Deep patch-clamp recordings in awake mice from medial septal neurons during hippocampal sharp-wave ripples

Abstract

Objectives

The medial septum modulates hippocampal oscillations, including ripples, which are critical for memory consolidation. While the role of the medial septum in theta rhythms is well-established, its specific contribution to hippocampal ripple activity remains poorly understood. This study sought to investigate the relationship between medial septal activity and hippocampal ripples in vivo.

Methods

This study aimed to characterize the in-vivo membrane potential dynamics of putative medial septal neuron subtypes and their contribution to hippocampal ripples in awake mice. We performed in-vivo whole-cell patch-clamp recordings from medial septal neurons in head-fixed, awake mice, while simultaneously acquiring hippocampal local field potentials.

Results

Medial septal neurons were classified into glutamatergic, cholinergic, and GABAergic subtypes using hierarchical clustering based on their intrinsic electrophysiological properties. We analyzed the firing rates and subthreshold membrane potential dynamics of these neurons during hippocampal ripple events and examined their correlations with ripple parameters (duration, frequency, and power). Our results revealed subtype-specific responses. Notably, putative glutamatergic neurons exhibited a slight decrease in firing rate, yet displayed a pronounced depolarization of their membrane potential approximately 100 ms before ripple onset, peaking at the initiation of ripples. This depolarization was inversely correlated with subsequent ripple amplitude and power. In addition, membrane hyperpolarization was positively correlated with ripple duration.

Conclusion

These findings elucidate the contribution of glutamatergic medial septal neurons to hippocampal ripple dynamics and suggest a tightly regulated interaction between the medial septum and hippocampus in shaping ripple activity.

Introduction

Hippocampal neuronal oscillations play a vital role in information processing. Theta rhythms (4–12 Hz) are closely linked to attention, spatial navigation, and memory encoding [1–4], whereas hippocampal ripples (150–250 Hz) are associated with memory consolidation [5]. Both are regulated by the medial septum, a subcortical region comprising GABAergic, cholinergic, and glutamatergic neurons that project to the hippocampus via the fornix [6,7].

Reciprocal septo–hippocampal connections modulate and sustain theta rhythms [8–11]. Recent work demonstrated that theta and ripple events are predominantly observed during distinct behavioral states [12]. Cholinergic medial septal neurons suppress hippocampal ripples [7], whereas the relationship between GABAergic medial septal neurons and hippocampal ripples has not yet been elucidated. Nevertheless, medial septal neuronal firing markedly decreases during ripple episodes [13].

Although prior studies have explored medial septum–hippocampus interactions, the roles of medial septal glutamatergic neurons remain poorly characterized. Most investigations have depended on single-unit extracellular recordings or ex-vivo slice patch-clamping, leaving subthreshold dynamics and functional roles of neuron subtypes during ripples unexamined. To bridge this gap, we performed in-vivo whole-cell patch-clamp recordings from medial septal neurons in head-fixed, awake mice, while simultaneously monitoring hippocampal local field potentials (LFPs). This approach allowed direct assessment of membrane potential changes in medial septal neuron subtypes, classified by electrophysiological properties, and their contributions to ripple activity in vivo.

Methods

Ethical approval

All animal experiments, approved by the University of Tokyo Animal Experiment Committee (P4-2), conformed to its Guidelines for the Care and Use of Laboratory Animals. Protocols additionally adhered to the ‘Fundamental Guidelines for Proper Conduct of Animal Experiments and Related Activities in Academic Research Institutions’ (Ministry of Education, Culture, Sports, Science and Technology, Japan, Notification No. 71, 2006), the ‘Standards Relating to the Housing and Keeping, and Pain Relief of Experimental Animals’ (Ministry of the Environment, Japan, Notification No. 88, 2006), and the ‘Guidelines on Methods of Disposal of Animals’ (Prime Minister of Japan, Notification No. 40, 1995).

Animals

Male Institute for Cancer Research mice (28–40 days old; Japan SLC Inc., Shizuoka, Japan) were housed under a 12-h light/dark cycle (lights on at 7 : 00 a.m.) at 22 ± 1 °C, with ad libitum access to food and water.

