynaptic feedforward inhibition, with the net effect of depolarizing principal neurons. In contrast with this conventional doctrine, here we report that a glutamatergic projection from the hypothalamic supramammillary nucleus (SuM) largely evades postsynaptic pyramidal neurons (PNs), but preferentially target interneurons in the hippocampal CA3 region to predominantly provide feedforward inhibition. Using viral-based retrograde and anterograde tracing and ChannelRhodopsin2 (ChR2)-assisted patch-clamp recording in mice of either sex, we show that SuM projects sparsely to CA3 and provides minimal excitation onto CA3 PNs. Surprisingly, despite its sparse innervation, the SuM input inhibits all CA3 PNs along the transverse axis. Further, we find that SuM provides strong monosynaptic excitation onto CA3 parvalbumin-expressing interneurons evenly along the transverse axis, which likely mediates the SuM-driven feedforward inhibition. Together, our results demonstrate that a novel long-range glutamatergic pathway largely evades principal neurons, but rather preferentially innervates interneurons in a distant brain region to suppress principal neuron activity. Moreover, our findings reveal a new means by which SuM regulates hippocampal activity through SuM-to-CA3 circuit, independent of the previously focused projections from SuM to CA2 or dentate gyrus.
SIGNIFICANCE STATEMENT The dominant mode of neuronal communication between brain regions is the excitatory synaptic transmission mediated by long-range glutamatergic projections, which form en passant excitatory synapses onto both pyramidal neurons and interneurons along its path. Under normal conditions, the excitation onto postsynaptic neurons outweighs feedforward inhibition, with the net effect of depolarization. In contrast with this conventional doctrine, here we report that a glutamatergic input from hypothalamic supramammillary nucleus (SuM) largely evades PNs but selectively targets interneurons to almost exclusively provide disynaptic feedforward inhibition onto hippocampal CA3 PNs. Thus, our findings reveal a novel subcortical-hippocampal circuit that enables SuM to regulate hippocampal activity via SuM-CA3 circuit, independent of its projections to CA2 or dentate gyrus.
Introduction
The dominant mode of neuronal communication between distant brain regions is the excitatory synaptic transmission mediated by long-range glutamatergic axonal projections emitted from excitatory neurons. Typically, excitatory neurons in a specific brain region send out glutamatergic axonal projections to its targets and form en passant excitatory synapses along the path onto different cell types, including both excitatory and inhibitory neurons. For example, in the hippocampus, CA3 pyramidal neurons (PNs) project to CA1 via Schaffer collaterals, which make numerous en passant excitatory synapses onto both PNs and different types of interneurons in CA1 (X. G. Li et al., 1994). Stimulation of Schaffer collaterals rapidly elicits suprathreshold spikes in certain CA1 interneurons, such as parvalbumin (PV)+ interneurons, which in turn inhibit CA1 PNs (Pouille and Scanziani, 2001; Pelkey et al., 2017). This is known as disynaptic feedforward inhibition (Buzsaki, 1984; Pelkey et al., 2017), whose function is thought to improve the temporal precision of synaptic depolarization and spikes in postsynaptic CA1 PNs by narrowing the window for temporal summation of synaptic potentials (Pouille and Scanziani, 2001). Under physiological conditions, the monosynaptic excitatory drive from Schaffer collaterals onto CA1 PNs generally outweighs the feedforward inhibitory drive. Thus, the net effect of Schaffer collateral activation is to depolarize or excite downstream CA1 PNs (Pouille et al., 2009; Isaacson and Scanziani, 2011). This mode of neuronal communication is almost universally true across most brain areas, including the hippocampus and neocortex (Lawrence and McBain, 2003; Torborg et al., 2010; Isaacson and Scanziani, 2011; Pelkey et al., 2017). In contrast with this conventional doctrine, we report here that a long-range hypothalamic glutamatergic input largely evades PNs but selectively targets interneurons in the hippocampal CA3 region. As a result, this glutamatergic input leads to minimal excitation, but instead, nearly exclusively provides feedforward inhibition onto CA3 PNs.
We focused on the CA3 region, as it is often viewed as a key hippocampal region that is vital for memory formation and intrahippocampal information processing (Witter, 2007; Rebola et al., 2017). Previous studies, including ours, on CA3 circuit function have largely focused on its recurrent excitatory connections and mossy fiber projections from the dentate gyrus (DG) (Miles and Wong, 1986; Henze et al., 2000; Nakazawa et al., 2002; Nicoll and Schmitz, 2005; McHugh et al., 2007; Witter, 2007; Le Duigou et al., 2014; Guzman et al., 2016; Rebola et al., 2017; Sun et al., 2017). By contrast, the extrahippocampal inputs to CA3 remain relatively understudied. Although neural tracing studies have provided anatomic evidence that CA3 receives projections from several extrahippocampal regions, including supramammillary nucleus (SuM) in the hypothalamus (Haglund et al., 1984; Vertes, 1992, 2015; Lin et al., 2021), the physiological function of SuM-CA3 pathway remains unknown. Indeed, previous studies on SuM-hippocampal circuit have exclusively focused on the projections from SuM to DG and CA2 (Carre and Harley, 1991; Pan and McNaughton, 2004; Soussi et al., 2010; Vertes, 2015; Hashimotodani et al., 2018; Ito et al., 2018; Chen et al., 2020; Y. Li et al., 2020; Ajibola et al., 2021; Farrell et al., 2021; Oliva, 2021; Robert et al., 2021; Y. D. Li et al., 2022; Tabuchi et al., 2022). In comparison, the SuM-to-CA3 projection has been largely neglected.
In this study, we conducted cell type-specific retrograde and anterograde tracing and ChannelRhodopsin2 (ChR2)-assisted ex vivo patch-clamp recordings to examine the connectivity between SuM and CA3. Our results show that, unlike DG and CA2 that receive direct glutamatergic excitation from SuM (Hashimotodani et al., 2018; Chen et al., 2020; Robert et al., 2021; Tabuchi et al., 2022), SuM's projection to CA3 PNs is dominated by feedforward inhibition as a result of its selective targeting of CA3 interneurons over PNs.
Materials and Methods
Animals
All mouse lines, including C57BL/6J (stock #000664), Grik4-Cre (stock #006474), PV-Cre (stock #017320), somatostatin-Cre (SOM-Cre, stock #013044), and Ai9 tdTomato reporter (stock #007909), were purchased from The Jackson Laboratory. Mice were housed and bred in the Animal Resource Center of Case Western Reserve University School of Medicine on a 12/12 h, light/dark cycle with ad libitum access to food and water. Both male and female mice (2-5 months old) with C57BL6 background were used in the experiments. The procedures described were conducted in accordance with National Institutes of Health regulations and approved by the Institutional Animal Care and Use Committees of Case Western Reserve University.
