Abstract
Autapses are self-synapses formed by a single neuron. They selectively form in a subpopulation of neocortical glutamatergic pyramidal cells (PCs) where autaptic transmission provides strong feedback regulation of self-activity in individual neurons. PCs in the hippocampal formation (HPF) possess morphological and electrophysiological characteristics similar to neocortical PCs; it remains unclear, however, whether they form functional autapses. We performed whole-cell recording from HPF PCs in acute slices obtained from mice of either sex and found surprisingly that none of the recorded PCs in CA1–3 show autaptic responses; only a subpopulation of PCs (∼50%) in the subiculum forms functional autapses, particularly those targeting to the nucleus accumbens. Further experiments reveal that the autaptic responses in subicular PCs are mediated solely by AMPA receptors but not NMDA receptors and occur much earlier than those of the medial prefrontal cortex during early development. Together, the results indicate that functional autapses selectively form in a considerable subset of subicular PCs but are completely absent from PCs in the hippocampus proper, suggesting a key role of autapses in regulating the self-activity of subicular PCs and thus the main output signals of the hippocampus.
Significance Statement
Unlike conventional synapses, autapses are self-synapses providing feedback regulation of a neuron’s own activity. We find that autapses selectively form in a subpopulation of subicular pyramidal cells (PCs) but are entirely absent in other hippocampal formation (HPF) PCs, suggesting a key role of autapses in regulating the self-activity of individual subicular PCs and thereby the primary output of HPF. Additionally, autapses of subicular PCs emerge earlier than those in medial prefrontal cortex, and their time course correlates well with developmental changes in neuronal morphology. Therefore, the selective and early formation of autapses in subicular PCs may play a vital role in early-life cognitive functions.
Introduction
Synapses are fundamental structures in the nervous system, connecting neurons to form neural circuits and transmit information. Typically, a synapse involves two neurons, where the presynaptic neuron's axon closely contacts the postsynaptic neuron. A unique type of synapses, known as autapses or self-synapses, occur when a single neuron forms synaptic contacts between the axon and its own cell body or dendrites. Autapses were first identified in the rabbit neocortex using Golgi staining in the 1970s (Van Der Loos and Glaser, 1972). Later other morphological studies have identified autaptic contacts in various brain regions, including the neocortex (Lübke et al., 1996), hippocampus (Cobb et al., 1997), substantia nigra (Karabelas and Purpura, 1980), and striatum (Park et al., 1980; Preston et al., 1980). Given that autapses are abundant in cultured neurons but relatively sparse in the intact brain, they were suspected of wiring errors or redundant structures (Bekkers and Stevens, 1991a; Segal, 1994, 1991). However, autapses tend to selectively form in certain types of neurons and play important roles in regulating their self-activities, indicating that they could be a fundamental building block of the brain.
In the neocortex, GABAergic inhibitory autapses are selectively found in fast-spiking (FS) interneurons but completely absent in low-threshold spiking (LTS) cells (Thomson et al., 1996; Tamás et al., 1997). In FS cells, autapses provide self-inhibition and regulate the temporal precision and the reliability of subsequent action potential (AP) generation (Bacci et al., 2003; Manseau et al., 2010; Jiang et al., 2012). While morphological studies have confirmed the presence of autapses in neocortical glutamatergic excitatory pyramidal cells (PCs; Cobb et al., 1997), their physiological properties and functions remain poorly understood. Using a two-photon axotomy method in single Layer 5 PCs to disconnect the presynaptic axon from the postsynaptic somatodendritic compartment, we have demonstrated that individual APs evoked by axonal stimulation elicit unitary autaptic currents approximately five times greater than those of regular PC→PC synapses (Yin et al., 2018). To facilitate the detection of autaptic responses, we perfused acute cortical slices with a bath solution containing Sr2+, which can desynchronize and thus prolong synaptic vesicle release (Bekkers and Clements, 1999), facilitating the identification of autaptic PCs. In these experiments, we observed barrages of individual autaptic currents long after the burst stimulation in Layer 5 PCs, particularly those projecting to subcortical regions but very few in PCs of superficial layers (Yin et al., 2018 ), indicating a selective formation of autapses in certain types of neorcortical PCs.
The hippocampal formation (HPF) is a critical structure in the brain and play crucial roles in learning, memory formation, and spatial navigation (Scoville and Milner, 1957; Morris, 2006; Fenton, 2024). It is composed of three major parts: the hippocampus proper, the dentate gyrus (DG), and the subiculum (van Strien et al., 2009). Principal/projecting neurons in the hippocampus proper and subiculum are also PCs with morphological and electrophysiological features similar to neocortical PCs. In the classic DG→CA3→CA1 trisynaptic pathway, DG granule cells receive highly integrated multimodal sensory information from the entorhinal cortex and send output to CA3 PCs via mossy fibers; CA3 PCs send axons to synapse onto CA1 PCs through Schaffer collaterals. CA2 PCs integrate synaptic inputs from the entorhinal cortex and CA3 and send outputs to CA1 neurons, linking cortical inputs to hippocampal CA1 neuronal outputs (Chevaleyre and Siegelbaum, 2010). The subicular PCs receive inputs from CA1 and convey hippocampal output signals by projecting their axons to various cortical and subcortical structures (McNaughton, 2006; Matsumoto et al., 2019). Considering the similarities of HPF PCs to neocortical PCs, we speculate that functional autapses may also occur in HPF PCs.
We performed whole-cell recordings from HPF PCs in Sr2+-containing bath solutions. Surprisingly, we found that autaptic responses are present only in a subpopulation of subicular PCs, completely absent in CA1–3 PCs, and the timing of autapse emergence in these PCs during early postnatal development agrees well with the spatial overlap of their dendrites and the axon.