Surgical procedures and surgical preparation

Anesthesia was confirmed throughout all surgical procedures by the absence of pedal withdrawal, whisker movement, and blink reflexes. Following anesthesia, the mouse’s head hair was removed, and the animal was placed in a stereotaxic frame. For patch-clamp and LFP recordings, two craniotomies were performed: (a) a 0.5 mm² opening, 2.5 mm caudal, and 0.2 mm lateral to bregma, and (b) a 1.0 mm² opening, 1.0 mm rostral, and 1.0 mm lateral to bregma. The dura mater was carefully removed. A 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate-coated tungsten electrode (#42364; Sigma-Aldrich, Missouri, USA) was inserted into craniotomy (a), targeting the CA1 region of the hippocampus (1.5 mm rostral, 0.2 mm lateral to bregma, and 1.0 mm ventral to the dura), after removing brain tissue from the surface to the medial septum. The exposed brain surface and skull were covered with approximately 2 mm-thick 1.7–2.0% agarose.

In-vivo electrophysiological recordings

Blind in-vivo patch-clamp recordings were obtained from putative neurons in the medial septum. Recording pipettes, pulled from borosilicate glass capillaries (1.0 mm optical density) using a micropipette puller (P-1000; Sutter Instrument, California, USA), had a resistance of 4–7 MΩ when filled with an intracellular solution containing (in mM): 135 K-gluconate, 4 KCl, 10 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, 10 creatine phosphate, 4 MgATP, 0.3 Na₂GTP, 0.3 ethylene glycol tetraacetic acid, and 0.2% biocytin (pH: 7.2–7.3, osmolarity: 285–295 mOsm/kg). Recordings utilized an Axopatch 700B amplifier and a Digidata 1440A digitizer (Molecular Devices, Foster City, California, USA) with data acquired at 20 kHz using Clampex 10.7 software. Only recordings with a liquid junction potential-corrected resting membrane potential below –45 mV and an action potential peak above –20 mV were included.

Hippocampal LFPs were recorded using a DAM80 amplifier (World Precision Instruments, Hertfordshire, UK) and a Digidata 1440A digitizer. Data acquisition was at 20 kHz using Clampex 10.7 software.

Hippocampal ripple detection

LFP traces were downsampled to 2 kHz and band-pass filtered between 150 and 250 Hz. Ripples were detected using a threshold of four times the SD of the baseline noise [14]. Detected events were visually inspected and manually rejected if incorrectly identified. Ripple frequency was calculated using complex Morlet wavelet analysis of the LFP. Ripple power was calculated as the squared difference between the maximum and minimum LFP values between ripple onset and offset.

Medial septal membrane potential trace analysis

For each ripple, the mean membrane potential (V_m) was calculated from –200 to +200 ms around ripple onset. ΔV_m was defined as the difference between V_m at any given time point within this window and the mean V_m. ΔV_m traces from all ripples were averaged, aligned to ripple onset, to generate onset-triggered averages. To estimate the probability level, 10 000 randomly selected 400-ms ΔV_m traces were also calculated, defining the 95% confidence interval as the range between the upper 2.5% and lower 2.5%. The Z-score was derived from the V_m at its minimum.

Electrophysiological properties are listed in Table 1. To assess whether peri‑ripple spike probabilities exceeded chance, we generated neuron‑specific surrogate data from the same membrane potential trace. We first removed the initial and final 0.6 s of each recording and excluded all samples within ±0.6 s of any ripple onset, yielding a set of ripple‑free time points. From these, we randomly selected control time points (10 000 times the number of ripples for each neuron). For each ripple onset and each control time point, we extracted a ±0.6 s window and constructed binary spike rasters aligned to the event. For each peri‑event time bin, we then counted the number of ripple‑aligned and surrogate windows that contained at least one spike, formed 2 × 2 contingency tables (spike present/absent × ripple‑aligned/surrogate), and tested for deviations of spike occurrence probability from chance using Chi‑square tests. This procedure allowed us to determine whether peri‑ripple modulation of firing exceeded that expected from the neuron’s own background activity.

Table 1.

The electrophysiological properties of medial septal neurons derived from in-vivo patch-clamp recordings

Properties	GABAergic	Cholinergic	Glutamatergic
AP	61.5 ± 7.78	13.0 ± 1.0	10.25 ± 5.06
Adaptation	2.65 ± 1.85	1.16 ± 0.44	0.26 ± 0.17
CV (step)	12.0	12.67 ± 1.15	11.0 ± 1.15
AP threshold (mV)	54.15 ± 0.14	54.06 ± 0.01	34.89 ± 23.94
Rm (Ω)	0.14 ± 0.02	0.14 ± 0.06	0.14 ± 0.05
Theta power (mV²)	0.27 ± 0.06	14.52 ± 16.52	5.67 ± 3.94
Vm (mV)	−55.82 ± 11.90	−52.45 ± 4.43	−50.17 ± 8.95
FSL (ms)	11.88 ± 5.98	26.58 ± 5.83	98.75 ± 100.80
Sag (mV)	−7.85 ± 1.91	−5.04 ± 1.39	−1.90 ± 2.12

AP, action potential; CV, coefficient of variation; FSL, first spike latency.

