The precise timing of presynaptic and postsynaptic activity results in synaptic modifications, which depend on calcium influx. [Ca2+] transients in the spines of spiny neurons in layer 4 (L4) of the somatosensory barrel cortex of young rats were investigated in thalamocortical brain slices by two-photon excitation microscopy to determine the spike timing dependence of the Ca2+ signal during near-coincident presynaptic and postsynaptic activity. [Ca2+] transients evoked by backpropagating action potentials (bAPs) were mediated by voltage-dependent Ca2+ channels and were of comparable size in a spine and adjacent dendritic shaft. They decreased with the distance of the spine from the soma. EPSP-evoked [Ca2+] transients were restricted to spine heads and were mediated almost entirely by Ca2+ influx through NMDA receptors (NMDARs). Their amplitude was independent of the position of the spine along the dendritic arbor. bAPs interacted with EPSPs to generate sublinear or supralinear Ca2+ signals in a spine when EPSP and bAP occurred within a time window of ∼50 msec. Synaptic stimulation, coincident with a bAP, evoked a large postsynaptic Ca2+ influx that was restricted to a single spine, even after EPSPs were blocked by the AMPA receptor antagonist NBQX that rendered synapses effectively “electrically silent.”
We conclude that the spines of L4 cells can act as sharply tuned detectors for patterns of APs occurring in the boutons of the afferents to L4 cells and the spines of L4 cell dendrites. The readout for near-coincident presynaptic and postsynaptic APs is a large transient Ca2+ influx into synaptically active spines mediated by the brief unblocking of NMDARs during the dendritic bAP.
- two-photon microscopy
- coincidence detection
- spiny stellate cell
- barrel cortex
- NMDA receptor
- backpropagating action potential
Ca2+ is an intracellular second messenger in signaling cascades of neurons, constituting a “chemical code” for their electrical activity. Ca2+ influx through voltage-dependent Ca2+ channels (VDCCs) (Westenbroek et al., 1990; Magee and Johnston, 1995) and glutamate gated ion channels (Bekkers and Stevens, 1989; Spruston et al., 1995a) located in the postsynaptic membrane of excitatory synapses result in a transient elevation of intracellular Ca2+ during neuronal activity (Markram and Sakmann, 1994; Spruston et al., 1995b; Helmchen et al., 1996). These Ca2+ channels can interact in a complex way, facilitating nonlinear Ca2+ signaling at a synaptic contact (Yuste and Denk, 1995; Köster and Sakmann, 1998). So far, the precise timing dependence of this nonlinear Ca2+ signaling is unknown.
Spiny stellate neurons in layer 4 (L4) of the barrel cortex receive whisker-specific synaptic input from cells located in the thalamic ventral posterior medial nucleus (VPM). Stellate neurons constitute the major recipient neurons for thalamocortical input and are thought to amplify and then relay excitation via their axonal arbor to other cortical layers, mostly restricted to within a barrel column (Feldmeyer et al., 1999, 2002; Lübke et al., 2000; Petersen and Sakmann, 2000, 2001; Silver et al., 2003). The dendritic arbors of stellate cells respect the barrel borders, and they are, compared with pyramidal cells, electronically and morphologically more compact. In pyramidal neurons of layer 2/3 and layer 5 of the neocortex and in pyramids of the hippocampal CA1 subfield, which have an elongated apical dendrite and a tufted dendritic arbor, the axonally initiated action potentials (APs) propagate back into the dendritic arbor and elicit a transient rise in calcium concentration ([Ca2+]) (Markram et al., 1995; Yuste and Denk, 1995; Köster and Sakmann, 1998; Yuste et al., 1999). One of the functions of backpropagating APs (bAPs) in pyramidal cells is presumably to act as an induction signal for long-term changes in synaptic efficacy, depending on the coincident occurrence of presynaptic and postsynaptic APs (Markram et al., 1997a; Feldman, 2000), resulting in an highly localized transient increase in dendritic Ca2+ influx (Yuste and Denk, 1995; Köster and Sakmann, 1998).
Whether backpropagation of APs and the associated rise in dendritic and spinous [Ca2+] is also a property of spiny stellate cells is an open question. It has been suggested that spiny stellate cells show different Ca2+ dynamics than cortical pyramidal cells (Egger et al., 1999). Synaptic connections between layer 4 spiny neurons are characterized by a relatively high NMDA receptor (NMDAR to AMPA receptor (AMPAR) ratio, which would render these connections particularly sensitive to Ca2+ influx during coincident bAPs and subthreshold EPSPs (Feldmeyer et al., 1999).
We have used two-photon excitation microscopy (Denk et al., 1990, 1996) to characterize the [Ca2+] transients in spines and dendritic shafts of spiny stellate cells. Ca2+ influx into spines evoked by a bAP paired with an EPSP can sum linearly, sublinearly, and supralinearly, depending strongly on the interval between bAP and EPSP and the order of occurrence.
The results suggest that in excitatory relay neurons of layer 4 of the somatosensory cortex the large [Ca2+] transients in single spines are part of a potential mechanism that detects the occurrence of coincident presynaptic APs in thalamic VPM or other presynaptic L4 neurons and postsynaptic APs in L4 spiny stellate neurons, respectively (Brecht and Sakmann, 2002a,b).
