Calcium dynamics associated with action potentials in single nerve terminals of pyramidal cells in layer 2/3 of the young rat neocortex


  • Helmut J. Koester,

    Corresponding author
    1. Abteilung Zellphysiologie, Max-Planck Institut für medizinische Forschung, Jahnstraße 29, D-69120 Heidelberg, Germany
      Corresponding author
      H. J. Koester: Abteilung Zellphysiologie, Max-Planck Institut für medizinische Forschung, Jahnstraße 29, D-69120 Heidelberg, Germany. Email:
    Search for more papers by this author
  • Bert Sakmann

    1. Abteilung Zellphysiologie, Max-Planck Institut für medizinische Forschung, Jahnstraße 29, D-69120 Heidelberg, Germany
    Search for more papers by this author

Corresponding author
H. J. Koester: Abteilung Zellphysiologie, Max-Planck Institut für medizinische Forschung, Jahnstraße 29, D-69120 Heidelberg, Germany. Email:


  • 1Calcium dynamics associated with a single action potential (AP) were studied in single boutons of the axonal arbor of layer 2/3 pyramidal cells in the neocortex of young (P14-16) rats. We used fluorescence imaging with two-photon excitation and Ca2+-selective fluorescence indicators to measure volume-averaged Ca2+ signals. These rapidly reached a peak (in about 1 ms) and then decayed more slowly (tens to hundreds of milliseconds).
  • 2Single APs and trains of APs reliably evoked Ca2+ transients in en passant boutons located on axon collaterals in cortical layers 2/3, 4 and 5, indicating that APs propagate actively and reliably throughout the axonal arbor. Branch point failures are unlikely to contribute to differences in synaptic efficacy and reliability in the connections made by layer 2/3 pyramidal cells.
  • 3AP-evoked Ca2+ transients in boutons were mediated by voltage-dependent Ca2+ channels (VDCCs), predominantly by the P/Q- and N-subtypes.
  • 4Ca2+ transients were, on average, of significantly larger amplitude in boutons than in the flanking segments of the axon collateral. Large amplitude Ca2+ transients in boutons were spatially restricted to within ≤ 3 m of axonal length.
  • 5Single AP-evoked Ca2+ transients varied up to 10-fold across different boutons even if they were located on the same axon collateral. In contrast, variation of Ca2+ transients evoked by successive APs in a given single bouton was small (coefficient of variation, c.v. ≤ 0.21).
  • 6Amplitudes of AP-evoked Ca2+ signals did not correlate with the distance of boutons from the soma. In contrast, AP-evoked Ca2+ signals in spines of basal dendrites decreased slightly (correlation coefficient, r2= -0.27) with distance from the soma.
  • 7Measurements with the low-affinity Ca2+ indicator Magnesium Green suggest that the volume-averaged residual free [Ca2+]i in a bouton increases on average by 500 nM following a single AP. Higher concentrations of indicator caused, on average, a decrease in the amplitude and an increase in the decay time constant of Ca2+ transients. Assuming a single-compartment model the concentration dependence of decay time constants suggests a low endogenous Ca2+ binding ratio close to 140, indicating that of the total Ca2+ influx (≈2 fC) less than 1 % remained free.
  • 8The indicator concentration dependence of decay time constants further suggests that the residual free Δ[Ca2+]i associated with an AP decays with a time constant of about 60 ms (35°C) reflecting a high Ca2+ extrusion rate of about 2600 s−1.
  • 9The results show that AP-evoked volume-averaged Ca2+ transients in single boutons are evoked reliably and, on average, have larger amplitudes than Ca2+ transients in other subcellular compartments of layer 2/3 pyramidal cells. A major functional signature is the large variation in the amplitude of Ca2+ transients between different boutons. This could indicate that local interactions between boutons and different target cells modify the spatiotemporal Ca2+ dynamics in boutons and cause target cell-specific differences in their transmitter release properties.

Action potentials (APs) initiated in the axon of nerve cells propagate actively into the elaborate axonal arbor and its presynaptic terminals to evoke transmitter release via a transient rise in the intracellular Ca2+ concentration ([Ca2+]i). Domains of higher [Ca2+]i activate the exocytotic fusion of vesicles with the plasma membrane and the release of neurotransmitter (for review see Katz, 1969; Neher, 1998). In pyramidal neurons the AP also back propagates into the dendritic arbor (Stuart et al. 1997) and may serve as a reference signal for the detection of coincident pre- and postsynaptic activity (Magee & Johnston, 1997; Markram et al. 1997b) presumably via non-linear postsynaptic Ca2+ signalling (Yuste & Denk, 1995; Koester & Sakmann, 1998) mediated by the N-methyl-D-aspartate (NMDA) receptor channel and dendritic VDCCs (Schiller et al. 1998).

The electrotonic isolation and small diameter of most axonal boutons (< 2 μm) in the central nervous system (CNS) prevented direct electrophysiological studies of presynaptic terminals with the exception of the calyx of Held in the medial nucleus of the trapezoid body (MNTB; Borst & Sakmann, 1996), cerebellar basket cell terminals (Southan & Robertson, 1998) and hippocampal mossy fibre terminals (Geiger & Jonas, 1999). Indirect approaches like high-resolution optical fluorescence microscopy of presynaptic Ca2+ transients in boutons of cortical neurons under physiological conditions were prevented by the image degradation of optical methods in opaque tissue. Thus, optical studies of presynaptic Ca2+ dynamics in the CNS have been limited either to large terminals (Llinas et al. 1992; Helmchen et al. 1997), to averages over a large number of terminals (Regehr & Tank, 1991; Wu & Saggau, 1994a; Regehr & Atluri, 1995; Sinha et al. 1997), to cerebellar basket cell terminals (Llano et al. 1997; Tan & Llano, 1999) or to neurons in primary cell culture (Mackenzie et al. 1996). We used fluorescence imaging with femtosecond two-photon excitation (Denk et al. 1990), which reduces the limitation mentioned. AP-evoked Ca2+ signals were measured in individual boutons of pyramidal cells in layer 2/3 of the juvenile rat neocortex. The axon collaterals of these neurons form en passant synapses with dendritic spines or the dendritic shaft of different classes of target cells such as pyramidal cells and interneurons. Paired recordings from pyramidal cells and different classes of target cells have shown that the release properties of pyramidal cell terminals that govern the efficacy, reliability and short term plasticity of these connections can be target-cell specific (Markram et al. 1998b; Reyes et al. 1998) and possibly reflect differences in presynaptic Ca2+ dynamics (Rozov et al. 2001). One prerequisite to understand mechanistically such differences in the release properties is to establish the size and time course of AP-evoked Ca2+ inflow into individual boutons and the subsequent clearance and to find out whether differences in Ca2+ inflow and extrusion exist between the different boutons of a pyramidal cell axon arbor.


Slice preparation

Acute slices from the somatosensory cortex were prepared from 13- to 15-day-old (P13-15) Wistar rats. Briefly, a rat was decapitated and the brain was quickly removed, in accordance with local ethical committee guidelines. The sagittal side of the right hemisphere was glued to a tilted (approximately 20 deg towards the pia) block. In ice-cold extracellular solution, sagittal slices (300 μm thick) of the cortex were cut from the pia to the white matter using a vibratome. Brain slices were incubated at 37°C for 30 min prior to experiments and then stored at room temperature. Experiments were performed at a bath temperature of 35°C. Layer 2/3 pyramidal neurons in the somatosensory cortical area were identified using gradient-contrast (Dodt et al. 1998) video-IR imaging. The extracellular solution contained (mM): 125 NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 1 MgCl2, 25 glucose and 2 CaCl2 (Biometra, Goettingen, Germany).


Patch pipettes (3-6 MΩ resistance) for whole-cell voltage recording were filled with 125 mM potassium gluconate, 20 mM KCl, 10 mM Hepes, 4 mM ATP-Mg, 10 mM sodium phosphocreatine, 0.3 mM GTP and a Ca2+ indicator, Oregon Green 488 BAPTA-1 (OGB-1, 200 μM, Kd= 190 nM, Molecular Probes, USA), Oregon Green 488 BAPTA-5N (500 or 2000 μM, Kd= 20 μM, Molecular Probes) or Magnesium Green (100-2000 μM, Kd= 6 μM, Molecular Probes). Initial access resistances were < 15 MΩ. Cells were loaded with the indicator for >20 min before fluorescence measurements to allow sufficient equilibration of the indicator concentration between pipette and cytoplasm. Electrical recordings were made with a patch-clamp amplifier (EPC-7, List-Electronics, Darmstadt, Germany or Axoclamp-2A, Axon Instruments, Foster City, CA, USA) operated in current-clamp mode. The voltage signal was digitized by an AD-converter (ITC-16, Instrutech, Great Neck, NY, USA) controlled by a program (Pulse, Heka Elektronik, Lambrecht, Germany) on a Macintosh computer (Apple, Cupertino, CA, USA). Analyses were performed off-line using commercial software (IGOR WaveMetrics, Lake Oswego, OR, USA) with in-house algorithms.