Hierarchical clustering of medial septal neurons

All extracted electrophysiological features were standardized using Z-score normalization. Pairwise Euclidean distances were computed to quantify dissimilarity between individual cells based on their standardized electrophysiological profiles, yielding a distance matrix. Hierarchical clustering was then applied to this matrix using the complete linkage method.

Data analysis software

All data were analyzed utilizing MATLAB software (MathWorks, Massachusetts, USA).

Results

Hierarchical clustering of medial septal neurons based on their electrophysiological characteristics identifies neuronal subtypes

In this study, we established a method for recording membrane potential dynamics from medial septal neurons in awake mice using in-vivo whole-cell patch-clamp techniques. To achieve this, we performed a craniotomy and carefully aspirated the overlying cortical tissue above the medial septum, followed by thorough rinsing with PBS to eliminate tissue fluid and blood. This preparation minimized electrode travel distance and enhanced recording stability. We further optimized the electrode insertion angle, determining that a 20° angle minimized vascular damage and improved success rates. Despite these optimizations, the success rate remained low at 3.94% (10 out of 254 trials), underscoring the technical difficulty of the procedure (Fig. 1a and b).

Fig. 1.

The hierarchical clustering facilitates the classification of medial septal neuronal subtypes, based on in-vivo patch-clamp recordings data. (a) The experimental design for patch-clamp electrophysiological recordings from medial septal neurons in vivo in awake mice, and a schematic representation of patch-clamp electrode insertion into the medial septum. (b) The representative V_m recording obtained from a medial septal neuron. (c) Hierarchical clustering analysis of the electrophysiological properties of medial septal neurons. Columns represent individual recorded neurons, while rows delineate distinct electrophysiological properties. Each parameter is visualized within a Z-scored pseudocolor map. The right-hand dendrogram facilitates the classification of putative neuronal subtypes. (d) An exemplary trace of the V_m for each neuronal subtype during current injection. The applied current pulse is represented by the lower black waveform. Red color denotes GABAergic neurons, black is used for undetermined neurons, blue for cholinergic neurons, and green for glutamatergic neurons. AP, action potential; CV, coefficient of variation; FSL, first spike latency.

During the recordings, the electrophysiological properties of medial septal neurons were characterized using step current injections. Previous studies have highlighted the heterogeneity of medial septal neurons in terms of gene expression [15,16] and electrophysiological characteristics. Specifically, cholinergic neurons exhibited a slow firing frequency, whereas GABAergic neurons displayed higher firing rates, and glutamatergic neurons showed characteristic burst firing [17–20]. We extracted nine intrinsic electrophysiological parameters from membrane potential responses during current injections, and these parameters were hierarchically clustered (Fig. 1c, details in the Methods section). As a result, we categorized the data into three subtypes: putative cholinergic, glutamatergic, and GABAergic (Fig. 1d). This classification is consistent with previous acute slice patch-clamp studies of medial septal neurons [21].

Depending on their subtypes, medial septal neurons exhibit different firing pattern changes during hippocampal ripples

Previous studies on the medial septum have primarily focused on its contribution to hippocampal theta waves, with the extensive debate concerning the phase synchronization of medial septal GABAergic and cholinergic neurons with theta waves [21,22]. However, few studies have explored the relationship between medial septum neurons and hippocampal ripples. While the previous work revealed changes in the overall medial septal activity before and after hippocampal ripples [23], the specific alterations in the activity of individual medial septal neuronal subtypes in response to ripples remain unclear. To investigate the relationship between hippocampal ripple activity and the medial septum’s firing patterns, we performed in-vivo whole-cell patch-clamp recordings from the medial septum of awake mice while simultaneously recording hippocampal LFP (Fig. 2a). Hippocampal LFP analysis confirmed that ripples, rather than theta waves, were reliably observed in the hippocampus of mice during our recording sessions (Fig. 2b).

Fig. 2.

