The Kinetic Profile of Intracellular Calcium Predicts Long-Term Potentiation and Long-Term Depression

Individual cells respond differently to pairing

Figure 1 depicts the effect of conditioning on the peak amplitude of the evoked EPSPs in response to stimulation of layer 6 and with the calcium dye fura-4F in the recording pipette. A net potentiation of the EPSPs (1.26 ± 0.06 SEM; Sig) occurs (Fig. 1A1). Interestingly, when plotted as a frequency distribution histogram (Fig. 1A2), distinct groupings in outcomes are apparent. The histogram is best fit with a trimodal Gaussian distribution (Fig. 1A2). Of the 110 cells, the EPSPs of 48 (44%) significantly potentiated (Fig. 1B1, time plot of an individual example; Fig. 1B2, time plot of grouped results for all of the cells that potentiated) (mean = 1.89 ± 0.03 SEM; Sig), 23 (21%) did not change (Fig. 1C1, time plot of an individual example; Fig. 1C2, time plot of the grouped results for all of the cells that did not change) (mean = 1.06 ± 0.02 SEM; NS), and 39 (35%) depressed (Fig. 1D1, time plot of an individual example; Fig. 1D2, time plot of the grouped results for all of the cells that depressed) (mean = 0.61 ± 0.02 SEM; Sig). An additional control group (n = 11) (Fig. 1E1,E2) received only the depolarizing pulses, with no effect on their EPSPs (mean = 1.07 ± 0.02 SEM; NS).

Differential plasticity outcomes are biologically based

This tripartite outcome suggests either experimentally induced variability or some intrinsic difference between cells. To assess the former, we analyzed the outcomes with respect to the resting membrane potentials, the strength of afferent stimulation (using the initial EPSP amplitude as an index), the level to which the postsynaptic membrane was depolarized during application of the pairing pulses, the magnitude of the depolarizing pulses, any changes in the input resistance throughout the protocol, and the age of the individual animal from which each recording was obtained. The results of these analyses are shown in Figure 2. None of these parameters differ among the groups according to plasticity outcome (NS for all group comparisons; see legend for details).

Other factors that may account for the outcome include variable susceptibility to washout of plasticity factors caused by whole-cell recording, variation in calcium buffering by fura-4F, activation of different synaptic inputs, or variable effects of blocking synaptic inhibition. We evaluated these possibilities in a series of control experiments. In experiments that used whole-cell recording without fura-4F (Fig. 3A1-A5) or perforated-patch recording (also without fura-4F) (Fig. 3B1-B5), similar plasticity outcome distributions occurred (Fig. 3A2,B2), as they did in the whole-cell experiments with fura-4F (Fig. 1A2). With whole-cell recording and no fura-4F in the pipette, a net potentiation of the EPSPs occurs (mean = 1.20 ± 0.08 SEM; n = 39; Sig) (Fig. 3A1), as it does with the perforated-patch recording, (mean = 1.20 ± 0.14 SEM; n = 21) (Fig. 3B1). The fractions of cells exhibiting each plasticity outcome were as follows: LTP with whole-cell recording and no fura-4F = 17 of 39 or 44% (mean = 1.72 ± 0.03 SEM; Sig) (Fig. 3A3); LTP with perforated-patch recording = 9of 21 or 43% (mean = 1.92 ± 0.05 SEM; Sig) (Fig. 3B3); NC with whole-cell recording and no fura-4F = 8 of 39 or 21% (mean = 1.05 ± 0.02 SEM; NS) (Fig. 3A4); NC with perforated-patch recording = 4 of 21 or 19% (mean = 1.05 ± 0.01 SEM; NS) (Fig. 3B4); LTD with whole-cell recording and no fura-4F = 14 of 39 or 36% (mean = 0.70 ± 0.02 SEM; Sig) (Fig. 3A5), and LTD with perforated-patch recording = 8 of 21 or 38% (mean = 0.68 ± 0.02 SEM; Sig) (Fig. 3B5). Thus, the different outcomes between individual cells are not caused by either variable calcium buffering by the fura-4F or variable washout of plasticity factors attributable to the whole-cell recording condition.

