Abstract
Neurons typically generate action potentials at their axon initial segment based on the integration of synaptic inputs. In many neurons, the axon extends from the soma, equally weighting dendritic inputs. A notable exception is found in a subset of hippocampal pyramidal cells where the axon emerges from a basal dendrite. This structure allows these axon-carrying dendrites (AcDs) a privileged input route. We found that in male mice, such cells in the CA1 region receive stronger excitatory input from the contralateral CA3, compared with those with somatic axon origins. This is supported by a higher count of putative synapses from contralateral CA3 on the AcD. These findings, combined with prior observations of their distinct role in sharp-wave ripple firing, suggest a key role of this neuron subset in coordinating bi-hemispheric hippocampal activity during memory-centric oscillations.
Significance Statement
Neurons fall into multiple subtypes, many with distinct morphologies which likely correlate with their distinct functions. Hippocampal CA1 pyramidal cells are mostly treated as a homogeneous group that forms transient functional ensembles related to memory formation and consolidation. A large fraction of these neurons has an unusual axon origin from a basal dendrite. This feature facilitates participation during network states with strong perisomatic inhibition around the soma, a property indicative of a prominent role in memory-forming networks. We now report that pyramidal cells with axon-carrying dendrites receive particularly strong input from the contralateral hemisphere. This distinct connectivity links neuronal morphology with network connectivity and points toward the existence of a specialized subpopulation of hippocampal pyramidal cells for interhemispheric communication.
Introduction
Principal neurons in the mammalian cortex receive tens of thousands of synaptic inputs at their dendrites. These inputs propagate toward the soma and axon. The positioning and relative timing of these active synapses determine how excitatory and inhibitory inputs integrate at the axon initial segment (AIS), the site of decision-making for action potential generation (Kole et al., 2008; Bender and Trussell, 2012). Notably, in a significant portion of hippocampal pyramidal cells (over 50% in central CA1), the axon originates from a basal dendrite rather than the soma (Yakovlev, 1967; Thome et al., 2014). This unique morphology gives the axon-carrying dendrite (AcD) a privileged status: it can induce action potentials without signals first passing fully through the soma. The advantage of this mechanism is especially evident during perisomatic inhibition where synaptic inputs are strongly shunted at the soma. Consequently, inputs at AcD branches have a higher propensity to initiate action potentials compared with standard dendritic branches (Thome et al., 2014). Recent studies support this mechanism and have shown that AcD cells fire preferentially during hippocampal sharp-wave ripple oscillations, a network state marked by pronounced perisomatic inhibition (Hodapp et al., 2022). Thus, axon positioning correlates with the state-dependent participation of pyramidal cells in active neural assemblies. However, to understand its significance at a network level, we must explore the origin and strength of inputs to the AcD.
Synaptic inputs from multiple brain regions target different layers and subcellular compartments of CA1 pyramidal cells (Witter et al., 2000). The primary excitatory input to CA1 arises from the Schaffer collaterals (axon bundles from both ipsilateral and contralateral CA3 regions, constituting 45% of all presynaptic cells) and neurons in layer III of both lateral and medial entorhinal cortex (12%; Tao et al., 2021). Other sources, such as the amygdala, medial septum, and thalamic nuclei, contribute minimally to the total input (Nakashiba et al., 2008; Tao et al., 2021). Entorhinal cortex inputs primarily target the distal dendritic tufts in stratum lacunosum–moleculare (Doller and Weight, 1982; Witter et al., 2000; van Groen et al., 2003; Suh et al., 2011), which renders AcD branches unlikely recipients, since they are mostly formed by basal dendrites in stratum oriens (98%). CA3 axons, in contrast, traverse stratum radiatum and stratum oriens, where they predominantly innervate the dendritic tree of CA1 pyramidal neurons. Within this scheme, ipsilateral and contralateral CA3 fibers have different preferences: while axons from ipsilateral CA3 mostly target dendrites in stratum radiatum, contralateral CA3 innervate basal dendrites in stratum oriens (Shinohara et al., 2012).
While the coupling of pyramidal cell firing to the ripple frequency was found to depend on phasic inhibition provided by local basket cells (Gan et al., 2017), it remains unclear which presynaptic partners provide the excitatory drive to trigger their activity (Maier et al., 2011). We recently demonstrated that AcD cells have a higher propensity to fire during sharp-wave ripple oscillations compared with cells with canonical, somatic axon origins (Hodapp et al., 2022). Thus, identification of the presynaptic partners of AcD cells in CA1 may provide important insights into their functional role during network oscillations. Therefore, we investigated the strength of synaptic inputs arriving at AcD and nonAcD cells from ipsilateral and contralateral CA3, the two main input regions to the basal dendrites of CA1.
Materials and Methods
Animal handling and slice preparation
Experiments were approved by the animal welfare officer of the Interfaculty Biomedical Facility (IBF) at Heidelberg University. Animal maintenance and experimental procedures were performed in strict accordance with the guidelines of the European Community Council and the state government of Baden-Wurttemberg (animal protocol 35-9185.81/G-206/20). We used male individuals of wild-type C57BL/6 (JAX) mice from Charles River. Animals were anesthetized by CO2 and decapitated after the loss of the righting reflex. The brain was quickly removed and transferred into ice-cold (1–2°C) artificial cerebrospinal fluid (ACSF) containing the following (in mM): 124 NaCl, 3 KCl, 1.8 MgSO4, 1.6 CaCl2, 10 glucose, 1.25 NaH2PO4, saturated with 95% O2/5% CO2 and calibrated to pH 7.4. The cerebellum and the first third of the frontal brain were removed, and 350 µm horizontal slices were cut using a vibrating blade microtome (Leica VT1000S, Leica Biosystems). We used the medial portion of the hippocampus just before it curves into the dorsal compartment. The section is characterized by an oval appearance of the hippocampus and a “V”-shaped dentate gyrus. Brain hemispheres were separated and incubated in a submerged storage chamber for 30 min in ACSF at 32°C. Subsequently, the storage chamber cooled down to room temperature. For experiments, individual slices were moved into a submerged type recording chamber, restrained by a platinum-weighted harp, and constantly perfused with ACSF.
