Multibranch activity in basal and tuft dendrites during firing of layer 5 cortical neurons in vivo

Abstract

Layer 5 pyramidal neurons process information from multiple cortical layers to provide a major output of cortex. Because of technical limitations it has remained unclear how these cells integrate widespread synaptic inputs located in distantly separated basal and tuft dendrites. Here, we obtained in vivo two-photon calcium imaging recordings from the entire dendritic field of layer 5 motor cortex neurons. We demonstrate that during subthreshold activity, basal and tuft dendrites exhibit spatially localized, small-amplitude calcium transients reflecting afferent synaptic inputs. During action potential firing, calcium signals in basal dendrites are linearly related to spike activity, whereas calcium signals in the tuft occur unreliably. However, in both dendritic compartments, spike-associated calcium signals were uniformly distributed throughout all branches. Thus, our data support a model of widespread, multibranch integration with a direct impact by basal dendrites and only a partial contribution on output signaling by the tuft.

Pyramidal neurons feature extensive dendritic arborizations that receive and process widespread synaptic input. Much interest has been placed on the role of individual dendritic branches in determining neuronal output (1). The single branch has been hypothesized to function as a unit of plasticity (2), protein synthesis (3), and synaptic integration (4). Moreover, single dendritic branches have been shown to generate spatially restricted regenerative events, the so-called dendritic spikes, which can process and amplify local input (5). These features raise the question of whether the input–output relation of a cortical neuron is influenced primarily by single dendrites or by multibranch activity.

Layer 5 (L5) pyramidal neurons serve as the major output cell type of neocortex. Their morphology is characterized by a set of basal dendrites as well as a set of distal tuft dendrites that extend into layer 1 and are separated from the soma by a long apical trunk (6). These compartments are known to be specialized in terms of the type of synaptic input that they receive (7). Furthermore, in vitro studies of dendritic physiology have revealed several active dendritic signals in these compartments such as back-propagation of action potentials (bAPs) (8, 9), NMDA spikes resulting from clustered synaptic activation (5), and calcium spikes initiated in the apical trunk (10, 11). Despite the presence of these dendritic specializations, little is known about the importance of these features for in vivo function. Previous in vivo studies were limited by the technical challenges of recording from deep basal structures in L5 (12) or of recording multiple individual dendrites simultaneously from the same neuron (13). A complete understanding of dendritic integration in L5 pyramidal neurons requires in vivo data from both the basal and tuft compartments with single dendritic branch resolution.

Here we address the question of multibranch activity in L5 pyramidal neurons in vivo. We focus on the whisker region of primary motor cortex (wM1), which lacks a granular layer and thus brings L5 to a depth superficial enough for two-photon imaging (14). We use simultaneous dendritic calcium imaging and whole-cell somatic electrical recordings (15–17) to compare dendritic activity to neuronal output. The questions we address are: (i) What is the pattern of calcium activity in single dendrites during subthreshold activity and AP firing? (ii) How uniform is the calcium signal between different dendritic branches of the same neuron? (iii) How do these characteristics differ between the basal and tuft dendritic compartments?

Results

We examined L5 pyramidal neurons of wM1 in anesthetized adult mice. Cell somata were located between 400 and 650 μm below the pial surface (n = 25). Electrodes containing Oregon Green 488 1,2-bis(o-aminophenoxy)ethane-N,N,N,N-tetraacetic acid (BAPTA–1; OGB-1) were used to fill neurons with calcium indicator and to record the membrane potential in the whole-cell configuration. Spontaneous calcium activity was monitored in individual dendritic branches at multiple depths and compared with the concurrently recorded somatic membrane potential and AP activity. Due to the relatively superficial location of L5 in mouse motor cortex, we were able to image tuft, apical trunk, and basal dendrites in the same neuron (altogether 43 tuft dendrites, 35 apical trunks, and 23 basal dendrites).

Dendritic Calcium Signaling During Subthreshold Activity.