The temporal dynamics of peri-ripple firing rates across distinct neuronal subtypes within the medial septum. (a) The experimental design for simultaneous patch-clamp electrophysiological recordings from medial septal neurons and LFP acquisition from the hippocampus in vivo in awake mice, and a schematic diagram of the insertion sites for each electrode. (b) Upper panel: an exemplary V_m trace obtained from a medial septal neuron. The second panel from the top: an exemplary LFP recording obtained from the hippocampus. Third panel from the top: the hippocampal LFP, band-pass filtered within the 150–250 Hz frequency range. Lowermost panel: the hippocampal LFP, band-pass filtered within the 4–12 Hz frequency range. (c) Illustrative traces depicting concurrently acquired medial septal V_m and hippocampal LFP are displayed. Red arrows indicate the hippocampal ripples. Upper panel: putative glutamatergic neuron. Middle panel: putative cholinergic neuron. Lowermost panel: putative GABAergic neuron. (d) The Raster plot illustrates neuronal firing patterns, and the corresponding traces depict the temporal modulation of firing rate with hippocampal ripples. The red line delineates the mean peri-ripple firing rate, while the gray line represents the corresponding surrogate data. Statistical significance is denoted by *P < 0.05 (using the specified Chi-square test); data were obtained from n = 2, 3, and 4 neurons originating from two, three, and four mice, respectively. Upper panel: putative glutamatergic neuron. Lower panel: putative cholinergic neuron. HPC, hippocampus; LFP, local field potential.

After detecting hippocampal ripples and clustering the subtypes of recorded neurons (Fig. 2c), we analyzed the firing rates of medial septal neurons before and after ripples for each subtype, except for GABAergic neurons, because of the limited sample size. The firing rate of glutamatergic neurons exhibited a significant decrease after ripples (Fig. 2d: top, P = 0.014, Chi-square test, n = 4 neurons from four mice). In contrast, the results demonstrated no significant difference in the firing rate of cholinergic neurons before and after ripples (Fig. 2d: bottom, P = 0.052, Chi-square test, n = 3 neurons from three mice). Although we were unable to analyze GABAergic neurons because of limited recordings, previous studies reported decreased firing before and after ripples [23], a finding consistent with the tendency observed in our data. This study provides initial observations suggesting a possible reduction in the firing rate of glutamatergic neurons, similar to that reported for cholinergic neurons. These findings suggest that medial septal neurons exhibit distinct, subtype-specific response patterns to ripples.

Medial septal glutamatergic neurons show depolarization preceding ripples

We analyzed membrane potentials, focusing on the activity of putative glutamatergic neurons (Fig. 3a), during hippocampal ripples. Although these putative glutamatergic neurons exhibited a slight decrease in firing during the peri-ripple periods, their membrane potentials simultaneously showed significant depolarization (Fig. 3b: P = 3.89 × 10⁻¹⁷, paired t-test, n = 275 ripples from four mice). This depolarization commenced approximately 100 ms before ripple onset, peaked at ripple onset, and swiftly reverted to resting potential. To ascertain the influence of this antecedent depolarization or hyperpolarization on the ensuing ripple, we examined the correlation between three key ripple parameters – duration, frequency, and power (Fig. 3c and d) – and the magnitude of glutamatergic membrane potential changes immediately preceding the ripple (Fig. 3c: paired t-test, n = 128 ripples from four neurons from four mice; Fig. 3d: paired t-test, n = 147 ripples from four neurons from four mice). The findings demonstrated an inverse correlation between the magnitude of depolarization within glutamate neurons and the amplitude and power of ensuing ripples (Fig. 3c). This inverse correlation was particularly evident for the power of ensuing ripples. Conversely, when glutamatergic neurons showed hyperpolarization, a positive correlation was observed between the magnitude of hyperpolarization and the duration of the subsequent ripples (Fig. 3d). These results indicate that increased excitability of glutamatergic neurons attenuates subsequent ripples, whereas their inhibition shortens ripple duration.

Fig. 3.

Pre-ripple membrane potential shifts in medial septal glutamatergic neurons modulate hippocampal CA1 ripple properties. (a) CA1 LFP trace and four representative V_m traces from medial septal glutamatergic cells, including noticeable pre-ripple depolarizations, are aligned to the ripple onsets. (b) A total of 275 V_m traces in four cells were averaged relative to the ripple onsets. The blue area indicates the 95% CI of V_m traces. (c) The relationship between preripple depolarizations and the duration (left), frequency (middle), and power (right). The line of best fit was determined using the least-squares method from 128 depolarizations across four cells. Significance was determined using a t-test of the correlation coefficient. (d) The same as (c), but for the 147 preripple hyperpolarizations from four cells. CI, confidence interval; HPC, hippocampus; LFP, local field potential.

