To isolate mouse hippocampal dendritic spines, we adapted the method previously described by Kiebler et al. (1999) to obtain rat mossy fiber hippocampal DS, with several modifications. Briefly, young adult 5–7-week-old C57BL/6 male mice (Charles River, Laboratories Inc., Wilmington, MA, United States) brains were dissected. The brains were washed with ice-cold saline solution HB, containing 300 mM sucrose, 15 mM Na⁺, N-Tris-(hydroxymethyl-2-aminoetahnesulfonic acid, Na⁺-TES), adjusted at pH 7.4, and 1 mM MgSO₄ (Terrian et al., 1988). In this condition, mouse hippocampi were obtained. Single DS were isolated from hippocampi following a method previously reported (Kiebler et al., 1999), with modifications, as outlined in Figure 1A.

Briefly, mouse hippocampi (n = 8) were dissected and immediately washed with HB solution. The isolation procedure was conducted three separate times. Hippocampal tissue was manually homogenized and stored in HB (including Mg^{2 +} to stabilize large membrane structures) using a glass-Teflon Dounce-type homogenizer with a clearance of 0.15 mm. The material was centrifuged at 900g with a Sorvall SS34 rotor for 10 min at 4°C without filtering through nylon filters to remove larger aggregates and blood vessels, as indicated (Kiebler et al., 1999). The resulting pellet was washed, resedimented, and resuspended in 0.25 volume of HB, adjusted to 25% Optiprep (Gibco BRL, Life Technologies GmbH, Eggenstein, Germany). The step gradient was conducted by loading 8 ml of 25% Optiprep/pellet solution in an SW28 ultraclear centrifuge tube containing 9, 12.5, 15, and 25% Optiprep solution in HB (8 ml/each). The tube was spun in an SW28 Beckman ultracentrifuge at 18,000g for 20 min. A typical gradient (Figure 1A) yielded two bands (O1–O3) as well as a pellet (O4). Bands O1–O2 were not distinguishable as in Kiebler et al. (1999) for the rat material, although it is likely that the amount of initial tissue played a role in the thickness of the bands observed. Bands O1–O2 were removed with a pipette and visualized and further tested by immunocytochemistry and electrophysiological techniques. The preparation contained abundant DS, which were individually revealed with DIC optics (Figure 1B) and after phalloidin labeling (Figure 1C). Little contamination was observed from either the plasma membrane or ER, as disclosed by PSD-95 labeling (Figure 2A, Top).

C57Bl mice of 5–7 weeks old were killed by cervical dislocation and decapitated according to IACUC guidelines. Hippocampi were dissected out into ice-cold Ca²⁺-free medium Hibernate A (BrainBits, Springfield, IL, United States), and minced into small pieces. Flushing a few times through a fire-polished Pasteur pipette further dispersed tissue. Undispersed pieces were allowed to settle by gravity for 1 min, while the supernatant containing the dissociated hippocampal neurons was transferred to a new tube, which was centrifuged at 200g for 1 min. The cell pellet was resuspended in NbActive4 medium (BrainBitz, Springfield, IL, United States) and seeded onto poly-L-lysine-coated glass coverslips. Hippocampal cells were incubated at 37°C, in a wet incubator gassed with 5% CO₂, 20% O₂. Hippocampal neurons were kept alive for up to 2 weeks with NbActive4 medium changes every 5 days.

Cultured hippocampal cells were fixed for 15 min in freshly prepared para-formaldehyde (4%) and sucrose (2%) in phosphate buffer saline (PBS). Cells were rinsed (×3) with PBS and blocked with BSA (1%) for 30 min prior to exposure to primary antibodies, which are a mouse anti-PSD-95 (1:400) or a mouse anti-GluR2 (1:400) antibody (Neuromab, Davis, CA, United States). After incubation for 1 h in primary antibodies, cells were incubated in a 1:200 donkey anti-mouse Cy3-conjugated secondary antibody (Santa Cruz Biotechnology, Santa Cruz, CA, United States). The cells were then incubated with a 1:250 dilution of either TRITC or FITC-conjugated phalloidin to stain actin filaments. Isolated DS were stained without fixation, so that they could be subject to electrophysiological analysis. Briefly, isolated DS were seeded on a coverslip, and incubated with 1:400 mouse anti-PSD-95 antibody, 1:200 donkey anti-mouse Cy3-conjugated secondary antibody (Santa Cruz Biotechnology, Santa Cruz, CA, United States) and 1:250 FITC-conjugated phalloidin. After 30 min, isolated DS were attached to coverslip and were washed with PBS to reduce background fluorescence. Images were captured with an inverted Olympus microscope (IX71, Olympus America, Center Valley, PA, United States) connected to a digital CCD camera (C4742-80-12AG, Hamamatsu Photonics, Bridgewater, NJ, United States). Images were collected and analyzed with the IPLab 4.0 (Scanalytics, Vienna, VA, United States) acquisition and analysis software.