We first examine calcium transients observed in the absence of AP firing. Under isoflurane anesthesia, the subthreshold electrical activity of a L5 pyramidal neuron exhibits frequent transitions between up- and down-states (Fig. 1A). Although down-states are characterized by the quiescence of fluctuations in membrane potential, up-states are characterized by barrages of synaptic input. Calcium transients were observed in all dendritic compartments during subthreshold up-state activity, although never during down-state activity (Fig. 1B). The amplitudes of these events were similar in each dendritic compartment (Fig. 1C, mean subthreshold calcium transient amplitude ± SE, 13.4 ± 0.5% ΔF/F in tuft, 12.7 ± 1.8% ΔF/F in apical trunk, 13.6 ± 0.6% ΔF/F in basal dendrites), and the differences in mean amplitude were not statistically significant [P > 0.05, Kolmogorov-Smirnov (K-S) test applied pair-wise to all compartments]. As these amplitudes are near the threshold for event detection (typically ΔF/F ∼ 5%), it is possible that smaller amplitude calcium transients went undetected.

Fig. 1.

Small-amplitude subthreshold calcium signals are observed in all compartments. (A) Example data from a spontaneously active tuft dendrite of a L5 pyramidal neuron filled with 100 μM OGB-1 via a whole-cell recording pipette. Upper trace is the calcium fluorescence signal recorded at 40 frames per second. Lower trace is the simultaneously recorded membrane potential. Horizontal lines indicate baseline fluorescence for the calcium signal and mean down-state voltage for membrane potential. Gray boxes indicate epochs where up-states and significant calcium transients occur concurrently. Pink boxes indicate epochs where up-states occurred without a calcium transient. (B) Two-photon images and example data from different compartments of the same L5 pyramidal neuron as in A. (Left) Max projection of a Z-stack along the XZ plane obtained at the end of the experiment. Soma was located at ∼450 μm below the cortical surface. (Center) Imaging planes from example dendrites from the tuft (Top), apical trunk (Middle), and basal dendrites (Bottom). Solid lines indicate the ROI that was processed to form a calcium signal. Dashed red lines are placed at imaging depth for the indicated dendrite. At right of each imaging plane is an example calcium transient and the concurrently recorded membrane potential. (C) Histograms of the amplitude of subthreshold calcium transients for all dendritic branches recorded from the labeled compartment across all experiments (n = 43 apical tuft dendrites, 35 apical trunks, and 23 basal dendrites). Bin width is 2.5% ΔF/F.

To explore the spatial distribution of subthreshold calcium transients, we divided each region of interest (ROI) of a dendritic branch into 10 μm segments and compared the calcium signals within each segment. We found that the spatial and temporal profile of subthreshold events within a dendritic branch was heterogeneous. Typically a subthreshold calcium transient observed in a whole branch was composed of calcium transients in a discontinuous subset of its segments (Fig. 2 A and B). In 13% of subthreshold calcium events, we observed calcium transients in a length of dendrite encompassing at least three contiguous 10 μm ROIs. The amplitudes of these contiguous events were indistinguishable from those of more distributed events (P > 0.05 K-S test; Fig. 2C). Furthermore, although most calcium events were time-locked to the onset of the up-states (Fig. S1), a fraction of 25% (n = 228 events total from 22 dendrites) occurred at variable delays greater than 500 ms after the up-state’s onset, in line with recent single spine observations in the auditory cortex (18). These observations are consistent with distributed subthreshold events arising from the activation of single synapses, whereas contiguous events arise from multiple neighboring synapses. Furthermore, we found that dendritic calcium events cease in the presence of blockers of synaptic activation, specifically the NMDA-receptor antagonist D-(-)-2-amino-5-phosphonopentanoic acid (APV; 500 μM) and the AMPA-receptor antagonist 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide disodium salt (NBQX; 100 μM) (n = 8 cells; example shown in Fig. 2D).