Intracranial injections of viral vectors
Injections were performed at postnatal day 28 (P28). Animals received 0.1 mg/kg buprenorphine hydrochloride (Temgesic; Indivior) for analgesia subcutaneously 30 min before surgery. The depth of the anesthesia was observed and adjusted according to breathing rhythm and toe pinch response. Anesthesia was maintained by mask inhalation of vaporized isoflurane at concentrations between 1.5 and 2.5%. Mice were kept warm on a 37°C heating pad while being fixated in the stereotax (Kopf Instruments). Body temperature was continuously measured on the abdominal skin. The fur on the head was partly removed; the skin was disinfected and incised along the sagittal midline for ∼1 cm. The skull surface was exposed and dried with cotton sticks. Using the anatomical landmarks bregma and lambda, the skull was leveled in the lateral–medial and the anterior–posterior axes. For unilateral injection into ventral CA3, we used the following coordinates referenced to bregma: A–P, −2.4; L–M ± 2.6; and D–Vl, −2.9, −2.7, and −2.5. At each location in the dorsoventral axis, we pressure-applied 100 nl viral solutions through a Hamilton syringe at 200 nl/min. After injection, the skull surface was moisturized with 0.9% NaCl, and the skin was stitched. Animals recovered from surgery under infrared light for 1 h. Buprenorphine analgesia (0.1 mg/kg) was administered 3 h post-surgery. During subsequent incubation time of 3–4 weeks, mice were checked daily. Correct expression of the virus in CA3 was verified post hoc by confocal imaging of fixed acute slices. ChR2 expression was present in CA3 and the lower blade of dentate gyrus, but never reached the border of CA2. The position of CA2 was verified by immunofluorescence staining against the CA2 pyramidal cell marker PCP4 in some slices. It colocalized with the area ± ∼100 µm from the end of stratum lucidum. Granule cells of the dentate gyrus project exclusively to CA3 and should thus not contaminate the specificity of our local light stimulation (Amaral et al., 2007).
Patch-clamp recordings
Patch pipettes consisted of a hollow glass electrode containing an AgCl-coated silver wire and filled with an artificial intracellular solution containing the following (in mM): voltage clamp, 144 Cs-gluconate, 10 HEPES, 4 MgATP, 0.3 NaGTP, 10 phosphocreatine, equilibrated to pH 7.3 with CsOH, and current clamp, 126 K-gluconate, 10 HEPES, 0.3 EGTA (ethylene-glycol-bis-tetraacetic acid), 4 KCl, 0.3 GTP (guanosine-5-triphosphate), 10 phosphocreatine-disodium-salt hydrate, 4 ATP (adenosine-5-triphosphate), pH adjusted at pH 7.3 with KOH (Technical White Paper: Electrophysiology, October 2017 v.5; URL, https://help.brain-map.org/download/attachments/8323525/CellTypes_Ephys_Overview.pdf). Cells were visualized with a 60× water immersion objective using infrared oblique illumination optics. Recordings were performed with an NPI-ELC-03XS amplifier. Data were filtered at 10 kHz and sampled at 50 kHz with a CED 1401 mk2 (Cambridge Electronic Design). Patch pipettes (borosilicate glass; outer diameter 2 mm, inner diameter 1.16 mm; Science Products) were pulled on a DMZ Zeitz puller. Tip diameter was assessed electronically and ranged from 2.8 to 4.5 MΩ in the bath. Series resistances were compensated using the bridge circuity of the amplifier. The liquid-junction potential created by solutions of different ion compositions was not corrected for, but calculated to be +14.5 mV. In current clamp, membrane potentials were adjusted by constant current application to values around −70 mV. To ensure the quality of recordings, only cells with stable access resistances below 22 Mohm and <10% variation were used for analysis. Pulses of 5 mV and 20 ms were used to calculate series resistance. In one set of experiments, action potential propagation was blocked by bath application of the voltage-gated sodium channel blocker tetrodotoxin (TTX, 1 µM) and the potassium channel blocker 4-aminopyridine (4-AP, 100 µM). Slices were incubated for >15 min to ensure full blocking. Stimulations were repeated six times. Rise and decay times were calculated from 20 to 80% of the current amplitude maximum, respectively. Half-widths were calculated from 50 to 50% of the maximum amplitude.
Optical stimulation
Cells were clamped to membrane potentials of −70 mV. Blue light pulses (470 nm LED; Thorlabs) were directed through the objective onto the respective stimulation site. A light pulse of 174 mW and 0.5 ms was established as the shortest duration triggering roughly 80% maximal cell responses in a set of exploratory experiments. The stimuli were applied at 0.025 Hz for 5 min. The input strength was assessed as the mean amplitude of six excitatory postsynaptic currents (EPSCs). We applied weak test pulses of 43 mW intermittently to the stimulation pulses. Proximal stimulation sites were determined orthogonally from the pyramidal cell band either at 100 µm distance in stratum oriens or 150 µm distance in stratum radiatum measured from the center of the recorded cell. Upstream recording sites were positioned 50 µm upstream in the direction of CA3 from the proximal site.
Immunofluorescence and confocal imaging
During the electrophysiological recordings, cells were filled with 1% biocytin. Slices were fixed in 2% paraformaldehyde (PFA; Sigma-Aldrich) diluted in 0.1 M phosphate buffer, pH 7.4, at room temperature for 90 min. After fixation, slides were stored in phosphate-buffered saline (PBS; in mM, 137 NaCl, 2.7 KCl, 8 NaH2PO4, pH 7.2) at 4°C. The subsequent staining procedure was performed as follows: slices were washed in PBS, blocked for 2 h in blocking solution [5% normal goat serum (NGS; Vector Laboratories,), 0.5% Triton X-100 (Merck) in PBS] and incubated in primary antibody solution (0.2% Triton X-100 and 1% NGS in PBS) overnight at room temperature. Slices were again washed and incubated for 2 h in the secondary antibody solution. All washing steps were done three times for 15 min in PBS. After the final washing step, slices were mounted on microscope slides using Dako Fluoromount. Fluoroshield with DAPI (Sigma-Aldrich) was used in some experiments to visualize the cell band.
Slices were imaged with a Nikon A1R confocal microscope. We used objectives with magnification factors 10 (Nikon Plan Apo λ 10× NA 0.45), 40 (Nikon Plan Fluor 40× NA 1.3 oil immersion), and 60 (Nikon N Apo 60× NA 1.4 λ oil immersion). Fluorophores were imaged using laser lines for 488 nm and 640 nm. Confocal stacks had a z-distance of 0.5 µm to determine proximal morphology and 0.132 µm for spine and synapse analysis. Images were analyzed in ImageJ (Wayne Rasband, NIH, open-source) or Imaris (Oxford Instruments).