For quantitation studies, immunolabeling of the various sample preparations was conducted with two different monoclonal antibodies, namely K28/74 (PSD-95 MAGUK scaffold protein, dilution 1:10), and N327/95 (Glutamate receptor ionotropic, NMDA2A, dilution 1:10) from Developmental Studies Hybridoma Bank, University of Iowa (United States) to confirm the enrichment of DS in our preparation. The labeling was made on the fresh samples by incubation with both primary and secondary antibodies and FITC-phalloidin, from which co-labeling was explored in each case. The criteria for identifying a positive DS were the co-labeling of either PSD-95 and actin or Glutamate receptor and actin, always in combination with a DIC image consistent with a DS by shape and size (see Supplementary Appendix 1, 2). A quantitative assessment of enrichment was conducted as follows (see Supplementary Appendix 2):

Briefly, we marked in an aliquot of the pellet corresponding to step 1 of the centrifugation (Figure 1A) (“whole supernatant”, Supplementary Appendix 2) both actin and PSD-95, and we counted by field the number of times that rounded-shaped vesicles in DIC overlapped with the double marking (points with three positive criteria). We simultaneously counted the number of rounded structures in DIC, the number of phalloidin-labeled dots (green dots), and the number of PSD-95-labeled dots (red dots). Thus, we calculated for each aliquot of the pellet (prior to centrifugation in step 2) the following relationship:

We performed the same procedure for band O2 of step 2 of the centrifugation (Figure 1), (“DS fraction”, Supplementary Appendix 2), which is the band that would contain the fraction enriched in DS, thus obtaining Ratio₂. In each case, measurements were made from 4 aliquots, and the ratios were averaged. The enrichment was calculated as:

In this way, we obtained the following values that were subjected to the t test:

Ratio 1 = 0.0995 ± 0.0398, Ratio 2 = 0.4520 ± 0.0781, Enrichment = 4.59, p = 0.007.

High-resolution electron microscopy with a negative staining protocol of isolated DS was conducted at the Microscopy Core of the Membrane Biology Program at MGH. A sample (10 μl drop) of the original isolated DS preparation was applied to a formvar-coated gold-200 mesh grid (Electron Microscopy Sciences, Hatfield, PA, United States) for one minute. The sample was drawn off and replaced with a drop of 2.0% aqueous phosphotungstic acid (EMS, 10 s), which was drawn off and rinsed with drops of double distilled water and dried. Grids were examined in a JEOL JEM 1011 transmission electron microscope at 80 kV. Images were collected using an AMT (Advanced Microscopy Techniques, Danvers, MA, United States) digital imaging system.

Dendritic spines electrophysiological data were acquired with an Axon Patch 200B amplifier, low pass filtered at 10 kHz, and digitized with 1400A Digidata (Axon Instruments, Union City, CA, United States). The pCLAMP 10.0 software (Axon Instruments) was used to acquire and analyze the data. Patch pipettes were pulled from borosilicate glass (Garner Glass Co, Claremont, CA, United States) with a two-stage Narishige PB-7 vertical puller and then fire-polished to a resistance between 8–15 MΩ using a Narishige MF-9 microforge (Narishige International USA, East Meadow, NY, United States). For the cell-attached patch on DS of cultured hippocampal neurons, the bathing solution contained (in mM): 135 NaCl, 5 KCl, 1.2 CaCl₂ and 10 HEPES (free acid), pH was adjusted to 7.4 with NaOH (cultured cells). For patches of isolated DS, the bathing solution contained (in mM): 135 KCl, 5 NaCl, 1 EGTA, and 10 HEPES, pH was adjusted to 7.4 with NaOH. Patch pipettes were filled with a solution containing (in mM): 135 NaCl, 5 KCl, 1.2 CaCl₂, 10 HEPES (free acid), 100 μM glycine and 10 μM NMDA, pH was adjusted to 7.4 with NaOH. Tight seals were established on DS of cultured hippocampal neurons or isolated DS preparation under ×100 oil lens of the inverted Olympus (IX71) microscope. DS signals were recorded, as indicated, in the presence or absence of 10 μM NMDA and 100 μM glycine to stimulate NMDA receptors. Whenever indicated, the DS-attached pipette contained instead 20–60 μg/ml of the actin filament disrupter cytochalasin D. The DS membrane potentials were recorded under current-clamp mode without current injection. Whenever dual in situ DS-dendrite recordings were conducted, a second patch clamp amplifier (Dagan 3900A) was connected through a second pipette, which was kept in current mode throughout the experiment. The dendrite-attached pipette contained a saline solution with following composition (in mM): 135 KCl, 5 NaCl, 10 HEPES, 1 EGTA, pH 7.4, always in the absence of activating ligands.