Another possibility that may explain the different outcomes is potential activation of different populations of inputs to different cells. Thus, we analyzed the individual responses of the EPSPs evoked by layer 6 stimulation for each cell (Fig. 4) and did an additional series of control experiments in which the stimulating electrode was positioned directly below the recorded cell in layer 4 (Fig. 5) to restrict the population of afferents activated. A series of 30 individual control EPSPs elicited from an individual cell are illustrated in Figure 4A; their average and the fitted smoothing function (see Materials and Methods) are illustrated in Figure 4B, indicating how the EPSP onset, peak amplitude, latencies to EPSP onset and peak, rising phase slope, and decay-phase time constant were measured. The individual measures of latency and its associated variance (SD) for each cell in the three outcome groups also are illustrated (Fig. 4C,E and D,F, respectively). These results make several points. There is very little jitter in the individual EPSPs for a single cell (example in Fig. 4A and plot of variances for EPSP onset latencies for each cell and each outcome group are illustrated in Fig. 4D), consistent with a major monosynaptic input component; the mean latencies to the EPSP onset for the individual cells (5-7.5 msec) (Fig. 4C) are consistent with monosynaptic inputs when the conduction distance of ∼1.5 mm for afferents and synaptic delay at 25°C are taken into account; the decay of the EPSP is well fitted by a single exponential (Fig. 4B), which is indicative of an EPSP that is not “contaminated” by an underlying IPSP; and the differences in the plasticity outcomes (LTP vs NC vs LTD) are not attributable to variance in the latency (as might be expected if one outcome group represented a monosynaptic set of inputs and another represented a disynaptic pathway) (Fig. 4D). Because our stimulating electrode is placed in layer 6 and the concentration of bicuculline used may only partially block GABA_A-mediated synaptic inhibition, however, we also performed a series of experiments with stimulation in layer 4, directly below the layer 2/3 neuron being recorded and without bicuculline. These results are illustrated in Figure 5, with the grouped data for the entire sample illustrated in Figure 5A (mean = 1.13 ± 0.10 SEM; NS) and the outcome distribution represented in Figure 5B. With stimulation in layer 4, the net change in EPSPs for the entire sample is only +13% and not significant; however, although this is a small sample, the distribution forms three clusters (see details of trimodal Gaussian fit in legend) with an LTP group, an NC group, and an LTD group, in which the plasticity for the individual EPSPs in the LTP and LTD groups is significant, as are the grouped results for the LTP and LTD subsamples. The fractions of EPSPs exhibiting each plasticity outcome are as follows: LTP = 6 of 14 or 43% (mean = 1.53 ± 0.01 SEM; Sig) (Fig. 5C); NC = 3 of 14 or 27% (mean = 1.00 ± 0.01 SEM; NS) (Fig. 5D); LTD = 5 of 14 or 36% (mean = 0.73 ± 0.01 SEM; Sig) (Fig. 5E).

The responses from the layer 6 and layer 4 stimulation experiments also were analyzed for effects of the pairing on the slope of the EPSP rise (Fig. 6). Interestingly, the relative changes in both parameters for the three outcome groups are similar for both experimental groups; however, the variance in the postpairing versus prepairing EPSP slope ratio is less for the layer 4 stimulation condition, suggesting that in this experiment a more homogeneous group of synapses is activated. Also, in support of an individual cellular basis for the differential plasticity outcomes is the observation that the different outcomes (LTP, LTD, or NC) occur in different slices from the same animal as well as in neighboring cells from the same slice.