Virus and antibodies
We used the following recombinant adeno-associated viruses in our experiments: for optogenetic stimulation of principal neurons, a humanized ChR2 with H134R mutation fused to EYFP and expressed under the CaMKIIa promoter, AAV5-CaMKIIa-hChR2(H134R)-EYFP, and for synapse analysis, AAV5-hSyn-DIO-EGFP and AAV1.CaMKII 0.4.Cre.SV40 (diluted 1:500 in 0.9% NaCl, both Addgene). The viral expression was visualized with the EYFP signal.
The following primary antibodies were used: rabbit anti-βIV-spectrin (1:1,000; self-made, corresponding to amino acids 2,237–2,256 of human βIV-spectrin; Gutzmann et al., 2014), Alexa 647 anti-rabbit 1:1,000 (Molecular Probes), as well as Avidin 546 1:1,000 (Molecular Probes).
Morphological analysis
CA1 pyramidal cells were filled with biocytin via patch-clamp pipettes. Cells were classified as AcD cells when the dendrite and axon shared a common root that was at least 2 µm long and longer than its mean diameter (Fig. 1D). The length of apical dendrites was measured from the base of the somatic envelope until the end of the furthest dendritic branch using confocal pictures at 20× magnification. Dendritic spines were counted at secondary branches at a 50–200 µm distance from the somatic base in ImageJ (Wayne Rasband, NIH, open-source) using the simple neurite tracer package (SNT; Arshadi et al., 2021). The length and complexity of basal dendritic trees were measured using the Filament Tracer Package of the Imaris Cell Imaging Software (Oxford Instruments). Potential synaptic contacts were measured as approximations, defined as regions where axonal boutons and dendritic spines approached each other in confocal sections <2 µm using Imaris (Sik et al., 1995). For this analysis, afferent axons were sparsely labeled by EGFP-encoding adenovirus (AAV5-hSyn-DIO-EGFP and AAV1.CaMKII 0.4.Cre.SV40). Approximations were counted as clustered synapses if they had at least two other approximations within 20 µm on the same dendritic fiber (Kastellakis et al., 2015; Kastellakis and Poirazi, 2019; Kirchner and Gjorgjieva, 2021).
Contralateral CA3 provides stronger input to CA1 cells with dendritic axon origin. A, Diagram showcasing unilateral virus injection in the mouse's ventral CA3 hippocampal region. Under the CaMKIIa promoter, both Channelrhodopsin-2 (ChR2) and the EYFP reporter were expressed. Recordings took place 3 weeks post-injection. The right panel displays EYFP-tagged cells and fibers in both ipsilateral and contralateral hippocampal sections. B, Fluorescence intensity of axonal fibers in stratum radiatum and stratum oriens of hippocampal area CA1 (normalized to the alveus background, devoid of EYFP-positive fibers). The ipsilateral CA3 indicates a preferential fluorescence in stratum radiatum, while contralateral CA3 leans toward stratum oriens (paired t test, numbers in the graph correspond to the number of animals). C, Confocal microscopy image of biocytin-marked CA1 pyramidal neurons within hippocampal slices. The left image presents the ChR2 signal (green) in stratum radiatum and oriens, paired with biocytin-filled pyramidal neurons (magenta). The right highlights neurons labeled for the axon initial segment (βIV-spectrin), highlighting axons arising from the soma (indicated by blue arrows) and those from a dendritic branch (red arrows). D, Classification criteria for AcD cells. The AcD stem dendrite had to be at least 2 µm long and a width-to-length ratio favoring length (Thome et al., 2014; Hodapp et al., 2022). E, Confocal image of recorded neuron (biocytin) and stimulated fibers from contralateral CA3 (green). The circle illustrates the area of stratum oriens illuminated with 50–100% LED intensity (470 nm, 174 mW, 0.5 ms). F, G, Light stimulation of ChR2-expressing fibers from contralateral CA3 in stratum oriens triggers stronger EPSCs and EPSPs in AcD cells compared with nonAcD cells. F, Provides the average traces for currents (top) and potentials (bottom). Bar plots in G represent mean ± SEM of both currents (left) and potentials (right). H, Analysis of the input strength upon contralateral CA3 stimulation shows no link between cell position along the pyramidal layer's axial axes and EPSP amplitude. Multiple regression analysis determined a significant correlation solely with AcD morphology and not the cell location. The inset visualizes cell location categorized by AcD morphology. The numbers displayed on the bars provide the total cell and animal count.
Statistical tests
The investigator was blind to cell morphology during recording. EPSC analysis was automated using custom-written Python scripts. Cell anatomy was analyzed before electrophysiological data were assigned to respective cells to ensure an unbiased evaluation. Statistical analysis was performed in GraphPad Prism 8 (GraphPad Dotmatics). Bar plots and values are given as mean ± standard error of the mean (SEM). The number of cells, branches, and animals is indicated within bars. Some groups failed the Anderson–Darling normality test; thus, all tests were two-tailed nonparametric Mann–Whitney U tests if not stated otherwise in text and figure legends. Multiple groups were tested using the nonparametric Kruskal–Wallis ANOVA tests. If p < 0.05, individual differences between AcD and nonAcD cells were tested with Dunn's multiple-comparisons test. Sholl analysis data were tested using one-way ANOVA. Multiple linear regression analyses were conducted using the least squares method. Sample size, mean, median, standard deviations, interquartile ranges, normality test results, and parametric test results are provided in Table 1. If not stated directly, significance thresholds are indicated as follows: *p < 0.05, **p < 0.01, and ***p < 0.001.
Statistical parameters and tests
Results
Enhanced input from contralateral CA3 to AcD cells relative to nonAcD cells
The highest density of AcD cells was reported in the CA1 area of the intermediate/ventral hippocampus (Thome et al., 2014). Thus, we focused on this region for our study. Utilizing virus injections, we labeled axons from ipsilateral and contralateral CA3 pyramidal neurons (Fig. 1A). These CA3 neurons predominantly project to the stratum oriens and radiatum of the CA1 areas in both hemispheres. Fluorescence intensity assessments post-virus-mediated EYFP expression in CA3 neuron axons confirmed the primary innervation of stratum radiatum by axons from ipsilateral CA3 cells. In contrast, fibers from contralateral CA3 demonstrated a bias toward stratum oriens, the primary residence of most AcD branches (Fig. 1B; Shinohara et al., 2012).