Materials and Methods
Slice preparation. Thalamocortical slices were prepared from the somatosensory cortex of 13- to 15-d-old (P13-P15) Wistar rats as described previously (Feldmeyer et al., 1999). All experimental procedures were in accordance with the animal welfare guidelines of the Max-Planck Society. Briefly, the rat was decapitated and the brain was quickly removed. The brain was then placed on a ramp with a 10° slope and a vertical cut at an angle of 50° was made as described by Agmon and Connors (1991). The brain was glued with the cut surface facing down onto the chilled stage of the tissue slicer. The upper part of the brain was discarded by cutting off a 3 mm thick slice. Then 350 μm thick slices were cut, incubated at 37°C for 30 min, and subsequently maintained at room temperature before recording. All experiments were performed at physiological temperatures (34-36°C).
The extracellular solution contained the following (in mm): 125 NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 1 MgCl2, 25 glucose, and 2 CaCl2, bubbled with 95% O2 and 5% CO2. A 10 μm concentration of glycine was always added to the extracellular solution (Johnson and Ascher, 1987).
Electrophysiology. Patch pipettes (5-7 MΩ) for whole-cell voltage recording were filled with a solution containing the following (in mm): 125 K-gluconate, 20 KCl, 10 HEPES, 4 ATP-Mg, 10 Na-phosphocreatine, and 0.3 GTP and a Ca2+ indicator, Oregon Green 488 BAPTA-1 (OGB1; 200 μm; Molecular Probes, Eugene, OR). The barrel field of the rat somatosensory cortex and individual spiny neurons were visualized using infrared (IR) gradient contrast videomicroscopy (Dodt and Zieglgänsberger, 1998). The barrels appeared in layer 4 as evenly spaced dark structures, separated by lighter septa (Feldmeyer et al., 1999; Lübke et al., 2000; Petersen and Sakmann, 2000). The somata of spiny neurons in layer 4 tended to cluster at the barrel borders and were identified by their shape and size (10-15 μm in diameter) and by the star-like initial parts of their dendrites lacking a prominent apical dendrite. Recordings were made from cells located between 50 and 100 μm below the slice surface (Stuart et al., 1993) in the whole-cell (wc) recording configuration with a patch-clamp amplifier (EPC-9/2; List-Electronics, Darmstadt, Germany) operated in current-clamp or voltage-clamp mode. The voltage signals were filtered at 3 kHz and digitized at 10 kHz by the integrated analog-to-digital converter. Initial access resistance was <15 MΩ. Cells were filled for at least 20 min with OGB1 before fluorescence measurements to allow equilibration of the indicator concentration between the patch pipette and the cytoplasm.
Pharmacological compounds. Na+ channels were blocked by tetrodotoxin (TTX; 1 μm), VDCCs by Cd2+ (100 μm), AMPA receptors by 1,2,3,4-tetrahydro-6-nitro-2,3-dioxo-benzoquinoxaline-7-sulfonamide (NBQX; 10 μm), NMDA receptors by d-5-amino-phosphonopentanoic acid (APV; 25 μm) and group II metabotropic glutamate receptors by 2-(S)-α-ethylglutamic acid (EGlu; 50 μm). All compounds were purchased from Tocris (Bristol, UK).
Calcium imaging. A galvanometer scanning unit (TCSNT; Leica Microsystems, Mannheim, Germany) was attached to an upright microscope (DMLFS; Leica) equipped with a 63× objective (HCX APO W63x UVI; numerical aperture, 0.9; Leica). For two-photon excitation, short pulses of 170-200 fsec at 76 MHz from a Ti:Sa-Laser (MIRA 900F; Coherent, Santa Clara, CA) at 870-890 nm pumped by a solid-state laser (Verdi 5W; Coherent) were used. Line-scan imaging was performed with high temporal resolution (2.2 msec/line). External non-descanned detection behind the objective and the condenser was used, yielding a high signal-collection efficiency. Transmission and epifluorescence signals were recorded by photomultiplier tubes (R6357; Hamamatsu Photonics, Herrsching, Germany) and summed offline (Köster et al., 1999). Line-scan imaging and electrical recordings were synchronized by a hardware trigger supplied by the scanner electronics.
The distance from the soma was extrapolated from small linear segments placed online along the dendritic branches. This tracing method underestimated the exact distance, derived from a reconstruction of the dendritic length from a three-dimensional fluorescence stack, by <10% (n = 4).
Simultaneously with full-frame fluorescence image acquisition, IR-scanning gradient contrast (IR-SGC) images were recorded (Rathenberg et al., 2003). The transmitted IR laser light was imaged through a gradient-contrast tube (Dodt et al., 1999) onto a photomultiplier.
Data analysis. Fluorescence line-scan images were analyzed using custom software macros based on the Leica confocal software. A line was scanned every 2.2 msec, and two lines were averaged offline, resulting in a temporal resolution of 4.4 msec. The fluorescence for each time point was averaged for regions of interest (ROIs) enclosing the dendrite or spine examined. Before stimulation, fluorescence was averaged for 100 msec to obtain the basal fluorescence F0. A region distant to any indicator-containing structure was chosen to determine the background fluorescence FB. Background fluorescence was independent of the size and position of the background ROI, indicating a homogeneous background along the scan line. Relative fluorescence changes were calculated as ΔF/F(t) = (F(t) - F0)/(F0 - FB). Assuming a constant background, ΔF/F is slightly underestimated but independent of the size of the neuronal compartment (Sabatini et al., 2002). Single exponential fits to the decay of the fluorescence transients yielded the peak amplitude, denoted as A, and the decay time constant τ. Fluorescence traces are averages of two to five trials. Data are represented as means ± SD. Differences between groups were evaluated using Student's t test (either paired or two-tailed unpaired) and statistical significance was asserted for p < 0.05. The number of spines, each of which was measured in a different cell, is given by n.