Cultured hippocampal neurons were bathed in an extracellular saline solution containing (in mM): 135 NaCl, 5 KCl, 1.2 CaCl₂, 10 HEPES, pH 7.4, and were viewed under objectives of ×60 and ×100. The dendrites were patched using a pipette filled with (in mM): 135 KCl, 5 NaCl, 10 HEPES, 1 EGTA, pH 7.4. The membrane voltages on dendrites were recorded under current clamp using Axopatch 200B amplifier (Axon instruments). At the same time, the DS from the same dendrites were patched using a pipette filled with the extracellular saline solution. In some cases, the saline solution also contained 10 μM NMDA and 100 μM Glycine to activate NMDA receptor, or 20–60 μM cytochalasin D (CD) to modulate the cytoskeletons. The electrical signals in DS were recorded in voltage clamp mode using a Dagan 3900A amplifier. The signal from both dendrites and DS were digitized with a 1440A Digidata (Axon instrument) and analyzed with pClamp10.0 software (Axon instrument).

For the reconstitution of isolated DS in a lipid bilayer system, a lipid mixture was prepared with a content of 70% 1-palmitoyl-2-oleoyl-phosphatidyl-choline and 30% 1-palmitoyl-2-oleoyl phosphatidyl-ethanolamine in n-decane (∼20–25 mg/ml) (Figures 3A,B). Isolated DS were mixed with the above lipid solution at 1:1 ratio followed by brief sonication to form liposomes. The liposomes were painted with a glass rod to the aperture (150 μm diameter) of a polystyrene cuvette (CP13-150) that fits in a lipid bilayer chamber (model BCH-13, Warner Instruments Corp.). The cis side of the lipid bilayer was bathed with 150 mM KCl and 10 mM HEPES (pH 7.4), while the trans side of the bilayer was bathed with 15 mM KCl and 10 mM HEPES (pH 7.4). Channel activity and membrane oscillations were recorded in voltage-clamp mode.

All chemicals and drugs were obtained from Sigma-Aldrich (St. Louis, MO, United States) unless otherwise stated.

Values throughout the manuscript were expressed as the mean ± SEM, where n = number of individual experiments. Statistical significance was achieved with a p < 0.05. Single channel current-to-voltage relationships were fitted with a generalized Goldman-Hodgkin-Katz (GHK) equation of the form:

Where “i” is the type of ion, K⁺ and Cl^–, z is the valence, F is the Faraday constant, R is the ideal gas constant, V (voltage) is the actual holding potential, T is temperature, C_iⁱⁿ and C_i^out are the ionic concentrations on either side of the membrane. In BLM experiments, the inside represents the cis compartment, and the outside is the trans compartment of the hemi-chambers.

Dendritic spines contained abundant actin structures and glutamate receptors, as indicated by phalloidin and anti-GluR2 antibody staining, respectively (Figure 2A) and NMDA2A (Supplementary Appendix 1). This finding also suggested that the isolation procedure primarily damaged the structured cytoskeleton, confirmed by electron microscopy (Figure 2B), where we observed multiple DS, both in dissociated and aggregated forms. It is important to note, however, that DS clustering may be enhanced by high electron density negative staining. DS were often observed as saccular vesicles, either open or closed, with postsynaptic density (PSD), and cleaner lumen (see for comparison Gray, 1959; Blackstad and Kjaerheim, 1961; Peters and Kaiserman-Abramof, 1970; Peters et al., 2008; García-López et al., 2010; Frotscher et al., 2014). Contrary to presynaptic components, which are most commonly axon terminals characterized by synaptic vesicular content that often mingle with mitochondria; the apposing postsynaptic elements in the CNS have features similar to any part of a neuron (Peters et al., 2008). The images in Figure 2 show freshly single DS, which were labeled with TRITC-phalloidin, displaying an abundant cytoskeleton. Isolated DS were of different sizes, and some of them appeared open. Many DS contained PSD (Figure 2C). The enriched DS preparation was assessed without any fixation or further procedures, such that any morphological features may be directly correlated with electrophysiological information. The material amounted to 58 μg protein/ml.