We next investigated whether the synaptic strength of CA3 inputs differs between AcD and nonAcD cells. We employed voltage-clamp recordings of CA1 pyramidal cells in acute ex vivo slices, combined with optogenetic stimulation of ChR2-expressing presynaptic fibers. The stimulation was targeted to the center of the basal dendritic tree at ∼100 µm distance from the soma in stratum oriens. Input strength was assessed as the mean amplitude of six EPSCs triggered by blue light pulses (470 nm, 174 mW, 0.5 ms, 0.025 Hz). After recording, biocytin-filled neurons were reconstructed and classified as AcD or nonAcD cells (Fig. 1C–E and Materials and Methods). Notably, optogenetic stimulation of contralateral CA3 fibers around the basal dendritic tree elicited stronger EPSCs in AcD cells compared with cells with somatic axon origin (Fig. 1F,G; mean amplitude of EPSCs nonAcD cells: 155 ± 18 pA, n = 18; AcD cells: 263 ± 34 pA, n = 20; p = 0.004). While EPSC amplitudes are reliable indicators of synaptic input strength, they do not capture the extent of membrane depolarization provided by the input. We thus measured the amplitude of excitatory postsynaptic potentials (EPSPs) in current-clamp configuration at membrane potentials of −70 mV. Again, optogenetic stimulation of contralateral CA3 fibers led to larger EPSPs in AcD compared with nonAcD cells (Fig. 1F,G; EPSP amplitude: nonAcD cell, 2.9 ± 0.6 mV, n = 14; AcD cell, 5.1 ± 0.7 mV, n = 31; p = 0.0192). Previous findings suggest input efficiencies vary across the CA1 superficial to deep axis (Valero et al., 2015), and AcD cells predominantly occupy superficial positions (Thome et al., 2014). Implementing a multiple linear regression model, we discerned that under our experimental conditions, only cell morphology but not position associated significantly with the input strength from contralateral CA3 (Fig. 1H).
AcD cell-specific input strength is restricted to the basal dendrites
The majority of contralateral CA3 axons target the apical dendrites in stratum radiatum (Shinohara et al., 2012). Therefore, we examined the responses from contralateral CA3 after optogenetic activation in this layer. Here, synaptic contacts are mainly found on the laterally branching oblique dendrites. We selected two distinct stimulation sites: around the apical dendritic stem and a location further upstream closer to CA3 (Fig. 2A, stimulation sites “3” and “4”). Stimulations at both the proximal and upstream CA3 fiber locations in stratum radiatum revealed comparable input strengths for both AcD and nonAcD cells (Fig. 2B). Specifically, the mean EPSC post-proximal radiatum stimulation was 330 ± 43 pA for nonAcD cells (n = 17) and 378 ± 60 pA for AcD cells (n = 17; p = 0.7596). After upstream radiatum stimulation, the EPSC values were 264 ± 42 pA for nonAcD cells (n = 16) and 357 ± 54 pA for AcD cells (n = 16; p = 0.1964). To maintain consistency in our methods, we also stimulated further upstream of the stratum oriens site to target more distant fibers (Fig. 2A, stimulation site “2”). Notably, there was no significant difference in synaptic current amplitudes between AcD and nonAcD cells (Fig. 2C; mean EPSC nonAcD cells: 286 ± 38 pA, n = 17; AcD cells: 344 ± 31 pA, n = 19; p = 0.1139). We speculated that this stimulation site might capture axonal fibers traversing stratum pyramidale to innervate the apical dendrite in stratum radiatum. Our EYFP-labeled sections confirmed the presence of such fibers (Fig. 2D,E; see Table 1 for statistics).
Specific innervation is confined to the basal dendritic input area. A, Diagram depicting optogenetic stimulation sites of contralateral CA3 fibers. Proximal sites were positioned orthogonally from the pyramidal cell layer, either 100 µm basally in stratum oriens (sites 1 and 2) or 150 µm apically in stratum radiatum, measured from the center of the recorded cell (sites 3 and 4). Upstream sites (2 and 4) are 50 µm nearer to CA3. In select experiments (F), action potential propagation was inhibited using a bath application of 1 µM TTX and 100 µM 4-AP. B, Optogenetic stimulation around apical dendrites reveals no distinction between AcD and nonAcD cells in terms of mean EPSC amplitudes. C, Stimulation 50 µm upstream of the basal dendritic stimulation site shows no notable difference between AcD and nonAcD cells. D, A substantial number of axon collaterals traverse between stratum oriens and radiatum. The representative confocal image visualizes EGFP-positive fibers from contralateral CA3, with cell layer borders demarcated by dotted lines. E, Axonal fibers from contralateral CA3 traverse and cross stratum pyramidale. Quantitative analysis of individually traced segments of contralateral CA3 fibers in CA1. F, The application of TTX (1 µM) and 4-AP (100 µM) to block action potential propagation reinstates asymmetry between AcD and nonAcD cell responses in basal dendrites. Stimulation of oblique fibers in stratum oriens, with the action potential generation blocked, exhibits no difference between the two cell morphologies. Bar plots present mean ± SEM and numbers provide the total cell and animal count.
When we restricted activation to the immediate vicinity of the stimulus site by inhibiting long-range action potential propagation using TTX and 4-AP (Yamawaki et al., 2016), the input specificity of contralateral CA3 fibers in stratum oriens was re-established (Fig. 2F dist oriens: mean EPSC nonAcD cells, 286 ± 38 pA, n = 17; AcD cells, 380 ± 46 pA, n = 20; p = 0.074). However, input efficiencies in stratum radiatum remained consistent across both cell types. These findings indicate that the unique synaptic strengths from contralateral CA3 to AcD and nonAcD cells in the CA1 region are primarily linked to the basal dendritic compartment.