Another possible explanation for the different plasticity outcomes between cells is a difference in the timing of the evoked postsynaptic spike during the applied depolarization with respect to the synaptic response. Indeed, spike-timing-dependent plasticity is a potentially important variable. Thus, we analyzed all of our results with respect to the temporal relationship of the evoked spikes to the EPSPs. Figure 7 illustrates this analysis. A series of 30 of the 60 traces (every other trial) during the 10 min pairing period for an example cell is overlaid (Fig. 7A). Note the regularity of the first and second spikes in each depolarization, with subsequent spikes in the five spike train showing some increasing variable latency. The averaged evoked EPSP (obtained during the last 5 min of the control period) for this cell is illustrated in Figure 7B, with the EPSP onset (triangle and dashed line) and peak (circle and dotted line) indicated for analysis of spike timing relative to these parts of the EPSP. These results (Fig. 7C-F) show that the spike timing with respect to the EPSP was consistent between cells and not related to the plasticity outcome, whether evaluated for the first (Fig. 7C,E) or second spike (Fig. 7D,F) and with respect to the EPSP onset (Fig. 7C,D) or peak (Fig. 7E,F). Similar results (data not shown) occur for the subsequent spikes. Thus, reasons for the different plasticity outcome groups must be found elsewhere.

Calcium imaging strategies and responses to individual pairings

Considering the pivotal role of [Ca²⁺]_i for induction of plasticity (Mulkey and Malenka, 1992; Yasuda and Tsumoto, 1996; Hansel et al., 1997; Yang et al., 1999; Zucker, 1999; Cormier et al., 2001; Lisman, 2001), we investigated the calcium signals generated during the pairing protocol in the three different outcome groups. We used two approaches: recording of [Ca²⁺]_i transients in response to individual pairings during plasticity induction (fast imaging) and recording of cumulative sustained [Ca²⁺]_i levels over the course of the entire experiment (slow imaging). The timing diagram of the optical data acquisition (blue and green lines) for both approaches are shown with respect to the electrophysiological stimulation paradigm (red and black lines) in Figure 8A and with an expanded time scale in B and C. The pseudocolor ratiometric images taken with the slow and fast-imaging methods are also illustrated (Fig. 8, D and E, respectively). We randomly applied our fast calcium imaging method to approximately one-fourth of the sample (n = 26 of 110 cells). LTP occurred in 11 of 26 of those cells (42%), LTD in occurred in 9 of 26 of those cells (35%), and there was no change in 6 of 26 of those cells (23%), a distribution similar to the entire sample of 110 cells tested. In addition, we applied this fast calcium imaging protocol to another eight control cells that received only the depolarizing pulses during the conditioning protocol. Examples (Fig. 9) of the calcium transients (top rows) and electrophysiological responses (bottom rows) during individual pairings are illustrated for a representative cell from each group [LTP (Fig. 9A), NC (Fig. 9B), LTD (Fig. 9C), depolarizing pulses only (Fig. 9D)]. The magnitude of the calcium rise is highest in response to the first pairing-depolarization and decays over the course of conditioning.

Amplitude and time course of dendritic calcium transients during individual pairings predict plasticity outcome

The analysis of the [Ca²⁺]_i transients in the dendrites and somata in response to individual pairings-depolarizations is summarized with respect to plasticity outcome (Fig. 10A,B). The peak transient [Ca²⁺]_i during the pairings is greater both in the apical dendrites (1040 ± 50 nm) and in the somata (630 ± 20 nm) for the LTP group (n = 11; upward triangles) versus the LTD group (641 ± 33 nm in the dendrite, Sig; 414 ± 32 nm in the somata, Sig; n = 9; downward triangles) and versus the no change group (454 ± 41 nm in the apical dendrites, Sig; 376 ± 28 nm in the soma, Sig; n = 6; circles). Furthermore, the average peak Ca²⁺ levels in the LTD group in both the apical dendrites and the somata are different (Sig) from the no change group. The peak transient [Ca²⁺]_i during pairing for the NC group does not differ from the depolarizing pulses only group (465 ± 32 nm in the apical dendrites, NS; 334 ± 26 nm in the somata, NS; n = 8; squares).