Materials and Methods
Acute brain slice preparation
Experiments were carried out on mice (C57BL/6J) in accordance with protocols approved by the Animal Care and Use Committee at the Fudan University. The animals were housed under a 12 h light/12 h dark cycle with unrestricted access to food and water. Mice of either sex were randomly assigned to each experimental group. Mice from Postnatal Day (P)5 to 21 were rapidly decapitated after deep anesthesia with sodium pentobarbital (50 mg/kg, i.p.). The brains were immediately dissected out and immersed in an ice-cold aerated (95% O2 and 5% CO2) sucrose-based slicing solution, which contained the following (in mM): 213 sucrose, 2.5 KCl, 2 MgSO4, 2 CaCl2, 26 NaHCO3, 1.25 NaH2PO4, and 25 dextrose (315 mOsm), pH 7.4. The brain tissue containing the HPF or the medial prefrontal cortex (mPFC) was cut into horizontal slices with a thickness of 300 μm using a vibratome (Leica VT1200S). The slices were then collected and incubated at 34.5°C in aerated artificial cerebrospinal fluid (ACSF). The chemical composition of the ACSF was similar to the sucrose-based solution, except that the sucrose was replaced by 126 mM NaCl. After 30 min of incubation, the slices were then maintained at room temperature (∼25°C) until use.
Electrophysiological recording
Individual slice was transferred to the recording chamber and perfused with aerated normal ACSF at a temperature (35–36°C) close to physiological conditions. Cells of the slice were visualized under an upright infrared differential interference contrast microscope (BX61WI, Olympus) equipped with a 40× water-immersion objective. Patch pipettes used for somatic recordings had an impedance of 4–6 MΩ when filled with the K+-based internal solution containing the following (in mM): 140 K-gluconate, 3 KCl, 2 MgCl2, 0.2 EGTA, 10 HEPES, and 2 Na2ATP (285–295 mOsm), pH 7.2–7.3. Biocytin (0.2%) was added to the internal solution for post hoc biocytin–avidin staining.
In the Sr2+ experiments examining autaptic connections of PCs with whole-cell recording, we added 8 mM SrCl2 to the ACSF but reduced the concentration of CaCl2 and MgSO4 to 1 mM. Trains of 20 voltage pulses at 100 Hz (1 ms in pulse duration, 100 mV in amplitude) were applied every 20 s to the recorded cell through the somatic recording pipette (filled with the K+-based internal solution) to evoke action currents. The occurrence of desynchronized synaptic currents during and after the train stimulation reflects the existence of functional autapses. PCs with axons shorter than 80 μm were excluded from data analysis because their axons were cut before branching.
In another set of Sr2+ experiments examining whether autapses express NMDA receptors, we performed similar experiments as described above but in ACSF with the following modifications: (1) omit MgSO4 but add 10 μM glycine or 100 μM D-serine to ensure activation of NMDA receptors at a holding potential of −70 or −90 mV; (2) omit CaCl2 but added 8 mM SrCl2 to cause asynchronous neurotransmitter release; (3) add tetrodotoxin (TTX; 0.5 μM) to block AP generation and thus AP-dependent synaptic transmission (i.e., pharmacologically isolate the recorded PC from the neural network); (4) add picrotoxin (PTX; 50 μM) to block miniature inhibitory postsynaptic currents (mIPSCs) mediated by GABAA receptors. In addition, we employed a Cs+-based pipette solution (in mM): 138 CsMeSO3, 3 CsCl, 2 MgCl2, 2 Na2ATP, 10 HEPES, and 0.2 EGTA (285 mOsm), pH 7.2, to block K+ conductances, ensuring that the cell membrane is electrically compact and thus allowing voltage pulses reach the autapses without a profound decrease in amplitude and trigger autaptic responses. At a holding potential of −70 or −90 mV in a voltage-clamp mode, we applied voltage pulses (100–200 mV in amplitude, 10–1,000 ms in duration) to the soma and monitored the occurrence of synaptic events after each pulse that reflect asynchronous release at autapses. Properties of the evoked autaptic events were then compared with those in the presence of 50 μM APV, an NMDA receptor antagonist. For analysis, we selected individual nonoverlapping poststimulation autaptic events as well as the baseline miniature excitatory postsynaptic currents (mEPSCs) that occur before each voltage pulse. These experiments were performed at room temperature.
Voltage and current traces were low-pass filtered at 10 kHz and sampled at 25 kHz using the Spike2 software (Cambridge Electronic Design). The liquid junction potential was not corrected for Vm values shown in the main text and figures.
Two-photon axotomy together with dual soma–axon recording
For two-photon axotomy together with dual soma–axon recording experiments in PCs of the subiculum, we perfused the HPF slices with normal ACSF and obtained whole-cell recording using the K+-based internal solution. We also added Alexa Fluor 488 and 594 (100 μM) to this internal solution for somatic and axonal recording, respectively. Following successful somatic recording and sufficient diffusion of the fluorescent dye to distal neurite compartments (normally 3–5 min), axonal recordings was achieved from the axon bleb of the same PC. In the voltage-clamp mode, axonal stimulation (brief voltage pulses, ∼1 ms in pulse duration) would evoke action currents at the soma, reflecting the arrival of backpropogating APs. Two-photon laser axotomy was then performed at a location on the axon initial segment (AIS) 40–60 μm distal to the soma (corresponding to the distal end of the AIS). Two-photon axotomy disconnected the axon from the soma and thus separated the presynaptic axon compartments and the postsynaptic somatodendritic compartments. Under these conditions, backpropagating APs could no longer be detected at the somatic recording site, while autaptic connections remained intact and thus autaptic EPSCs (aEPSCs) could be detected. Two-photon imaging of the recorded cell and axotomy procedures were performed using a two-photon laser scanning microscope equipped with a 40× water-immersion objective and a laser with excitation wavelength of 920 nm. Brief two-photon laser pulses (50–200 ms each) were applied to the selected axonal site until successful axotomy was achieved, reflected by the disappearance of somatic action currents. If there was a small residue EPSC-like current, the recorded PC was then considered as an autaptic PC. Otherwise, the recorded PC was considered nonautaptic. Imaging data were acquired using a Fluoview FV 3000 (Olympus) and further analyzed with ImageJ.