Calcium imaging

A modified galvanometer scanning unit (TCS 4D, Leica Microsystems, Heidelberg, Germany) was adapted to an upright microscope (BX50WI, Olympus Optical Co., Hamburg, Germany) equipped with a × 40 objective (LUMPlanFL 40×W0.9IR, Olympus). For two-photon excitation (TPE), we used short pulses of 170-200 fs at 890 nm (determined in the focal plane, for method see Koester et al. 1999) at 76 MHz from a Ti:Sa-Laser (MIRA 900F, Coherent, Santa Clara, USA) pumped by a solid-state laser (Verdi 5W, Coherent). Line scan imaging was performed at the highest magnification, resulting in a high duty cycle. Both scan directions and external detection behind objective (NA 0.8) and condenser (oil immersion, NA 1.4) were used for signal collection resulting in a high signal collection efficiency. Transmission- and epifluorescence signals were recorded by photomultiplier tubes (PMTs, R6357, Hamamatsu Photonics, Herrsching, Germany, selected for high quantum efficiency) and summed off-line. To obtain a high signal while minimizing the photo damage we performed a detailed analysis of photo damage with respect to the average excitation laser power (Koester et al. 1999). This determined the excitation intensities that allowed examination of axonal structures for a few seconds with little changes of the observed Ca2+ transients by photo damage. To resolve fast transients the imaging system was used in line-scan mode in order to achieve a high temporal resolution (0.5 or 4.5 ms). The bandwith of the fluorescence recordings (>250 kHz) was much higher than the sampling rate. When averaging several sweeps to improve the S/N ratio the intersweep interval was always >10 s.

Histological procedures

For biocytin staining of axonal and dendritic arbors a protocol similar to that described by Markram et al. (1997a) was used. Cells were filled during recordings with biocytin (3 mg ml−1). Immediately after the experiments cells were fixed for >24 h using 4 % paraformaldehyde in 0.1 M PBS (phosphate-buffered saline, pH 7.4). Between all chemical treatments slices were rinsed several times in 0.1 M PBS. Slices were incubated in 3 % H2O2 for 20 min to block endogenous peroxidases followed by a treatment (1 h) with 2 % Triton X-100 (Sigma) for membrane permeabilization. Slices were then incubated in biotinylated horseradish peroxidase (HRP) conjugated to avidin (ABC elite, Vector labs, Peterborough, UK) for >2 h. Finally, slices were developed with 0.07 % diaminobenzidine-0.01 % H2O2 under visual control for 1-5 min. The reaction was stopped by washing with PBS. After washing slices were mounted on a slide using Mowiol (Hoechst, Frankfurt, Germany).

Sensitivity of optical recording

The basal fluorescence of recordings from boutons was on average only 13 ADUs (analog-digital converter units) higher than the background (n= 10; randomly selected samples). On average 44 pixels covering the bouton examined were averaged to obtain one data point in time of the fluorescence F (t). From experiments with fluorescence solutions (Koester et al. 1999) it was estimated that the gain of the system for a PMT supply voltage of 920 V was about 4 ADU (photon)−1. This estimate is based on the assumption that the signal electrons generated in the PMT are Poisson distributed as are the fluorescence photons. Furthermore, it is assumed that the noise of the amplification is significantly lower than the variation caused by the statistical nature of the photon distribution. Based on this, it can be estimated that the number of photons collected from the bouton under study was about 13 × 44/4 = 143 photons per sample. A fluorescence change that deviated more than 2 s.d. from the noise of the base line can be detected with high reliability. Two standard deviations (s.d.) correspond to an increase of about 24 photons per sample.

Data analysis

Fluorescence line scan images were analysed using custom software. A line was scanned every 2.27 ms. Pixels of two lines between two positions enclosing an axon or dendrite were averaged to obtain one time point. This resulted in a temporal resolution of 4.5 ms. Stimulation protocols began 150 ms after the start of the line scan (512 lines = 1162 ms). Before stimulation, fluorescence was averaged for 100 ms in order to obtain the basal fluorescence, F0. A region distant to any indicator-containing structure was chosen for determination of the background fluorescence, FB, which was subtracted. Relative fluorescence changes were calculated as ΔF/F(t) = (F(t) –F0)/(F0FB), and were fitted with a single exponential function using a least-squares fit routine (IGOR WaveMetrics). This fit yielded the amplitude of the fluorescence increase and decay time constant τ. All results are quoted as means ±s.d. unless otherwise indicated. P < 0.01 indicates significance using Student's t test.

Single-compartment model

A single-compartment model as described by Neher & Augustine (1992) was used for estimating the endogenous Ca2+ binding ratio. Several assumptions underlie the single-compartment model: (1) Ca2+ influx is instantaneous; (2) diffusional equilibration is assumed at the peak of the response; (3) Ca2+ binding ratios are constant; and (4) the Ca2+ clearance mechanism is assumed to be linear and non-saturable. The validity of these assumptions is discussed by Helmchen et al. (1997). Briefly, since the Ca2+ inflow was short compared with the time resolution of fluorescence measurements, the approximation of instantaneous Ca2+ influx is probably valid. Further, for a given Ca2+ load the Ca2+ binding ratio is constant. However, the finding that a part of the boutons showed a clear deviation from a single-exponential decay is an indication either that the Ca2+ clearance mechanism is not linear and not non-saturable or that steady-state conditions are not reached at the peak of the response (see Discussion).

The differential binding ratio for the binding of Ca2+ to the buffer B (forming the complex CaB):


is a function of the calcium load to the compartment. To be able to apply the single-compartment model we used the incremental binding ratio. This has a constant value for a given calcium load.

The incremental binding ratio is given by (Neher & Augustine, 1992):


The decay time constant τ in the single-compartment model can be calculated as:


where γ denotes the calcium clearance rate and κS the endogenous binding ratio.

For low-affinity dyes, fluorescence measurements were converted to absolute Ca2+ amplitudes using (Jaffe et al. 1992):


where (ΔF/F)max denotes the maximal fluorescence change for saturating Ca2+ levels and Kd the dissociation constant for the binding of Ca2+ to the fluorescent indicator.

Pharmacological compounds

All voltage-dependent Ca2+ channels (VDCCs) were blocked with 100 μM Cd2+; P/Q-type channels with 0.1 μM ω-agatoxin IVA (Calbiochem, San Diego, CA, USA), N-type channels with 1 μM ω-conotoxin GVIA (Tocris, Bristol, UK), L-type channels with methoxyverapamil (D-600, 100 μM, Sigma) and low-voltage-activated (LVA) Ca2+ channels with 50 μM Ni2+.


Axonal arbor of layer 2/3 pyramidal neurons observed by fluorescence imaging and biocytin-HRP staining

Figure 1A shows dendrites and axon collaterals of a pyramidal neuron in layer 2/3 visualized by fluorescence imaging with two-photon excitation. This neuron was loaded during whole-cell recording with the high-affinity Ca2+ indicator Oregon Green 488 BAPTA-1 (OGB-1). The main axon was readily identified as a single neurite emerging from the soma, orthogonal to the pia, extending down towards the deeper cortical layers. The arbor of axon collaterals spread over a wider range than the dendritic arbor, both in the horizontal and vertical directions. The higher-order axon collaterals were smaller in diameter than dendrites and lacked spines. The morphology of the axon and its collaterals observed by fluorescence imaging corresponded well to the morphology that became apparent after biocytin filling and visualization by the horseradish peroxidase reaction (Fig. 1B).

Figure 1.