Dye saturation. Peak [Ca2+] transient amplitudes might be underestimated when the high-affinity Ca2+ indicator OGB1 is used because of saturation. The linearity of the fluorescence increase, as reported by the dye, was evaluated from [Ca2+] transients evoked by two bAPs separated by different time intervals Δt (n = 8). These sequences resulted in [Ca2+] transients with two peaks corresponding to each bAP. The peak amplitude, A2APs, was derived from a single exponential fit to the decay of the [Ca2+] transient. A2APs increased with shorter time intervals (supplementary Fig. 1A) to 0.95 ± 0.28 at Δt = 10 msec. This value was much smaller than the maximal increase in fluorescence (2.55 ± 0.52) evoked by a train of 20 APs at 100 Hz.
Normalizing A2APs, derived from the AP-AP sequence, to the linear sum of a [Ca2+] transient evoked by a single bAP (AAP = 0.46 ± 0.16 and τAP = 490 ± 190 msec) gave a measure for the linearity of the dye (denoted as the “nonlinearity factor”). The nonlinearity factor was close to unity for all time intervals tested between 10 and 300 msec, indicating linear summation of the [Ca2+] transients evoked by two bAPs (supplementary Fig. 1B). Comparing A2APs to the predicted linear sum for each measurement showed no significant difference (paired t test; p > 0.1; n = 6). Thus, at least up to A2APs = 1, the nonlinearity factor faithfully reported the linear summation of two AP-evoked [Ca2+] transients.
Identification of active synaptic contacts in spiny neurons
Spiny stellate neurons in layer 4 of the barrel cortex were loaded with the Ca2+ indicator Oregon Green 488 BAPTA-1 via the patch pipette in the wc recording configuration. After filling, their morphology became visible in the two-photon excitation (TPE) fluorescence image (Fig. 1A). The tip of an extracellular stimulation pipette was positioned in close proximity to a dendritic branch (∼5 μm) under visual control (Fig. 1B) using the overlay of the IR-scanning gradient contrast (IR-SGC) image with the TPE fluorescence image of the indicator loaded cell.
Active synaptic contacts were identified in the full-frame, continuous image acquisition mode (500 msec/frame) by a localized, transient increase in fluorescence after synaptic stimulation. Spines responding with a [Ca2+] transient (Fig. 1C,D) were located at distances ranging between 30 and 210 μm from the soma (79 ± 35 μm; n = 46). Putative shaft synapses were found only infrequently (n = 2), and [Ca2+] transients in these contacts were not different from spine synapses.
The stimulation strength was set to evoke reliable EPSPs, with peak amplitudes between 1 and 11 mV (6 ± 3 mV; n = 29). This reflected the activation of, on average, four connections with a total of ∼12 synaptic contacts during an EPSP [assuming a mean EPSP amplitude of 1.6 mV and three synaptic contacts per layer 4 connection (Feldmeyer and Sakmann, 2000)]. In the vicinity of an active spine (∼20 μm) no other spines were observed that responded, suggesting that the other synaptic contacts contributing to the EPSP were located far from the particular spine under investigation. Once an active spine was found, almost every stimulus evoked a transient rise in [Ca2+]. The failure rate was <10% (n = 10 spines), which is in agreement with results reported for connected pairs of spiny stellate cells (Feldmeyer et al., 1999).
[Ca2+] transients in spines and shafts
Figure 1D shows an EPSP (top trace) and the simultaneously recorded [Ca2+] transient evoked by synaptic stimulation in the spine and shaft shown in Figure 1C. The Ca2+ signal was almost completely restricted to the spine head. In a plot of the peak [Ca2+] transient amplitudes (AEPSP) in shafts versus peak amplitudes in spine heads, all data points fell clearly below unity (Fig. 2A). The peak amplitude of the small signal in the shaft was correlated with the peak amplitude in the spine (linear regression; r = 0.75). On average, the peak amplitude in the shaft was 26 ± 12% of the peak amplitude in the spine head (n = 9; paired t test; p < 0.0001).
The [Ca2+] transients evoked by bAPs were measured in the same spines and shafts (n = 6). In Figure 1E, the bAP-evoked peak [Ca2+] transient amplitudes (AAP) were similar in the shaft and spine. The peak amplitudes in shafts plotted versus the peak amplitudes in spine heads fell almost exactly on the unity line (Fig. 2B). The mean ratio of the peak amplitude in the shaft compared with that in the spine was 105 ± 10% (paired t test; p > 0.2). The bAP-evoked [Ca2+] transient was almost completely blocked (7 ± 5% of control; n = 4) after bath application of Cd2+ (100 μm), indicating Ca2+ influx through VDCCs. The bAP-evoked [Ca2+] transients represent a global dendritic signal common to spines and shafts, whereas EPSP-evoked [Ca2+] transients are spatially confined to the spine head. The small shaft Ca2+ signal recorded after synaptic stimulation was presumably caused by the diffusion of Ca2+ between the spine and the shaft (Svoboda et al., 1996), although most of the Ca2+ in a spine was cleared by other mechanisms, such as Ca2+ pumps.