Hippocampal slice preparation
Transverse hippocampal slices were prepared from 2- to 5-month-old mice as described previously (Sun et al., 2014, 2017, 2020). In brief, animals were anesthetized with isoflurane and killed by decapitation in accordance with institutional regulations. Both left and right hippocampi were immediately dissected out from the mouse brain on ice, and the entire hippocampus was embedded in a block of premade agar (3%-4%). The transverse hippocampal slices (400 µm thick) were then cut from dorsal to ventral on a Leica VT1000 vibratome (Leica Biosystems) in ice-cold dissection solution containing the following (in mm): 10 NaCl, 195 sucrose, 2.5 KCl, 10 glucose, 25 NaHCO3, 1.25 NaH2PO4, 2 Na-pyruvate, 0.5 CaCl2, and 7 MgCl2. Only dorsal slices (dorsal 20%-50%) were used in this study. Slices were incubated at 32°C in regular ACSF as follows (in mm): 125 NaCl, 2.5 KCl, 20 glucose, 25 NaHCO3, 1.25 NaH2PO4, 2 Na-pyruvate, 2 CaCl2, and 1 MgCl2) for 20-30 min and then kept at room temperature for at least 1.5 h before transfer to the recording chamber. Dissection solution and recording ACSF were both saturated with 95% O2 and 5% CO2 (pH 7.4). All electrophysiological recordings were performed at 31°C-32°C.
Transverse division along the CA3-CA2 axis
To determine the proximodistal position of CA3 PNs, we performed biocytin-based staining in most recorded neurons. The relative position of the recorded CA3 PNs along the proximodistal axis was normalized, with 0 assigned to the most distal CA3 at the end of mossy fiber near CA2 and 1 assigned to the most proximal CA3. We evenly divided CA3 into three segments along the proximodistal axis (from proximal to distal: CA3c, CA3b, and CA3a), as we previously described (Sun et al., 2017, 2020). CA2 and CA3a PNs were first evaluated based on their input resistance, and were then further verified using post hoc biocytin staining, based on the absence or presence of thorny excrescences, respectively.
Whole-cell patch-clamp recording
Whole-cell recordings were obtained from CA2 and CA3 PNs using the “blind” patch-clamp technique under a 4× objective, as described previously (Sun et al., 2014, 2017). Recordings from PV-tdTomato+ cells were obtained using conventional fluorescence guided visualized patch-clamp technique under a 40× water-immersion objective. The patch pipettes were pulled from borosilicate capillary glass using a P-1000 Micropipette Puller (Sutter Instruments) and had resistances ranging from 4 to 6 mΩ. The pipettes were filled with an intracellular solution of the following chemicals (in mm): 135 K-Gluconate, 5 KCl, 0.1 EGTA-Na, 10 HEPES, 2 NaCl, 5 ATP, 0.4 GTP, 10 phosphocreatine (pH 7.2; 280-290 mOsm). For IPSC measurements under voltage clamp, Cs+ based intracellular solution was used: 135 Cs-methanesulfonate, 5 KCl, 0.1 EGTA-Na, 10 HEPES, 2 NaCl, 5 ATP-Mg, 0.4 GTP-Na2, 10 phosphocreatine-Na2 (pH 7.2; 280-290 mOsm); 0.2%-0.5% biocytin was routinely included in the intracellular solution. Liquid junction potential was not corrected. Series resistance was monitored throughout each experiment; neurons with a series resistance >30 mΩ were excluded from analysis. In some experiments, multiple CA3 PNs at different transverse positions in the same slice were sequentially recorded (in random order), as previously described (Sun et al., 2017, 2020). During voltage-clamp recording, neurons were held at –75 or 0 mV for EPSC or IPSC measurements, respectively, whereas postsynaptic potentials (PSPs) and actional potentials were measured at resting potential in current-clamp mode. In some current-clamp recording experiments, a constant positive current was injected into the soma to maintain membrane potential at ∼−50 mV to reveal hyperpolarizing responses mediated by inhibition. Input resistance was calculated by measuring the ratio of net change of peak membrane voltage in response to a 1 s, −100 pA constant current injection under current clamp.
Optogenetics
The 470 nm blue light pulses were delivered via a 40× objective placed above the slices through an LED light source (LED4D294, Thorlabs) driven by pClamp11 software. Photostimuli consisting of 2 ms blue light pulses of maximum light intensity were delivered to evoke light response. In some experiments, the long duration light pulses (5-10 ms) were used.
Stereotaxic viral injection
Viral injections were performed as described previously (Sun et al., 2017). Briefly, mice were anesthetized with isoflurane (1%-3%) and placed in a digital stereotaxic apparatus (RWD Life Science, catalog #68045). The glass micropipettes pulled from a P-1000 puller (Sutter Instrument) were used for viral injection. For AAVretro tracing, a pAAVretro-hSyn-mCherry (AAVretro-mCherry, Addgene) was injected into CA3 of WT mice unilaterally, using the coordinates (AP: −1.90 mm; ML: 2.00 mm; DV: −2.20 mm). Animals were killed for histology after 3 weeks.
For rabies virus retrograde tracing, AAV2-hSyn-FLEX-TVA-P2A-eGFP-2A-oG (helper virus) and EnvA G-deleted rabies-mCherry were purchased from the Salk Viral Vector Core; 200-400 nl helper virus was injected into CA3 of Grik4-Cre, PV-Cre, or SOM-Cre mice unilaterally, using the coordinates of AP: −1.90 mm; ML: 2.00/2.20 mm; DV: −2.20/−2.30 mm, to target CA3 PNs, PV interneurons, or SOM interneurons, respectively. Two weeks later, EnvA G-deleted rabies-mCherry (100-400 nl) was injected into CA3 using the same coordinates to target the cells infected by helper virus. The rabies virus was allowed to replicate and retrogradely spread from the starter cells to first-order presynaptic neurons for 1 week before the animals were killed.
For SuM anterograde tracing, 100 nl AAV8-hSyn-ChR2-EYFP (Salk Viral Vector Core) was injected into SuM of WT mice bilaterally using the coordinates (AP: −2.70 mm; ML: ±0.50 mm; DV: −4.80/−5.00 mm). Three weeks later, animals were killed for histology.