Comparison of layer 2/3 pyramidal cell dendritic and axonal arbor in fluorescence and bright-field image after biocytin-HRP staining

A, fluorescence image of a layer 2/3 pyramidal neuron using two-photon excitation imaging. The image is a patchwork of highlight projections through two stacks of 23 frames. The neuron was filled via a somatic patch pipette with the fluorescence indicator OGB-1 (200 μM). B, bright-field micrograph of the same neuron that was also filled during whole-cell recording with biocytin (3 mg ml−1) via the patch pipette, fixed and developed afterwards with biocytin-HRP conjugated to avidin.

Comparison of fluorescence and bright-field images of micrographs of biocytin-labelled cells at higher magnification (Fig. 2A-F) showed that the axonal and dendritic morphology seen by the two methods correlated well (11 neurons). In Fig. 2A and B and 2C and D, the brighter varicosities (arrows) correspond to axonal en passant boutons which were strictly colocalized in fluorescence and HRP bright-field images. Collaterals that branched from the main axon had bright varicosities arranged in a characteristic ‘beads on a string’ pattern (Fig. 2D). These had a somewhat larger diameter than the axon segments flanking them and a higher basal fluorescence.

Figure 2.

Comparison of subcellular axon and dendrite structures as observed with fluorescence imaging and bright-field microscopy of biocytin-HRP stained cells (same cell as in Fig. 1).

A and B, axon collateral branching from main axon (left). Fluorescence image (A) and micrograph of biocytin-HRP stained axon (B) of the middle region as indicated by frames in Fig. 1. The main axon (left) gives rise to a secondary branch and a branch point for third-order branches (arrows). C and D, en passant varicosities on axon collateral. Fluorescence image (C) and photomicrograph of biocytin-stained (D) axon collateral with several clearly visible en passant varicosities (arrows). Note that the brighter structures of the fluorescence image in C correspond to the ‘beads on a string’ structures of the biocytin staining in D. E and F, basal dendrite and dendritic spines. Panels show the regions (left frames) indicated in Fig. 1 at higher magnification. Spine-like protrusions colocalize in fluorescence and bright-field images.

The higher fluorescence intensity was, presumably, a consequence of the higher overlap of the point-spread function of the excitation light and indicator-filled structure and presumably did not indicate a different resting intracellular calcium or indicator concentration.

The close correlation between bright spots in fluorescence images and axon varicosities suggest that the bright spots in fluorescence images correspond to axonal boutons of the ‘en passant’ type. A similar correspondence in fluorescence and bright-field images was found for spines of the dendritic tree (Fig. 2E and F).

Calcium transients in boutons evoked by a single action potential

Upon initiating a single AP by brief somatic current injection (300-600 pA, 4-10 ms), a transient increase in the fluorescence of axonal varicosities was observed (Fig. 3A and B, left panels) using a high-affinity calcium indicator (OGB-1, 200 μM). The amplitude of the AP-evoked Ca2+ fluorescence transient (given as ΔF/F, see Methods) was, on average, almost twice as large in boutons than in the adjacent axon collaterals (ΔF/F= 1.32 ± 0.68, n= 95 and ΔF/F= 0.74 ± 0.25, n= 13, respectively), this difference being statistically significant (P < 0.01). Further, the amplitude of the Ca2+ transients in boutons was often larger than the amplitude of Ca2+ transients in dendritic spines (Fig. 3A andB, right panels).

Figure 3.

Comparison of Ca2+ transients evoked by an AP in a single bouton and a single spine of a layer 2/3 neuron

A, optical detection of boutons (left) and spines (right) in a pyramidal neuron. Scans of fluorescence change were made repetitively along the horizontal lines indicated. B, APs evoked by somatic current injection (400 pA, 5 ms) are shown in the upper traces. Lower traces show Ca2+ transients recorded in the bouton (left) and the spine (right) using OGB-1 (200 μM) as the indicator. Peak amplitudes were ΔF/F= 1.9 and 0.7, respectively. The decay of the transient in the bouton was fitted satisfactorily by superposition of two exponential functions (continuous lines, black and blue). The respective decay time constants were 29 and 290 ms. The transient decay in the spine was fitted satisfactorily with a single time constant of 370 ms.

The Ca2+ transients increased rapidly to a peak value that could be as large as ΔF/F∼3.5 and then decayed to the resting level over a few hundreds of milliseconds (Fig. 3B, left) comparable with the time course of Ca2+ transients in spines (Fig. 3B, right). The decay often had two components (see below). To estimate the rise time of the fluorescence transients, line scans were made at a higher temporal resolution of 0.5 ms (all other measurements being at 4.5 ms). Calcium transients evoked by a single AP in a bouton (Fig. 4A and B) had, on average, a rise time of 1.2 ± 0.1 ms (n= 13, mean ± s.e.m). In dendritic spines (Fig. 4A and C) the rise time was 1.3 ± 0.1 ms (n= 10, mean ±s.e.m.). The real rise time in [Ca2+]i presumably is much faster and these values are upper limits (see Discussion).

Figure 4.

Fast rise time of AP-evoked Ca2+ transients in a single bouton

A, AP evoked by brief somatic current injection. B, fluorescence increase evoked by a single AP in a bouton of a layer 2/3 pyramidal cell loaded with OGB-1 (200 μM). Scan was made at a time resolution of 0.5 ms. Continous lines show the exponential fits to the rise- and decay time course. The rise time constant was 1.11 ms and the decay time constant was 590 ms. C, fluorescence increase evoked by single AP in a spine. The fitted rise time constant was 1.05 ms; the decay time constant was 660 ms.

Large AP-evoked Ca2+ transients in the axonal arbor were restricted to boutons. Within a few micrometres (< 3 μm) away from the centre of a bouton the amplitude dropped to the average amplitude found for segments of the axon collateral between boutons (Fig. 5A and B). In contrast, the amplitude of AP-evoked Ca2+ transients in dendritic spines was only slightly and not significantly higher (P= 0.07; Fig. 5C and D) than that recorded in the dendritic shaft.

Figure 5.

Spatial extent of Ca2+ transients in boutons

A, fluorescence image of an en passant bouton of a layer 2/3 pyramidal cell axon collateral. B, amplitudes of AP-evoked Ca2+ transients in five different boutons. Individual recordings were made at centres of a boutons and at distances of 3 and 6 μm (○). The respective averages are shown as •. C, fluorescence image of spines of a basal dendrite. D, amplitudes of AP-evoked Ca2+ transients in centres of spines and at distances of 2 and 6 μm from the spine (○). The respective averages are shown as •.

Calcium inflow is mediated by P/Q- and N-type VDCCs

Ion channels mediating voltage-dependent Ca2+ inflow into boutons were characterized pharmacologically by bath application of compounds known to reduce Ca2+ influx by blocking VDCCs. The AP amplitude and time course and the resting membrane potential (with the exception of baclofen application), measured at the soma, were not measurably affected by these compounds. The degree of block effected by each compound was measured by comparing averaged Ca2+ transients (using OGB-1, 200 μM, intersweep interval >10 s, three sweeps), before and after (wash-in time of 10-30 min) bath application (Fig. 6).

Figure 6.

Effect of specific Ca2+ channel subtype blockers on AP-evoked Ca2+ fluorescence transients in boutons and dendrites

A, averaged records of Ca2+ transients in boutons (left family of traces) and shafts of basal dendrites (right traces) before and after application of various VDCC subtype blockers. Time of AP is indicated by arrows. In each pair of traces the record with the lower amplitude is that recorded in the presence of blocker. Blockers used were the same as in Table 1. B, average effect of VDCC blockers in boutons (□) and shafts of basal dendrites ( inline image). Error bars indicate s.d.*Significant difference between the effect of the various blockers on boutons and dendrites. The values listed in Table 1 are illustrated here in bar graphs.

Cadmium (100 μM) effectively abolished AP-evoked Ca2+ transients in boutons as well as in spines (to 2 ± 1 % of control, n= 10), indicating that Ca2+ influx is mediated by VDCCs. The P/Q-subtype channel blocker ω-agatoxin IVA (0.1 μM) reduced boutonal Ca2+ transients to 42 ± 25 % (n= 9) vs. 62 ± 34 % (n= 10) in spines, the difference between boutons and spines not being statistically significant (P= 0.15). The N-type channel blocker ω-conotoxin GVIA (1 μM) reduced Ca2+ transients to 49 ± 16 % (n= 12) vs. 90 ± 27 % in spines (n= 10). This difference was statistically significant (P < 0.01). The L-type channel blocker methoxyverapamil (D-600, 100 μM) reduced Ca2+ influx in boutons to 81 ± 17 % (n= 6) and in spines to 64 ± 26 % (n= 5) (not significant: P= 0.23). Large differences between boutons and spines were also found for the effect of Ni2+, known to block at low concentrations predominantly LVA-Ca2+ channels. In boutons application of Ni2+ (50 μM) reduced influx only to 76 ± 20 % (n= 7), much less than in spines (48 ± 12 %, n= 3; P < 0.01). Collectively, these results (Table 1) indicate that in pyramidal neurons of layer 2/3 in the neocortex a mosaic of several VDCC subtypes mediates Ca2+ inflow in boutons. The neuron is polarized between boutons and spines with respect to the subtypes of VDCCs that are present in its axonal and dendritic arbor.