The peak [Ca2+] transient evoked by an EPSP in a spine head was larger than the corresponding transient evoked by a bAP (Fig. 2C) in 70% (24 of 35) of the spines. The mean peak amplitude evoked by an EPSP was AEPSP = 0.58 ± 0.27 (n = 47), compared with AAP = 0.44 ± 0.20 in the case of a bAP-evoked [Ca2+] transient (n = 36), which was significantly larger (paired t test; p < 0.01). The decay time constants were not significantly different for EPSP-evoked (τEPSP = 384 ± 203 msec) and bAP-evoked (τAP = 432 ± 209 msec) [Ca2+] transients (paired t test; p > 0.4).
Dependence of [Ca2+] transients in spines on the distance from the soma
The relative size of [Ca2+] transients evoked by either an EPSP or a bAP showed different dependencies on the distance of the spine from the soma. The EPSP-evoked peak [Ca2+] transient was independent of the distance from the soma (Fig. 3A) (r = 0.16; n = 47). In contrast, the peak [Ca2+] transient amplitude evoked by a bAP in spines decreased with the distance from the soma (Fig. 3B) (linear regression; r = -0.44; n = 49), in agreement with the distance dependence measured for bAP-evoked [Ca2+] transients in dendritic shafts (Fig. 3C) (r = -0.55). Comparing the mean AAP values close to the soma (distance <50 μm) to those in the distal segments (distance >150 μm) showed a reduction in the bAP-evoked [Ca2+] transients to 41% of the soma value (unpaired t test; p < 0.005).
To test whether the [Ca2+] transients were evoked by actively backpropagating APs, the spread of the [Ca2+] transient within the dendritic arbor was measured in voltage clamp for control and after bath application of TTX (1 μm). Somatic AP waveforms, recorded in current clamp, were used as voltage commands in voltage clamp (Stuart and Sakmann, 1994). In this recording configuration, [Ca2+] transients were acquired along the dendritic arbor as controls. Bath application of TTX abolished the [Ca2+] transients. Waveforms were scaled by a factor of 1.5-2 to elicit [Ca2+] transients close to the soma with the same amplitude as in the control measurements (Kaiser et al., 2001). This procedure ensured that the applied AP waveform activated VDCCs in the same way as under control conditions. [Ca2+] transients in the presence of TTX were strongly attenuated compared with control (Fig. 3D).
The decay time constants of EPSP-evoked (r = 0.08) and AP-evoked (r = 0.08) [Ca2+] transients in spines were independent of the distance of the spine from the soma (data not shown).
These results indicate that dendritic [Ca2+] transients are evoked by actively backpropagating APs.
[Ca2+] transients evoked by pairing EPSP and bAP depend on the interstimulus interval
To examine the interaction of synaptic potentials with bAPs, EPSPs and bAPs were paired at different time intervals. This pairing resulted in very different [Ca2+] transients with linear, sublinear, and supralinear summation of the Ca2+ signals, depending on the interval between the two stimuli and on their order of occurrence (Fig. 4).
EPSP- and bAP-evoked [Ca2+] transients in a spine were recorded individually as controls, and exponential fits to the decay of the [Ca2+] transients yielded the peak [Ca2+] amplitudes AEPSP and AAP and the corresponding decay time constants τEPSP and τAP, respectively (Fig. 4B). Then EPSP and bAP were paired at an interstimulus interval Δt (measured between the onset of the extracellular stimulus and the peak of the AP), resulting in [Ca2+] transients with two peaks. A single exponential fit to the decay time course of the [Ca2+] transient, evoked by the paired stimuli, yielded the peak amplitude Apaired and the decay time constant τ. Apaired normalized to the linear sum of the control stimuli for the corresponding time interval is referred to as the nonlinearity factor. A value of 1 corresponds to linear summation of both Ca2+ signals.
Supralinear summation was found when the EPSP preceded the bAP by 10 msec (Fig. 4D). The peak [Ca2+] transient amplitude for this pairing sequence was, on average, 1.7 ± 0.2-fold larger than the expected linear sum (n = 13). In contrast, when the EPSP followed the AP with a delay of 50 msec, the [Ca2+] transients summed sublinearly (i.e., the peak [Ca2+] transient amplitude was smaller than the expected linear sum) in all spines tested (Fig. 4C). The nonlinearity factor for the AP-EPSP pairing sequence was, on average, 0.86 ± 0.08 (n = 7).
Peak [Ca2+] transient amplitude dependence on interval between EPSP and bAP
The interval between bAP and EPSP was varied between -300 and 300 msec and yielded a plot for the spike-timing dependence of the peak [Ca2+] amplitude. Negative time intervals (Δt < 0) denote the pairing sequence AP-EPSP, positive intervals (Δt > 0) denote the pairing sequence EPSP-AP. Figure 5A shows averages of peak [Ca2+] amplitudes Apaired plotted versus the time interval Δt between the two stimuli. Comparing Apaired for adjacent time points revealed a significant increase in the [Ca2+] transient from Δt = 1 to Δt = 10 msec (unpaired t test; p < 0.005) and a significant decrease from Δt = 50 to Δt = 100 msec (p < 0.05). The average peak [Ca2+] amplitude in this interval was Apaired (10 msec ≤ Δt ≤ 50 msec) = 1.52 ± 0.46. For all other intervals the amplitudes were not significantly different. Comparing Apaired for any pair of time points for Δt <10 msec and Δt >50 msec also did not indicate significant differences (p > 0.05). Pooling the peak amplitudes in these intervals yielded Apaired = 0.91 ± 0.30. Thus, only when an EPSP is followed by a bAP within 10-50 msec is a significantly increased Ca2+ influx generated (unpaired t test; p < 10-6).