Fig. 2.

Subthreshold calcium signals are not uniform within a dendritic branch. (A) Example local calcium signals in tuft dendritic branch from a L5 pyramidal cell. Soma was located at 520 μm below the pia. At left is a two-photon fluorescence image of a tuft dendritic branch. Colored lines show ROIs segmented at 10 μm intervals along the length of the branch. At right are traces for four example subthreshold events. Colored traces are calcium signals recorded from the ROI of the same color as in the fluorescence image. Dashed circles indicate the largest calcium transient associated with each subthreshold event. Rightmost example shows a contiguous subthreshold calcium event. Black trace is the membrane potential recorded in the whole-cell patch configuration. Bottom row shows amplitude for each calcium trace as given by the height of a bar of the same color. (B) Same as A but for a basal dendritic branch from a L5 pyramidal cell with soma located at 480 μm below the pia. A vertical line was placed in one example to facilitate comparison of differences in timing of signals in different ROIs. An arrow indicates a transient that begins later than in other ROIs from the same branch. (C) Histogram of the amplitude of categorized subthreshold calcium transients. Events were defined as contiguous if there was significant concurrent activity in at least three contiguous 10 μm ROIs, and otherwise were classified as distributed. Data were combined from different types of branches (n = 2 tuft dendrites, 1 oblique dendrite, and 4 basal dendrites). (D) Example traces from an experiment where 100 μM NBQX and 500 μM APV were injected below the cortical surface by applying positive pressure through a micropipette.

Spine Calcium Activity in Tuft and Basal Dendrites During Subthreshold Activity.

If subthreshold calcium transients arise from the activation of individual synapses, we expect spines to show localized calcium transients during subthreshold activity. Therefore, we imaged the calcium activity of individual spines in a separate set of experiments (n = 7 cells with 66 tuft spines and 35 basal spines). We observed that calcium transients in the spines of tuft and basal dendrites would at times either directly match the calcium activity in the branch or act independently from it (Fig. S2). These isolated spine calcium transients were similar between the tuft and basal dendrites in terms of amplitude (mean ± SD relative to adjacent shaft, 34 ± 16% ΔF/F for tuft and 32 ± 14% ΔF/F for basal; Fig. S2C) and frequency of activity (mean frequency ± SD, 2.3 ± 2.1 events per minute in tuft spines and 3.6 ± 2.3 events per minute in basal spines; Fig. S2D). Importantly, the amplitude of the shaft signal during isolated spine activity (mean amplitude ± SE, 14.5 ± 1.3% ΔF/F) was comparable to the shaft signal associated with subthreshold transients in whole-cell experiments (mean amplitude ± SE, 13.4 ± 0.8% ΔF/F). This indicates that spine calcium transients may underlie subthreshold dendritic calcium activity.

Relation Between Evoked APs and Dendritic Calcium Activity.

We now turn our attention to the relation between AP activity and calcium signals in tuft and basal dendrites. As a first approach, we used current injection in the whole-cell configuration to evoke APs from the soma (500–900 pA lasting 50 ms; Fig. S3). We found that bAPs largely failed to evoke calcium activity in the tuft during the down-state (20% success rate, n = 8 tuft dendrites; Fig. S3 B and D), but that significant responses were often seen during up-state current injections (56% success rate). These up-state responses had small amplitudes (13 ± 2.5% ΔF/F), and their relation to evoked APs was ambiguous due to the possibility of ongoing subthreshold calcium activity. These data are consistent with the notion that synaptic input is required for APs to affect tuft calcium activity, as demonstrated previously in vitro (11). In contrast, calcium responses in the basal dendrites and apical trunks were insensitive to cortical state. Nearly every stimulation elicited a large calcium transient (99%, n = 11 dendrites). We also observed that the amplitude of the calcium signal increased linearly with the number of evoked APs (Fig. S3C). These data suggest that suprathreshold calcium responses in the basal and apical trunk dendrites are a direct result of bAP invasion. We further investigated the relation between spike activity and calcium by examining spontaneously occurring APs.