To characterize the electrical properties of DS without interference from soma, dendritic shaft or other DS, we first reconstituted DS in a BLM system (Figure 3A) to characterize single channel currents. This standard technique has been extensively used in the reconstitution of ion channels of various origins, and the basics can be found in classic texts on the subject (see Miller, 1986; Hanke and Schlue, 1993). Briefly, the DS preparation was sonicated with lipids as described in Methods and incorporated in the bilayer membrane (Figure 3B) by painting the preparation with a glass rod. We observed NMDAR single-channel currents in the reconstituted DS, as expected (Figure 3C) in the presence of NMDA (10 mM) and Glycine (100 mM) in the trans chamber. The channels had a single channel conductance of 53.8 ± 5.2 pS (Figure 3D, mean ± SEM, n = 4) and were blocked by 1 mM Zn²⁺ applied through the cis chamber (Figure 3E, n = 3).

We also observed spontaneous activity in the form of electrical oscillations after reconstitution of DS membranes in the BLM reconstitution system (Figures 4A,B, n = 36), which were further explored with membrane-attached patch clamping both in vitro and in situ. The oscillations observed in the BLM system suggested that DS membranes contain the required active electrical components to support the generation of active electrical signals. The Fourier spectrum revealed ion-channel activity (Figure 4A, Right). Although the amplitude of these changes in membrane conductance was mainly unchanged, their frequency was strongly dependent on the holding potential, with almost complete inhibition at –20 mV (Figure 4C). To further explore the nature of these oscillations, we added Ca²⁺ (1 mM) to the bathing solution after the spontaneous generation of the oscillations, to mimic maximal Ca²⁺-channel activation. We observed a dramatic shift in the frequency and “reversal potential” of the spontaneous activity, which almost disappeared at positive, but not negative potentials (Figure 4C). The data suggest the presence of Ca²⁺-dependent oscillatory behavior, which is intrinsic to DS function. This, we believe is the first demonstration of such behavior, and raises the interesting possibility of being the electrical properties of DS not previously revealed.

We further assessed the electrical properties of isolated DS by membrane-attached patch clamping in the presence of NMDA (10 μM) and Glycine (100 μM) applied through the patch pipettes (Figures 5A,B). Voltage oscillations were often observed at diverse frequencies and amplitudes (Figure 5C, n = 12). These electrical oscillations were absent when NMDA was omitted from the pipettes (Figure 5C, “Control”, n = 4), suggesting that NMDAR activation might be required for this phenomenon to occur. This was further supported by the inhibitory effect of Zn²⁺ applied to the bath (1 mM, an NMDAR blocker), which eliminated the electrical oscillations (Figure 5D, n = 4). The DS electrical oscillations were also Ca²⁺-dependent, as an increase in intracellular Ca²⁺ (1–5 mM) activated oscillations in silent DS patches (Figure 5D, n = 7). Conversely, Ca²⁺ depletion by addition of EGTA (5–10 mM) to the bath eliminated the activated oscillations (Figure 5D, n = 4). Fourier analysis of the oscillations disclosed a peak frequency at 0.3 Hz in the presence, but not the absence of Ca²⁺ (Figure 5E).

Spontaneously sealed isolated DS (Figure 6A) were also patched to obtain “whole cell-like” voltage and current clamp information. The “whole-cell” voltage clamp configuration (Figure 6B) allowed the calculation of the isolated DS conductance (Figure 6C), which was only slightly dependent on the holding potential. Analysis of the tail currents (Figure 6D) after subtraction of the linear components showed voltage dependent conductance contributions consistent with both Na⁺ and K⁺ currents. This experimental approach also allowed current clamp manipulations, which elicited action potentials (Figure 6E).

These in vitro studies clearly demonstrated the active properties of isolated DS. To further characterize the electrical properties of DS and to evaluate the possible role of connecting dendrites and soma, we patched DS in situ using cultured adult mouse hippocampal neurons maintained in culture medium for up to two weeks (Figure 7A). DS appear in different shapes and forms based on the relative sizes of the spine head and neck (Peters and Kaiserman-Abramof, 1970), including mushroom spines with a large head and a narrow neck, thin spines with a smaller head and a narrow neck, and stubby spines without constriction between the head and the attachment to the shaft. Neurons were labeled for actin, and PSD-95 to visualize DS, and DAPI counterstaining to label the nucleus (Figure 7B). Patch pipette was approach to DS in situ (Figure 7C) and obtained seals and electrical information.