The decay of the [Ca2+] transients evoked by paired stimuli did not depend on the stimulus interval, or on the order of occurance of EPSP and AP. The average decay time constant for the paired stimuli was τ = 375 ± 119 msec, similar to the decay times of the single stimulus-evoked transients (p > 0.2).
Nonlinear summation of Ca2+ signals
The increased peak [Ca2+] transient amplitudes for the sequence EPSP-AP between 10 and 50 msec was the result of nonlinear summation of Ca2+ signals. Figure 5B shows the nonlinearity factor as a function of the interstimulus interval. It reveals the time domains of linear, sublinear, and supralinear summation. Comparing the peak amplitude Apaired to the predicted linear sum for each time interval yielded a significant supralinear summation for the time intervals Δt = 10 msec (p < 0.0001; n = 13), 20 msec [nonlinearity factor (20 msec) = 1.5 ± 0.3; p < 0.0001; n = 13], and 50 msec [nonlinearity factor (50 msec) = 1.3 ± 0.1; p < 0.005; n = 5]. A significant sublinear summation was found for Δt = -50 msec (paired t test; p < 0.005; n = 7). For all other interstimulus intervals, deviations from the linear sum were insignificant (p > 0.2).
Thus, supralinear summation of Ca2+ signals in single spines evoked during pairing of EPSPs and bAPs was restricted to a relatively short time window of 50 msec and to the sequence EPSP-AP. This time window can, for spiny stellate cell synapses, be denoted as a “coincidence time window,” during which the occurrence of presynaptic and postsynaptic APs is encoded by an increased Ca2+ influx as opposed to the uncorrelated occurrence of presynaptic and postsynaptic APs outside of this time window. Reversing the order of stimulation (AP-EPSP) resulted either in linear summation of Ca2+ signals or in sublinear summation for Δt = -50 msec.
Time course of supralinear summation
The Ca2+ influx evoked by a bAP, if preceded by an EPSP in the coincidence time window, was larger than the Ca2+ influx evoked by a bAP alone. Normalizing the rise in [Ca2+] evoked by the bAP (A′AP) in the sequence EPSP-AP to the rise in [Ca2+]by a control bAP (AAP) is a measure for the EPSP-timing-dependent potentiation of the bAP-evoked [Ca2+] transient (“EPSP-AP potentiation”). The potentiation (A′AP/AAP) was 2.3 ± 0.5, 2.1 ± 0.7, and 1.4 ± 0.3 for Δt = 10, 20, and 50 msec, respectively (Fig. 6A).
The EPSP-AP potentiation (with a maximum at Δt = 10 msec) decayed along an exponential time course, with a time constant of 32 msec (Fig. 6A). This value matches the decay time constant of the NMDAR-mediated component of EPSCs in spiny stellate cells [τNMDAR = 35 ± 15 msec (Feldmeyer et al., 1999)] and suggests that the mechanism generating the supralinear Ca2+ influx during near-coincident presynaptic and postsynaptic APs is determined mostly by the closing kinetics of synaptic NMDAR channels. A bAP following an EPSP relieves the Mg2+ block of the synaptic NMDAR channels, which are in an open state after the binding of glutamate, but still remain Ca2+-impermeable because of the voltage-dependent Mg2+ block. With an increasing interstimulus interval between EPSP and bAP, fewer NMDARs are in an open (but blocked) state, and the effect of the membrane depolarization during the bAP is correspondingly smaller.
Additional evidence that the bAP-evoked supralinear [Ca2+] transient is “probing” the number of open NMDARs was the value of the EPSP-AP potentiation at Δt = 1 msec. At this interval (assuming fast propagation of the AP into the dendritic arbor within <1 msec) the number of open NMDARs was not yet maximal because of the slow opening kinetics of NMDARs [reflected in the rise time of 5-10 msec of the NMDAR-mediated EPSC (Feldmeyer et al., 1999)]. The potentiation was smaller at Δt = 1 than at Δt = 10 msec, when more NMDARs were open.
The EPSP-AP potentiation for the sequence EPSP-AP at Δt = 10 msec scaled with the size of the EPSP-evoked peak [Ca2+] transient amplitude (r = 0.8; n = 13) (Fig. 6B). Again, this is consistent with the view that the main source of the supralinear Ca2+ influx is synaptic NMDAR channels.
Dependence of spine [Ca2+] transients evoked by coincident activity on distance from soma
The decrease of the peak [Ca2+] amplitude evoked by a bAP with the distance from the soma suggested that coincidence effects in spines might be sensitive not only to the order and interval between EPSP and bAP but might depend, in addition, on the location of a spine. Apaired of [Ca2+] transients evoked by the sequence EPSP-AP at Δt = 10 msec plotted against the distance of a spine from the soma showed no correlation (r = -0.002) (Fig. 6C). The nonlinearity factor was also independent of the distance of a spine from the soma (r = 0.1) (Fig. 6D).
Thus, the size of the Ca2+ influx and the nonlinearity factor during near-coincident presynaptic and postsynaptic APs were independent of the position of the synaptic contact along the dendritic arbor. The membrane depolarization caused by a bAP was sufficient for all spines to relieve the Mg2+ block of NMDARs. Although the Ca2+ influx through VDCCs decreased with the distance from the soma, the influx through NMDARs counteracted this effect, resulting in a stereotyped, almost constant, spine Ca2+ signal during coincident EPSPs and APs.