Relation Between Spontaneous APs and Dendritic Calcium Activity.

Calcium transients were time-locked to spontaneous spike activity in all dendritic compartments, although with less reliability and a somewhat smaller amplitude in tuft dendrites (Fig. 3, average amplitudes ΔF/F ± SE are 18.7 ± 0.3% in tuft, 24.5 ± 0.7% in apical trunk, 26.9 ± 0.4% in basal dendrite; n = 1,055, 672, and 605 calcium transients, respectively). In general, suprathreshold calcium responses were larger than those associated with subthreshold activity (Fig. 3 E and F) and more frequent (percent of detected calcium transients accompanied by APs: 83% in tuft, 97% in apical trunk, and 91% in basal dendrites). In addition, calcium transients during AP activity were typically present throughout the dendritic branch (Fig. S4), which is in contrast to what we observed during subthreshold activity (Fig. 2). For dendrites of 30 μm length or more, we examined the calcium signal of every 10 μm segment when the overall branch showed a significant AP-associated calcium transient. Calcium transients were observed in all segments for 89% of events in the tuft (n = 345 events) and 95% of events in the basal dendrites (n = 320 events). These data suggest that AP-associated calcium transients in the tuft and basal dendrites are propagated along the entire dendritic branch.

Fig. 3.

Relation between AP activity and dendritic calcium signaling. (A) Two-photon images of a spontaneously firing L5 pyramidal neuron filled with 100 μM OGB-1 via a whole-cell recording pipette. The soma was located ∼550 μm below the pia. Shown are imaging planes from example dendrites from the tuft (A, a), apical trunk (A, c), and basal dendrites (A, d). Solid lines indicate the ROI that was processed to form a calcium signal. Also shown is a max projection of a Z-stack along the XZ plane obtained at the end of the experiment (A, b). Dashed red lines indicate the imaging depth for imaging planes A, a; A, c; and A, d. (B) Example data from a spontaneously active dendrite indicated in A, a. Upper trace is calcium signal recorded at 40 frames per second. Lower trace is the simultaneously recorded membrane potential. Gray boxes indicate epochs where calcium activity and APs occur concurrently. Pink boxes indicate epochs where APs fired without calcium activity. (C) Same as B but using data from apical trunk shown in A, c. (D) Same as B but using data from basal dendrite shown in A, d. (E) Example data comparing the amplitude of calcium transients associated with subthreshold and superthreshold electrical activity. (Left) A two-photon fluorescence image of a basal dendrite from a L5 pyramidal cell with soma located at 520 μm below the pia. (Right) Two calcium traces above the concurrently recorded membrane potential. Example at left shows spike activity and at right shows subthreshold activity. The number of APs in the burst is indicated. Conventions are otherwise same as in A–C. (F) Scatterplot of the largest calcium transient associated with suprathreshold (empty squares) and subthreshold (filled circles) activity for all dendrites recorded in our study (n = 101 branches). Relative sparseness of subthreshold data points is due to exclusion of dendritic branches that lacked any detectable calcium transients in nonspiking intervals.