In the two-photon axotomy experiments, to minimize energy accumulation and potential thermal damage, we carefully adjusted the exposure time of each laser pulse at 40 mW to prevent abrupt axonal disconnection. Such sudden disconnection would disrupt somatic/axonal whole-cell recordings, as evidenced by immediate, large increases in holding currents. The exposure time should be set such that multiple laser pulses were required to sever the axon. Following each pulse, we waited several seconds to assess axotomy success before proceeding with additional pulses when necessary. In most experiments, we observed small inward currents preceding individual EPSC-like events postaxotomy. These small currents represent axonal APs propagating through the incompletely severed axon.
Stereotaxic injections
Stereotactic injections were performed in mice aged P12–14. After deep anesthesia with isoflurane, the head of a mouse was secured in a stereotactic frame (RWD Life Science). A solution containing Retrobeads (78R170, Lumafluor) was injected into the medial shell of the nucleus accumbens (NAc) or the basolateral amygdala (BLA) at a rate of 20 nl/min using a syringe pump (World Precision Instruments). The total injection volume ranged from 50 to 100 nl. Coordinates (in mm) for injection at NAc are AP +1.5, ML ±0.75, and DV −4.1 and at BLA are AP −1.1, ML ±2.8, and DV −4.6.
Cell staining and reconstruction
After the electrophysiological recordings, the brain slices were fixed in a 4% paraformaldehyde solution for avidin staining. The slices were kept in the fixative for >12 h and then stained with Alexa Fluor 488-conjugated avidin (Invitrogen, Thermo Fisher Scientific). The z-stack images were acquired with an interval of 1–2 μm using a confocal microscope (LSM 900, Carl Zeiss) equipped with 20× and 63× objective. We used the Fiji plugin SNT toolbox (https://imagej.net/plugins/snt/index) to achieve the 3D morphology reconstruction of the labeled cells.
Data analysis
Data analysis was performed using the Prism 9, Fiji, MATLAB, and Spike2 software, with all measurements taken from distinct, individual samples. Unless otherwise stated, data are represented as mean ± SEM. The error bars in the figures are also SEM. For comparison of paired data, two-tailed paired Student's t test was used to assess the significance of differences between normally distributed groups (p > 0.05; Shapiro–Wilk test). For non-normally distributed data, the Wilcoxon matched-pairs signed–rank test was applied. For comparisons between two independent groups, unpaired two-tailed Student's t test was used for data that passed the normality test. Differences were considered statistically significant if p < 0.05 (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant).
Data and code availability
All data reported in this paper are available upon request. This study does not generate any new code.
Results
Autapses selectively occur in subicular PCs
To investigate whether the HPF PCs possesses autapses similar to those in the neocortex, we conducted whole-cell recording from PCs in different subregions of the hippocampus proper and subiculum in acute horizontal HPF slices from mice P14–21 (Fig. 1a). The slices were continuously perfused with a bath solution containing 8 mM SrCl2. The inclusion of Sr2+ would facilitate the detection of autaptic PCs because of the occurrence of desynchronized and long-lasting autaptic events immediately after a train of stimulation (Yin et al., 2018).
Subicular PCs form functional autapses. a, A representative mouse HPF slice with a glass pipette recording from a PC in the subiculum. b, Left, An example whole-cell recording in a voltage-clamp mode from a subicular PC in a slice perfused with a bath solution containing 8 mM SrCl2. Note the occurrence of barrages of putative aEPSC events (arrows) after a train of voltage pulse stimuli (20 pulses at 100 Hz). Right, An example recording from a nonautaptic PC. Note the absence of aEPSCs. c, Group data comparing the EPSC frequency pre- and poststimulation in autaptic (n = 48 PCs; 11 mice) and nonautaptic PCs (n = 45 PCs; 11 mice). An autaptic PC is identified if its post-train EPSC frequency is 1.5 times higher than the baseline frequency (pre). d, The percentage of autaptic PCs and nonautaptic PCs in the subiculum. e, An example image showing dual soma–axon recording and the site of two-photon axotomy (white arrow) in a subicular PC. Patch pipettes for the soma and the axon were filled with internal solutions containing Alexa Fluor 488 and 594 dyes, respectively. Inset, A representative trace showing spiking responses of the recorded cell to a step current injection (200 pA, 500 ms). f, A representative dual soma–axon recording in a voltage-clamp mode. Before axotomy, axonal stimulation elicited action currents at the soma (holding potential, −70 mV). After axotomy (black arrows), AP backpropagation was blocked as evidenced by the absence of somatic action currents. Middle inset, expanded action current and aEPSC for comparison. Right inset, Overlay of individual aEPSCs evoked by single stimulus at the axon (current traces from the gray box). g, Left, Decay time constant of aEPSCs in subicular PCs (n = 7 PCs; 3 mice). Right, Peak amplitude. Data were presented as mean ± SEM, Student's t test. ***p < 0.001; ns, not significant. Also see Extended Data Figure 1-1.
Figure 1-1
Autaptic responses can be induced by 20-Hz train stimulation in subicular PCs. (a) Three consecutive current traces showing autaptic responses of a subicular PC to 20-Hz stimulation (5 pulses). Black arrowheads indicate individual aEPSC events. (b) Comparison of the number of aEPSC events with 20-Hz and 100-Hz stimulation in the same cell (n = 11 PCs, 6 mice). Data were presented as mean ± SEM, Student’s t-test. *** P < 0.001. Download Figure 1-1, TIF file.