The larger intracellular Ca²⁺ signal in the LTP group can be partly ascribed to the differences in time course of the responses over the conditioning period. In the LTP group (upward triangles; n = 11), the peak amplitudes of the sequential somatic (Fig. 10C) and apical dendritic (Fig. 10D) Ca²⁺ signals decline in a nearly linear manner with slopes of -6.7 ± 0.8 and -14.0 ± 0.8 pm/sec, respectively. The decline in amplitude of the sequential peaks of the individual somatic and apical dendritic Ca²⁺ transients over the course of the pairing protocol in the LTD group (downward triangles) and NC group (circles) are each fitted by a single exponential decay (Fig. 10C,D). The average of the decay time constants that characterizes the progressive diminution of the peak calcium transients in response to successive pairings is slower for the LTD group (τ_dendritic = 360 ± 52 sec SEM; τ_somatic = 320 ± 18 sec SEM; n = 9) than for the no change group (τ_dendritic = 95 ± 2 sec SEM; τ_somatic = 32 ± 13 sec SEM; n = 6). Thus, the magnitude of the individual Ca²⁺ transients in the apical dendrites and somata remains elevated in cells that will undergo LTP versus those that will not change or undergo LTD. The average of the decay time constants that characterize the progressive diminution of the peak calcium transients in response to successive pairings in response to application of the depolarizing pulses alone (squares) also is best fitted to a single exponential decay with time constants comparable with those in the cells that exhibit no change in synaptic efficacy in response to the pairing (τ_dendritic = 44 ± 6 sec SEM; τ_somatic = 63 ± 6 sec SEM; n = 8).

Over the course of the 60 pairings, the time constants that fit the decaying phase of the individual calcium transients of each cell associated with each sequential pairing (or depolarizations alone) lengthen for each group (Fig. 10E,F). Notably, the decay time constants of the individual calcium transients during the very first pairing for the cells in the groups that undergo plasticity (τ_dendritic for the LTP group = 155 ± 5 msec SEM; τ_dendritic for the LTD group = 154 ± 4 msec SEM; τ_somatic for the LTP group = 310 ± 21 msec SEM; τ_somatic for the LTD group = 319 ± 25 msec SEM) are significantly shorter (p < 0.01 for the LTP vs NC group, for the LTP vs depolarizing pulses only group, for the LTD vs NC group, and for the LTD versus depolarizing pulses only group) than for those that did not undergo plasticity. Interestingly, the decay time constants of the first calcium transients are not different between the LTP versus LTD groups (p > 0.1). Moreover, although the decay time constants of the individual calcium transients increase with successive pairings (or depolarizing pulses alone) throughout the course of the protocol for all groups, those of the LTP and LTD groups remain below the level of the NC and depolarizing pulse only groups over the full 10 min of conditioning.

Individual 0.1 Hz trials of afferent stimulation alone do not cause any detectable calcium rises as evaluated with our fast-frame imaging method (data not shown). Thus, our measurements are not sensitive enough to detect the individual EPSP-mediated calcium signaling that is likely confined to dendritic spines; however, the calcium signals that occur in the proximal apical dendrite and soma in response to pairing in cells that subsequently undergo plasticity are greater than those elicited by depolarizing pulses alone or in response to pairing that does not induce plasticity. These signals thus correlate with the plasticity induction cascade and are suggestive of an intrinsic difference of the responses of those cells to the pairing. We also tested the possibility that the decline in the sequential Ca²⁺ transient peaks and the concomitant increase in the decay time constants of each transient over the course of the pairing protocol might reflect a change in the electrophysiological activation of the neuron by analyzing the spiking response from each cell in the different outcome groups; however, neither the number of spikes nor the average instantaneous firing frequency changes for the sequential pairings-pulses in the conditioning train for each of the groups (Fig. 10G,H), and there is no difference between groups (NS).

