Voltage compartmentalization in dendritic spines in vivo

Abstract

Dendritic spines mediate most excitatory neurotransmission in the nervous system, so their function must be critical for the brain. Spines are biochemical compartments but might also electrically modify synaptic potentials. Using two-photon microscopy and a genetically encoded voltage indicator, we measured membrane potentials in spines and dendrites from pyramidal neurons in the somatosensory cortex of mice during spontaneous activity and sensory stimulation. Spines and dendrites were depolarized together during action potentials, but, during subthreshold and resting potentials, spines often experienced different voltages than parent dendrites, even activating independently. Spine voltages remained compartmentalized after two-photon optogenetic activation of individual spine heads. We conclude that spines are elementary voltage compartments. The regulation of voltage compartmentalization could be important for synaptic function and plasticity, dendritic integration, and disease states.

Dendritic spines are small protrusions that cover the dendrites of neurons (1) and mediate most excitatory connections in the brain (2–4). Aside from connecting neurons, they must play an additional role because excitatory axons specifically target spines, avoiding dendritic shafts (5). Spine morphologies, with a small (~1 μm in diameter) head connected to the dendrite by a thin (~140 nm in diameter) neck, suggest that they isolate synapses from the dendrite. Indeed, spines are calcium compartments (6, 7) that biochemically isolate synaptic inputs, enabling input-specific plasticity (8, 9). However, calcium compartmentalization also occurs in aspiny dendrites (10, 11), so spines may implement additional functions. One possibility is that spines are also electrical compartments (12–15). Although dendritic voltages should fully invade the spine, synaptic potentials could attenuate as they propagate from the spine toward the parent dendrite (13, 16, 17). If spines were electrical compartments, the local depolarization at the spine caused by synaptic potentials could be substantially higher than what is measured at the dendrite or soma (18).

The introduction of optical methods has enabled measurements of the membrane potential of spines in vitro using organic dyes or genetically encoded voltage indicators (GEVIs) (19). Although the invasion of action potentials (APs) into spines has been demonstrated (20–23), optical measurements of the electrical responses of spines to synaptic inputs have been inconsistent, with some studies finding similar potentials in spines and dendrites (21) and others finding larger depolarizations in spines (22, 23). Experiments with two-photon photobleaching or glutamate uncaging have been used to estimate the degree of electrical compartmentalization of spines, with a variety of results (24–31). Consistent with voltage compartmentalization, nanopipette recordings from spines in brain slices reveal large amplitude synaptic potentials (32). However, these experiments examined spines in vitro, so the electrical behavior of spines during physiological states remains unexplored.

To investigate the electrical function of spines in vivo, we developed a GEVI that could be efficiently excited with two-photon illumination and expressed it in pyramidal neurons from layer 2/3 mouse somatosensory cortex. This sensor, postASAP (short for “postsynaptic ASAP”), used an ASAP (accelerated sensor of APs) GEVI backbone (33), modified with mutations for enhanced sensitivity and a PSD95.FingR nanobody domain to enrich its expression in spines (34, 35) (Fig. 1A; see supplementary text). To express postASAP in vivo, we used in utero electroporation to achieve sparse, yet robust, expression and measured the coexpression of postASAP and red fluorescent protein (RFP) (Fig. 1B and fig. S2A). The soluble RFP showed higher fluorescence in larger volume compartments [fluorescence ratio soma/spine = 3.02 ± 1.89, soma/dendrite = 2.09 ± 1.84, dendrite/spine = 1.45 ± 1.21 (mean ± SD); p < 0.0001, Kruskal-Wallis test; fig. S2B]. Meanwhile, postASAP fluorescence levels were evenly distributed among neuronal compartments [fluorescence ratio soma/spine = 0.98 ± 0.59, soma/dendrite = 1.04 ± 0.58, dendrite/spine = 0.95 ± 0.62 (mean ± SD); p = 0.40, Kruskal-Wallis test], indicating membrane targeting, with robust labeling of dendrites and spines. To investigate the biophysical properties of postASAP, we first analyzed the fluorescence-voltage (F-V) relation with two-photon excitation in cultured ND7/23 cells (Fig. 1C). F-V curves were fitted with a Boltzmann function and displayed linearity in the range of −100 and −40 mV. To calibrate postASAP signals in vivo, we used whole-cell patch-clamp recordings and, simultaneously, two-photon imaging of postASAP in neuronal somata (Fig. 1D), finding a clear correspondence between electrical activity and fluorescence (Fig. 1E). Changes in fluorescence correlated with subthreshold depolarizations, rather than with APs (Fig. 1, E and F; see supplementary text), as expected from the low-pass behavior of the voltage sensor and imaging rate (~16 ms per frame). Within subthreshold depolarizations, the relation between voltage and changes in fluorescence was fitted by a linear function with a slope of 1.71 ± 0.02 mV per percent −ΔF/F [mV/(−ΔF/F%)] (slope ± SEM, p < 0.0001) (Fig. 1G), similar to the linear range of the F-V curve in cultured cells [1.67 mV/(−ΔF/F%)].