Kinetics of optogenetically evoked postsynaptic currents are similar in AcD and nonAcD cells
To offer a thorough characterization of the CA3 inputs to distinct dendritic segments, we took a closer look at the kinetics of optogenetically induced EPSCs at all four stimulation sites, both in the presence and absence of TTX/4-AP (Fig. 3). Notably, optogenetic stimulation of axonal fibers by ChR2 (type H134R) causes a slow and prolonged release of glutamate, resulting in nonphysiological postsynaptic current kinetics (Boyden et al., 2005; Lin, 2011; Schoenenberger et al., 2011; Malyshev et al., 2015). However, these events are still influenced by the synaptic count, presynaptic release probability, and the makeup of postsynaptic ion channels. Our kinetic analysis, spanning different stimulation sites and cell types, did not identify any significant differences in terms of EPSC rise, half-width, or decay times (Fig. 3A–C). This suggests that the intrinsic attributes of the synapses between contralateral CA3 and CA1 are likely consistent, irrespective of whether they are AcD or nonAcD cells.
Kinetics of EPSCs triggered by contralateral CA3 stimulation show no differences between AcD and nonAcD cells. A, The illustration representation stimulation sites, as well as the specific parameters analyzed in the responses. The top panel refers back to Figure 2A for additional clarity. The bottom panel delineates the kinetics of the EPSC. These kinetics are defined in terms of rise and decay times that occur between the 20 and 80% thresholds of peak EPSC amplitudes, along with half-widths that occur at the 50% threshold of the maximal amplitude. B, C, The kinetics of the EPSC reveal no consistent patterns when juxtaposed against diverse stimulation sites, cell structures, and conditions of AP propagation. Specifically, B represents conditions in standard ACSF, while C portrays conditions in the presence of TTX and 4-AP. Bar plots present mean ± SEM and numbers provide the total cell and animal count.
Ipsilateral CA3 shows no input preference to either AcD or nonAcD cells
The vast majority of AcD branches in CA1 are basal, rather than apical, dendrites (Thome et al., 2014) and thus reside in stratum oriens. Our observations from histology support this, revealing that 98% (108 out of 110) of AcD branches were basal and located in stratum oriens. However, the primary excitatory inputs to CA1 come from ipsilateral CA3 fibers which predominantly target stratum radiatum. We thus studied the innervation strength of ipsilateral CA3 fibers onto AcD and nonAcD cells (Fig. 4A,B) and found stronger responses for stimulation of ipsilateral CA3 axons within stratum radiatum than in stratum oriens (Fig. 4C,D, compare with Figs. 1, 2). Using the same stimulation protocol as for contralateral CA3, we found no difference in synaptic input strength or EPSC kinetics between AcD and nonAcD cells. This observation persisted regardless of the stimulation sites, whether in stratum radiatum or oriens, and both for proximal and distal positions. Thus, ipsilateral inputs from CA3 to CA1 pyramidal neurons do not show a pronounced preference for AcD cells, in contrast to contralateral inputs.
Synaptic inputs from ipsilateral CA3 show uniform strength in AcD and nonAcD cells. A, Confocal image of the ventral hippocampus following virus injection in the ipsilateral CA3, with CA1 cells filled with biocytin (yellow). B, Somatic ChR2 expression (yellow) was only observed in CA3 proximal to CA2, visualized by PCP4 expression (magenta). Inset shows whole hippocampal structure. C, D, Average EPSC triggered by optogenetic stimulation of the ipsilateral CA3 fibers do not exhibit any pronounced bias toward cells with either dendritic or somatic axon origins (positions illustrated in Fig. 2A). Additionally, there were no discernible disparities in EPSC kinetics between the two cell morphologies. Bar plots present mean ± SEM and numbers provide the total cell and animal count.
Morphological characteristics are conserved across AcD and nonAcD cells
Morphological features of neurons, especially their dendritic structure, are vital components that determine the impact of synaptic inputs as well as their functional roles in neural circuits. Recent studies did not reveal significant differences between AcD and nonAcD cells regarding overall synaptic input strength or dendritic structure (Thome et al., 2014; Hodapp et al., 2022) in CA1 pyramidal cells, but some specializations in the apical dendrite of neocortical neurons (Hamada et al., 2016). Along with these findings, our present data further affirm the similarity between these two cell types in CA1. Both cell types exhibited comparable soma size, input resistance, and apical dendrite length (Fig. 5A,B).
AcD cells and nonAcD cells have similar morphological and electrophysiological characteristics. A, Graphical representation of evaluated morphological parameters, including apical dendrite length, combined lengths of basal dendritic branches, count of primary dendritic branches, and synaptic spine density. Other measures are represented in Figure 1D, such as AIS distance from the somatodendritic compartment and measurements of the AcD stem dendrite's length and diameter from the somatic outline to the axonal arbor. B, Parameters that passively influence neuronal excitability, like soma size (ascertained as cross-sectional area), input resistance (determined in voltage-clamp mode), apical dendrite diameter (calculated as an average of measurements taken 5, 10, 15, 20, and 25 µm above the soma), and apical dendrite length (defined as the distance between the soma and the apex of the dendritic tree), are similar between AcD and nonAcD cells. C, Basal dendritic trees of both AcD and nonAcD cells show no difference when assessed for size and complexity. The left panel presents the aggregated length of basal dendrites per cell and the count of primary basal branches for each cell. A visual comparison of AcD and nonAcD branches within AcD cells is offered using orange and green bars, respectively. The right panel depicts dendritic complexity, evaluated as the number of intersections in a Sholl analysis. D, The left panel depicts the size (total length) of individual primary branches within AcD cells, highlighting that AcD branches are typically more extensive than individual nonAcD branches. The right panel, employing Sholl analysis, illuminates the greater complexity of AcD branches, particularly evident within 30–100 µm Sholl radii. Mean values are represented by lines, with shaded areas indicating SEM. E, Detailed analysis of AIS and AcD stem morphology. The left panel showcases the varied morphologies of AcD stems in recorded cells (mean length and diameter of 6.48 ± 0.47 µm and 2.02 ± 0.09 µm). The middle segment indicates that both nonAcD and AcD cells share similar AIS lengths. The right panel contrasts AIS distances (βIV-spectrin signal) from the somatodendritic compartments (soma in nonAcD cells, axonal bifurcation point in AcD cells). This distance is greater in nonAcD cells compared with AcD cells. F, Dendritic spine densities are similar between AcD and nonAcD cells as well as AcD and nonAcD branches within AcD cells. A representative image from a biocytin-filled CA1 pyramidal cell is included. Bar plots present mean ± SEM and numbers provide the total cell and animal count.