Tuft and basal dendrites showed clear qualitative differences in the reliability of their signals with respect to AP activity (cf. Fig. 3 B–D). Calcium transients in the basal dendrites and apical trunk showed a one-to-one linear relationship with number of APs fired (Fig. 4 B and C). The amplitude of the transients scaled with the size of the burst, and nearly every burst of APs was accompanied by a calcium transient (Fig. 4D). In contrast, the amplitude of AP-associated calcium transients in the tuft did not show a clear relationship with the number of APs in a burst (Fig. 4A). Tuft dendrites also showed more failure events where bursts of APs did not lead to a detectable calcium transient (Fig. 4 A and D). To quantify these observations, we calculated the linear correlation coefficient between the amplitude of a calcium transient and the number of APs that were fired simultaneously by the cell (Fig. 4E). The mean correlation observed in the basal dendrites was 0.84 ± 0.02 SE, indicating that most of the variability in calcium amplitude was due to the number of APs in a burst. The value in the apical trunk was similar (0.83 ± 0.03 SE). These data are consistent with AP-associated activity in the basal dendrites and the apical trunk being dominated by linearly summed bAPs. However, the tuft dendrites showed a significantly smaller mean correlation of calcium amplitude and AP count (0.51 ± 0.04 SE, P < 0.001 K-S test; Fig. 4E). This smaller correlation value for the tuft arose from two sources. First, 26% of spike bursts did not yield any detectable calcium transient in the tuft compared with 3% for the basal dendrites (P < 0.0001, binomial test; Fig. 4D). Second, even when a calcium transient did occur, the relation of its amplitude to AP count was weak (Fig. 4A).

Fig. 4.

Basal and trunk dendrites but not tuft dendrites show strong linear correlation between AP firing and calcium activity. (A) Analysis of relation between spike activity and calcium activity in tuft dendrites. (A, a) Examples of spontaneous calcium activity in a single tuft dendrite when different numbers of spikes are fired. Upper trace is calcium signal, and lower trace is concurrently recorded membrane potential. The number of spikes fired in a burst is indicated at right in red. (A, b) Scatterplot of amplitude of calcium transients as a function of the number of spikes fired for a single dendritic branch. Spiking that was unaccompanied by a calcium transient is not included in this analysis. Red line is a best linear fit to the data. (A, c) Best linear fit of calcium amplitude versus number of spikes for all analyzed tuft branches. Gray lines show data from individual tuft dendrites. Dendrites that had less than 10 calcium transients or burst events were excluded from analysis. Red line indicates best linear fit from averaging all data. (B) Same as A but for data recorded from apical trunks. (C) Same as A but for data recorded from basal dendrites. (D) Bar graph of percent of spike events that were accompanied by a detectable calcium transient for different classes of dendrites. Black vertical lines indicate 95% confidence interval. (E) Scatterplot of correlation coefficient between number of spikes fired and amplitude of calcium signal as a function of the depth of the dendritic branch below the pia. Color of data point indicates class of dendrite recorded. All spike events were included in this analysis regardless of whether there was a detectable calcium transient in the imaged dendrite.

Multibranch Calcium Activity and Spontaneous APs.

To investigate the contribution of individual dendrites to the generation of AP activity, we next examined the uniformity of calcium activity between multiple dendritic branches. We examined fields of view that contained multiple dendrites and compared the amplitude of simultaneously recorded calcium transients on different branches. We found that for both the tuft and basal dendrites, calcium transients occur synchronously throughout branches in the field of view (Fig. 5 A and C). In addition, the amplitude of the calcium transients tracked each other so that the absolute difference in amplitude between transients on different branches was small (mean of 2.6% ΔF/F for tuft dendrites and 2.3% ΔF/F for basal dendrites; Fig. 5 B, D, E, and F) and comparable to levels of background noise (mean of 2.3% ΔF/F for tuft dendrites and 3.6% ΔF/F for basal dendrites), where background noise was calculated as the difference in calcium amplitude between branches during randomly selected intervals without AP firing. If in these experiments neuronal firing were driven by an individual dendritic branch, we would expect to see an increased amplitude in that branch because its calcium activity would be the sum of bAP activity and its unique synaptic input. However, we do not see such differences, and therefore, we find no evidence that certain dendrites have a preferential role in driving APs.

Fig. 5.