Fig. 1. Characterization of postASAP.

(A) Construction of postASAP. D, Asp; G, Gly; H, His; L, Leu; N, Asn; R, Arg; S, Ser; T, Thr. (B) Two-photon imaging of neurons expressing RFP-p2a-postASAP. Somatic expression (scale bar, 15 μm) is shown on the left. Dendritic expression (scale bar, 3 μm) is shown on the right. (C) Sensitivity of postASAP in ND7/23 cells to 500-ms voltage steps (mean ± SD; n =5 cells). The red area corresponds to a linear range: y = 0.6x, coefficient of determination (R²) = 0.9973, and p = 0.0014. (D) Experimental design. (E) On the left, a patched neuron expressing postASAP is shown. The red lines show the pipette outline, and the white dotted square shows the region of interest (ROI) for fluorescence measurement (scale bar, 5 μm). On the right, a representative optical trace of postASAP (light green, raw fluorescence; black, 10-Hz low-pass filtered) and electrical recording (red) are shown. (F) Average somatic electrical subthreshold signals (red, six events) and simultaneous fluorescence changes (black and gray, mean ± SD). (G) Correlation of peak postASAP fluorescent changes and subthreshold electrical amplitude [mean ± SD; n = 317 subthreshold events, 14 cells, and 8 animals; linear regression (red): y = 0.58x, confidence interval = 0.55 to 0.61, R² = 0.7156, and p = 0.0003].

We then used two-photon imaging in vivo to measure the voltage dynamics experienced by basal dendrites and their spines from neurons expressing postASAP while performing simultaneous somatic whole-cell recordings (Fig. 2A; see supplementary text). During spontaneous activity, we found three spatiotemporal patterns of depolarizations in dendritic trees (Fig. 2B and movie S1). In the first spatial pattern (“AP”), which occurred during trains of APs, dendrites and spines were synchronously depolarized (Fig. 2, B and C). Peak depolarizations during APs were similar in spines and adjacent dendrites, confirming the functional expression of postASAP in spines and AP invasion into spines without failures or decrement (6, 7, 20, 36, 37) [Fig. 2, D and E; spine = 20.0 ± 9.5 mV, dendrite = 21.3 ± 9.9 mV (mean ± SD); p = 0.24, Mann-Whitney test]. In the absence of APs, during subthreshold potentials or even in the absence of pronounced somatic depolarization in the whole-cell recordings, two other spatial types of depolarization were observed (Fig. 2, B and C; fig. S8; and supplementary text). In a second spatial pattern (“Dendrite+Spines”), localized segments of dendrites with associated spines were depolarized together [Fig. 2F; spine = 15.9 ± 5.8 mV, dendrite = 15.3 ± 5.6 mV (mean ± SD); p = 0.3767, Mann-Whitney test]. Interestingly, we also observed a third pattern (“Spine-only”), where individual spines became depolarized independently, or with a reduced depolarization of the parent dendrite, presumably reflecting isolated synaptic events [Fig. 2G; spine = 15.0 ± 7.2 mV, dendrite = 6.97 ± 4.2 mV (mean ± SD); p < 0.0001, Mann-Whitney test]. Spine voltage peaks reached by Spine-only events ranged from 5.0 to 39.8 mV, although this might represent an overestimate owing to the difficulty of measuring smaller depolarizations with sufficient signal-to-noise ratio (fig. S4A and supplementary text).

Fig. 2. Spine and dendritic voltage dynamics in vivo during spontaneous activity.

(A) In vivo two-photon imaging and somatic whole-cell recording of a neuron expressing postASAP is shown at the top (red lines, pipette outline). Imaged dendrites (43 μm from center of the image to cell body) of a patched cell are shown at the bottom (scale bars, 5 μm). (B) A somatic electrical recording of the neuron in (A) is shown at the top. AP, train of APs; Sub, subthreshold depolarization; RMP, resting membrane potential. Simultaneous fluorescence changes of numbered spines and adjacent dendrites in (A) are shown at the bottom. (C) Representative image with peak fluorescence changes in dendrites and spines during three conditions in (B). (D) Depolarization during APs, generated by three 100-ms current pulses (300 pA). Somatic imaging and electrophysiological recording are shown at the top (scale bar, 5 μm). Representative fluorescence changes in a dendrite are shown at the bottom [average three trials; spine at 48 μm from cell body; scale bar, 5 μm; color scale same as (B)]. (E) Peak spine and dendrite fluorescence changes during AP trains (n = 125 spines, 37 dendrite segments, 5 cells, and 4 animals; linear regression: y = 0.93x, R² = 0.823, and p < 0.0001). Box and whiskers represent median (line), 25th to 75th percentiles (box), range (whiskers), and mean as a “+.” (F) Examples of Dendrite+Spines patterns are shown on the left (average 10 events; scale bar, 5 μm). Peak fluorescence changes in spine heads and adjacent dendrites are shown on the right (n = 221 spines, 90 dendritic segments, 13 cells, and 7 animals). n.s., not significant. (G) Same as (F) for Spine-only pattern (n = 116 spines, 90 dendritic segments, 13 cells, and 7 animals; scale bar, 5 μm). ****p < 0.0001.