To selectively target ChR2-EYFP into SuM neurons that project to CA3, an AAVretro carrying Cre recombinase (AAVretro-EF1α-mCherry-IRES-Cre, 100 nl, Addgene) was unilaterally injected into CA3 of WT mice first, using the coordinates (AP: −1.90 mm; ML: 2.20 mm; DV: −2.30 mm). Cre would retrogradely spread to presynaptic SuM neurons that project to CA3. After 2 weeks, a second Cre-dependent AAV (AAVdj-EF1a-DIO-hChR2-EYFP-WPRE-pA, 200 nl, Salk Viral Vector Core) was injected into SuM bilaterally using the coordinates (AP: −2.80 mm; ML: ±0.40 mm; DV: −4.80 mm) to target CA3-projecting SuM neurons. Animals were killed for histology after 2 weeks.
For optogenectic stimulation of SuM fibers, 200-400 nl of AAV8-hSyn-ChR2-EYFP (Salk Viral Vector Core) was injected bilaterally into SuM in the WT or PV-tdTomato mice (AP: −2.70 mm; ML: ±0.50 mm; DV: −4.80 mm). Animals were killed 3-5 weeks after viral injection for electrophysiological recording.
Biocytin staining
Biocytin staining was performed as described previously (Sun et al., 2017). Briefly, slices that underwent whole-cell recordings were fixed at 4°C for 24-48 h in 4% PFA in PBS, pH 7.3. The fixed slices were subsequently treated with PBS containing 5% normal goat serum and 0.5% Triton for 2 h at room temperature. Slices were then incubated in PBS containing streptavidin, AlexaFluor-594 or -647 conjugate (1:500) and 0.1% Triton for 2 d at 4°C. Subsequently, slices were rinsed in PBS several times and embedded in VECTASHIELD Antifade Mounting Medium (Vector Laboratories), followed by fluorescent imaging using a confocal microscope (LSM 800; Zeiss).
Immunohistochemistry and confocal microscopy
Mice were deeply anesthetized, perfused with precooled PBS, followed by 4% PFA in PBS. Brains were dissected out and postfixed with 4% PFA overnight at 4°C. Coronal sections (45-µm-thick) were cut using Leica 1000-S vibratome. Floating sections were first rinsed 3 times in 1× PBS and then blocked in 1× PBS with 0.3%-1% Triton X-100 and 3%-5% normal goat serum for 1 h at room temperature. Incubation with primary antibodies was performed at 4°C overnight in 1× PBS with Triton X-100. The following primary antibodies were used: chicken anti-GFP (1:500, Aves Lab, catalog #GFP-1020); rabbit anti-PV (1:1000, Abcam, catalog #ab11427); and rabbit anti-PCP4 (1:200, Sigma, catalog #HPA005792). Sections were then washed 3 times in 1× PBS and incubated with secondary antibodies (goat anti-chicken 488, 1:1000, Invitrogen, catalog #A11039; goat anti-rabbit 594, 1:1000, Invitrogen, catalog #A11037; or goat anti-rabbit 647, 1:1000, Invitrogen, catalog #A21245) for 1-2 h at room temperature. Sections were then washed 3 times in 1× PBS and incubated with DAPI (Invitrogen, catalog #D1306, 0.5 µg/ml) for 10-20 min at room temperature. Sections were washed 3 times in 1× PBS, mounted and embedded in VECTASHIELD Antifade Mounting Medium (Vector Laboratories). The fluorescence images were obtained using a laser scanning confocal microscope (LSM 800; Carl Zeiss) and analyzed in ImageJ.
Experimental design and statistical analyses
Student's t test, one-way ANOVA, or two-way ANOVA test followed by a Tukey test for multiple comparisons were used for statistical analysis. Statistics were analyzed using Excel with Microsoft 365 (Microsoft), Igor Pro 8 software (WaveMetrics), or OriginPro 2022b (OriginLab). Only regular-spiking CA3 PNs were included in this study. All data are expressed as mean ± SEM. p < 0.05 is considered as statistically difference.
Results
Anatomical evidence of the monosynaptic projection from SuM to CA3
To explore the connectivity between SuM and CA3, a retrograde AAV (AAVretro) carrying mCherry was unilaterally injected into dorsal CA3 (Fig. 1A–C) (Tervo et al., 2016). We found an abundant expression of mCherry+ cells in SuM (Fig. 1A–C), indicating that SuM neurons project to CA3. To further address whether CA3 PNs receive monosynaptic projection from SuM, we next performed cell type-specific rabies virus-based retrograde tracing to determine the brain regions that send monosynaptic projections to CA3 (Fig. 1D–F). We selectively targeted CA3 PNs using a Grik4-Cre line, in which Cre expression is restricted to CA3 PNs and DG granule cells (Nakazawa et al., 2002). A Cre-dependent AAV-DIO-TVA-EGFP-oG helper virus was unilaterally injected into dorsal CA3, followed by a second injection of pseudo-typed rabies-mCherry 2 weeks later. Mice were killed for histology 7 d after rabies virus injection (Fig. 1D). We observed mCherry+ cells in SuM in 4 of 5 Grik4-Cre mice tested (Fig. 1D–F), indicating the monosynaptic projections from SuM to CA3 PNs.
Retrograde and anterograde tracing reveals the monosynaptic projection from SuM to CA3. A-C, Identification of SuM-CA3 projection using AAVretro-mCherry. Repeated in 3 mice. A, Schematic view of SuM-CA3 projection and viral injection site in dorsal CA3 in WT mice. B, Sample confocal image represents the injection site of AAVretro-mCherry in dorsal CA3. C, Sample confocal image represents mCherry+ cells observed in SuM (left). An expanded view shown in right. f, Fornix. D-F, Pseudotyped rabies tracing in Grik4-Cre mice reveals the monosynaptic projection from SuM to CA3. Repeated in 5 mice. D, Schematic view of the experimental procedure and timeline. E, Sample confocal image represents the injection site of rabies virus in dorsal CA3. F, Sample confocal images represent mCherry+ cells observed in SuM (left) and an expanded view (right). G-K, Anterograde tracing represents the SuM axonal projection patterns in the hippocampus. Repeated in 2 mice. G, Schematic view of SuM-CA3 projection and injection of AAV-ChR2-EYFP into SuM in WT mice. H, Sample confocal image showing the injection site in SuM and the expanded view (inset). I-K, ChR2-EYFP+ fiber distribution in the hippocampus. Note the dense ChR2-EYFP+ fibers in DG, CA2, and distal CA3 (CA3a), and the relatively sparse expression of ChR2-EYFP+ fibers observed in proximal CA3 (CA3b/c). Purkinje Cell Protein 4 (PCP4, red) is a marker for CA2. s.o., stratum oriens; s.p., stratum pyramidale; s.l., stratum lucidum; s.r., stratum radiatum; s.l.m., stratum lacunosum moleculare.