In a voltage-clamp mode, trains of voltage pulses (20 pulses at 100 Hz) were delivered to the recorded cell with an intertrain interval of 15 s to successfully evoke individual action currents that reflect the generation APs (Fig. 1b). In a subpopulation of subicular PCs, we detected barrages of EPSCs immediately after the train stimulation and lasted for hundreds of milliseconds, while such responses were not recorded in other PCs (Fig. 1b). According to previous characterization in neocortical slices (Yin et al., 2018), these EPSCs should be mediated by the release of glutamate at autapses. We then compared the frequencies of EPSCs before and after the train stimulation and found a substantial increase in EPSC frequency in a subpopulation of subicular PCs. If cells with post-train EPSC frequency 50% higher than the baseline are considered autaptic PCs, 51.61% of subicular PCs (n = 48/93 PCs; 11 mice) formed autapses (Fig. 1c,d). In a subset of experiments, we stimulated subicular PCs with a lower intensity (five pulses at 20 Hz), a protocol similar to that described previously (Yin et al., 2018), and also detected an increase in EPSC events after the train stimulation, but this increase was less than that evoked by the high stimulation intensity (20 pulses at 100 Hz; Extended Data Fig. 1-1). To ensure the detection of autaptic PCs, we utilized the high-intensity stimulation protocol in the following experiments.
Meanwhile, considering that the detected synaptic events after the train stimulation in Sr2+-based ACSF might differ mechanistically from synchronous monosynaptic events (i.e., AP-triggered monosynaptic EPSCs) in normal ACSF, we performed dual soma–axon whole-cell recording experiments together with two-photon axotomy as described previously (Yin et al., 2018) to isolate the autaptic events in single subicular PCs (Fig. 1e). We observed action currents at the somatic recording site in response to individual axonal stimulus before axotomy. Following axotomy, the somatic action currents disappeared, reflecting the failure of AP backpropogation to the soma. With close examination of the residue somatic currents, we observed autaptic EPSCs (aEPSCs) in a subpopulation of subicular PCs (Fig. 1f), similar to our previous findings in mPFC PCs (Yin et al., 2018). The small inward currents preceding individual aEPSCs most likely represent axonal APs propagating through incompletely severed axons (see Materials and Methods). The evoked aEPSCs showed an average peak amplitude of −132.03 ± 16.05 pA and a decay time constant of 5.48 ± 0.45 ms (n = 7 subicular PCs; three mice; Fig. 1g). These results indicate that subicular PCs form functional autapses, further confirming our findings in Sr2+-based ACSF.
Next, through recording in Sr2+-based ACSF, we examined whether PCs in CA1–3 regions also form autapses (Fig. 2a–c). Surprisingly, none of the recorded PCs showed an increase in EPSC frequency after the train stimulation (Fig. 2d–f), indicating that PCs in these hippocampal subregions do not form functional autapses. Together, these results demonstrate that autapses are selectively formed in PCs of the subiculum, but not those in hippocampus proper subregions.
PCs in CA1–3 subregions do not form autapses. a–c, Example consecutive current traces obtained from PCs in hippocampal CA1–3 subregions. Note that there is no apparent increase in EPSC frequency after the train stimulation. d–f, Group data comparing the EPSC frequency pre- and poststimulation in PCs of CA1 (n = 21 PCs; 5 mice), CA2 (n = 17; 4 mice), and CA3 (n = 17; 5 mice). Data were presented as mean ± SEM, Student's t test. *p < 0.05; ns, not significant.
Subicular PCs show higher axonal arborization near the soma
Although PCs in different HPF subregions share similar morphological features such as the pyramid-shaped cell body, the thick apical dendrites, and the thin basal dendrites, the selective formation of autapses in subicular PCs could be attributable to their unique morphological properties. We therefore examined the morphology of HPF PCs and performed sholl analysis of the axons and dendrites (Fig. 3). The recorded cells were filled with biocytin during the whole-cell recording and subjected to post hoc avidin staining.
Subicular PCs possess heavily branched axons and basal dendrites near the soma. a, An example mouse HPF slice showing recorded subicular PCs with post hoc avidin staining (green). Red, NeuN staining. b, Left, An example PC with avidin staining in the subiculum. The arrowhead indicates the axon, while the white arrows indicate the putative autaptic contacts. Right, Single-plane images of the indicated putative contacts. c, Top, Measurement of the fold change in the number of individual EPSCs within a period of 0.7 s before and after the train of voltage pulse stimuli (20 pulses at 100 Hz). Bottom, Plot of the fold change versus the number of putative autaptic contacts. The data were fitted by linear regression. d, Example PCs from hippocampal CA1–3 subregions. White arrows indicate the axons. e, Group data showing the locations of putative autaptic contacts with axonal (left) or dendritic (right) paths relative to the soma (n = 66 contacts from 15 subicular PCs). f, Distribution of dendritic and axonal path lengths of putative autaptic contacts. g, Left, Sholl analysis of the axon and its collaterals (SUB n = 10 PCs; CA1 n = 5; CA2 n = 6; CA3 n = 5). Right, Sholl analysis of the basal dendrites. h, Comparison of axonal branch numbers between subicular PCs and CA1–3 PCs. Data were presented as mean ± SEM, Student’s t test. **p < 0.01.