Next, we wondered what spatial patterns of depolarizations were found after sensory stimulation. We used air puffs to activate the whiskers while simultaneously imaging postASAP in dendrites and spines (Fig. 3A; see supplementary text) and also found AP (fig. S7, C and D, and movie S2), Dendrite+Spines, and Spine-only patterns (Fig. 3, B and C). After sensory stimulation, Dendrite+Spines events [Fig. 3D; spine = 13.7 ± 5.5 mV, dendrite = 13.0 ± 5.4 mV (mean ± SD); p = 0.1299, Mann-Whitney test] were detected more frequently, whereas Spine-only events did not change significantly (fig. S8 and supplementary text). Spine-only patterns [Fig. 3E; spine = 12.8 ± 5.4 mV, dendrite = 7.5 ± 4.5 mV (mean ± SD); p < 0.0001, Mann-Whitney test] had similar peak amplitudes to those during spontaneous activity. All dendritic and spine depolarizations were reversibly blocked by synaptic antagonists (fig. S9 and supplementary text), confirming their physiological nature.

Fig. 3. Spine and dendritic voltage dynamics in vivo after sensory stimulation.

(A) Experimental design. (B) Representative image of dendrites and spines expressing postASAP (scale bar, 5 μm). (C) Simultaneous fluorescence changes of numbered spines and adjacent dendrites in (B). (D) Example image with peak fluorescence changes in dendrites and spines in the time point indicated in (C) is shown on the left. Peak fluorescence changes in spine heads and adjacent dendrites for the Dendrite+Spines pattern are shown on the right (n = 255 spines, 49 dendrite segments, and 5 animals). (E) Same as (D) for Spine-only events (n = 181 spines, 49 dendrites, and 5 animals). ****p < 0.0001.

We then turned our attention to the Spine-only patterns, where spines activated independently. To explore this, we activated spine heads or dendritic segments using two-photon optogenetics (38) with ChrimsonR (39) (Fig. 4A; see supplementary text) while simultaneously measuring their voltages with postASAP (Fig. 4B). For calibration, we imaged postASAP in the soma during two-photon optogenetic activation (Fig. 4C) and found a gradual fluorescence response, proportional to laser power (fig. S12A). In vivo patch-clamp recordings indicated reliable optogenetic responses and allowed us to measure the currents generated by the photostimulation (Fig. 4D). We then photostimulated small segments of dendrites while measuring postASAP fluorescence in dendrites and spines and found similar depolarizations in dendrites and adjacent spines [Fig. 4E; spine = 8.1 ± 4.2 mV, dendrite = 7.6 ± 4.0 mV (mean ± SD); p = 0.55, Mann-Whitney test]. Finally, we photostimulated spine heads to mimic synaptic potentials and found depolarizations that decreased as they spread into the parent dendrites [Fig. 4F; spine = 11.2 ± 3.6 mV, dendrite = 6.0 ± 2.2 mV (mean ± SD); p < 0.0001, Mann-Whitney test]. These optogenetic results were in agreement with our measurements during spontaneous or sensory-evoked activity (Figs. 2 and 3), confirming voltage attenuation from spine to dendrite but no attenuation from dendrite to spine. This asymmetric electrical behavior is expected from cable properties (13, 16).

Fig. 4. Two-photon optogenetics and voltage imaging in vivo.