Discussion

The findings of this study provide novel insights into the electrophysiological dynamics of medial septal neurons in awake mice and their intricate relationship with hippocampal ripples. Utilizing in-vivo whole-cell patch-clamp recordings, we classified medial septal neurons into cholinergic, glutamatergic, and GABAergic subtypes based on their electrophysiological signatures. This classification aligns with studies utilizing acute brain slice preparations [21], thereby reinforcing hierarchical clustering as a robust method for characterizing neuronal heterogeneity within the medial septum. Despite substantial technical challenges (recording success rate: 3.94%; 10 out of 254 trials), our methodological optimizations highlight the demanding nature of these experiments and the need for continued refinement of in-vivo recording techniques.

A key finding of this study is the subtype-specific firing patterns of medial septal neurons during hippocampal ripples. This domain has been relatively underexplored compared with the well-established role of the medial septum in theta oscillations. In the two GABAergic neurons recorded, a statistically significant increase in the firing rate was observed preceding ripple onset (P = 0.009). While this finding must be interpreted with caution because of the small sample size, it raises the possibility that these neurons contribute not only as pacemakers for theta activity but also as facilitators of ripple events via inhibitory control of hippocampal circuits. Furthermore, the observed reduction in glutamatergic neuron firing, along with a previously reported decrease in cholinergic activity before and after ripples [23], implies that these subtypes may exert suppressive or regulatory influences on ripple generation, either directly or indirectly.

While the truncation of distal dendrites may affect dendritic integration and burst generation, the somatic firing frequency and rheobase current are predominantly regulated by the axon initial segment and proximal somatic conductances. Therefore, despite the potential loss of distal dendritic inputs, the recorded intrinsic properties provide a valid representation of the neuron’s proximal excitability profile [24,25].

These divergent response patterns highlight the functional specialization of medial septal neuronal subtypes and suggest a coordinated interaction between excitatory and inhibitory mechanisms in orchestrating hippocampal oscillatory dynamics. However, because of the limited sample size – particularly for cholinergic neurons – these conclusions should be considered preliminary. Future investigations incorporating larger datasets and targeted recordings or post-hoc visualizations based on genetic markers will be essential to elucidate the precise contributions of each neuronal subtype.

Of particular interest are the subthreshold membrane dynamics observed in glutamatergic medial septal neurons, which exhibited a gradual depolarization beginning approximately 100 ms before ripple onset, peaking at ripple initiation, and rapidly returning to baseline. The inverse correlation between the magnitude of this depolarization and the amplitude and power of subsequent ripples suggests that increased excitability of glutamatergic neurons may attenuate ripple intensity. In contrast, a direct correlation between hyperpolarization magnitude and ripple duration indicates that the inhibition of these neurons may also shorten ripple events. These findings imply that glutamatergic neurons in the medial septum modulate ripple characteristics through a dynamic balance of excitation and inhibition, functioning as a regulatory gate. This study demonstrates that glutamatergic neurons, similar to the cholinergic neurons as reported previously [7], are involved in hippocampal ripple activity.

Finally, these electrophysiological observations offer valuable insights into the functional interplay between the medial septum and hippocampus in the context of memory processing. The distinct activity patterns of medial septal neuronal subtypes during ripples suggest that the medial septum contributes to memory consolidation not only through theta synchronization but also via direct modulation of ripple characteristics. In summary, this study delineates the specific responses of medial septal neuronal subtypes to hippocampal ripples and underscores the complex modulatory role of glutamatergic neurons in shaping ripple dynamics. Although constrained by technical limitations and small sample sizes, these findings lay the foundation for future research aimed at uncovering the mechanisms by which the medial septum orchestrates hippocampal oscillatory states and supports memory-related functions. Addressing these questions will be pivotal for advancing our understanding of the neural circuitry underlying learning and memory.

Conclusion

In conclusion, these findings suggest that medial septal glutamatergic neurons dynamically modulate ripple characteristics: increased excitability attenuates ripple amplitude and power, whereas inhibition shortens ripple duration. This indicates that medial septal glutamatergic neurons may act as a regulatory gate, providing novel insights into the medial septum’s direct role in shaping hippocampal ripple dynamics and memory processes.

Acknowledgements

This work was supported by JST ERATO (JPMJER1801), AMED-CREST (24wm0625401h0001, 24wm0625502s0501, 24wm0625207s0101, and 24gm1510002s0104), the Institute for AI and Beyond of the University of Tokyo, and JSPS Grants-in-Aid for Scientific Research (22K21353).

Conflicts of interest

There are no conflicts of interest.