Suprathreshold calcium activity is global within the tuft and basal dendritic compartments. (A) Example data recorded simultaneously from multiple tuft dendrites. (Left) Fluorescence image of tuft dendrites from a L5 pyramidal cell with soma located 520 μm below the pia. Yellow dashed lines represent the path of dendrites located just outside the imaging plane. Colored solid lines are ROIs used to extract calcium signals. (Right) Example traces from this imaging plane. Colored traces are calcium signals that correspond to dendritic branches in ROIs at left. Vertical gray boxes denote a significant transient on at least one of the three traces. Black trace is the simultaneous membrane potential recorded in the whole-cell patch configuration. Bottom trace shows the amplitude (% ΔF/F) of the three calcium signals at each transient. Asterisk indicates a transient that appeared significantly larger than in the other ROIs. (B) Histograms of the difference in calcium amplitude between each pair of branches shown in A for all spike events (n = 133). (C) Same as A but for an imaging plane containing multiple basal dendrites. Soma is located at a depth of 460 μm. (D) Same as B but for data from dendrites shown in C (n = 36 events). (E) Data representing the observed differences in amplitude of calcium activity measured between pairs of simultaneously imaged tuft branches during spike activity (n = 13,791 events from 72 pairs of tuft dendrites). Data have been smoothed with a boxcar filter of width 1%. (F) Same as E but for basal dendritic branches (n = 40,657 events from 251 pairs of basal dendrites).

Discussion

Here we present in vivo results comparing basal and tuft dendritic calcium activity in L5 pyramidal neurons. Our data include in vivo recordings of basal dendrites and spines in L5 neurons as well as combined recordings of somatic electrical activity and L5 basal dendrite activity. Previous studies have investigated the function of dendrites on L5 pyramidal neurons in vivo but restricted imaging to superficial dendrites of the tuft (12, 13). In these studies, imaging was performed in barrel cortex, where L5 begins at 600 μm or more below the pial surface (14). This depth is below the reach of conventional two-photon calcium imaging, and special techniques developed to image this deep have only been used to record somatic calcium activity (19). We circumvented this issue by focusing on the whisker representation area in wM1. This cortical region lacks a granular layer and so has a relatively superficial location of L5 that begins at 260 μm below the pia (14). Given the high convergence of projections from other brain regions onto wM1 (20-22), the importance of wM1 in tactile behavior and learning (23–25), and the capability to perform two-photon imaging in multiple layers, this system is ideally suited for studies of dendritic integration in vivo.

As there is no prior work characterizing L5 basal dendrites in vivo, we compare our data to studies of basal dendrites of L2/3 pyramidal cells. Calcium imaging has previously demonstrated that bAPs propagate efficiently into L2/3 basal dendrites (26, 27). We extend this result to L5 basal dendrites and show that the invasion of bAPs does not depend on ongoing synaptic activity. In addition, we found that the amplitude of calcium activity is highly similar between different basal dendrites. This property indicates the output of the neuron is propagated robustly and uniformly throughout the entire basal compartment (9). Previous studies have further characterized subthreshold calcium activity in L2/3 basal dendrites as being organized into localized hotspot activity (26) that arises from single spine synaptic inputs (15, 28). Our data confirm that this is also the case for L5 pyramidal neurons despite differences in this cell type’s function, location, and morphology. Basal dendrites of L5 pyramidal cells therefore appear functionally similar to basal dendrites of other cortical pyramidal cell types.

Previous work (12) has characterized calcium signals in L5 tuft dendrites in vivo. They found that calcium transients in the tuft accompany complex dendritic electrical spikes. These complex spikes occurred primarily during somatic AP activity but sometimes fail to occur even during bursts of somatic APs. These observations are similar to the unreliable calcium transients we observed in the tuft during AP firing. It is therefore possible that our tuft signals also result from complex dendritic spikes. However, we additionally observed that these calcium transients occur uniformly throughout the dendritic tuft. We note that global tuft activity has been shown to be behaviorally relevant in the primary somatosensory cortex (S1) of the awake animal (29). Therefore, the synchronized tuft activity described here may be a highly significant feature of dendritic processing. Furthermore, we found that subthreshold activity in the tuft is characterized by small-amplitude, highly localized calcium transients, similar to basal dendrites. Taken together, our data support a model where tuft activity is partially decoupled from neuronal output and yet becomes highly synchronized during AP activity.