(A) Experimental design. (B) Construct and representative postASAP fluorescence changes in soma (light green, raw fluorescence; black, 10-Hz low-pass filtered) during 500-ms stimulation trials (red, 100 mW power). (C) Representative soma (top) and peak fluorescence response (bottom) during stimulation trials (×10, 500 ms, 100 mW). The dotted circle shows the stimulation area (scale bar, 10 μm). (D) Representative in vivo voltage-clamp recordings during optogenetic stimulation of proximal dendrites (100 mW, 100 ms) are shown on the left. Peak currents are shown on the right; −22.7 ± 11.3 pA (mean ± SD), 10 trials (n = 7 cells and 4 animals). (E) Representative peak fluorescence changes during optogenetic activation of dendritic shafts are shown on the left [stimulation ROIs are indicated by white dotted circles; 10 trials, 100 ms, 100 mW; color bar same as (C); scale bar, 5 μm]. Peak fluorescence changes in stimulated dendritic shaft (dendrite stim), adjacent dendritic spine (spine), and unstimulated dendritic shaft (dendrite no stim) are shown on the right (n = 34 dendrites and 9 animals). (F) Same as (E) during optogenetic activation of spines (n = 35 spines and 12 animals; scale bar, 5 μm). ****p < 0.0001. (G) Simplified electrical model. (H) Resistance ratio (R_n to R_den) of Spine-only events during spontaneous (Spon, n = 116), sensory stimuli (Sen, n = 181), and optogenetic spine stimulation (Opto, n = 35). (I) Values of spine neck resistance for stimulated spines; median = 213.7 megohms (n = 35 spines). In (D), (H), and (I), boxes and whiskers represent median (line), 25th to 75th percentiles (box), range (whiskers), and mean as a “+.”

To estimate the electrical properties of the spines, we used a simplified steady-state electrical equivalent circuit (30, 40) (Fig. 4G), with a very small (<0.01 pF) spine capacitance, so that synaptic currents flow through the spine neck resistance (R_n). By defining the spine input resistance (R_sp) as the sum of R_n and the dendritic input resistance (R_den), we obtain

$$R_{\text{sp}}=R_{\text{n}}+R_{\text{den}}$$ (1)

According to Ohm’s law, R_sp depends on the current flowing across the synapse (I_syn) and the voltage in the spine (V_sp):

$$R_{\text{sp}}=\frac{V_{\text{sp}}}{I_{\text{syn}}}$$ (2)

Applying the voltage divider equation, one finds

$$\frac{R_{\text{n}}}{R_{\text{den}}}=\frac{V_{\text{sp}}}{V_{\text{den}}}-1$$ (3)

Because we obtained a similar average resistance ratio (R_n/R_den) for Spine-only events during spontaneous activity and sensory stimuli, and also for optogenetically stimulated spines [Fig. 4H; 1.7 ± 2.8, 2.1 ± 5.9, and 1.1 ± 0.8, respectively (mean ± SD); p = 0.34, Kruskal-Wallis test], R_n would be equivalent to R_den for most spines (29). To estimate R_n, we solved Eqs. 1 to 3, obtaining

$$R_{\text{n}}=\frac{V_{\text{sp}}-V_{\text{den}}}{I_{\text{syn}}}$$ (4)

and assuming an I_syn = 22.7 pA for photostimulated spines (Fig. 4D), we obtained an average R_n of 226.6 ± 128.8 megohm (mean ± SD), ranging from ~0 to 530.8 megohm (Fig. 4I; see supplementary text), in line with previous estimates with different methods (29, 30, 32).

Our results demonstrate that measurements of spine depolarization in vivo are feasible with two-photon GEVI imaging, providing an initial glimpse into the rich spatiotemporal patterns of depolarizations of dendrites and spines during spontaneous activity or sensory stimulation. Consistent with previous reports (6, 7, 20–22), we found synchronous depolarization of dendrites and spines during back-propagation of axonal APs into dendrites and spines (AP pattern). In addition, we detected local and transient depolarizations of dendritic segments and their spines (Dendrite+Spines pattern), likely corresponding to a combinatory of synaptic potentials and dendritic spikes (41–44). Importantly, we also found individual spines activating independently (Spine-only pattern), even in the absence of dendritic or somatic activity. These spine-independent depolarizations likely represent individual synaptic potentials, owing to their occurrence during subthreshold synaptic inputs, their spatial restriction, and their sensitivity to synaptic blockers (6, 7, 45, 46). These synaptic depolarizations of spines can be large in amplitude, at least of tens of millivolts (16, 32, 47–49). The presence of spine-independent depolarization during spontaneous and sensory-evoked activity implies that spines can compartmentalize voltage in physiological states in vivo. Indeed, using two-photon optogenetic activation of individual spines, we demonstrated this directly while also revealing that dendritic potentials propagated into spines faithfully. Thus, our data indicate that the spines act asymmetrically, with no attenuation of dendritic potentials or AP but significant attenuation of synaptic potentials (13, 16), maintaining a voltage gradient with the dendrite (50, 51). The mechanisms underlying spine voltage compartmentalization could involve geometrical and structural factors, passive cable properties, voltage-sensitive ion channels, or synaptic receptors (18). Regardless of the mechanism, our results show that dendritic spines, in addition to serving as biochemical compartments, are also elementary electrical compartments for synapses. The regulation of the voltage compartmentalization of synaptic inputs by spines could be important for synaptic function (52), synaptic plasticity (8, 53), and dendritic integration (54) and be affected in mental and neurological diseases (55).

Data and materials availability:

All data are available in the manuscript and the supplementary materials.