Table 1.  Effect of calcium channel blockers on Ca2+ influx
Reduction toCd2+P/Q-typeN-typeLVA-typeL-typeBaclofen
  1. Pharmacological dissection of VDCC subtypes and GABA effect. All numbers are presented as means ± S.D. (% of control). Cd2+ (100 μM) was used to block all VDCC types, P/Q-type channels were blocked using 0.1 μM ω-agotoxin IV A, N-type channels using 1 μM ω-conotoxin GVIA, LVA-channels using 50 μM Ni2+, L-type channels using 100 μM methoxyverapamil (D-600) and GABA receptors using baclofen (100 μM).

Boutons2 ± 142 ± 2549 ± 1676 ± 2081 ± 1751 ± 16
Spine2 ± 162 ± 3490 ± 2748 ± 1264 ± 26
Basal dendrite2 ± 163 ± 2774 ± 4747 ± 12

Transmitter release from nerve terminals of layer 2/3 pyramidal cells is reduced by activation of GABAB-receptors. Application of the GABAB-receptor agonist baclofen (100 μM) caused hyperpolarization (5-10 mV) of the resting membrane potential at the soma and the Ca2+ influx was reduced to one half (51 ± 16 %; n= 6), consistent with the strong reduction by baclofen of unitary EPSPs recorded in one class of target cells (bitufted, somatostatin positive cells) of pyramidal cell axons in layer 2/3 (Zilberter et al. 1999).

Action potentials propagate reliably into the axonal arbor

Calcium transients in en passant boutons were measured also in response to repetitive pyramid cell stimulation to estimate the reliability of AP propagation from the soma into the axonal arbor of these cells. Two methods and fluorescence indicators were used for the frequency ranges f≤ 10 Hz and 10 Hz ≤f≤ 100 Hz, respectively.

With the high-affinity indicator (OGB-1) the Ca2+ transients evoked by single APs were reliably resolved in that fraction of boutons that showed large transients (ΔF/F > 0.5). For AP frequencies ≤ 10 Hz the reliability of AP propagation at branch points was examined by testing boutons with large transients on each ‘daughter’ branch for AP-evoked Ca2+ signals (Fig. 7). Propagation through a branch point was classified as reliable if both boutons on the two higher-order branches showed no failure in AP-evoked Ca2+ signals (≥ 30 APs). In this way in four neurons, two to four different axon collaterals branching off from the main axon and all their higher-order branches were examined (n= 32 branch points). Axon collaterals branched off from the main axon in layer 2/3, and layers 4 and 5. For frequencies up to 10 Hz all APs propagated reliably through all branch points examined.

Figure 7.

Action potentials propagate reliably in the axonal arbor at low frequency

Left panel, fluorescence image of an axonal branch point. Distal to an axonal branch point on both higher-order branches a bouton was selected (circles) where an AP-evoked Ca2+ transient was detected. Right panel, top trace shows somatic recording of APs evoked by brief current injection. At low AP frequency the fluorescence increase evoked by single APs was detectable in both boutons (lower two traces). Cell was loaded with 200 μM OGB-1.

In three of the neurons a few higher-order branches (< 8 % of all examined) were identified that had abnormal morphology, showing bright, bubble-like structures that suggested that these axon collaterals had been lesioned, presumably during preparation of the slice. Boutons distal to this lesion showed no Ca2+ responses to APs. These axonal branches were also located near the slice surface (≤ 30 μm).

For AP frequencies between 10 and 100 Hz a low-affinity indicator (Magnesium Green, 500 μM) was used to avoid saturation by the large Ca2+ influx occurring at higher AP frequency. A bouton on an terminating branch of an axon collateral was selected (Fig. 8A). The Ca2+ transients evoked by a single AP or by a train of ten APs were compared (Fig. 8B). To test whether AP-evoked Ca2+ transients are similar for each AP in a train, the Ca2+ transient amplitude expected for the high frequency train (Ac) was calculated by superimposing ten times the single-exponential function fitted to the fluorescence response to a single AP:

Figure 8.

Action potentials propagate reliably in the axonal arbor at high frequency

A, bright-field image of a biocytin-stained pyramidal cell, illustrating the bouton (arrow) that was examined for fluorescence transients. B, a single AP (somatic recording, upper left trace) or a train of 10 APs (100 Hz) (upper right trace) evoked fluorescence transients (lower left and lower right traces). C, the fluorescence amplitude expected for the train of spikes (Ac) was calculated from the fluorescence response to a single AP. Ratio of Am/Ac was 0.91. D, boutons in different cells were tested for responses to AP trains of 20, 50, 75 and 100 Hz. The graph shows the ratio of the measured amplitude (Am) and the expected amplitude (Ac) for different frequencies (○). • represent the average ratio for a given AP frequency. Ratios Am/Ac were: 0.97 ± 0.24 (n= 3) for 20 Hz; 1.14 ± 0.12 (n= 3) for 50 Hz; 0.99 ± 0.21 (n= 5) for 75 Hz and 0.79 ± 0.17 (n= 7) for 100 Hz.


In this equation ten single exponential functions of the amplitude Asingle are summated with a time shift of Δt= (1/f). The parameters were selected such that at t= 0 the last exponential (concomitant with the last stimulus) begins (Fig. 8B). Hence the expected amplitude Ac corresponds to the highest peak reached during the last stimulus. The amplitude (Asingle) and decay time were derived from the single AP-evoked Ca2+ fluorescence transient measured in the same bouton.

The expected amplitude for the spike train (Ac) was compared with the measured amplitude (Am) of the Ca2+ transient of the AP train (Fig. 8B and C). In all boutons tested (n= 18), Am was similar to Ac (Fig. 8D, the ratio Am/Ac was 0.94 ± 0.21 (n= 18, average over all frequencies)). If there had been a significant fraction of propagation failures one would have expected a lower amplitude Am, and a ratio Am/Ac significantly smaller than 1. The ratio Am/Ac close to 1 indicates that on average most or all APs propagate through the entire axonal tree. These results demonstrate that with AP frequencies of up to 100 Hz, most APs propagate reliably throughout the axonal arbor in the neocortex of layer 2/3 pyramidal cells.

Calcium accumulation during 10 Hz trains of APs

Synaptic connections between layer 2/3 pyramidal cell axons and their target cells show frequency-dependent depression of EPSP amplitudes while others show facilitation depending on the identity of the target cell (Reyes et al. 1998). Facilitation is dependent on a long lasting rise in [Ca2+]i of nerve terminals (Rozov et al. 2001). We have measured the Ca2+ signal in boutons during 10 Hz trains of APs. To reduce the alteration of the physiological Ca2+ signal by the added indicator we used a low-affinity indicator (Magnesium Green, 500 μM). The low S/N ratio of Ca2+ signals when using this indicator did not allow reliable measurements of the Ca2+ accumulation evoked by AP trains in single boutons. Fluorescence recordings from several boutons in different cells were therefore averaged to get an approximate estimate of the time course of the Ca2+ accumulation.

Figure 9 shows the time course of the Ca2+ signal (average over several boutons and cells) evoked by a train of ten APs at 10 Hz. The basal level of Ca2+ increases with a time course slower than the decay time constant of the individual AP-evoked Ca2+ transients. The time constant of reaching a steady-state Ca2+ level, as indicated by the fluorescence signal, during 10 Hz AP trains was 157 ms (at 35°C), the time constant for the decay of an AP-evoked Ca2+ transient at this indicator concentration was about 70 ms (Fig. 15).

Figure 9.

Time course of summation of AP-evoked Ca2+ transients and steady-state [Ca2+]i levels

Average Ca2+ fluorescence signal evoked by 10 Hz trains of APs made from recordings in 19 boutons of three pyramidal cells. The level of the fluorescence before each transient in the train is indicated by a dot. Steady state is reached with a time constant of 157 ms (single- exponential fit, continuous line).