Furthermore, the basal dendritic characteristics, including its length, the number of primary branches, and its overall complexity, did not show any marked differences between the two groups (Fig. 5C). Notably, AcD cells possess an axon-carrying branch that constitutes a significant portion of their total basal dendritic tree (Fig. 5C, orange bars). Moreover, when comparing the AcD branches to the nonAcD branches within the same cell, the former typically displayed increased size and complexity (Fig. 5D). Specifically, the AcD branches measured an average of 714 ± 60 µm in length (n = 38), while the nonAcD branches averaged 403 ± 33 µm (n = 75), indicating a significant difference (p < 0.0001). This size distinction is especially apparent in bifurcations located 30–100 µm away from the soma, suggesting that AcD branches have enhanced potential for forming connections with passing axonal fibers in stratum oriens.
To delve deeper into the AIS morphology, we utilized immunostaining against βIV-spectrin, an essential scaffolding protein of this compartment (Ogawa and Rasband, 2008; Gutzmann et al., 2014). Both cell types displayed identical AIS lengths (Fig. 5E). However, in AcD cells, the AIS origin was closer to the onset of the axon compared with nonAcD cells. This observation is consistent with previous findings (Thome et al., 2014). The stem segment of the AcD had dimensions of 5.6 ± 0.5 µm in length and a diameter of 1.9 ± 0.1 µm, aligning with earlier studies (Thome et al., 2014; Hodapp et al., 2022). To estimate the number of potential excitatory synaptic connections, we examined the density of synaptic spines. Aligning with prior data (Hodapp et al., 2022; Wahle et al., 2022), spine density was consistent across AcD and nonAcD cells on their basal dendrites, as well as between AcD and nonAcD branches within AcD cells (Fig. 5F). This provides further evidence for the conservation of morphological characteristics across the two cell types.
Differential axon–spine approximations in AcD versus nonAcD cells and branches
We observed that AcD cells receive stronger synaptic inputs from contralateral CA3 fibers than their nonAcD counterparts (Fig. 1G). To investigate potential structural correlates of this discrepancy, we quantified potential synaptic contacts (termed approximations) between sparsely labeled axonal fibers of contralateral CA3 and individual dendritic branches of pyramidal cells. Approximations were identified in confocal images after deconvolution and computer-guided morphological reconstruction (Fig. 6A, extended data Movie 1). In line with the results from electrophysiological experiments, we found that excitatory fibers from contralateral CA3 form more putative synapses with AcD cells than those with nonAcD cells (Fig. 6B, left bar pair; nonAcD cells: 25 ± 3, n = 10; AcD cells: 37 ± 2, n = 17; p = 0.0069). Since AcD and nonAcD cells have similar sizes of basal dendritic trees (Fig. 5C; Hodapp et al., 2022), we hypothesized that the stronger input efficiency of contralateral CA3 to AcD cells may result from a higher number or density of synaptic connections. Indeed, we found that AcD cells have a higher density of approximations with contralateral CA3 axons than nonAcD cells (Fig. 6C, left bar pair; nonAcD cells: 1.77 ± 0.22, n = 10; AcD cells: 2.48 ± 0.89, n = 17; p = 0.0404).
Differential axon–spine approximations in AcD versus nonAcD cells and branches. A, Depiction of axon–spine approximations between biocytin-labeled CA1 pyramidal cell dendrites (red) and sparsely virus-labeled contralateral CA3 axons (cyan). Criteria for approximation: dendritic and axonal signals within a 2 µm distance in the 3D reconstruction. See also extended data Movie 1. B, Comparison of approximation numbers and densities in AcD versus nonAcD cells and branches (within AcD cells). Contralateral CA3 fibers have significantly more approximations with AcD cells (left pair). Within AcD cells, most approximations are located at nonAcD branches (sum, middle pair). If approximations were assigned to individual branches, the AcD branch exhibited the greatest number. C, Axon–spine approximations are more concentrated on basal dendrites of AcD cells. Within AcD cells, the densities are equal among AcD and nonAcD branches. D, Mean distances of axon–dendrite approximations from the somatic envelope are not different between AcD and nonAcD cells (left pair) or branches (right pairs). E, AcD cells show a higher number of clustered approximations (defined as ≥3 approximations within 20 µm of dendrite; see confocal image) compared with nonAcD cells. Among AcD cells, nonAcD branches in sum contain more such clusters than individual AcD branches. F, For ipsilateral CA3 fibers, there is no discernible difference in the number or density of axon–spine approximations between biocytin-labeled CA1 pyramidal cell dendrites and virus-labeled contralateral CA3 axons, using the same 2 µm distance criterion for approximations in 3D reconstructions. Bar plots present mean ± SEM and numbers provide the total cell and animal count.
Confocal imaging of CA1 dendrites and contralateral CA3 axons. The movie showcases biocytin-labeled dendrites of CA1 pyramidal neurons (red) and GFP-tagged axons from contralateral CA3 (green) in the stratum oriens, visualized using Imaris software. [View online]
Next, we compared whether there was a preference for AcD or nonAcD branches within AcD cells. Indeed, within individual AcD neurons, the axon-carrying branch showed more approximations than single nonAcD branches (Fig. 6B, right bar pair; nonAcD branches: 8 ± 1, n = 44; AcD branches: 16 ± 3, n = 17; p = 0.0137). However, compiling all nonAcD branches revealed a larger overall approximation number than a singular AcD branch (Fig. 6B, middle bar pair). The increased number of putative synaptic contacts on AcDs could, in principle, be caused by a higher density of approximations or by the larger size of AcD branches compared with “normal” dendrites (compare Fig. 5D). We therefore measured the density of approximations of CA3 fibers and found that there was no difference between axon-carrying and nonaxon-carrying branches (Fig. 6C). This suggests that the increased number of approximations reflects the increased surface area of the AcD.
Finally, we measured the spatial distribution of CA3 axon approximations along the dendritic tree and found uniform approximation distances to the somatic envelope, regardless of cell group or branch type (Fig. 6D). It is noteworthy that individual axonal fibers occasionally form multiple proximate contact sites on a single dendritic branch (Fig. 6E). Moreover, AcD cell approximations exhibited a higher tendency to cluster than those on nonAcD cells, both in AcD and nonAcD branches in roughly equal measures (compare Fig. 6B). Thus, the more frequent occurrence of clustered approximations likely result from the larger size of AcD branches compared with single nonAcD branches (compare Fig. 5D; Kastellakis et al., 2015; Kastellakis and Poirazi, 2019; Kirchner and Gjorgjieva, 2021).