To examine the SuM fiber projections in the hippocampus, we injected AAV-Syn-ChR2-EYFP into SuM in WT mice (Fig. 1G–K). Consistent with the previous reports (Chen et al., 2020; Haglund et al., 1984; Vertes, 1992), we observed a dense expression of ChR2-EYFP+ fibers in DG and CA2 areas (Fig. 1I–K). Notably, we also detected the expression of ChR2-EYFP+ fibers in CA3 (Fig. 1I–K), confirming our retrograde tracing results above (Fig. 1A–F). We note that ChR2-EYFP+ fibers were significantly sparser in CA3 than those in DG and CA2 (Fig. 1I–K), which may explain the omission of SuM-CA3 pathway in previous studies (Chen et al., 2020; Tabuchi et al., 2022). Furthermore, we found that ChR2-EYFP+ fibers were distributed throughout the transverse axis in CA3, with relatively dense expression in stratum pyramidale and stratum oriens in distal CA3 (CA3a) and sparse expression in the other layers and proximal CA3 (CA3b/c) (Fig. 1I–K).
To more rigorously examine the SuM-CA3 projections, we next selectively targeted ChR2-EYFP into SuM neurons that project to CA3. We targeted Cre recombinase into CA3-projecting SuM neurons by injecting an AAVretro-Cre-mCherry into CA3 (Fig. 2A–C). After a 2 week waiting period to allow for Cre expression, a second Cre-dependent AAV-DIO-ChR2-EYFP was injected into SuM (Fig. 2A–C). Thus, CA3-projecting SuM neurons were labeled with ChR2-EYFP (Fig. 2A–C). Mice were killed 2 weeks after the second injection to examine ChR2-EYFP expression. We confirmed the sparse expression of ChR2-EYFP in a subset of SuM neurons, which were colabeled by mCherry (Fig. 2A–C). Importantly, we found expression of ChR2-EYFP+ fibers along the transverse axis in CA3 (Fig. 2D,E). We also observed ChR2-EYFP+ fibers in DG, CA2 (Fig. 2D), and other brain regions (Fig. 3). Together, we conclude that CA3-projecting SuM neurons send collaterals to CA2, CA3, and DG in the hippocampus.
Selective targeting of ChR2-EYFP into CA3-projecting SuM neurons. A, Experimental procedure and timeline. B, Sample confocal image represents the expression of Cre-mCherry in dorsal CA3, the first injection site of AAVretro-Cre-mCherry. C, Left, Sample confocal image represents the expression of ChR2-EYFP in SuM, the second injection site of Cre-dependent AAV-DIO-ChR2-EYFP. Right, Expanded view represents colabeling of Cre-mCherry and ChR2-EYFP in SuM cells. D, E, The distribution of ChR2-EYFP+ fibers in the hippocampus. Note the dense ChR2-EYFP+ fibers in DG, CA2, and CA2/CA3a border, and the relatively sparse expression in CA3b/c. CA3SR, CA3 stratum radiatum. Repeated in 3 mice.
CA3-projecting SuM neurons project to the regions outside of the hippocampus. Experimental procedure is the same as in Figure 2. SuM fibers are detected in other SuM targetregions: A, lateral septum (LS), medial septum (MS), and diagonal band of Broca (DB); B, lateral hypothalamus (LH); C, entorhinal cortex (EC); D, reuniens thalamic nucleus (Re); and E, mediodorsal thalamic nucleus (MD).
SuM provides minimal excitation, but strong feedforward inhibition onto CA3 PNs
The anatomic tracing results above raise the possibility that SuM may make functional connections with CA3 PNs. Thus, we performed ChR2-assisted whole-cell voltage-clamp recording to examine the functional connections between SuM and CA3 (Fig. 4). AAV-Syn-ChR2-EYFP was injected into SuM bilaterally, and acute hippocampal slices were prepared 3-5 weeks following viral injection (Fig. 4). Because there is no clear-cut border between CA2 and CA3a (Kohara et al., 2014; Sun et al., 2017), we used input resistance and the absence or presence of thorny excrescences to differentiate CA2 versus CA3a PNs in the CA2/CA3 bordering region (Fig. 4A), as we described previously (Sun et al., 2017). Consistent with our previous finding, we found that input resistance of CA3a PNs more than twofold greater than CA2 PNs, with almost no overlap between two groups (CA3a: 112.5 ± 10.4 mΩ, ranging from 70.2 to 169.2 mΩ, n = 11 cells; CA2: 51.7 ± 3.8 mΩ, ranging from 40.2 to 64.8 mΩ, n = 7 cells; p = 0.00035, unpaired t test).
Optogenetic simulation of SuM input leads to minimal excitation, but strong feedforward inhibition onto CA3 PNs. A, Sample confocal images represent biocytin-filled CA2 and CA3a PNs from a transverse hippocampal slice that underwent whole-cell recording. Expanded views (right) represent the presence and absence of thorny excrescences (arrowheads) in CA3a and CA2 PNs, respectively. B, Membrane voltage responses of CA2 (left) and CA3a (right) PNs in response to 1 s somatic current injections. C, Voltage-clamp recording represents the light-evoked EPSC responses in PNs in CA2 and CA3a, but not in CA3b or CA3c. D, Percent of CA2 and CA3 PNs that receive light-evoked EPSCs along the transverse axis. Numbers of cells shown above bars. E, Group data of the amplitude of light-evoked EPSC of CA2 (n = 10 cells/7 slices/6 mice) and CA3a PNs (n = 7 cells/5 slices/5 mice). *p < 0.05 (unpaired t test). F, Sample confocal images represent biocytin-filled CA3 PNs from a hippocampal slice that underwent whole-cell recordings. At least three CA3 PNs from a, b, and c subregions were recorded in a same slice. G, Voltage-clamp recording represents the light-evoked IPSC (holding potential at 0 mV) from CA3a, b, and c. Sample IPSC traces of three CA3 cells were from the same slice. H, Summary data of light-evoked ISPCs in CA3a, b, and c. Each line indicates CA3 PNs from the same slice. n = 5 slices/3 mice. *p < 0.05 (one-way repeated-measures ANOVA followed by a Tukey test). I, Schematic illustration of the position of CA3 PNs that displayed light-evoked IPSC responses. Light-evoked IPSCs were detected in all CA3 PNs tested along the transverse axis. J, K, Sample traces (J) and group data (K) of light-evoked ISPCs in the absence or presence of DNQX (20 µm) and APV (50 µm). n = 5 cells/5 slices/3 mice. *p < 0.05 (paired t test). L, M, Sample traces (L) and group data (M) of light-evoked ISPCs in the absence or presence of DAMGO (100 nm). n = 5 cells/5 slices/2 mice. *p < 0.05 (paired t test).