Figure 15.

Decay time constants of the AP-evoked Ca2+ transients in single boutons loaded with different concentrations of Magnesium Green

A, AP-evoked calcium transients in single boutons. Each trace is an average of 20 recordings in single boutons of different cells filled with 100 μM (left), 500 μM (middle) or 2000 μM (right) Magnesium Green. Continuous lines represent single-exponential fits of the fluorescence decay. Time constants were 37, 62 and 283 ms, amplitudes (ΔF/F) were 0.25, 0.16 and 0.10, respectively. B, distribution of decay time constants at different Magnesium Green concentrations measured in single boutons as shown in A. The peaks of Gaussian fits (continuous lines) were 47 ± 32, 50 ± 28 and 167 ± 124 ms. Mean decay time constants were 45, 70 and 211 ms. C, averages of all fluorescence recordings for each indicator concentration (100 μM, n= 16 boutons in five cells; 500 μM, 23 boutons in eight cells; 2000 μM, n= 13 boutons in three cells). Recordings were averaged over several boutons and cells to obtain the decay time constant. Decay time constants were 37, 73 and 180 ms. The amplitude decreased from ΔF/F= 0.16 (100 μM) to ΔF/F= 0.05 (2000 μM).

Ca2+ transient amplitudes in the axonal arbor

The prominent bright varicosities in the axonal collaterals observed in fluorescence images corresponded to en passant boutons and to axonal branch points. To compare Ca2+ transients in boutons with those in other subcellular compartments of pyramid cells the peak amplitudes of Ca2+ transients evoked by single APs were measured (Fig. 10) when neurons were loaded with a high-affinity indicator (OGB-1, 200 μM).

Figure 10.

Comparison of Ca2+ transients in different subcellular compartments of a layer 2/3 pyramidal neuron

Left panel, fluorescence image of a layer 2/3 pyramidal neuron filled with OGB-1 (200 μM). Image is a highlight projection of a stack of 23 frames. The dark structure on the left is the whole-cell recording pipette. Right panel, Ca2+ transients evoked by a single AP. The top trace shows an AP evoked by brief somatic current injection. Traces below show the Ca2+ fluorescence transients evoked by a single AP in the different subcellular compartments as indicated. Each trace represents the average of 2-5 sweeps. The lines represent fits of single exponentials to the decay. Recovery in the bouton is fitted with two exponentials.

Calcium transients at axonal branch points had a similar amplitude and variation (ΔF/F= 0.7 ± 0.3, n= 9) as those in axon segments between boutons (ΔF/F= 0.74 ± 0.25, n= 13) and in the main axon, the average amplitude being ΔF/F= 0.53 ± 0.24 (n= 22). In comparison the Ca2+ transients in end and en passant boutons had a higher amplitude. Amplitudes were, however, distributed over a broader range. The peak amplitude was on average ΔF/F= 1.32 ± 0.68 (n= 95) and amplitudes ranged from ΔF/F as low as ≤ 0.2 up to 3.5. The low amplitude (ΔF/F≤ 0.2) measured in some boutons did not indicate a lack of Ca2+ inflow since 10 Hz trains of APs always evoked a detectable Ca2+ signal. The larger amplitude in boutons compared with other compartments was statistically significant (for all P < 0.01). For comparison, in the proximal main apical dendrite the average amplitude was ΔF/F= 0.61 ± 0.28 (n= 16), in the basal dendritic branches ΔF/F= 0.61 ± 0.15 (n= 37) and in the spines of basal dendrites ΔF/F= 0.68 ± 0.22 (n= 50).

The results summarized in Fig. 11 suggest that AP-evoked Ca2+ inflow in axon collaterals is highly localized to their boutons. In segments flanking boutons the inflow was significantly smaller. Their variation of Ca2+ transients in axon segments between boutons was also small, indicating that the density of VDCCs in collaterals is almost homogeneous.

Figure 11.

Range of peak amplitudes of Ca2+ transients in the different subcellular compartments of layer 2/3 pyramidal cells

A, distribution of peak amplitudes of Ca2+ transients evoked by single APs recorded in boutons. Continuous line represents a Gaussian fit (1.2 ± 1.0). B, distribution of AP-evoked peak amplitudes in spines of basal dendrites. Continuous line represents a Gaussian fit (0.6 ± 0.3). C, range of Ca2+ transients evoked by a single AP in the shaft of the proximal apical dendrite (Apical), in the shaft of basal dendrites (Basal), in spines of basal dendrites (Spine), in the main axon (Axon), in axon collateral segments in between boutons (Coll) and in boutons. • indicate the average amplitude with error bars ± 1 s.d. and × the range of amplitudes.

Ca2+ transients in single boutons

The large differences in the amplitude of Ca2+ transients between different boutons contrasts with the small variation of the Ca2+ signal of an individual bouton during successive trials. Variability was tested by recording Ca2+ transients in 63 boutons (three sweeps each) using OGB-1 (200 μM). Amplitudes recorded in one particular bouton were normalized to the mean. The normalized recordings of all boutons were pooled and the standard deviation of this distribution was compared with the variation of the average background noise. The distribution was Gaussian (Fig. 12; mean, 1.0) with a variation (s.d., 0.21) that was, on average, not different from the noise variation of the fluorescence recordings (the noise of the basal fluorescence, root mean square ∼3 ADU corresponding to s.d. of 0.2 in ΔF/F).

Figure 12.

Variation of AP-evoked Ca2+ transients in boutons

Distribution of the normalized and pooled amplitudes of the fluorescence transients evoked by single APs recorded in 63 boutons (three sweeps each). Continuous line shows Gaussian fit. Mean, 1.0 ± 0.21.

Spatial profile of Ca2+ transients in the axonal and dendritic arbors

In the dendritic arbor of neocortical pyramidal cells the Ca2+ signals evoked by single APs varies with the distance from the soma (Schiller et al. 1995). However, the spatial profile of AP-evoked Ca2+ signals in the axonal arbor is not known. Therefore, we recorded AP-evoked Ca2+ fluorescence transients in boutons as a function of their geometric distance from the soma (Fig. 13A and B). The peak amplitude was not correlated with the geometric distance between soma and bouton (r2= 0.05). Moreover, up to 5-fold differences in the peak of Ca2+ transients were observed between different boutons of the same axon collateral even when both boutons were in close proximity (5-10 μm). In contrast, AP-evoked Ca2+ transients in dendritic spines showed a slight decrease in amplitude with distance from the soma (r2= -0.27).

Figure 13.

Spatial profile of AP-evoked Ca2+ fluorescence transients in the axonal and basal dendritic arbor

A, fluorescence image of a layer 2/3 pyramid with a long horizontal axon collateral. AP-evoked Ca2+ transients in single boutons (shown as traces below the image) were mapped along this branch as indicated by the circles. B, amplitude of the AP-evoked Ca2+ transient in boutons of axon collaterals. Graph shows amplitudes of the Ca2+ fluorescence transients evoked by one AP plotted against the geometric distance from soma, indicating lack of correlation (r2= 0.05). Note that axonal length is longer than geometric distance. C, amplitude of the AP-evoked Ca2+ signal in basal dendrites. Amplitude of AP-evoked Ca2+ transients in spines of a basal dendrite. The amplitude decreased with distance (r2= -0.27).

Decay time course of Ca2+ transients in single boutons

The differences in the decay of the AP-evoked Ca2+ transients when using OGB-1 as the indicator suggested that boutons fall into two broad classes. Figure 14A and B shows that Ca2+ transients recorded in different boutons of the same axon collateral can differ substantially in their decay time courses. In one bouton the decay was fitted satisfactory only by the sum of two exponentials, in the other bouton it was fitted by a single exponential. The decay was classified as two-exponential if a fit with the sum of two exponential functions yielded two time constants differing by more than 2-fold. On average 65 % (n= 26 vs. n= 14) of the boutons had double-exponential decays. The longer time constant was similar to the time constant of transients in boutons with single-exponential decays (429 and 495 ms, respectively). The shorter time constant of double-exponential decays was 60 ± 50 ms (n= 26). The peak amplitude in these boutons was, on average, significantly higher than in boutons with single-exponential decays (Fig. 14C and D, P < 0.01).

Figure 14.