Supralinear Ca2+ signals are restricted to the spine head
The large Ca2+ influx evoked by pairing an EPSP with a bAP was restricted to the spine head. Figure 7A shows a [Ca2+] transient measured in a spine and shaft (shown in Fig. 1) that was evoked by the sequence AP-EPSP, with an interval of 20 msec (Δt = -20 msec). The peak amplitude in the shaft corresponded well with the bAP-evoked [Ca2+] transient, and it was substantially smaller than the [Ca2+] transient in the spine head. The reverse sequence EPSP-AP at the same interval (Δt = 20 msec) evoked an even larger, supralinear Ca2+ signal, which was also restricted to the spine head. On average, the peak [Ca2+] transient in the shaft during coincident EPSP and bAP was 32 ± 10% of the [Ca2+] transient in the spine head (n = 3) and corresponded to the bAP-evoked [Ca2+] transient in the shaft. The supralinear Ca2+ signal was highly localized and restricted to the site of the active synaptic contact, where it originated with only a small accumulation of Ca2+ in the adjacent shaft. Thus, spines compartmentalize Ca2+ also during coincident activity and briefly generate a large [Ca2+] gradient between the synaptic spine head and the adjacent dendritic segment.
Influence of ionotropic glutamate receptors on EPSP-evoked [Ca2+] transients
Spine [Ca2+] transients evoked by EPSPs were strongly reduced in the presence of the AMPAR antagonist NBQX (10 μm) (Fig. 8A). The distribution of the remaining [Ca2+] transients after bath application of NBQX, compared with control, varied over a wide range, between 0 and 60%. In 3 of 11 spines the [Ca2+] transient was blocked to <8% or even abolished (Fig. 8B). On average, NBQX reduced the Ca2+ signal to 27 ± 18% (n = 11; p < 0.0001) and the EPSP to 34 ± 28% (p < 0.00005) of control (at -71 ± 3 mV) (Fig. 8C). The additional application of APV (25 μm) abolished the remaining [Ca2+] transient completely and reduced the EPSP amplitude to 7 ± 1% of control (n = 3). The remaining EPSP (0.6 ± 0.2 mV) was presumably evoked by inhibitory inputs, which were depolarizing at a resting membrane potential of -71 mV and an internal [Cl-] of 20 mm.
APV (25 μm) applied without NBQX blocked the Ca2+ signal to 4 ± 3% of control (n = 4; paired t test; p < 0.02), whereas the EPSP amplitude was reduced much less, to 66 ± 17% of control (p < 0.02).
Pharmacological dissection of supralinear Ca2+ influx
To dissect the mechanism causing supralinearity, the [Ca2+] transients during coincident presynaptic and postsynaptic APs were measured in the presence of specific antagonists of glutamate receptor subtypes.
Figure 9A illustrates the effect of NBQX and APV on supralinear spine [Ca2+] transients. In this experiment, the Ca2+ influx during synaptic stimulation was blocked by NBQX (same experiment as that illustrated in Fig. 8A). When synaptic stimulation was paired with a single bAP at an interstimulus time interval of 10 msec, a large [Ca2+] transient was recorded again (middle trace; Apaired, NBQX = 1.1). Its size was similar to the transient measured under control conditions (Apaired, control = 0.9). Subsequent application of APV reduced the Ca2+ influx to a level that was similar to the bAP-evoked [Ca2+] transient (right trace).
NBQX had no influence on the peak [Ca2+] transient evoked by pairing EPSP and AP at Δt = 10 msec (reduction, compared with control, 90 ± 23%; p > 0.1; n = 5) (Fig. 9B). The peak [Ca2+] transient for the sequence AP-EPSP at Δt = -10 msec corresponded to the linear sum of the AP-evoked and the reduced EPSP-evoked [Ca2+] transients (nonlinearity factor = 0.9 ± 0.2; p > 0.2; n = 5).
Figures 8A and 9A thus demonstrate that in a spine with a small or no AMPAR-dependent EPSP, which did not respond with a [Ca2+] transient during synaptic stimulation, a full-sized Ca2+ influx was evoked when synaptic stimulation was near-coincident with a bAP. This behavior is reminiscent of “silent synaptic contacts” (Liao et al., 1995).
APV reduced the Ca2+ influx evoked by any pairing sequence (-20 msec < Δt < 20 msec) to the size of the corresponding bAP-evoked [Ca2+] transient (n = 7). Thus, by blocking NMDAR channels, the effect of the order of pairing EPSP and AP on Ca2+ influx became insignificant.
Next, we tested the hypothesis that the bAP relieved the Mg2+ block of NMDARs, resulting in supralinear Ca2+ influx. [Ca2+] transients evoked by pairing an EPSP with a bAP at Δt = ±10 msec were measured in 1 mm Mg2+ and in Mg2+-free solution. The EPSP-evoked [Ca2+] transient increased fourfold after the washout of Mg2+ (n = 2). To avoid saturation of the indicator dye, a subsaturating concentration of APV (5 μm) was added to the bath, resulting in a reduction of the EPSP-evoked [Ca2+] transient to 70% of control in 1 mm Mg2+ (n = 4). The peak [Ca2+] transient amplitudes for the sequences AP-EPSP and EPSP-AP at Δt = ±10 msec became indistinguishable in 0 Mg2+ (Apaired = 0.7 ± 0.2 in both cases) (Fig. 9C,D) and corresponded to the expected linear sum. The nonlinearity factor for both sequences was 1.0 ± 0.2. Thus, the relief of NMDAR channels from the Mg2+ block abolished the supralinear summation of Ca2+ signals.