Analogous to our optogenetic stimulation paradigm, we also investigated the number and density of putative synapses from ipsilateral CA3 to CA1. Similar to our electrophysiological observations, we did not find any statistical differences in the number or density of axon–spine approximations for either AcD and nonAcD cells or AcD and nonAcD branches (Fig. 6F). In summary, CA1 pyramidal cells with AcD morphology receive stronger synaptic input from contralateral CA3 than canonical cells with somatic axon origin. The observed disparity might be caused, at least in part, by higher numbers of synaptic contacts onto the basal dendritic tree of these cells with contralateral CA3 fibers.
Discussion
Our findings reveal an asymmetric interhemispheric input to AcD versus nonAcD cells, which likely affects their integration into coordinated network activity. Cells with dendritic axon origin are more excitable and convey a privileged input channel for action potential generation (Thome et al., 2014; Hodapp et al., 2022). It was hitherto unknown whether this special dendrite also has specific presynaptic partners, which would add a further distinction in information processing. In the present study, we investigated the connection from contralateral and ipsilateral CA3 to CA1 pyramidal cells. We found that fibers from contralateral CA3 trigger stronger inputs in the basal dendritic tree of AcD cells. The morphological analysis demonstrated that neurons with axon-carrying basal dendrites exhibit more axon–spine approximations with contralateral CA3 fibers and that these approximations are most numerous at AcD branches. Combined with the electrophysiological recordings, our results indicate that AcD cells receive stronger synaptic inputs from contralateral CA3 fibers, due to a higher number of excitatory synaptic connections at AcD branches. We did not find any differences upon the stimulation of fibers in stratum radiatum, though AcD axons originate very rarely from the apical dendritic tree (Lorincz and Nusser, 2010; Hofflin et al., 2017; Wahle et al., 2022). Our findings demonstrate that cells with dendritic axon origins receive specialized synaptic input. This reveals a direct link between neuronal structure and network integration, and it might provide clues about the functional role of AcD morphology in the living brain.
Potential causes of stronger CA3 input
We found that excitatory input from contralateral CA3 to the basal dendritic tree of CA1 pyramidal cells is considerably stronger in cells with dendritic axon origin compared with that in nonAcD cells (Figs. 1F,G, 2F). There are multiple potential explanations for such differences, of which we discuss five in the following section. Firstly, cells with stronger responses may form more synaptic contacts with contralateral CA3. Indeed, AcD cells featured more axon–spine approximations than nonAcD cells (Fig. 6B). Although we did not verify that morphological approximations comprise functional synapses, the correlation between this structural feature and the observed difference in synaptic strength is strongly suggestive of an increased number of synaptic inputs to basal dendrites of AcD cells. Furthermore, the lack of AcD specific input strengths from ipsilateral CA1 coincides with the equal number of axon–spine approximations found in our histological analysis (Fig. 6F). Secondly, synapses might be located closer to the soma, causing larger postsynaptic potentials in somatic electrophysiological recordings. We found no differences in the size of the dendritic tree or spine densities between both cell morphologies (Fig. 5) in line with previously published data (Hodapp et al., 2022; Wahle et al., 2022). Moreover, the distribution of distances of synaptic approximations was similar between AcD and nonAcD cells (Fig. 6D), excluding that different dendritic distances of synapses contribute to our results. Thirdly, contralateral CA3 to CA1 synapses may be stronger due to higher release probability, higher numbers of postsynaptic glutamate channels, or dendritic spikes. We did not directly measure the pre- and postsynaptic properties of contralateral CA3 to CA1 connections. Optogenetic stimulation triggers calcium entry with different dynamics and amplitude than electrical stimulation or physiological activation of synapses (Lin, 2011). This is probably also one of the reasons why we did not see any dendritic spikes in our current-clamp recordings. However, the half-widths, rise times, and decay times of the EPSCs were not different between AcD and nonAcD cells (Fig. 3B,C), indicative of similar synaptic properties. Fourthly, an equal number of fibers might target AcD and nonAcD cells in CA1, but the preparation of horizontal slices might cut selectively more fibers targeting nonAcD cells. There is no indication that fibers targeting AcD or nonAcD cells might traverse the hippocampus at different angles, although we cannot exclude such differences at the present state. However, such differences would mostly affect the number of long-range connections in a slice, while our optogenetic stimulation only activates the distal parts of axons. Thus, we would still expect an intact activation of synapses. The assumption of differences in the number of preserved fibers would also be difficult to reconcile with the increased number of axon–spine approximations in AcD cells (Fig. 6B). A fifth explanation might arise from the nonrandom distribution of AcD cells in CA1, which coincides with known biases in input strength across the superficial to deep axes of the pyramidal layer (Valero et al., 2015). However, the location of the patched cells did not significantly differ between AcD and nonAcD cells, and we found no correlation between cell location in the superficial–deep axis and the amplitude of EPSPs triggered by optogenetic stimulation of CA3 fibers (Fig. 1H).
In summary, our data can exclude several alternative explanations of the different input strengths (options 2–5). However, the higher absolute numbers and densities of axonal approximations on AcD and nonAcD (axon-free) branches of AcD neurons suggest that AcD cells as a whole are more strongly connected to contralateral CA3 across their basal dendritic tree. Approximations were often clustered, with one fiber forming multiple putative contact sites with an individual dendritic branch. We found that such clusters were approximately three times more common in AcD cells compared with those in nonAcD cells. Again, we found no evidence for different numbers of clusters on the axon-carrying and axon-free dendrites of the same AcD cells. Spatially and temporally synchronized inputs are particularly prone to trigger superlinear input summation, including dendritic spikes (Polsky et al., 2004). Though they were observed more frequently in AcD branches in a previous study (Thome et al., 2014), we could not observe any clear dendritic spike using our optogenetic stimulation in either AcD or nonAcD cells. Instead we compared the maximum EPSP slope per EPSP amplitude as a measure for the contribution of active dendritic sodium conductances but found no differences between both groups (data not shown). Thus, the clustering of contralateral CA3 inputs may contribute to the particularly strong excitation of AcD cells but cannot explain all observed effects. It would also be important to know how axons from different CA3 neurons converge on single CA1 cells, but this question cannot be solved with our approach of sparse labeling.