Double- and single- exponential decay time course of AP-evoked Ca2+ transients in boutons

A, AP-evoked Ca2+ transient recorded in a bouton loaded with OGB-1 (200 μM). The decay could be fitted satisfactory with the sum of two exponentials (time constants were 552 and 40 ms). Histograms on the right show the distribution of the slow (τ1) and fast (τ2) time constants of recordings in 68 boutons. The Gaussian fits yielded 353 ± 190 ms and 22 ± 48 ms. B, calcium transient in a bouton with a single-exponential decay. The time constant is 519 ms. Histogram on the right shows the distribution of recordings in 16 boutons, where the decay was fitted with a single exponential. Mean, 402 ± 214 ms. C and D, amplitude distribution of Ca2+ transients with a double-exponential decay (C) and amplitudes of transients with a single-exponential decay (D).

Endogenous Ca2+ binding ratio of boutons

The endogenous Ca2+ binding ratio and the Ca2+ clearance rate determine the global (volume-averaged) Ca2+ dynamics in a bouton. To estimate the size of AP-evoked Ca2+ influx, the binding ratio and the clearance rate we loaded cells with different concentrations of Ca2+ indicators. The indicators used did not allow ratiometric calibration of the fluorescence recordings and prevented direct determination of the binding ratio in a single bouton. If a low-affinity indicator (Magnesium Green or Oregon Green 488 BAPTA-5N) with a dissociation constant that is much higher than the resting [Ca2+]i is used and a single-compartment model (see Methods) is assumed, the exogenous incremental Ca2+ binding ratio κ‘B (of the added indicator) can be approximated by κ‘B=[B]/Kd, where [B] is the concentration of the exogenous added buffer and Kd its dissociation constant. We assumed Kd= 6 μM for Magnesium Green and Kd= 20 μM for Oregon Green 488 BAPTA-5N as listed in the provider's data sheet.

Pyramidal cells were loaded with Magnesium Green at concentrations of 100 μM (n= 16 boutons, 5 cells, κ‘B= 17); 500 μM (n= 23 boutons, 8 cells, κ‘B= 83) and 2000 μM (n= 13 boutons in 3 cells, κ‘B= 333) or with Oregon Green 488 BAPTA-5N at concentrations of 500 μM (n= 17 boutons, 4 cells, κ‘B= 25) and 2000 μM (n= 23 boutons in 4 cells, κ‘B= 100). Decays of AP-evoked Ca2+ transients recorded in boutons were fitted with a single exponential (Fig. 15A). All fluorescence recordings for a particular indicator and indicator concentration were averaged across different boutons and neurons and the decay time course of these averaged signals was fitted with a single exponential (Fig. 15C). Determining the average decay time constant for a given buffer concentration in a different way, by fitting individual responses with a single exponential and averaging the decay time constants, yielded comparable values (Fig. 15B).

Table 2 summarizes the measured decay time constants. Time constants plotted as a function of the exogenous incremental Ca2+ binding ratio, κ‘B, demonstrate that decays are longer at higher indicator concentrations (Fig. 16). A linear regression extrapolates to the value of the endogenous Ca2+ binding ratio of the boutons (i.e. the ratio of bound Ca2+vs. free Ca2+, after initial equilibration, eqns (1) and (2)), which was 141. The decay time constant of an AP-evoked Ca2+ transient in a bouton without exogenous buffer, derived from the Y-axis intercept, was 56 ms and the Ca2+ clearance rate was calculated (eqn (2)) to be about 2600 s−1. For basal dendritic shafts a decay time constant of 57 ms, an endogenous Ca2+ binding ratio of 112 and a Ca2+ clearance of 2000 s−1 were derived. Spines of basal dendrites had very similar decay time courses as the dendritic shafts for all indicators and indicator concentrations used. This suggests that dendritic spines have a comparable endogenous Ca2+ binding ratio and clearance rate as the dendritic shafts.

Table 2.  Decay time constants of Ca2+ transients for different indicators
 Indicator/concentration100 μM500 μM2000 μM
  1. Summary of decay time constants (given in ms) at different indicators, indicator concentrations and compartments. The decay time constants for Ca2+ fluorescence transients evoked by a single AP were measured in different compartments with vairous concentrations of the Ca2+ indicators Magnesium Green and Oregon Green 488 BAPTA-5N.

BoutonsMagnesium Green37 ± 473 ± 5180 ± 17
 Oregon Green 2N73 ± 7141 ± 22
Basal dendritesMagnesium Green64 ± 1370 ± 26222 ± 140
(Shaft)Oregon Green 5N65 ± 7149 ± 22
Apical dendrites (shift)Oregon Green 5N112 ± 5
Figure 16.

Estimate of the average endogenous Ca2+ binding ratio of boutons in axon collaterals and basal dendrites of layer 2/3 pyramidal cells

A, the decay time constants of Ca2+ transients in boutons loaded with different indicator concentrations (•, Magnesium Green; ○, Oregon Green) are plotted against the exogenous incremental Ca2+ binding ratios κB . The straight line represents a linear regression. The decay time constant of Ca2+ transients in a naive bouton is on average 56 ms, the endogenous Ca2+ binding ratio is 141, from X- and Y-axis intercepts of regression line. B, decay time constants of Ca2+ transients in basal dendrites (•, Magnesium Green; ○, Oregon Green) plotted against the exogenous Ca2+ binding ratios κ′B. The decay time constant of Ca2+ transients in basal dendrites is 57 ms and the endogenous Ca2+ binding ratio is 112.


The results indicate that AP-evoked calcium fluorescence signals can be measured in single varicosities of the axonal arbor of layer 2/3 pyramidal cells. The main axon that lacked varicosities originated from the soma opposite the apical trunk and branched into collaterals of smaller diameter. These second- and higher-order axonal branches had brighter varicosities of larger diameter. Examination of such varicosities in biocytin-labelled pyramidal cells by electron microscopy shows that they contain vesicles (K. M. M. Kaiser, J. Lübke, Y. Zilberter & B. Sakmann, personal communication), and most likely correspond to functional release sites. In a previous study of neocortical cells in primary cell culture, sites of vesicular release were attributed to similar structures (Mackenzie et al. 1996). Further evidence for varicosities being boutons derives from the fact that Ca2+ transients in the axon segments flanking varicosities were, on average, of smaller amplitude, in accordance with other studies (Mackenzie et al. 1996; Llano et al. 1997; Tan & Llano, 1999). Adjacent to a varicosity with a large amplitude Ca2+ signal, the amplitude decreased steeply with distance (< 3 μm) to a size that was also detected in collaterals lacking varicosities. Finally, the rise time of Ca2+ fluorescence transients in varicosities was short (about 1 ms). The rise time of the fluorescence signal is limited by the temporal resolution of the line scan and does not reflect the time course of Ca2+ entry which is substantially faster (Borst & Sakmann, 1996; Borst & Sakmann, 1998). The rise time may, however, represent an upper estimate of the time needed for the equilibration of Ca2+ with the endogenous Ca2+ buffers and the indicator in a bouton. In contrast, the rise time of AP-evoked fluorescence transients was longer (about 7 ms) in axons of cerebellar basket cells (Tan & Llano, 1999).

Collectively, the findings suggest that single boutons of layer 2/3 pyramidal cell collaterals were unequivocally identified and that the AP-evoked fluorescence transients represent the volume-averaged increase in [Ca2+]i and its decay in single boutons. The determinants of these Ca2+ transients will be discussed with respect to synaptic transmission in pyramidal cell synapses (Reyes et al. 1998; Rozov et al. 2001).

Reliable propagation of APs into the axonal arbor

Conduction of APs along an excitable neurite may fail at branch points. Branch point failures were suggested as one mechanism contributing to variability in synaptic transmission (Lüscher & Shiner, 1990; Stoney, 1990) and such failures were reported to occur after hyperpolarizing pulses in hippocampal synapses (Debanne et al. 1997). In pyramidal cell boutons, at low frequencies of ≤ 10 Hz branch point failures were not observed when measuring the Ca2+ inflow evoked by APs in boutons located on the sub-branches distal to a branch point. When boutons were close to a branch point (e.g. as in Fig. 7) passive AP spread might have accounted for activation of boutonal VDCCs. However, the measurements in higher-order branch points also implied that distant boutons (> 100 μm distal to the soma) were tested. Further, at AP frequencies >10 Hz, boutons on highest-order branches were identified that responded reliably with Ca2+ signals to APs without indication of failure. AP propagation seemed to fail only in branches that showed morphological signs of damage presumably inflicted during preparation of the brain slice. A high reliability of AP propagation into the axonal arbor was also found in cultured cortical cells (Mackenzie et al. 1996) and in layer 5 pyramidal cells (Frenguelli & Malinow, 1996).