In line with previous reports (Chen et al., 2020; Robert et al., 2021), we detected light-evoked EPSC responses in CA2 PNs (100%, 10 of 10 cells). By contrast, we only detected light-evoked EPSCs in a fraction of CA3a PNs (30.4%, 7 of 23 cells), and did not detect any excitatory responses in proximal CA3 (CA3b/c, 0%, 0 of 28 cells) (Fig. 4C,D). Furthermore, the light-evoked EPSC in CA3a was very small (−18.0 ± 2.5 pA, n = 7 cells), and was significantly weaker than CA2 (−41.7 ± 7.5 pA, n = 10 cells, p = 0.022, unpaired t test) (Fig. 4E).
Remarkably, in contrast with a negligible level of excitatory response, light pulses consistently produced large IPSCs (holding potential = 0 mV) in all CA3 PNs tested under voltage clamp (100%, 419.5 ± 106.1 pA, n = 21 of 21 cells, Fig. 4F–I). We have previously reported a proximal-to-distal increasing gradient of PV-mediated inhibition along the transverse axis in dorsal CA3 (Sun et al., 2017). Thus, we next compared light-evoked IPSCs in CA3a, b, and c (Fig. 4F–I). To control for potential variations of ChR2 expression across different animals or slices, we sequentially recorded and compared IPSCs of CA3 PNs at different transverse positions from the same slice (in random order), as described in our previous studies (Sun et al., 2017, 2020). Interestingly, we found that SuM-driven IPSCs also displayed a proximal-to-distal increasing gradient along the transverse axis (Fig. 4F–H). Furthermore, bath application of DNQX and APV (AMPA and NMDA receptor antagonists) reduced IPSCs to 2.2 ± 0.9% of baseline (baseline: 463.7 ± 124.8 pA vs DNQX/APV: 6.9 ± 2.6 pA, n = 5 cells, p = 0.022, paired t test, Fig. 4J,K). These data strongly suggest that the inhibition is driven by the glutamatergic input from SuM to CA3 interneurons, rather than direct GABAergic inhibition. To address the potential source of inhibition, we bath-applied DAMGO, a µ-opioid receptor agonist that is known to selectively suppress PV-mediated inhibition (Glickfeld et al., 2008). We found that DAMGO (100 nm) substantially reduced IPSCs by 87.3 ± 2.7% (baseline: 328.4 ± 101.7 pA vs DAMGO: 36.4 ± 9.0, n = 5 cells, p = 0.038, paired t test, Fig. 4L,M), indicating that PV+ neurons are likely a major source for feedforward inhibition. Together, these results suggest that activation of SuM-CA3 input resulted in minimal excitation, but led to strong feedforward inhibition, which is largely mediated by suprathreshold spikes in PV+ neurons.
SuM strongly excites PV+ neurons in CA3 along the transverse axis
Next, to provide direct anatomic evidence of monosynaptic connection between SuM and CA3 interneurons, we performed cell type-specific pseudo-typed rabies tracing on PV+ or somatostatin (SOM)+ neurons in CA3: the two most prominent types of interneurons. We injected Cre-dependent AAV-DIO-TVA-EGFP helper virus and rabies virus into CA3 in PV-Cre or SOM-Cre mice, respectively (Fig. 5). We found mCherry+ cells in SuM in all 3 PV-Cre mice tested (3.7 ± 0.9 cells, n = 3 mice, Fig. 5B,C), resulting in a connectivity rate of 0.046 ± 0.016 (defined as the ratio of the number of mCherry+ cells in SuM to the number of mCherry+/EGFP+ starter cells in CA3). By contrast, we did not detect any mCherry+ cells in SuM in SOM-Cre mice (repeated in 4 SOM-Cre mice, Fig. 5D,E). These data provide anatomic evidence that PV+ neurons in CA3 receive the monosynaptic input from SuM.
Pseudotyped rabies virus tracing reveals the monosynaptic projection from SuM to PV+ neurons in CA3. A, Experimental procedure and timeline. B, Top left, Sample confocal image represents the viral injection site in dorsal CA3 from a PV-Cre mouse. Top right, Expanded view represents a starter cell (mCherry+/GFP+) that is also immunopositive for PV staining (arrowhead). Bottom, Expanded views of the starter cell. C, Sample confocal images represent mCherry+ cells observed in SuM (top). Two mCherry+ cells in SuM shown in the expanded views (bottom). D, Sample confocal image represents the viral injection site in dorsal CA3 from a SOM-Cre mouse. E, No mCherry+ cells detected in SuM (repeated in 4 SOM-Cre mice).
To directly assess the functional connectivity between SuM and CA3 PV+ neurons, we generated PV-tdTomato mice by crossing a PV-Cre line with an Ai9 tdTomato reporter line, which enabled us to perform ChR2-assisted target patch-clamp recordings on PV+ neurons (Fig. 6). AAV-ChR2-EYFP was bilaterally injected into SuM, and hippocampal transverse slices were prepared 3-5 weeks after viral injection (Fig. 6A–C). We verified that tdTomato+ cells displayed fast-spiking firing pattern in response to 1 s somatic current injection (maximum firing rate = 204.2 ± 17.4 Hz, action potential half-width = 0.277 ± 0.020 ms, n = 26 cells), a characteristic electrophysiological feature of PV+ neurons (Fig. 6D–F). Remarkably, we found light stimulation resulted in strong excitatory responses in nearly half of PV+ neurons (−56.0 ± 9.6 pA, n = 14 cells). Furthermore, bath application of glutamatergic antagonists, DNQX/APV, reduced the light-evoked EPSCs to 4.3 ± 1.6% of baseline (baseline: −69.7 ± 14.7 pA vs DNQX/APV: −3.2 ± 1.5 pA, n = 4 cells, p = 0.017, paired t test), indicating that SuM-CA3 PV+ input is glutamatergic (Fig. 6G,H). To rule out the possibility of disynaptic excitation mediated by light-evoked spikes in DG and/or CA2, we recorded light-evoked EPSCs by bath application of TTX (0.5 μm) and 4-aminopyridine (4-AP, 1 mm; Fig. 6I) (Petreanu et al., 2009). Under this condition, light pulses can reliably elicit EPSP and EPSC under current clamp or voltage clamp (n = 3 cells, Fig. 6I), indicating the monosynaptic connection between SuM and CA3 PV+ neurons. In addition, we found no significant difference of EPSC latencies between CA2/3 PNs (1.54 ± 0.06 ms, n = 12 cells) and PV+ neurons (1.72 ± 0.18 ms, n = 16 cells, p = 0.88, one-way ANOVA followed by a Tukey test), whereas IPSC latency in CA3 PNs (3.66 ± 0.24 ms, n = 17 cells) was significantly longer than EPSC latencies of either PNs or PV+ neurons (p = 2.4 × 10−10, one-way ANOVA, Fig. 6J). These results strongly suggest that SuM forms monosynaptic synapses onto CA3 PV+ neurons and drives disynaptic feedforward inhibition onto CA3 PNs.