Under the fluorescence microscope, the thin axon and its collaterals could be easily distinguished from the relatively thicker and highly spiny dendrites. As shown in Figure 3, a and b, the subicular PC's axon branches extensively overlap with its basal dendrites and form putative autaptic contacts. In contrast, axons of PCs in CA1–3 show much less branching and arborization in their basal dendrite fields, leading to less overlap between the axon and the dendrite (Fig. 3d). In subicular PCs, further analysis of the number of putative autaptic contacts and the strength of autaptic responses (i.e., fold change in the number of EPSC events after the train stimulation) revealed a weak positive correlation (Fig. 3c). The average number of putative autaptic contacts was 4.94 ± 3.35 per cell (n = 18 PCs; nine mice), and the average axonal and dendritic path lengths to the soma were 261.99 ± 14.02 µm and 90.79 ± 4.2 µm, respectively (n = 15 PCs; eight mice; Fig. 3e,f).
Sholl analysis reveals that the complexity of both axon and dendrite in subicular PCs is significantly higher than those of other PCs. The number of axonal intersections near the soma (within 100 μm) in subicular PCs is the highest among HPF PCs, and the number of intersections with the basal dendrites is comparable in subicular PCs and CA1 PCs, with slightly more intersections in subicular PCs (Fig. 3g). Further analysis reveals that subicular PCs possess more axonal branches than those in CA regions (Fig. 3h). Together, these results suggest that the extensive overlap between the axonal arborization and the basal dendritic field would substantially increase the possibility of forming autaptic contacts in subicular PCs.
Autapses of subicular PCs emerge earlier than mPFC PCs
Synapse overproduction occurs in early developmental stages and is followed by progressive synapse elimination (also known as synapse pruning) as the brain matures (Yu, 2023). Indeed, we have previously demonstrated that Layer 5 PCs in mPFC form autapses in both juvenile and adult mice, with a higher proportion of autaptic PCs in juvenile mice and lower in adult mice, most likely reflecting a pruning process of autapses in mPFC (Yin et al., 2018). Considering that the neocortex develops earlier than the HPF (Altman and Bayer, 1990; Molyneaux et al., 2007), we speculate that autapses in subicular PCs may occur later than those of neocortical PCs. Therefore, we investigated the development of PC autapses in both the mPFC and the subiculum.
In our experiments, we examined the percentage of autaptic PCs in juvenile mice (from P5 to 20) and adult mice (P60 to 70; Fig. 4a–c). At P5–7, 58.57 ± 4.46% of subicular PCs showed post-train autaptic currents in response to trains of voltage pulses (20 pulses at 100 Hz per train) with amplitude sufficient to evoke action current for each pulse. However, to our surprise, none of the recorded mPFC PCs during this period produced autaptic currents (Fig. 4a,c). These results indicate that autapse formation occurs earlier in the subiculum than the mPFC. Further experiments revealed that the percentage of autaptic PCs in the subiculum decreased to ∼40% during P8–20 and further declined to ∼23% in young adult mice (Fig. 4c). In contrast, the percentage of autaptic PCs in mPFC showed a rapid increase from P5 to 20. In agreement with our previous findings (Yin et al., 2018), the proportion of autaptic PCs decreased to ∼30% in young adults (Fig. 4c). These results reveal a time window of autapse overproduction from P5 to 20 and the occurrence of autapse pruning in Layer 5 PCs of mPFC as the cells mature. In the subiculum, however, we only detected the pruning process; the overproduction period should occur earlier than P5.
Autapses of subicular PCs emerge earlier than those in mPFC during early development. a, Example current traces from two PCs at P7. Note the presence of poststimulation EPSC events in the subicular PC but not in the mPFC PC. b, Similar to a but for PCs at P10. Note the presence of poststimulation synaptic events in both cells. c, Plots of the percentage of autaptic PCs in mPFC and subiculum at different postnatal time windows and in young adults (P60–70; subiculum, P14–20 n = 5 mice; otherwise n = 3 mice; mPFC, P5–7 and P14–20 n = 5 mice; otherwise n = 3 mice). d, Example avidin-staining images of mPFC PCs within the four postnatal time windows and in young adults shown in c. The specific age of the animals is indicated in the images. e, Example subicular PCs. f, Sholl analysis of axons and basal dendrites of P5–8 subicular (n = 6; 2 mice) and mPFC PCs (n = 6; 2 mice). g, Sholl analysis of axons and basal dendrites of P11–13 subicular (n = 6; 2 mice) and mPFC PCs (n = 5; 2 mice). Data were presented as mean ± SEM. Also see Extended Data Figure 4-1.
Figure 4-1
Axon conduction fidelity and developmental changes in somatic AP waveform. (a-b) Representative traces showing simultaneous soma-axon recording from a mPFC PC (a, P14 mice) and a subicular PC (b, P14 mice) with a train of 100-Hz somatic stimulation. Note that in both recorded cells, axons could follow high-frequency somatic stimulations. (c-d) Group data showing the axonal AP probability at different somatic stimulation frequencies in P14-20 mPFC (n = 7 PCs, 2 mice) and subicular PCs (n = 6 PCs, 2 mice). (e) Overlay of somatic AP waveforms from mPFC PCs at P7, 10, and 20. (f) Group data comparing the resting membrane potential (RMP) and AP parameters of mPFC PCs at P7 (n = 21 PCs, 3 mice), P10 (n = 12 PCs, 2 mice), and P20 (n = 22 PCs, 4 mice). Data were presented as mean ± SEM, one-way ANOVA. ** P < 0.01; **** P < 0.0001. Download Figure 4-1, TIF file.
To examine whether differences in the fidelity of AP conduction along the axon affects the detection of autaptic PCs, we performed dual soma–axon recordings from both mPFC and subicular cells in P14–20 mice where the axon blebs were accessible for whole-cell recording. We found that in both types of PCs, APs evoked by somatic stimulation at frequencies from 10 to 100 Hz could reliably propagate to the axonal recording site (Extended Data Fig. 4-1). These observations suggest that the difference in the percentage of autaptic PCs between mPFC and subiculum in P14–20 is not attributable to differences in AP conduction fidelity. Considering that broadening of AP waveform may cause more Ca2+ influx and thus more neurotransmitter release, we therefore analyzed the waveform of somatic APs of mPFC PCs in P7, 10, and 20. We found no significant change in the AP amplitudes but a progressive decrease in AP half-width. These results suggest that the developmental increase in the proportion of autaptic PCs in mPFC may not result from AP waveform changes (Extended Data Fig. 4-1).