Thus, under brain slice recording conditions with reduced synaptic input and at AP frequencies of up to 100 Hz, branch point failures are unlikely to contribute substantially to the variability of synaptic transmission between layer 2/3 pyramidal cells and their target neurons in a cortical column.

Calcium influx and VDCC subtypes

Pharmacological dissection of VDCC subtypes indicates that P/Q-type channels (ω-agatoxin IVA sensitive) which are thought to be linked tightly to transmitter release (Reuter, 1996) are responsible for the largest fraction of the AP-evoked Ca2+ influx in boutons. N-type channels (ω-conotoxin GVIA sensitive), known to contribute to evoked transmitter release in other synapses (Sheng et al. 1998), contributed less. The P/Q- and N-type channels (ω-agatoxin IVA sensitive) mediated most (≥ 70 %) of presynaptic AP-evoked Ca2+ influx. In contrast, low-voltage-activated (LVA) channels (Ni2+-sensitive) dominated AP-evoked Ca2+ influx into dendrites and spines of L2/3 pyramidal cells, in contrast to hippocampal CA1 pyramidal cells (Yuste et al. 1999), where Ni2+ (50 μM) did not block a significantly dendritic Ca2+ influx (but see Christie et al. 1995; Magee et al. 1996). Previously, another study (Wu & Saggau, 1994b) also reported different effects of ω-agatoxin IVA and ω-conotoxin GVIA on hippocampal terminals of the guinea-pig (21 % block for ω-agatoxin IVA and 41 % for ω-conotoxin GVIA). The sum of the pharmacological effects exceeded 100 %. This can be explained by the observed decrease in amplitude of Ca2+ transients during the whole-cell recording due to rundown and/or long equilibration times of indicator.

The contributions of the VDCC subtypes to the total Ca2+ influx in the apical dendritic trunk of layer 5 pyramidal cells (Markram et al. 1995) were different from those reported here for basal dendrites of layer 2/3 pyramidal cells. While we found only a minor contribution of N-type channels (26 ± 47 %) to total dendritic Ca2+ influx measured in the shafts, Markram et al. (1995) reported a larger contribution (28 ± 3 %, ±s.e.m.). Also the results obtained for the P/Q-type channel blockers differ (38 ± 34 vs. 10 ± 3 %).

Variability of Ca2+ transients in single boutons

Analysis of the variation of amplitudes (ΔF/F) of relative fluorescence changes of Ca2+ transients in single boutons was limited by the effects of photo damage that restricted the exposure time to the recording of three to ten responses in a single bouton, depending on the excitation light intensity. Therefore, the recordings of a large number of different boutons had to be pooled. The distribution of the normalized and pooled relative amplitudes of Ca2+ transients was fitted by a Gaussian distribution where the variation of the normalized amplitudes (s.d.= 0.21) matched the variation of the background. This indicates that the variation of amplitudes of Ca2+ transients in a bouton is much lower than the variation of the background and the variation of Ca2+ influx evoked by different APs presumably contributes only little to the variation of postsynaptic responses in the target cells of L2/3 pyramidal cells. This is in contrast to results for cortical cells in tissue culture (Mackenzie et al. 1996) and layer V pyramidal cells (Frenguelli & Malinow, 1996).

Variability of Ca2+ transients in different boutons

AP-evoked Ca2+ transients in different boutons, even when they were located on the same axon collateral, varied over a wide range, as the amplitudes (ΔF/F) ranged from ∼0.2 up to 3.5. A comparable amplitude heterogeneity was observed in axon varicosities of cerebellar basket cells (Llano et al. 1997). Several mechanisms could account for this variability between different boutons.

Different surface-to-volume ratios may be one cause. To account for large differences the radius has to vary over the same wide range as the amplitudes because the surface-to-volume ratio changes with r−1 (where r= radius). The size of a bouton could only be determined in two dimensions with sufficient spatial resolution and no correlation between bouton size and amplitude of the Ca2+ transient was found for the boutons analysed. Further, the axonal shafts have a higher surface-to-volume ratio than the varicosities but APs evoked, on average, smaller Ca2+ transients.

Large differences in the amplitude of Ca2+ transients were observed even within a few micrometres of axonal length, comparing bouton with collateral segment and bouton with bouton on the same collateral. Within the short distances between boutons an AP could be significantly attenuated by impedance mismatch only when the diameter of the bouton is severalfold larger than the axonal shafts (see Lüscher & Shiner, 1990, for boutons close to branch points). The axonal AP waveform is changed by blocking potassium channels and this affects the Ca2+ influx into axons (Tan & Llano, 1999). However, it seems unlikely that differences in potassium channel density between individual boutons will greatly affect the AP waveform locally.

Differences in the endogenous Ca2+ binding ratio between boutons may be another cause for heterogeneity of Ca2+ transients in different boutons. To explain the large variation in the amplitudes between boutons the Ca2+ binding ratio of different boutons has to vary over the same wide range. To test this possibility the binding ratio has to be determined for individual boutons, which was not achieved.

The amplitude of the Ca2+ signal in individual boutons is presumably dependent on the local density and/or the subtype of VDCCs. As the size of Ca2+ transients is not correlated with distance, the variation in Ca2+ dynamics between different boutons could be determined by local interactions with the target cells, that for example change the density of presynaptic Ca2+ channels.

Calcium dynamics in boutons

Despite the large variation between boutons we attempted to obtain rough estimates of the determinants of residual free Ca2+ in a bouton following a single AP. We have used the average values of peak amplitudes and decay time constants of Ca2+ transients pooled over many (>100) boutons and the values derived are order of magnitude estimates.

Ca2+ clearance

When using a high-affinity indicator the prominent double-exponential decay of a fraction of all measured Ca2+ transients (> 50 %) indicates that the assumption of a single compartment may not be completely valid for model calculations. In addition in a single-compartment model the decay of Δ[Ca2+]i following a single spike and the increase in [Ca2+]i in a spike train have the same time constant, which was not observed (Fig. 9).

A biphasic decay may have different causes. If several buffer species with different Ca2+ binding characteristics are present in a bouton, a biphasic decay can occur (Markram et al. 1998a; Maeda et al. 1999; Lee et al. 2000). An additional Ca2+ pump with co-operative calcium binding or having a threshold which could be located in the plasma membrane or in an intracellular membrane (ER, mitochondria) may be another cause for the multi-exponential decays. The Na+-Ca2+ exchanger that contributes to Ca2+ clearance also can cause a non-stationary clearance rate (Fierro & Llano, 1996). Finally, since an en passant bouton is not a closed compartment, diffusion of Ca2+ into the adjacent segment of the collateral may also affect the Ca2+ clearance rate.

Endogenous Ca2+ binding ratio

The size of the volume-averaged Δ[Ca2+]i is governed by the endogenous Ca2+ binding ratio of a bouton. Fluorescence measurements were, however, not sufficiently sensitive to determine the binding ratio in individual boutons. Photo damage limited repeated optical recordings from the same bouton. Averages from pooled measurements with a low-affinity indicator in many boutons and different cells were used to describe the average behaviour of Ca2+ dynamics in boutons. With these restrictions a relatively low endogenous Ca2+ binding ratio of 141 and a fast decay time constant of 56 ms was estimated for the Ca2+ transient evoked by an AP in a naive bouton. Several assumptions underlie these estimates. Firstly, to calculate the incremental exogenous Ca2+ binding ratio, κ‘B, we assumed that the amplitude of the AP-evoked Δ[Ca2+]i is small compared with the maximal fluorescence change. Since the average fluorescence increase, ΔF/F, evoked by an AP did not exceed 20 % of the maximal fluorescence change, (ΔF/F)max, and was on average ∼10 %, when using Magnesium Green, this assumption was presumably valid. Secondly, we used as a first approximation a single exponential to fit the decay of fluorescence. This was necessary to use the single-compartment model. Further, we assumed a dissociation constant of 6 μM for the binding of Ca2+ to Magnesium Green and 20 μM for Oregon Green 488 BAPTA-5N. These values might differ for the experimental conditions used. For example in frog skeletal muscle fibres an apparent dissociation constant of 19 μM was measured for the binding of Ca2+ to Magnesium Green (Zhao et al. 1996). This is consistent with the observation that other indicators also have higher apparent dissociation constants in a cytoplasmic environment. Thus the value we estimated (κS= 141) may represent an upper limit for the average Ca2+ binding ratio of boutons. Interestingly, for the basal dendrites of layer 2/3 cells a very similar endogenous Ca2+ binding ratio of 112 was estimated. This could indicate that basal dendrites and axonal boutons have comparable endogenous Ca2+ buffers.