Other presynaptic partners
Our study was focused on inputs from contralateral CA3 due to its central position in the hippocampal circuit and preference for stratum oriens (Fig. 1B). However, proximal and distal stimulation protocols were also performed in slices with ChR2-expressing axons from ipsilateral CA3 (Fig. 4). While responses to this dominant input source were generally stronger, we found no difference between AcD and nonAcD cells for ipsilateral CA3 inputs, regardless of the stimulation location (distal–proximal, oriens–radiatum; Fig. 4). We do not know the full pattern of presynaptic partners of AcD versus nonAcD cells or AcD versus nonAcD dendritic branches. However, it is feasible that brain regions showing a preference for stratum oriens might also demonstrate an imbalance in the innervation of cells depending on their axon origin. Candidate regions that innervate stratum oriens comprise neurons from ipsilateral CA2 (Shinohara et al., 2012), medial septum (Colom et al., 2005), and recurrent connections between CA1 pyramidal cells (Yang et al., 2014). Thus, our study shows a difference in input strength of one marked synaptic connection, but we cannot exclude the existence of further asymmetries between AcD and nonAcD cells. Other prominent presynaptic partners, such as layer III of the lateral and medial entorhinal cortex as well as thalamic nucleus reuniens, innervate the distal dendritic tufts in stratum lacunosum (Doller and Weight, 1982; Deller et al., 1996; Witter et al., 2000; van Groen et al., 2003; Suh et al., 2011). Since AcD branches predominantly reside very distal from this region in stratum oriens, a preference for AcD cells of these inputs seems unlikely, but cannot be excluded.
Propagation of presynaptic stimulation
The stimulation of axonal fibers in stratum oriens triggers synaptic responses not only in its immediate vicinity but across all cells that receive input from these axons. Action potentials will propagate through the fibers in orthodromic and antidromic directions and might thereby reach targets in stratum oriens, pyramidale, and radiatum of CA1, CA2, and CA3, as well as subiculum, and septal nuclei (Swanson et al., 1981; Witter, 2007). However, most CA3 fibers project from CA3 to CA1 and usually remain either in stratum radiatum or stratum oriens (Fig. 2D,E). We have largely excluded network effects via multisynaptic pathways by analyzing only synaptic events with onset times smaller than 10 ms. By moving the stimulation site further from the recorded cell, we may recruit more fibers and thus more synaptic connections terminating on the recorded neurons, leading to stronger postsynaptic responses (compare Figs. 1G, 2C). However, this might inadvertently lead to stronger recruitment of synapses at the apical dendritic tree. We assumed that this recruitment reduced the specificity of contralateral CA3 to AcD cell connections (Fig. 2C). Indeed, when we inhibited action potential propagation by blocking voltage-gated sodium and potassium channels (TTX and 4-AP), the specificity observed at proximal locations in stratum oriens was recovered (Fig. 2F, left two graphs).
Formation of hemisphere preference
So far, there is little knowledge on how AcD morphology develops in the maturing brain. This leaves room for speculation about how specialized interhemispheric connections may form in AcD cells. There are several possible mechanisms. Axonal fibers from contralateral CA3 and dendritic spines on AcD branches may express specialized pre- and/or postsynaptic adhesion proteins that facilitate and strengthen synaptic connections (Sudhof, 2021). Such interactions were found to modulate synaptic connectivity between CA3a and distal CA1, as well as distal CA3 and proximal CA1 (Berns et al., 2018). However, to the best of our knowledge there is so far no study on the postsynaptic machinery at AcD branches. On the other hand, the preference of contralateral CA3 fibers for AcD cells may precede the formation of AcD morphology, which continues past P8 (Thome et al., 2014; Lehmann et al., 2023). CA1 neurons that receive particularly strong input from contralateral CA3 may release growth factors that increase the size and complexity of activated dendritic branches, which, in turn, may lead to changes in the location of the axon origin and the AcD morphology. Such dynamic changes of axon position along the soma–dendrite axis have been shown in vitro and indicated in vivo (Lehmann et al., 2023).
Impact on interhemispheric communication
One marked feature of AcD cells is their preferential firing during sharp-wave ripples when CA1 pyramidal cells receive strong phasic excitatory and inhibitory inputs. We showed previously that input to AcD branches largely evades perisomatic inhibition and is therefore able to trigger action potentials in the attached axon when cells with somatic origin are silenced (Hodapp et al., 2022). Whereas ripples are local events formed by local interneurons, sharp waves occur in close temporal correlation in both hemispheres (Ylinen et al., 1995; Maier et al., 2011; Gan et al., 2017). Thus, the strong excitatory connection between CA3 and the contralateral CA1 region may facilitate the coupling of neuronal assemblies between both hemispheres during sharp-wave ripple oscillations. Long-range projections from dorsal CA3 to medial CA1 were found to be important for long-term memory consolidation (Carr et al., 2012). Consequently, neurons with AcD morphology might play a key role in this interhemispheric interaction and additionally provide a hub for short-term and long-term changes to those connections (Petrus et al., 2019; Ferrier et al., 2020).
In summary, our work reveals a new structure–function relation for the preferential activation of neurons with dendritic axon origin. Previous studies have demonstrated that dendritic axon origins give rise to privileged synaptic input channels with unique computational properties (Thome et al., 2014) and that such cells are preferentially activated during sharp-wave ripples in vivo and in vitro (Hodapp et al., 2022). We now show that these cells are uniquely integrated into the hippocampal network circuitry, with a particularly strong activation from the contralateral hippocampus. These connections may guarantee interhemispheric communication during sharp-wave ripple complexes or, potentially, in other functional states. AcD cells may thus be essential for interhemispheric information transfer and synchronization of network activity.
Footnotes
This work was supported by the German Research Foundation (DFG BO 3512/2-1 to M.B. and DFG EN 1240/2-1 to M.E. and the Walter Benjamin Programme PN 458054460 to C.T.). The authors thank Camilo Faust Akl for lending his computer; Nadine Zuber for her contribution with staining and solution preparation; Jan Maximilian Jansen for his help with Imaris, Nadja Sharkov, and Tina Sackmann for assisting with image acquisition; Nadja Lehmann and Susanna Weber for method instruction; and the Nikon Imaging Center Heidelberg for providing microscopes.
The authors declare no competing financial interests.
- Correspondence should be addressed to Christian Thome at christian.thome{at}jku.at.