Prior work in vitro has reported that basal and tuft dendrites amplify their inputs via local dendritic NMDA spikes (5). It has been hypothesized that such events are the primary mechanism by which dendrites integrate their inputs and that single dendrites therefore play a primary role in firing the cell (1). NMDA spikes provide large calcium signals over a length scale of 10–30 μm (30) and are thought to arise from clustered synaptic inputs (5, 31). However, our dataset provides little evidence for local NMDA spikes during spontaneous activity in the anesthetized animal. Although larger regions of dendrites did occasionally show calcium transients, these events were small-amplitude and provide no evidence for an amplification mechanism. Generally, subthreshold dendritic events were localized to regions of less than 10–20 μm, in both tuft and basal dendrites.

In total, we found little evidence for a preferential role for single dendrites in firing the cell. We observed that differences in amplitude of AP-associated calcium transients between pairs of dendrites are quite small for both basal and tuft dendrites. If single dendrites were of primary importance in driving neuronal output, we would expect sizeable differences in the calcium activity of different dendrites. However, the recorded differences in amplitude between basal or tuft dendrites were similar in scale to the level of background noise. Therefore, our data support a “multibranch model” where synaptic input and dendritic activity are spread uniformly throughout all branches in either the tuft or basal dendritic trees. As our data were taken primarily under conditions of spontaneous activity in the anesthetized state, future work may reveal that dendritic amplification mechanisms play a more primary role under specific behavioral conditions.

Materials and Methods

All experimental procedures were performed in accordance with the recommendations of the animal ethics committee at the Technical University Munich and were approved by the state government of Bavaria, Germany.

Animal Preparation.

The animal preparation procedures were the same as described previously (16) (for details see SI Materials and Methods).

Whole-Cell Recordings and Electroporation.

In the majority of our experiments (n = 25), we used the whole-cell patch technique to load L5 pyramidal neurons with calcium dye and record their somatic membrane potentials. Due to the long diffusion time for dye to reach distal tuft dendrites, imaging began 40 min after loading began. Identity as a L5 pyramidal neuron was confirmed by the depth (>400 μm below the pia), the presence of an apical dendrite that terminated as a tuft in layer 1, and the presence of spines. In a subset of experiments (n = 6), neurons were filled with calcium dye via the electroporation technique as previously described (32).

Two-Photon Calcium Imaging of Single Branches.

Two-photon calcium imaging was performed on a custom-built setup similar to what has been described previously (28). In general, neurons were imaged at several focal planes distributed across the tuft, apical trunk, and basal dendrites. The width of the field of view was controlled by the amplitude of scanning mirror rotation and ranged between 100 and 300 μm to fit the layout of dendrites in the focal plane. Spontaneous activity was monitored at each focal plane for 5–10 trials of 30 s each. Z-stacks were taken at the end of each experiment using the largest possible field of view (300 × 300 μm) with a step size of 0.5 μm along the z-axis.

Two-Photon Calcium Imaging of Spines.

The same two-photon imaging setup was used for the imaging of single spines. For the implementation of the low-power temporal oversampling procedure (15), the scanner was configured for the following mode: the number of lines was reduced to 64, the number of pixels in each line was resampled to 256, and the imaging frame rate was increased to 200 Hz. The width of the field of view was accordingly reduced to the range of ∼40 μm. Laser power was reduced to 10–60 mW (depending on imaging depth) during spine imaging. The experimental protocol for the spine imaging was the same as for dendritic imaging.

Data Analysis.

Calcium imaging data were analyzed by custom-written programs in MATLAB (version 2010b; Mathworks). Whole-cell recordings were extracted to text files using Igor (Igor Pro v5.0.1.0; Wavemetrics) and then processed in MATLAB (for details see SI Materials and Methods).