SuM provides monosynaptic excitation into CA3 PV+ neurons. A, Genetic strategy to target tdTomato into PV+ neurons by crossing a floxed tdTomato reporter line (Ai9) with a PV-Cre line. B, Experimental procedure and timeline. C, Representative confocal image shows the hippocampal slice from a PV-tdTomato mouse that was injected with AAV-ChR2-EYFP into SuM. D, Schematic diagram of a likely SuM-to-CA3 circuit. Rec, Recording patch pipette. E, Sample confocal image represents a biocytin-filled tdTomato+ neuron in CA3 (arrows) that underwent whole-cell recording. F, Current-clamp recording shows the membrane voltage responses to 1 s somatic injections of indicated currents from a tdTomato+ neuron in CA3. G, H, Sample traces (G) and group data (H) of light-evoked EPSCs in the absence and presence of DNQX/APV. n = 4 cells/4 slices/3 mice. *p < 0.05 (paired t test). I, Sample light-evoked EPSP (top) and EPSC (bottom) under current-clamp and voltage-clamp recordings in the presence of TTX (0.5 µm) and 4-AP (1 mm). Repeated in 3 cells/3 slices/1 mice. J, Group data of latencies of EPSCs in CA2/3 PNs (PN-EPSC, n = 12 cells/8 slices/6 mice), EPSCs in CA3 PV+ neurons (PV-EPSC, n = 16 cells/12 slices/6 mice), and IPSCs in CA3 PNs (PN-IPSC, n = 17 cells/6 slices/3 mice). ***p < 0.001; n.s., Not significant; one-way ANOVA followed by a Tukey test. K, Schematic illustration of the position of the recorded tdTomato+ cells in CA3. The cells that received light-evoked EPSC responses were evenly distributed along the transverse axis. n = 36 cells/18 slices/7 mice. L, Amplitude of light-evoked EPSC of each cell plotted against their normalized position along the transverse axis, with “0” as the most distal and “1” as the most proximal. M, Group data of light-evoked EPSC amplitude from CA3a (n = 6 cells/6 slices/4 mice), CA3b (n = 2 cells/2 slices/2 mice), and CA3c (n = 6 cells/6 slices/4 mice). n.s., Not significant (one-way ANOVA).
Given that SuM fibers were significantly denser in distal CA3 (CA3a) than proximal CA3 (CA3b/c) (Figs. 1I–K, 2D,E), we next asked whether PV+ neurons in distal CA3 receive a stronger excitation from SuM, compared with those in proximal CA3. We found no significant difference of light-evoked EPSCs in PV+ neurons along the transverse axis (Fig. 6K–M).
Previous studies show that SuM provides a pure glutamatergic input onto CA2, but a mixed glutamatergic/GABAergic input onto DG (Hashimotodani et al., 2018; Tabuchi et al., 2022), raising a question of whether SuM-PV input is glutamatergic or mixed glutamatergic/GABAergic. To address this, we next performed voltage-clamp recording to measure both EPSC and IPSC while holding membrane potential at −75 and 0 mV, respectively (Fig. 7). Interestingly, we detected IPSCs in all CA3 PV+ neurons tested (5 of 5 cells). However, the IPSC latency was significantly longer than EPSC (Fig. 7B,C,E, EPSC: 1.72 ± 0.18 ms, n = 16 cells vs IPSC: 3.27 ± 0.26 ms, n = 5 cells, p = 0.00034, unpaired t test), indicating that IPSC is the disynaptic response. Moreover, IPSCs were completely blocked by bath application of DNQX/APV (Fig. 7B,D, baseline: 135.9 ± 34.2 pA, DNQX/APV: 4.8 ± 1.9 pA, n = 5 cells, p = 0.016, paired t test). These results strongly suggest that IPSCs in PV+ neurons were mediated by disynaptic inhibition driven by suprathreshold spikes of the neighboring PV neurons, arguing against a direct SuM-PV monosynaptic GABAergic input.
The absence of monosynaptic GABAergic connection between SuM and CA3 PV+ neurons. A, A schematic view of SuM-to-CA3 circuit and experimental configuration showing target recording of PV+ neurons. PV+ neurons provide mutual inhibition. B, Voltage-clamp recording traces from a PV+ neuron show light-evoked IPSC (hold potential = 0 mV) and EPSC (holding potential = –75 mV) in the absence (black) and presence (red) of DNQX/APV. C, Expanded view shown in B. IPSC onset is slower than EPSC onset. D, Group data of light-evoked IPSC amplitude in the absence and presence of DNQX/APV. n = 5 cells/5 slices/2 mice. *p < 0.05 (paired t test). E, Group data of light-evoked EPSC latencies (n = 16 cells/12 slices/6 mice) versus IPSC (n = 5 cells/5 slices/2 mice). ***p < 0.001 (unpaired t test).
Together, we concluded that, despite relatively sparse innervation in CA3, SuM uniformly provides a potent monosynaptic glutamatergic excitatory drive onto PV+ neurons distributed along the transverse axis in CA3.
Excitatory drive from SuM to PV+ neurons in CA3 is as strong as that to DG granule cells
Our tracing data show that the SuM fibers in CA3 were significantly sparser, compared with its dense innervation in DG and CA2 (Figs. 1, 2). This raises an interesting question as to whether its dense projection resulted in a more robust excitation in DG, compared with CA3 PV+ neurons. To address this, we recorded light-evoked EPSCs from DG granule cells by injecting ChR2 into SuM (Fig. 8). We found that, consistent with the previous reports (Hashimotodani et al., 2018; Chen et al., 2020; Tabuchi et al., 2022), a vast majority of DG granule cells (92%, 11 of 12 cells) received functional inputs from SuM, compared with 48.6% of CA3 PV+ neurons (17/35 cells, Fig. 8C,D). Interestingly, however, the light-evoked EPSCs in PV+ neurons were slightly, but not significantly, greater than DG granule cells (Fig. 8E). To control for the potential variations of ChR2 expression across different animals, we next performed a paired comparison of the mean EPSCs between DG granule cells and CA3 PV+ neurons from each individual animal (Fig. 8F). In line with the data from individual cells (Fig. 8E), the mean EPSCs in CA3 PV+ neurons from each animal also showed a trend of greater than that in DG granule cells (PV: −51.1 ± 7.8 pA vs DG: −28.8 ± 8.1 pA, p = 0.098, paired t test, n = 4 mice), although it did not reach statistical significance (Fig. 8F). We also compared short-term plasticity of light-evoked EPSCs between PV+ neurons and DG granule cells in response to a trains of 20 Hz stimulation. Both SuM-CA3 PV+ and SuM-DG synapses displayed short-term depression (Fig. 8G). However, the level of depression of SuM-DG synapses was significantly greater than SuM-CA3 PV+ synapses (Fig. 8G). Furthermore, although a moderate depolarizing membrane potential can be evoked by light stimulation in DG granule cells, we were not able to evoke spikes in any DG granule cells in response to a single or repetitive light stimulation (Fig. 8I,J), consistent with the previous reports (Hashimotodani et al., 2018). By contrast, light stimulation evoked suprathreshold spikes in a proportion of CA3 PV+ neurons (Fig. 8H,J). Together, we conclude that, despite significantly sparser innervation in CA3 compared with DG, SuM excitatory drive onto CA3 PV+ neurons is at least as strong as that onto densely innervated DG granule cells.