In agreement with the electrophysiological results showing developmental increase in the proportion of autaptic PCs in mPFC, the axons and dendrites underwent rapid growth, as reflected by increases in their complexity. As shown in Figure 4d, the basal dendrites of mPFC PCs become longer during the examined 3 postnatal weeks. In sharp contrast, the complexity of the axons and basal dendrites of subicular PCs in the first and the second postnatal week (Fig. 4e) was much higher than that of mPFC PCs. There was no obvious difference in the axonal and dendritic morphology between the third postnatal week and adult PCs in both mPFC and subiculum (Fig. 4d,e). Sholl analysis reveals increases in the number of both axon and dendrite intersections in mPFC PCs from P5–8 (Fig. 4f) to P11–13 (Fig. 4g), while those in subicular PCs decrease substantially.
Together, these results indicate that autapses in both mPFC and subiculum show developmental changes, with overproduction in mPFC PCs but elimination in subicular PCs during the 3 postnatal weeks. The changes in the percentage of autaptic PCs correlate well with developmental alterations in dendritic complexity and axon arborization.
Subicular PC autapses express AMPA receptors but not NMDA receptors
In general, the postsynaptic membrane of glutamatergic synapses expresses both AMPA and NMDA receptors, which work in concert to ensure rapid excitatory synaptic transmission. However, in mPFC PCs, we found that autaptic responses are solely mediated by AMPA receptors (Yin et al., 2018), suggesting no expression of functional NMDA receptors. It is of interest to know whether this is also true in autapses of subicular PCs.
In the following experiments, we removed Ca2+ and Mg2+ from the Sr2+-ACSF but added 10 μM glycine to enhance NMDA currents at negative membrane potential levels. Additionally, we blocked voltage-gated Na+ currents and GABAA receptors by adding 0.5 μM TTX and 50 μM PTX to the bath solution. We used a Cs+-based pipette solution to ensure the recorded cell electrically compacted (Fig. 5a). Under these experimental conditions, although there is no AP generation in the presence of TTX, we speculate that the huge voltage pulses (100–200 mV in amplitude, 10–1,000 ms in pulse duration) would reach the presynaptic compartment of autapses and trigger neurotransmitter release. Indeed, we observed barrages of autaptic EPSCs after individual voltage pulses, and these autaptic events could be completely abolished by the bath application of AMPA receptor blocker NBQX (10 μM; Fig. 5b,c). These results suggest that, similar to PCs in the neocortex, subicular PCs primarily express AMPA receptors rather than NMDA receptors.
Autaptic responses of subicular PCs are mediated solely by AMPA receptors. a, Schematic diagram of recording from a subicular PC in a modified bath solution with 0 Ca2+, 0 Mg2+, 8 mM Sr2+, and a cocktail of drugs that enhance NMDA responses (10 μM glycine), block AP generation (0.5 μM TTX), and abolish GABA-mediated mIPSCs (50 μM PTX). b, Example recording showing the asynchronous autaptic currents elicited by somatic voltage pulse stimulation, which can be largely diminished by 10 μM NBQX. Inset, A segment of current trace showing individual autaptic events. c, Group data comparing the standard deviation of current traces pre and poststimulation (left), before and after the application of NBQX (right; n = 5 PCs; 2 mice). d, Left, Average traces of individual autaptic EPSCs before (orange) and after (black) the application of APV (50 μM). Right, Group data comparing the decay time constants of autaptic EPSCs (n = 6; 3 mice). e, Left, Average traces of baseline mEPSCs (events occurred before the voltage pulse stimulation) and after the application of APV. Right, Group data comparing the decay time constants of mEPSCs before and after APV application (n = 6; 3 mice). f, P0–1 subicular PCs. Left, Average traces of baseline mEPSCs in control and in the presence of APV. Right, Group data comparing the decay time constants of mEPSCs before and after APV application (n = 6; 2 mice). g, Similar to d but in P10 mPFC PCs. Left, Average traces of individual autaptic EPSCs before and after the application of APV. Right, Group data comparing the decay time constants of autaptic EPSCs (n = 6; 2 mice). h, Similar to f, but in P7–9 mPFC PCs (n = 15; 3 mice). Data were presented as mean ± SEM, Student’s t test. **p < 0.01; ***p < 0.001; ns, not significant.
To further corroborate this finding, we examined the decay time of autaptic EPSCs before and after bath application of the NMDA receptor antagonist APV (50 μM). The baseline miniature EPSCs (mEPSCs) events before the train stimulation should mostly originate from regular recurrent glutamatergic synapses for their numerical superiority compared with autapses, while poststimulation EPSC events largely attributed to the activation of autapses (Yin et al., 2018). Bath application of APV significantly reduced the decay time of the baseline mEPSCs but not that of the evoked autaptic EPSCs (Fig. 5d,e), suggesting that autaptic EPSCs do not contain an NMDA receptor component, consistent with the complete blockade of autaptic EPSCs by NBQX. In addition, we examined mEPSCs with APV treatment in subicular and mPFC PCs at even earlier developmental stages. In subicular PCs at P0–1, APV significantly reduced the decay time of mEPSCs (n = 6 PCs; two mice; Fig. 5f). A similar effect was observed in mPFC PCs at P7–9 (n = 15 PCs; three mice; Fig. 5h). To investigate whether autapses express NMDA receptors at P10 when they begin to emerge in mPFC PCs, we measured the decay time of aEPSCs before and after the application of APV. We also found no significant change in the aEPSC decay time (Fig. 5g), suggesting that, although NMDA receptors are abundant in recurrent excitatory synapses (Fig. 5h), they are not expressed by autapses, even at the onset of autapse formation.