Kinetics of the cumulative somatic calcium rise predict plasticity outcome

In fast-frame calcium imaging experiments, it was apparent that the baseline intracellular calcium levels on which the sequential transient Ca²⁺ rises are imposed (Fig. 9, dashed lines connecting transients) also change for the LTP and LTD groups but not for the no change group. Thus, we measured these cumulative intracellular calcium signals with the slow-imaging method for 95 cells (84 conditioned with the pairing protocol and 11 with depolarizing pulses alone). The time-plot analyses of the sustained somatic [Ca²⁺] dynamics in the 84 cells (of 110 total) grouped according to synaptic plasticity outcome are illustrated in Figure 11 A-C. Records from a cell within each of the three outcome groups, respectively, are shown (Fig. 11 A1,B1,C1); the corresponding EPSPs from the same cells are shown in Figure 1, B1, C1, and D1. The time plots of the cumulative somatic calcium grouped results normalized to baseline [Ca²⁺]_i level are illustrated in Figure 11, B2, C2, and D2, respectively (LTP group, n = 37; no change group, n = 17; LTD group, n = 30). A consistent rise in the cumulative somatic [Ca²⁺]_i during the period of induction occurs only in the LTP and LTD groups. The trains of 60 depolarization pulses without synaptic pairing induce low and short-lived Ca²⁺ responses (Fig. 11 D1,D2), similar to those in the groups exhibiting no change in the EPSP amplitude in response to pairing (Fig. 11 B1,B2). The analysis of the relative EPSP amplitudes as a function of time course and level of the cumulative somatic calcium signals in response to the pairing is shown in Figure 12. The peak and amplitude (mean ± SD) of the sustained [Ca²⁺]_i rise (Fig. 12 A, B) in both the LTP group (128 ± 24 and 55 ± 11 nm, respectively; Sig; n = 37) and the LTD group (123 ± 23 and 52 ± 12 nm, respectively; Sig; n = 30) are significantly higher than in the no change group (95 ± 28 and 28 ± 12 nm, respectively; Sig; n = 17); however, the magnitude of these cumulative [Ca²⁺]_i responses does not differ between the LTP and LTD groups (NS). The absolute values (mean ± SD) of the initial or residual [Ca²⁺]_i levels (Fig. 12C,D) also do not differ among groups (NS) (LTP group: 73 ± 14 and 72 ± 14 nm, n = 37; NC group: 67 ± 19 and 68 ± 20 nm, n = 17; LTD group: 71 ± 13 and 76 ± 14 nm, n = 30, for initial and residual [Ca²⁺], respectively). The average time to peak (mean ± SD) for the cumulative [Ca²⁺]_i rise from baseline (Fig. 12 E) is less in the LTP group (190 ± 127 sec; Sig) than in the LTD group (579 ± 22 sec), however, and both differ from the no change group (417 ± 197 sec; Sig). The average time constants of decay (mean ± SD) (Fig. 12 F) are not different (NS) among the groups (576 ± 111 sec in the LTP group, 438 ± 185 sec in the no change group, and 520 ± 94 sec in the LTD group).

Plasticity outcomes

We have shown that in pyramidal neurons of visual cortex, repetitive low-frequency conjunctions of synaptic input and brief postsynaptic membrane depolarization produce a tripartite outcome (LTP, LTD, or NC) in different cells from the same animal at the same age. In addition, our results demonstrate that distinct spatiotemporal profiles of [Ca²⁺]_i are predictive of plasticity outcome in response to the same conditioning protocol.