Activation of SuM excites both CA3 PV+ neurons and DG granule cells. A, Sample confocal image represents a biocytin-filled DG granule cell (GC) that underwent whole-cell recording. B, Current-clamp recording shows membrane voltage responses to 1 s depolarizing or hyperpolarizing current injections in a DG granule cell. C, Voltage-clamp recording shows light-evoked EPSC recorded from a DG granule cell. D, Percent of CA3 PV+ neurons and DG granule cells that show light-evoked EPSC responses. Number of cells shown above bars. E, Group data of light-evoked EPSCs from individual CA3 PV+ neurons (n = 14 cells/10 slices/6 mice) and DG granule cells (n = 11 cells/7 slices/6 mice). n.s., Not significant (unpaired t test). F, Group data of light-evoked EPSCs from individual animals. Each circle represents the mean EPSC value from 1 animal. n = 4 mice; paired t test. G, Sample traces (top) and group data (bottom) of EPSC responses in CA3 PV+ neurons (n = 7 cells/6 slices/4 mice) and DG granule cells (n = 5 cells/4 slices/3 mice) in response to a 20 Hz train of 10 pulse light stimulation. **p < 0.01 (two-way repeated-measures ANOVA followed by a Tukey test). H, Sample traces of four trials represent suprathreshold spikes evoked by a 20 Hz train of 10 pulse light stimulation from a CA3 PV+ neuron. I, Sample voltage trace represents subthreshold membrane response evoked by a 20 Hz train of 10 pulse light stimulation in a DG granule cell. J, Probability of spikes evoked by light stimulation in CA3 PV+ neurons versus DG granule cells. Numbers of cells shown above bars.
Contrasting effect of SuM stimulation on CA3 PNs versus DG granule cells
Finally, we evaluated and compared the net effect of the SuM input onto CA3 PNs versus DG granule cells under current-clamp recording. As expected, we only detected a very small depolarization at sub-millivolt level (0.65 ± 0.08 mV, n = 7 cells) at resting potential in a fraction of CA3a PNs (7 of 16 cells), and did not detect any excitatory response in CA3b/c PNs (Fig. 9A). No suprathreshold spikes can be evoked in any CA3 PNs using maximum light intensity, repetitive stimulation (10-20 Hz/10 pulses), or light pulses with a longer duration at 10 ms (data not shown). By contrast, light stimulation resulted in a significantly greater depolarization in DG granule cells at resting potential, compared with CA3 PNs (DG: 3.82 ± 1.41 mV, n = 6 cells; CA3: 0.24 ± 0.07 mV, n = 19 cells, p = 0.000106, unpaired t test, Fig. 9A).
SuM simultaneously excites DG granule cells and suppresses CA3 PNs. A, Sample traces (left) and group data (right) of membrane voltage responses evoked by single light pulses under current clamp in CA3 PNs (n = 19 cells/11 slices/5 mice) versus DG granule cells (n = 6 cells/4 slices/3 mice) at resting potential. ***p < 0.001 (unpaired t test). B, Sample traces (left) and group data (right) of membrane voltage responses evoked by single light pulses in CA3 PNs (n = 19 cells/11 slices/5 mice) versus DG granule cells (n = 6 cells/4 slices/3 mice) while holding the membrane potential at –50 mV. ***p < 0.001 (unpaired t test). C, D, Sample traces (C) and group data (D) of membrane voltage responses evoked by a 10 Hz train of 10 pulse light stimulation at resting potential (n = 4 cells/3 slices/1 mouse) or at –50 mV (n = 7 cells/5 slices/3 mice) in CA3 PNs.
Resting potential is near reversal potential of inhibitory synaptic currents (∼−72 mV) based on our internal and external solutions. Thus, the inhibitory responses, even if they were present, were largely undetectable at resting potential under current clamp. In order to uncover feedforward inhibition, we next injected a small positive current to depolarize the membrane to −50 mV in both CA3 PNs and DG granule cells. At −50 mV, we revealed marked contrasting postsynaptic responses to light stimulation in CA3 PNs versus DG granule cells (Fig. 9B). Light pulses still caused membrane depolarization in DG granule cells (2.28 ± 0.93 mV, n = 6 cells) but resulted in pure hyperpolarization in all CA3 PNs tested (−2.37 ± 0.27 mV, n = 19 cells, p = 8.47 × 10−7, unpaired t test, Fig. 9B). Similarly, a train of 10 Hz light stimulation caused little responses at resting potential but resulted in hyperpolarizing postsynaptic responses at −50 mV in CA3 PNs (Fig. 9C,D). Together, we conclude that, while in DG, the net effect of SuM input is to provide an excitatory drive onto granule cells, its main action on the neighboring CA3 circuit is to suppress CA3 PN activity by eliciting strong feedforward inhibition.
Discussion
In this study, we provide anatomic and functional evidence that SuM selectively innervates interneurons to suppress PN activity in the hippocampal CA3 region. Our study provides a striking example that a long-range glutamatergic excitatory input selectively targets interneurons over principal neurons in a distant brain region to nearly exclusively provide feedforward inhibition. Moreover, our findings reveal a novel hypothalamic-hippocampal circuit by which SuM regulates hippocampal activity via CA3, independent of its dense projections to CA2 or DG.
SuM can regulate hippocampal circuit activity, independent of its projections to DG and CA2
The SuM is known to send long-range projections to the hippocampus to participate in a variety of behavioral tasks (Hashimotodani et al., 2018; Ito et al., 2018; Chen et al., 2020; Robert et al., 2021; Y. D. Li et al., 2022; Tabuchi et al., 2022). The previous studies on the SuM-hippocampus circuit have almost exclusively focused on the SuM-DG and SuM-CA2 inputs, whereas the function of SuM-CA3 input has been largely neglected. This omission likely results from the substantially sparser SuM fiber distributions in CA3 compared with CA2 or DG, as demonstrated by our tracing results and others (Vertes, 1992; Chen et al., 2020). Our patch-clamp recording results reveal that, despite very sparse SuM fibers in CA3, the SuM-CA3 input can have a substantial impact on CA3 network activity, as stimulation of SuM input led to robust disynaptic feedforward inhibition in all CA3 PNs along the transverse axis. Our results thus complement those previous studies by showing that SuM also suppresses CA3 PN activity by exciting CA3 interneurons, in addition to exciting DG and CA2. We propose that SuM can regulate hippocampal circuit activity through exciting CA3 interneurons, independent of its projections to CA2 and DG.