Influence of metabotropic glutamate receptors on [Ca2+] transients
Spiny neurons express group II metabotropic glutamate receptors (mGluRs), which contribute to the induction of changes in synaptic efficacy (Egger et al., 1999). The effect of the mGluR II antagonist EGlu (50 μm) on the [Ca2+] transients in spines during synaptic stimulation and coincident activity was tested to find out whether mGluRs shape spine [Ca2+] transients.
The [Ca2+] transient evoked by synaptic stimulation was not significantly reduced (82 ± 14% of control; p > 0.2; n = 4) after bath application of EGlu. The somatically recorded EPSP amplitude was also not reduced (92 ± 18% of control; p > 0.4) (Fig. 8C). [Ca2+] transients evoked by pairing an EPSP with a bAP at Δt = ±20 msec were also not affected by EGlu (Fig. 10A,B). The [Ca2+] transient for the sequence AP-EPSP was 89 ± 19% compared with control (paired t test; p > 0.05; n = 3). The [Ca2+] transient for the sequence EPSP-AP was 87 ± 12% compared with control (p > 0.05; n = 3). These results exclude a direct effect of mGluRs on generating [Ca2+] transients within a time window of ±20 msec.
In conclusion, blocking different GluR subtypes indicated that only the NMDAR contributes to the supralinear summation of synaptic Ca2+ signals evoked by EPSPs and bAPs during coincident presynaptic and postsynaptic APs.
Finally a supralinear Ca2+ signal was also observed when stimulation of the synaptic input elicited an AP, a condition reminiscent of the in vivo situation (Fig. 11) (n = 3). The large EPSP (10-15 mV) evoked an AP ∼15-25 msec after the EPSP onset as it is observed in adult spiny stellate neurons in vivo after whisker stimulation (Brecht and Sakmann, 2002b). The spine [Ca2+] transient evoked by suprathreshold synaptic stimulation in the acute slice preparation was much larger (Asuprathreshold = 1.6; τ = 225 msec) than the [Ca2+] transient after subthreshold stimulation (AEPSP = 0.5; τEPSP = 437 msec) measured in the same spine.
Backpropagating action potentials
Spiny cells in L4 of the barrel cortex belong to the class of neocortical neurons in which APs, when initiated in the soma-initial-axon segment backpropagate into the dendritic arbor and evoke a transient rise in dendritic [Ca2+] (Stuart and Sakmann, 1994; Markram et al., 1995; Spruston et al., 1995b). Thus, the AP output of a spiny stellate neuron in L4 that projects its axon collaterals within a column predominantly within L4, to L2/3, and to L5A (Feldmeyer et al., 2002; Lübke et al., 2003) is copied also to the dendritic arbor and encodes a transient elevation of [Ca2+] [“calcium code” (Magee and Johnston, 1995)]. Spiny stellate cells, like pyramidal neurons (Schiller et al., 1995; Spruston et al., 1995b; Köster and Sakmann, 1998) and inhibitory neurons in cortical layer 2/3 (Kaiser et al., 2001), which have long, electrically excitable dendrites, are endowed with this mechanism of feedback signaling to active synapses when presynaptic and postsynaptic APs are near coincident.
The spatial profile of the bAP-evoked [Ca2+] transient along the electrotonically compact dendritic arbor of spiny stellate cells is different from the spatial profile along the longer apical dendrite and tuft of L5 and L2/3 pyramidal neurons, where single APs do not propagate fully into the tuft and only more complex AP burst patterns result in full backpropagation and large [Ca2+] transients (Larkum and Zhu, 2002; Waters et al., 2003). The dendrites of spiny stellate neurons are, in this aspect, comparable to the basal dendrites of L2/3 pyramidal cells (Köster and Sakmann, 2000).
Spine-restricted [Ca2+] transients
In contrast to the global Ca2+ signal evoked by a bAP, the EPSP-evoked [Ca2+] transient was input-specific and restricted to an active spine head, similar to CA1 pyramidal neurons (Yuste and Denk, 1995) and L5 pyramids (Köster and Sakmann, 1998). This indicates that the Ca2+ accumulation in the spine head is much larger than in the adjacent shaft during synaptic activity and might be explained by negligible diffusion of Ca2+ across the spine neck (Svoboda et al., 1996; Yuste et al., 1999; Sabatini et al., 2002). Alternatively, the dendritic shaft might act as a large Ca2+ sink, which cannot accumulate Ca2+. The EPSP-evoked Ca2+ influx into a spine was mediated mainly by NMDAR channels because the [Ca2+] transient was blocked during bath application of APV. The peak EPSP recorded at the soma was reduced only slightly, excluding a major contribution of VDCCs to the Ca2+ influx during synaptic activity (Markram and Sakmann, 1994; Schiller et al., 1998; Kovalchuk et al., 2000). The EPSP-evoked [Ca2+] transient in spiny stellate neurons was much more sensitive to the block of AMPARs than EPSP-evoked [Ca2+] transients in L5 pyramidal neurons (Köster and Sakmann, 1998) and hippocampal CA1 pyramidal neurons (Kovalchuk et al., 2000). This difference might be caused by the resting membrane potential and different NMDAR/AMPAR channel ratios (Markram et al., 1997b; Feldmeyer and Sakmann, 2000).
Coincidence detection by spines
Spines of spiny neurons can act, as in other neocortical cells (Yuste and Denk, 1995; Köster and Sakmann, 1998; Yuste et al., 1999), as detectors for near-coincident presynaptic and postsynaptic APs that evoke a large [Ca2+] transient because of supralinear summation within a short time window of <50 msec. The pharmacological effects of GluR-channel antagonists strongly suggest that the major mechanism generating the large Ca2+ influx is unblocking of NMDAR channels by dendritic depolarization. This view is supported by the observation that removing Mg2+ from the bath abolished the supralinear summation of the Ca2+ signals and that the supralinearity decayed with a time constant comparable to the deactivation kinetics of the NMDAR channel in juvenile (P14) spiny stellate neurons.