Although synaptic input correlated with strong postsynaptic depolarization is generally considered to lead to LTP (Fregnac et al., 1994; Harsanyi and Friedlander, 1997; Chen et al., 1999; Bear, 2003), there are reports of heterogeneous outcomes (Debanne et al., 1999; Lei et al., 2003). In our study, these differences are not attributable to any of the following: variance in intrinsic electrophysiological properties or strength of conditioning stimuli (Fig. 2A-D), the response during conditioning (Figs. 9, 10G,H), cell-to-cell differences in calcium buffering by the dye (Fig. 3A), differential washout of plasticity factors (Fig. 3B), variable contributions of synaptic inhibition (Fig. 5), activation of different sets of synaptic inputs between cells (Figs. 4, 5, 6), timing differences between the EPSP and evoked spikes (Fig. 7), or cell class-slice-individual animal or age (Fig. 2F) differences. These results are consistent with a differential state (metaplasticity) (Abraham and Bear, 1996) for neurons determining their responses to conditioning stimuli. They may result from the history of the neuron or synapse before slice preparation (Abraham and Bear, 1996), unrecognized differences in neuronal subtypes (e.g., intrinsic calcium buffering) (Lee et al., 2000) among pyramidal, neurons and/or variable maturation between neighboring neurons at a chronological age (Murchison and Griffith, 1999).

The synaptic responses that we recorded were likely compound, reflecting a mixture of excitatory inputs from various afferent sources activated from layer 6, including monosynaptic geniculocortical afferents, disynaptic inputs routed through layer 4, and monosynaptic inputs from axon collaterals of infragranular neurons. Thus, it is possible that the differential plasticity outcomes reflect variable activation of these pathways for different cells; however, we consider this unlikely because analysis of the initial slopes of the EPSPs (reflecting monosynaptic inputs) (Fig. 6) results in the same outcome as layer 4 stimulation (Fig. 5). Moreover, similar outcomes, regardless of whether bicuculline is used, suggest that variable synaptic inhibition is unlikely to account for the differences. Because the dye used for calcium imaging buffers calcium, variable loading of the cell may contribute to different plasticity outcomes (although access resistances were low and similar between cells); however, this interpretation is negated by the results from the experiments without fura-4F during which similar plasticity outcomes occur.

There is considerable evidence demonstrating the importance of timing of spikes with respect to EPSPs for determining plasticity (Song et al., 2000; Sjostrom et al., 2001; Celikel et al., 2004; Kobayashi and Poo, 2004; Tzounopoulos et al., 2004; Yao et al., 2004) (for review, see Sjostrom and Nelson, 2002). Thus, during depolarization, the spike timing may have varied between cells; however, this was not the case (Fig. 7), implicating other factor(s). This is not to say that spike timing is not important, and indeed, if the experimental protocol were designed to alter spike timing, it might further modify the plasticity outcome.

Calcium transients in LTP versus LTD

Our results provide direct measurements of [Ca²⁺]_i during plasticity induction and maintenance. To our knowledge, this is the first demonstration of distinct combined amplitude and kinetic calcium profiles during induction of LTP versus LTD in the same cell type in response to the same induction protocol. Our observation that the amplitude component of the calcium transients is predictive of the outcome is consistent with previous theoretical considerations (Bienenstock et al., 1982; Bear et al., 1987), biochemical correlates for activation of kinase and phosphatase signaling pathways, respectively (Otmakhov et al., 1997; D'Alcantara et al., 2003), and experimental measures of relative changes in calcium levels during plasticity (Artola et al., 1990; Hansel et al., 1997; Yang et al., 1999; Cormier et al., 2001). The larger peaks and more rapid decay of the calcium transients during individual pairings in cells that undergo plasticity may reflect a more primed status of calmodulin or effective capture by target protein cascades (Mermelstein et al., 2001), more efficient calcium buffering-pumping by those cells causing a calcium profile that is optimally tuned (Murchison and Griffith, 1999), or more efficient calcium uptake that releases substantial calcium triggering plasticity after the pairings (Blaustein and Golovina, 2001). Although the amplitudes of calcium transients in the dendrites are predictive (Artola et al., 1990; Hansel et al., 1997; Yang et al., 1999; Cormier et al., 2001) and the kinetics of the calcium rises at synapses have been suggested potentially to play a role in the polarity of plasticity (Kato et al., 1999; Mizuno et al., 2001; Helmchen, 2002), such a distinct relationship between the rate of rise of the centripetal calcium signal and the plasticity outcome has not been demonstrated experimentally. Because we recorded at 25°C, the kinetics of the calcium responses do not precisely reflect the in vivo situation. There is the possibility that at elevated temperatures, different effects on synaptic responses may occur because of effects on certain currents; however, there are at least three reasons for the validity of the analysis at room temperature: (1) strong correlations of differences in the calcium profiles with respect to plasticity outcome (for individual calcium transients and for cumulative calcium rises); (2) distinct differences in the calcium profiles during the control experiments where depolarizing pulses alone are applied and when pairings do not induce plasticity; and (3) a significant difference in the individual calcium transient decay time constants (dendritic and somatic) that occur between the plasticity groups versus the NC groups on the very first pairing. Moreover, previous studies in slices of both basic synaptic physiology-plasticity (Keller et al., 1998; Yoshimura et al., 2000; Gonzalez-Islas and Hablitz, 2003; Hegg et al., 2003; Rae and Irving, 2004; Yamada et al., 2004) and calcium signaling using imaging methods (Miller et al., 1996; Pozzo-Miller et al., 1999; Kaiser et al., 2004) have been performed at room temperature of 22-25°C and are consistent with results from other studies at elevated temperatures ranging from 30°C to 37°C in many respects including synaptic properties, plasticity outcomes, and calcium signaling.