Our results show that, unlike CA3 PNs, nearly all DG granule cells and CA2 PNs receive excitatory input from SuM, and the net effect of SuM-CA2 and SuM-DG pathways is to enhance excitatory drive, consistent with previous reports (Hashimotodani et al., 2018; Chen et al., 2020; Ajibola et al., 2021; Robert et al., 2021; Tabuchi et al., 2022). Although CA2 and DG also receive the SuM-mediated feedforward inhibition (Hashimotodani et al., 2018; Ajibola et al., 2021; Robert et al., 2021), we found no evidence that SuM glutamatergic input evades DG granule cells or CA2 PNs. Thus, unlike the SuM-CA3 input, the role of SuM-CA2 and SuM-DG inputs appears to be more in line with the conventional innervation of most long-range glutamatergic pathways.
Extrahippocampal-to-CA3 connections
Area CA3 is traditionally thought to receive relatively limited excitatory inputs from extrahippocampal regions (Witter, 2007). The two most prominent glutamatergic inputs to CA3 arise from the hippocampal regions, including mossy fibers from DG and recurrent collaterals from CA3 PNs (Le Duigou et al., 2014; Rebola et al., 2017; Sun et al., 2017; Witter, 2007). The only known sizable extrahippocampal input to CA3 is the perforant path from the entorhinal cortex (Sun et al., 2017; Witter, 2007), which is closely associated with the hippocampus as a part of the medial temporal role memory system. Of note, like most conventional long-range glutamatergic pathways, all three pathways make excitatory synapses onto both PNs and interneurons in CA3, and their net effect is to depolarize CA3 PNs under physiological conditions (Miles and Wong, 1986; Torborg et al., 2010; Vyleta et al., 2016; Sun et al., 2017). Here we demonstrate that a novel long-range glutamatergic input from a hypothalamic nucleus also makes functional connections onto CA3. Strikingly, this pathway is markedly distinct from other three pathways with its preferential innervation onto interneurons over PNs. Furthermore, we provide anatomic and functional evidence that SuM strongly excites PV+ neurons in CA3. These results are consistent with the view that PV+ neurons are a major subtype of interneurons responsible for a large portion of feedforward inhibition in many brain areas (Buzsaki, 1984). We do not rule out the possibility that SuM may also innervate other types of CA3 interneurons, which requires further scrutiny.
Our findings add to a growing number of studies that demonstrate the heterogeneity of hippocampal CA3 circuit function (Lein et al., 2007; Hunsaker et al., 2008; Lee et al., 2015, 2020; Lu et al., 2015; Oliva et al., 2016; Sun et al., 2017, 2020; Cembrowski and Spruston, 2019). At the cellular and circuit level, previous studies, including ours, have, primarily focused on the diversity of somatodendritic morphology, intrinsic and dendritic membrane properties, and intrahippocampal connectivity of CA3 PNs (Ishizuka et al., 1990, 1995; X. G. Li et al., 1994; Kowalski et al., 2016; Sun et al., 2017, 2020; Hunt et al., 2018; Raus Balind et al., 2019; Lin et al., 2021; Mago et al., 2021). Yet, the heterogeneity of extrahippocampal inputs to CA3 PNs remains largely unexplored. This study provides the evidence that the heterogeneity of CA3 connectivity extends to the extrahippocampal-CA3 circuits. While a proportion of very distal CA3 PNs (near CA2) receive weak excitation from SuM, we did not detect any excitatory connections between SuM and proximal CA3 PNs. Furthermore, SuM-driven feedforward inhibition also displayed a proximal-to-distal increasing gradient. This result can be explained by our previous finding that CA3 PV+ neurons provide a proximal-to-distal increasing gradient of inhibition onto CA3 PNs (Sun et al., 2017).
Functional implication of the SuM-CA3 pathway
What is the functional advantage of this unique innervation pattern of the SuM-to-hippocampus circuit? We propose that activation of SuM may be well positioned to coordinate the CA2-CA3-DG circuit activity by exciting DG and CA2 but simultaneously suppressing background activity of the CA3 circuit. This action can potentially result in a better signal-to-noise ratio in the CA3 circuit. One potential scenario is that activation of SuM input, under certain behavioral paradigms, excites a population of DG granule cells to reach suprathreshold spikes and simultaneously reduces the overall background activity in CA3. Mossy fibers from DG granule cells are a major excitatory drive for CA3 PNs and are critical for pattern separation or memory discrimination (Leutgeb et al., 2007; McHugh et al., 2007). By contrast, CA3 recurrent network is thought to be important for pattern completion or memory generalization (Nakazawa et al., 2002, 2003). Thus, activation of SuM can potentially enhance behavioral discrimination by simultaneously activating and suppressing DG granule cells and CA3 PNs, respectively. Because the same SuM neurons appear to send collaterals to all three adjacent hippocampal subregions, it is experimentally challenging to test this idea in vivo in behaving animals. However, future studies using computational model hold the promise to test this idea.
A number of behaviors, including contextual and spatial memory and social behavior, have been identified to be attributed to either SuM-DG or SuM-CA2 projections (Chen et al., 2020; Y. Li et al., 2020; Y. D. Li et al., 2022). For example, SuM-CA2 projection is thought to be important for social memory, whereas SuM-DG projection is involved in contextual novelty detection and spatial memory (Chen et al., 2020; Y. Li et al., 2020; Oliva, 2021; Y. D. Li et al., 2022). Given that CA3 plays some different behavioral roles compared with CA2 and DG (Nakazawa et al., 2002; Kesner et al., 2004; Nakashiba et al., 2008; McHugh and Tonegawa, 2009; Rolls, 2013; Kesner and Rolls, 2015; Guzman et al., 2016; Oliva et al., 2020), our findings raise an interesting question of whether the SuM-CA3 input may also participate in specific behaviors.
Footnotes
This work was supported by National Institutes of Health K01MH117444, R01MH129294, and R01MH130367 to Q.S.; and institutional support from Case Western Reserve University to Q.S.
The authors declare no competing financial interests.
- Correspondence should be addressed to Qian Sun at qxs111{at}case.edu