Together, these results suggest that subicular PC autapses do not express functional NMDA receptors and the autaptic transmission is solely mediated by AMPA receptors.
Autapses preferentially form in certain types of subicular PCs
We demonstrated previously that autapses form selectively in a subpopulation of neocortical Layer 5 PCs projecting to subcortical brain regions such as the habenula (Yin et al., 2018). Therefore, we next investigated whether autapse formation in subicular PCs also shows target brain region specificity.
As the primary output of the HPF, axons of subicular PCs mainly project to the NAc and amygdala (Rosene and Van Hoesen, 1977; Friedman et al., 2002; Witter, 2006; Agster and Burwell, 2013). The subiculum–NAc circuit is involved in the regulation of reward and motivational behaviors, whereas the subiculum–amygdala circuit regulates emotion and stress (Fanselow and Dong, 2010; Boxer et al., 2023). We injected retrobeads into the NAc or the BLA of P12–14 C57 mice to selectively label subicular PCs projecting to the two distinct brain regions (Fig. 6a–c). Two days after the injection, subicular PCs containing retrobeads could be clearly identified under the fluorescence microscope. With whole-cell recording, we found surprisingly that subicular PCs projecting to the NAc (46.6%; n = 88 PCs; six mice) showed a much higher probability of forming autapses, as compared with those targeting the BLA (17.2%; n = 64; seven mice; Fig. 6d–f). As shown in the spatial map with the distribution of bead-labeled subicular PCs along both the proximal–distal and superficial–deep axes (Fig. 6g), PCs project to NAc and BLA located at the proximal region of the subiculum, a pattern consistent with previous findings (Kim and Spruston, 2012). However, in this spatial map, no obvious clustering of autaptic PCs projecting to either NAc or BLA was observed (n = 43 PCs; three mice projecting to NAc; n = 40 PCs; three mice projecting to BLA). These results suggest that autapses in the subiculum also show target brain region specificity, with NAc-targeting PCs showing a higher probability of forming autapses. These findings also suggest that autaptic PCs in the subiculum may contribute largely to reward-associated behaviors.
Autapse formation in subicular PCs is axon-projection specific. a, A schematic showing retrobeads injection into the shell of the NAc. b, A representative bright-field image showing the retrobeads injection site in the NAcSh. Horizontal section at the bregma −3.76 mm. c, A representative bright-field image showing the retrobeads injection site in the BLA. Horizontal section at the bregma −4.72 mm. d, Example recordings from PCs that target NAc (left) or BLA (right). e, Left, A representative image showing subicular cells labeled by retrobeads (injected to NAcSh). Blue, DAPI staining. Red, Retrobeads. Green, Avidin-stained PCs. Right, Images from the box region shown on the left. f, The bar graph comparing the percentage of PCs with autaptic responses in NAc-targeting PCs (n = 88 PCs; 6 mice) or BLA-targeting PCs (n = 64; 7 mice). g, A spatial map of recorded PCs with or without autaptic responses in the subiculum (n = 43 NAc-projecting PCs; 3 mice; n = 40 BLA-projecting PCs; 3 mice). Filled dots represent autaptic PCs. Hollow dots represent nonautaptic PCs. The X-axis represents the proximal–distal axis of the subiculum, and the Y-axis indicates the depth along the radial (superficial–deep) axis. The histograms show the percentage of autaptic PCs along the proximal–distal axis and superficial–deep axis.
Discussion
In this study, we reveal the formation of functional autapses in a subpopulation of PCs of the subiculum, but not in those of hippocampal proper including CA1–3 PCs. In agreement with the physiological evidence, subicular PCs show significantly greater axon complexity with much more axonal arborization within the field of their basal dendrites, as compared with PCs located at other subregions of the HPF. To our surprise, autaptic responses in subicular PCs show earlier emergence during postnatal development than those in mPFC. Moreover, autaptic responses are exclusively mediated by the activation of AMPA receptors but not NMDA receptors.
Previous studies carried out comprehensive investigations on morphological features and physiological functions of autapses in FS interneurons (Thomson et al., 1996; Tamás et al., 1997; Bacci et al., 2003; Manseau et al., 2010; Jiang et al., 2012). Because of their thin AP waveforms, autaptic currents could be easily identified as they immediately appear after each action current in the voltage-clamp recording mode (Bacci et al., 2003). In contrast, although neocortical PCs also form characteristic ultrastructure of autaptic contact under electron microscope (Lübke et al., 1996; Tamás et al., 1997), it had been unclear whether PC autapses are physiologically functional. In our previous studies (Yin et al., 2018; Li et al., 2021), we employed two methods to identify autaptic PCs in the voltage-clamp mode in neocortical slices: (1) after achieving dual soma–axon whole-cell recording from a single PC, apply two-photon axotomy near the AIS, and then stimulate the axon to evoke autaptic responses; (2) with recording alone at the soma with or without 8 mM Sr2+ in the bath, apply a train of stimulation at the soma to trigger autaptic asynchronous release that outlasts the stimulation. Using these methods, we found that both mouse and human neocortical PCs form functional autapses, which provide feedback self-excitation and thus facilitate burst firing, as well as promote coincidence detection of self-activity and synaptic inputs. In this study, we employed the second method by adding Sr2+ to the bath to examine the formation of autapses in PCs of the HPF but also the first method with two-photon axotomy to confirm the AP-induced monosynaptic autaptic responses in single subicular PCs.