Differences in calcium profiles throughout the conditioning protocol

It is notable that the differences in the peaks of the calcium transients (in proximal dendrites and somata) manifest between plasticity outcome groups within the first few pairings (Fig. 10C,D). It is equally striking that on the very first pairing the transients decay significantly faster in those cells that will undergo either type of plasticity (LTP or LTD) (Fig. 10E,F) versus in the NC or depolarization only groups. These results suggest that the calcium signal that leaves the sites of synaptic input after correlation of neurotransmitter binding and postsynaptic depolarization has already achieved the requisite amplitude and kinetic profile for induction of plasticity through interaction with other downstream calcium amplifying events (e.g., release of calcium from intracellular stores). The decline in the peak calcium transients with sequential pairings or depolarizing pulses alone suggests rundown of the calcium signals, probably through voltage-gated calcium channels (VGCCs) solely in the case of the depolarizing pulses alone. Our control experiments (Figs. 9, 10G,H) demonstrate that these changes are not caused by a declining level of depolarization of the neuron or its spike-generating capacity, so that the level of either passive depolarization or actively back-propagated spikes that travel to the synaptic sites is not the source of the change. It is more likely that this declination is caused by incomplete recovery of VGCCs (Patil et al., 1998) and/or enhanced sequestration-removal of calcium (Lee et al., 2000).

Our results are obtained from the calcium signals in dendritic shafts and somata versus dendritic spines; however, recent reports on synaptic inputs from layer 4 neurons onto layer 2/3 pyramidal neurons (Silver et al., 2003) demonstrate that many of these synapses are also on dendritic shafts, including apical and basilar dendrites (Feldmeyer et al., 2002; Silver et al., 2003), and located within 70 μm of the soma (Feldmeyer et al., 2002; Silver et al., 2003), where our calcium measurements were made. Thus, although many synapses are activated and the exact location of those inputs cannot be ascertained, the intradendritic calcium measures that have characteristic signatures for different outcomes are within the range of the monosynaptic input from layer 4 and may reflect either primary or secondary changes in [Ca²⁺]_i. Interestingly, Goldberg et al. (2003) demonstrated that calcium signals in dendritic shafts also are spatially constrained. The difference between the LTP and LTD groups in the kinetics of the cumulative somatic calcium rise in the first pairings is striking (Fig. 11A1,A2,C1,C2). Although the significance of this difference is unclear, specific temporal profiles of somatic calcium may contribute to selective modification of gene expression (West et al., 2001) implicated in longer term plasticity (Frey et al., 1993) requiring transcription and translation with delivery of newly synthesized proteins to “tagged” synaptic sites (Steward and Worley, 2001).