Ca2+ influx into boutons

When a low-affinity indicator is used for fluorescence measurements such as Magnesium Green, the amplitude of the Ca2+ transient in a bouton can be approximated by Δ[Ca2+]i=KdΔF/F/(ΔF/F)max (see eqn (3)), for transients with small amplitudes (ΔF/F < < (ΔF/F)max). This condition was fulfilled for transients evoked by single APs. Their amplitude (ΔF/F) was, on average, ∼10 % of the maximal fluorescence change (ΔF/F)max∼1.5 (as estimated from responses to high frequency AP trains (100-200 Hz) evoked at the end of an experiment). The amplitudes of the fluorescence transients evoked by a single AP using 100 μM Magnesium Green were on average ΔF/F= 0.125 ± 0.06 (n= 16) and ranged up to ΔF/F= 0.25. This corresponds to an increase in volume averaged [Ca2+]i of Δ[Ca2+]i= 500 ± 240 nM and a maximal increase of around 1 μM. Again the large s.d. reflects, presumably, the differences in amplitude between boutons. Collectively, the results suggest that the peak of volume averaged Δ[Ca2+]i, evoked by a single AP, is reached fast (rise time ≤ 1.2 ms) after the collapse of the Ca2+ domains around VDCCs, and that the peak residual Δ[Ca2+]i in boutons can reach a relatively high amplitude of up to 1 μM.

Ca2+ channel density in boutons

With an average amplitude of the AP-evoked Ca2+ transient of around 500 nM and assuming that the average volume (V) of a bouton is similar to that reported for hippocampal CA3- CA1 synapses (about 0.13 μm3; Shepherd & Harris, 1998), a Δ[Ca2+]i of 500 nM corresponds to nfree=VΔ[Ca2+]iNA= 40 free calcium ions. Here NA is Avogadro's number. The endogenous Ca2+ binding ratio was about 140 and thus the total number of Ca2+ entering the bouton during an AP would be ntotal= (1 +κ‘B)nfree∼ 5500. Assuming that the single Ca2+ channel current is i= 0.2 pA (Gollasch et al. 1992) at the maximum of the calcium current at around -30 mV membrane potential and the mean open time is Δt= 200 μs (Borst & Sakmann, 1996) the number of Ca2+ ions entering through a single VDCC during a single AP is nchannel=iΔt e½= 125 (where e is the electron charge). This yields an estimate of the number of Ca2+ channels opened in a bouton during a single AP of ntotal/nsingle= 44 channels. Assuming a spherical shape of the bouton, its surface A (in μm2) is given by:


and the density of VDCCs is ntotal/nsingle/A= 36 open channels μm2 of bouton membrane, assuming a homogeneous distribution of VDCCs in the boutonal membrane. Assuming an area of 0.2 μm2 for the active zone (AZ) this would yield about 6 channels per AZ. On the other hand, if all Ca2+ channels were localized in the AZ of a bouton, the density would be 220 Ca2+ channels μm2.

Comparison with the calyx of Held and other CNS terminals

Calyx of Held

Detailed measurements of residual [Ca2+]i dynamics in a CNS synapse have been made in a giant terminal (calyx of Held; Helmchen et al. 1997) in the midbrain of young (P8-11) rats. At this synapse the Ca2+ inflow is mediated mostly by the P/Q-type of VDCCs (Wu et al. 1999) blocked by ω-agatoxin IVA. They contribute 58 % of the total Ca2+ influx whereas the effect of the N-type channel blocker is smaller, contributing about 27 % to Ca2+ influx. These values are qualitatively similar to the effects of these toxins described here for the pyramid nerve terminals. In both terminals the P/Q-type specific blockers reduce the Ca2+ influx most effectively and they are also the most effective in reducing EPSP amplitudes (Wu et al. 1999; Rozov et al. 2001).

The estimated endogenous Ca2+ binding ratio of layer 2/3 pyramid nerve terminals, averaged over many boutons, is larger than that in the calyx of Held (141 vs. about 40, in Helmchen et al. 1997). The average decay time constants are similar (56 vs. 47 ms) whereas the estimated Ca2+ clearance rate is higher in layer 2/3 pyramid nerve terminals (2600 vs. 900 s−1). However, the total Ca2+ load is much smaller in pyramid nerve terminals (1.8 × 103vs. 1 pC). These differences could be due to differences in the geometry of the two types of nerve terminals and to developmental differences. In addition the values derived here for pyramid terminals are averages from many boutons that can have very different release properties and Ca2+ dynamics (Reyes et al. 1998; Rozov et al. 2001).

We estimated the number of VDCCs that are opened by a single AP to be close to 40. Assuming that there is about one active zone (AZ) per bouton (as for the CA3-CA1 pyramid synapse; Shepherd & Harris, 1998) this suggests that during an AP about 40 calcium channels are opened per release site. This value is comparable with the estimate in the calyx of about 30 channels per release site (as inferred from the number of AZs of a calyx, 450-550, authors’ observations) and the total Ca2+ influx during an AP (Borst & Sakmann, 1996). It also could suggest that exocytotic fusion of vesicles in both the calyx and pyramid terminals is controlled by overlapping Ca2+ domains, as was also inferred from the effects of intracellular EGTA on release from pyramidal cell terminals (Ohana & Sakmann, 1998; Rozov et al. 2001).

Hippocampal and cerebellar terminals

When recording an ensemble fluorescence signal from many terminals and their associated axons in the hippocampus (Sinha et al. 1997) and in cerebellar granule cell synapses (Regehr & Atluri, 1995) with low-affinity indicators loaded via the ester form some parameters of [Ca2+]i dynamics of small CNS terminals were derived. For granule cell terminals the rise in residual free [Ca2+]i evoked by one AP was estimated to be about 300 nM. This is somewhat lower than the average value we found for cortical terminals whereas the single-exponential decay time constant (150 ms) was substantially longer. On the other hand the estimate of the total Ca2+ charge entering a hippocampal pyramid cell bouton during an AP (< 1 fC) is comparable with the estimate for cortical pyramid cell boutons.

Functional significance of differences in Ca2+ dynamics

The experiments demonstrate directly that a single AP initiated in the axon, presumably in the vicinity of the soma of a layer 2/3 pyramidal cell, propagates in two directions – back into the dendritic arbor and forward into the axonal arbor to evoke activity-dependent rises in [Ca2+]i in spines and in boutons, respectively. Several differences in the AP-evoked Ca2+ signals are apparent between spines and boutons such as their average amplitude and decay time course, the Ca2+ channel subtypes that mediate the Ca2+ signal and the dependence on the distance from the soma. The most striking difference is, however, in their variability.

The variability is considerably smaller between spines and the decrease in amplitude with distance to soma seems to be caused by the extent of AP back propagation into the dendritic arbor. In contrast in the axonal arbor APs propagate reliably even at high frequencies and presynaptic branch point failures do not appear to contribute to the differences in synaptic transmission between layer 2/3 pyramidal cells and their different target neurons. However, the Ca2+ signals vary severalfold between individual boutons even of the same axon collateral. As the volume-averaged rise in [Ca2+]i is linked indirectly to the higher [Ca2+]i at the Ca2+ sensor at vesicle release sites, the differences in Ca2+ dynamics between boutons may underlie the differences in efficacy, reliability and short term plasticity of synaptic transmission in the connections formed by terminals of layer 2/3 pyramidal cell with different target cells (Reyes et al. 1998; Rozov et al. 2001).


We thank colleagues Drs Carl Petersen and Nathan Urban for critically reading the manuscript, Dr K. Kaiser for CCD-camera measurements with Magnesium Green on basal dendrites, M. Kaiser for technical assistance, Leica Microsystems (Heidelberg) for excellent support, Dr R. Uhl (Muenchen) and Dr K. Schaller for optical engineering, Dr W. Zinth (Muenchen) and collaborators for helping us to set up the laser system, H. Bohnet and C. Koch for programming the scanner and Dennis Kostka for programming the analysis algorithms.