Because the nonlinearity factor and the peak [Ca2+] transient amplitude for the sequence EPSP-AP at intervals of 10 msec is independent of the distance between the spine and the soma, the Ca2+ signals evoked during coincident activity represent a common and almost uniform signal for all active spines along the dendritic arbor. The large spinous Ca2+ accumulation during coincident activity is restricted to the active site representing a locally confined and input-specific signal.
The sublinear summation of the Ca2+ signals for the sequence AP-EPSP at intervals of -50 msec could reflect Ca2+-dependent inactivation of NMDAR channels (Legendre et al., 1993; Umemiya et al., 2001).
Because the postsynaptic [Ca2+] transients in spines were evoked by extracellular stimulation in L4, we do not know which presynaptic afferents were stimulated. It was estimated that 10-20% of the excitatory synaptic contacts onto spiny stellate cells are thalamocortical (Benshalom and White, 1986). Most of the rest are corticocortical connections, originating mainly from other spiny stellate cells (Feldmeyer et al., 1999). In either case the results for supralinear Ca2+ influx are likely to be the same because the properties of the NMDAR of spiny stellate cells are independent of the projecting afferents (Gil et al., 1999).
Connections between spiny stellate cells express long-term depression (LTD) independent of spike-timing (Egger et al., 1999), whereas the same pairing protocols resulted in induction of long-term potentiation and LTD depending on the order of stimuli in L5 pyramidal neurons (Markram et al., 1997a). We find that Ca2+-signaling in spines of L4 spiny stellates is similar to that in L5 pyramidal neurons (Köster and Sakmann, 1998). In both cell types, coincident activity results in a large rise in [Ca2+], which suggests that different cell types might decode increased [Ca2+] levels differently via NMDAR-independent pathways, such as mGluRs (Anwyl, 1999; Karmarkar and Buonomano, 2002).
Silent synaptic contacts
Putative electrically “silent synapses” are postulated to exist in connections of the developing postnatal (P6-P7) somatosensory (Isaac et al., 1995), and visual (Rumpel et al., 1998) cortex. Here, a fraction of the synaptic contacts are thought to possess mostly NMDARs but no functional AMPARs in their spine heads. At resting membrane potential, these synaptic contacts do not contribute (or contribute only very little) to EPSPs. Blocking the AMPARs in spiny stellate cells pharmacologically created a comparable experimental condition. The Ca2+ influx into individual spines evoked by subthreshold synaptic stimulation was strongly reduced, resulting in “Ca2+-silent” spines. Surprisingly, a large Ca2+ influx reappeared when synaptic stimulation was coincident with a bAP within a time interval of <20 msec. We conclude that electrically silent spines without functional AMPARs can nevertheless transmit synaptic signals by generating large [Ca2+] transients when coincident activation of other synapses evokes bAPs. This bAP-dependent spine Ca2+ signal could render silent synapses functional at least with respect to Ca2+-dependent dendritic mechanisms (Liao et al., 1995; Isaac et al., 1997; Malinow et al., 2000; Poncer and Malinow, 2001).
Physiological relevance of bAPs and spine Ca2+ transients for spiny cells in L4
Spiking of excitatory neurons in L4 of the barrel cortex is sparse, and whisker stimulation elicits only few APs in L4 neurons (Bruno and Simons, 2002). However, in vivo wc recordings from stellate cells in the barrel cortex indicate that a compound EPSP with >10 mV depolarization, when evoked by a whisker deflection, can trigger an AP within a few tens of msec (Brecht and Sakmann, 2002b). Thus, in vivo in those spines, which contribute to AP initiation, the EPSP-AP sequence will cause a large transient increase in [Ca2+]. Whether the coincidence of a compound EPSP and the bAP, when evoked by sensory stimulation, attenuates or enhances the efficacy of the thalamocortical (VPM-to-L4) or the corticocortical (L4-to-L4) connections is as yet unclear. Studies with CxNR1 knock-out mice indicate that NMDARs are essential for normal development and the anatomical patterning of the barrel cortex (Datwani et al., 2002).
Thalamocortical afferents from a single neuron in VPM can project to the principle whisker (PW) barrel and surround whisker (SuW) barrels (Arnold et al., 2001). The EPSPs evoked in L4 cells by PW or SuW deflection are delayed with respect to each other (Brecht and Sakmann, 2002b). Furthermore, SuW stimulation only infrequently evokes an AP in L4 cells. Thus, if large coincidence-dependent dendritic [Ca2+] transients were controlling synaptic strength in L4 spiny stellates, only those synaptic contacts that, when activated, contribute to the initiation of an AP would also be modified. Such a coincidence-based mechanism for altering synaptic efficacy could operate both during the development of barrel columns in the early postnatal time as well as in the adult cortex.
We thank H. J. Köster and V. Egger for initial help and D. J. Waters and R. M. Bruno for comments on this manuscript.
Correspondence should be addressed to Dr. Thomas Nevian, Abteilung Zellphysiologie, Max-Planck-Institut für medizinische Forschung, Jahnstrasse 29, D-69120 Heidelberg, Germany. E-mail:.
Copyright © 2004 Society for Neuroscience 0270-6474/04/241689-11$15.00/0