Elongating dendrites were often observed (Figures 8A,B), where stubby DS were observed to protrude from the dendritic shaft (Figure 8A). DS in cultured neurons were further identified by immunostaining, as revealed by high level expression of PSD-95 and the presence of actin networks (Figure 8A). Actin staining was especially prominent on large spines (Figure 8A, bottom panels). We performed the cell-attached patch clamping on DS from hippocampal neurons in situ (Figure 8B, bottom). Consistent with the findings in isolated DS, in the presence of NMDA (10 μM) and glycine (100 μM) in the pipette, and under current clamping conditions, we observed EPSP-resembling intrinsic oscillations on DS (Figure 8C, n = 11). Interestingly, these oscillations showed diverse amplitudes and frequencies, showing at least three fundamental frequencies around 4, 9, and 13 Hz (Figure 8D).

To study how the electrical signals generated by the DS back propagated along dendrites, an important mechanism for signal integration, we performed double-electrode patch clamping recordings (Figure 9A). One electrode was sealed against the DS and voltage clamp was performed. The second electrode was sealed against the dendrite from which the DS sprouted, which was current clamped. NMDA (10 μM) and glycine (100 μM) were usually included in the DS pipettes, to activate NMDAR and electrical oscillations. As a control, in the absence of NMDA, no oscillations were observed on either dendrites or DS (Figure 9B, n = 18). In contrast, we found NMDA-activated electrical oscillations in both DS and dendrites in the presence of NMDA and glycine (Figure 9C, 6 of 26). The oscillations were coupled between DS and dendrites, suggesting the signal on dendrites are propagated from DS. Similar to findings with other techniques, oscillations in both DS and dendrites showed diverse frequencies and waveforms (Figure 9C), including short periodic (Figure 9C, 1), burst of oscillations (Figure 9C, 2), and oscillations of reverse polarity (Figure 9C, 3). The signaling ratio between DS and dendrites was also of wide range and showed quick transitions between large and small amplifications (Figure 9C), with the underlying mechanism yet to be studied. There were usually very short delays (∼20 μs) between the DS-generated waveforms, and those evoked at the dendritic site (Figure 9D).

The actin cytoskeleton plays an essential role in the DS shape and function (Matus et al., 2000; Rao and Craig, 2000). In fact, it is thought that the actin cytoskeleton is instrumental in spinogenesis (Sekino et al., 2007; Zito et al., 2014). To date, however, information concerning a direct regulation by the actin cytoskeleton of DS ion channel activity and electrical function are unavailable. To determine whether the actin cytoskeleton regulates ion channel activity in DS, we explored the effect of the actin filament disrupter cytochalasin D (CD, 10 μg/ml) both in reconstituted DS with the BLM system, and membrane-attached conditions, in isolated DS. Addition of CD to the cis chamber increased instances of trans NMDA (10 μM) channel activation (Figures 10A,B). In the absence of NMDA, addition of actin (20 μM) and ATP (1 mM) to the cis chamber was sufficient to activate channel currents (Figure 10C, Left) with a current-to-voltage dependence. In the absence of NMDA, the addition of α-actinin (250 nM) to the cis (cytosolic) chamber induced time-dependent ion channel activation (Figure 10C, Middle). To further explore whether the actin cytoskeleton regulates DS ion channel activity, the effect of the actin severing protein gelsolin (10 μg/ml) was also assessed. Addition of gelsolin induced time-dependent channel activation (Figure 10C, Right), suggesting that dynamic changes in cytoskeletal organization help modulating DS electrical activity. The effect of gelsolin also confirms our original observations that the actin filament disrupter cytochalasin D modulates DS electrical activity and channel behavior.

To further evaluate the role of the actin cytoskeleton in the electrical activity of DS, we also performed membrane-attached patch clamping of isolated DS. Similar to experiments in the current clamp mode, the voltage-clamped DS showed intrinsic electrical activity (Figure 11) that was largely diminished (Figures 11A,B, n = 3) when we disrupted the actin cytoskeletal network with CD. This dramatic effect of CD confirmed the importance of the cytoskeleton on the intrinsic oscillatory function of DS, as shown by the spectral density in the Fourier analysis (Figures 11C,D).