It is well known that autapses are abundant in cultured hippocampal PCs (Segal, 1991, 1994; Bekkers and Stevens, 1991b), most likely those from the hippocampus proper. However, to our surprise, in the intact brain tissue, autapses selectively form in a subpopulation of PCs in the subiculum but are completely absent in PCs of other HPF subregions including CA1–3. Sholl analysis revealed more intersections of both axon and dendrite near the soma of subicular PCs, suggesting a higher probability of autaptic contact in the basal dendritic field (Fig. 3). In their corresponding subregions of CA1–3 PCs, axons show few axonal collaterals and normally converge into axon bundles ∼150 µm from the soma, with minimal overlap with their basal dendrites, leading to the lack of autaptic contacts. Similar to these results, LTS Martinotti cells do not form autapses because their axons show few arborization near the somatodendritic region, while most of the FS interneurons branch heavily in their somatodendritic field and form massive autaptic contacts (Tamás et al., 1997; Bacci et al., 2003; Jiang et al., 2012). Axo-axonic cells, which are also FS interneurons, however, are an exception, showing no autaptic contact (Cobb et al., 1997). These findings suggest that axon–dendrite crossing is necessary but not sufficient for autapse formation.
In general, synapses will undergo overproduction during development followed by elimination as the brain matures (Yu, 2023). Indeed, our results reveal a rapid increase in the proportion of autaptic PCs in mPFC within the initial 3 postnatal weeks (from P5 to 20) but a decrease as the cells mature (Fig. 4c; Yin et al., 2018). During early development, HPF PCs start being produced as early as Embryonic Day (E)10–15 and their synaptogenesis peaks during the first 2 postnatal weeks. As compared with hippocampal PCs, neocortical PCs begin neurogenesis (∼E12–18) and synaptogenesis slightly later and take a longer time to mature (Finlay and Darlington, 1995). In line with the developmental timelines, autapses of subicular PCs occur at least 1 week earlier than those in mPFC and show pruning during the 3 postnatal weeks and continuing into adulthood. The early formation of autaptic contacts and their pruning during the initial postnatal weeks suggest a vital role of these special synapses in early-life hippocampal functions such as learning and memory formation.
Our pharmacological experiments suggest that subiculum PC autapses express functional AMPA receptors but not NMDA receptors, similar to neocortical PC autapses (Yin et al., 2018). Considering the well matched timing of the arrival of pre- and postsynaptic APs at the autapses, we speculate an induction of spike-timing–dependent plasticity (STDP; Bi and Poo, 1998; Sjöström et al., 2008), a potential mechanism for the “giant” autaptic responses in neocortical PCs (Yin et al., 2018). Because functional NMDA receptors required for the induction of STDP are absent in PC autapses, it is possible that NMDA receptors might be expressed at new-born autapses and eliminated when the autaptic strength becomes strong and stable due to STDP (Durand and Konnerth, 1996; Pickard et al., 2000). However, our experiments reveal the lack of NMDA component at new-born mPFC autapses. Thus, the underlying mechanism for the induction of “giant” autaptic responses needs to be further examined.
Excitatory autapses can provide feedback self-excitation and promote burst firing in PCs, likely enhancing functional connectivity between brain regions (Ke et al., 2019). As compared with amygdala-targeting subicular PCs, a much higher proportion of NAc-targeting PCs show autaptic responses, implying that autapses may facilitate burst firing in these projecting cells and promote functional connectivity between the subiculum and NAc. Indeed, as the main output of the HPF, the subicular PCs send axons to innervate various cortical and subcortical areas and function as an interregional communication hub, integrating information from the hippocampus proper and enhancing functional connectivity via various network oscillations crucial for cognitive functions such as learning and memory (Matsumoto et al., 2019). Previous studies have shown that the subiculum–NAc circuit is associated with reward and motivation, while the subiculum–amygdala circuit regulates emotion and stress (Fanselow and Dong, 2010; Boxer et al., 2023). The bias of autapse formation in NAc-targeting PCs suggests a key role of autaptic PCs in regulating reward-related behaviors, which deserves further investigation.
Unlike spike bursts induced by the activation of ion channels, autapse-mediated bursts (or spike doublets) could be more susceptible to modulation because synaptic transmission requires the involvement of numerous key molecules and operates over longer timescales. A subgroup of subicular PCs also show bursting in response to step current injections, resulting from the activation of ion channels such as Ca2+ channels or the blockade of K+ channels (Stewart and Wong, 1993; Staff et al., 2000). In addition to these intrinsic conductances, autapse-mediated positive feedback may also contribute to the generation of spiking bursts.
Although the difference in axon arborization in the somatodendritic field can explain the presence and absence of autapse formation in PCs, the identity of molecules crucial for autapse formation in a subpopulation of subicular or neocortical PCs remains unclear. We speculate that certain types of cell adhesion molecules may contribute to the formation of autapses, while repulsive guidance molecules may prevent autaptic contacts. Using single-cell RNA sequencing to compare autaptic and nonautaptic neurons could uncover the mechanisms driving selective autapse formation in specific PC subpopulations. This approach may also identify molecular targets for genetic manipulation, enabling further investigation of physiological roles of autapses—from cellular and network dynamics to behavioral functions.
Footnotes
This study was supported by the National Natural Science Foundation of China (NSFC) Grants (32130044 and T2241002, Y.S.), STI2030-Major Projects (2021ZD0202500, Y.S.), Program of Shanghai Academic/Technology Research Leader (21XD1400100, Y.S.), and two other NSFC Grants (32200953 for W.K. and 32100930 for Q.H.).
The authors declare no competing financial interests.
- Correspondence should be addressed to Yousheng Shu at yousheng{at}fudan.edu.cn or Wei Ke at weike{at}fudan.edu.cn.
