The relation between spike-timing dependent plasticity and Ca2+ dynamics in the hippocampal CA1 network

Neuroscience 145 (2007) 80 – 87

THE RELATION BETWEEN SPIKE-TIMING DEPENDENT PLASTICITY AND Ca2ⴙ DYNAMICS IN THE HIPPOCAMPAL CA1 NETWORK T. AIHARA,* Y. ABIRU, Y. YAMAZAKI, H. WATANABE, Y. FUKUSHIMA AND M. TSUKADA

(STDP) induced in a cultured hippocampal network had an asymmetrical profile, an experimental breakthrough which led to the grounds for new computational theories. It was further revealed by Nishiyama et al. (2000) that the profile of STDP induced in the hippocampal CA1 network with inhibitory neurons is “symmetrical” for the relative timing of preand post-synaptic activation. It showed that there were two distinct windows for the induction of LTD at ⫺28 –16 ms and ⫹15 to ⫹20 ms, and the timing of those two LTD peaks were close to ⫹20 ms and ⫺20 ms in the paper. On the other hand, the requirements for the induction of a type of associative plasticity (STDP) regarding NMDA (N-methyl-D-aspartate) receptor activation and postsynaptic calcium increase are similar to the induction of homosynaptic LTD (Debance and Tompson, 1994). In particular, a large postsynaptic calcium entry that is temporally associated with a synaptic activation induces associative LTP, whereas a moderate and persistent elevation in calcium concentration decreases synaptic strength at the active synapse (Selig et al., 1995; Goda and Stevens, 1996; Feldman et al., 1998). It is commonly accepted that two separate thresholds exist for the induction of LTP and LTD (Lisman, 1989; Altola and Singer, 1993; Bear and Malenka, 1994; Hansel et al., 1997), and it has been demonstrated with direct elevations of calcium concentration in hippocampal neurons (Yang et al., 1999). Ca2⫹ transition for relative timing of pre- and postsynaptic activity (Koester and Sakmann, 1998; Yuste and Denk, 1995; Yuste et al., 1999; Schiller et al., 1998) and the dendritic location dependency of dynamics on Ca2⫹ transient (Connor and Cormier, 2000) has also been shown. Furthermore, it is known that inhibitory activity of interneurons gates back-propagation of APs (Tsubokawa and Ross, 1996; Buzsaki et al., 1996; Larkum et al., 1999). It is suggested that voltage-dependent calcium influx into the dendrites is regulated by inhibitory neurons and the temporal coincidence of synaptic depolarization and activation of voltage-dependent calcium conductance is critical for synaptic plasticity (Buzsaki et al., 1996). In our previous study (Tsukada et al., 2005), the profiles of STDP were classified into two types depending on their layer specific location along the dendrite in the CA1 network of the hippocampal slice, which include interneurons. One type was characterized by a symmetric time window observed in the proximal dendrite (PD) and the other was characterized by an asymmetric time window in the distal dendrite (DD). The bath-application of bicuculline (GABAA receptor antagonist) to hippocampal slices revealed that GABAergic interneuron projections were responsible for the symmetry of a time window.

Department of Intelligent Information Systems, Faculty of Engineering, Tamagawa University, 6-1-1, Tamagawa-gakuen, Machida, Tokyo 194, Japan

Abstract—In our previous study, spike timing dependent synaptic plasticity (STDP) was investigated in the CA1 area of rat hippocampal slices using optical imaging. It was revealed that the profiles of STDP could be classified into two types depending upon layer specific location along the dendrite. The first was characterized by a symmetric time window observed in the proximal region of the stratum radiatum (SR), and the second by an asymmetric time window in the distal region of the SR. Our methods involved the bath-application of bicuculline (GABAA receptor antagonist) to hippocampal slices, which revealed that GABAergic interneuron projections were responsible for the symmetry of a time window. In this study, the intracellular Ca2ⴙ increase of hippocampal CA1 neurons, induced by the protocol of timing between preand post-synaptic excitation (i.e. STDP protocol), was measured spatially by using optical imaging to investigate how the triggering of STDP is dependant on intracellular calcium concentration. We found that the magnitude of STDP was closely related to the rate of Ca2ⴙ increase (“velocity”) of calcium transient during application of induction stimuli. Location dependency was also analyzed in terms of Ca2ⴙ influx. Furthermore, it was shown that decay time constant of Ca2ⴙ dynamics during the application of STDP-inducing stimuli was also significantly correlated with STDP. © 2006 IBRO. Published by Elsevier Ltd. All rights reserved. Key words: STDP, Ca2ⴙ mobility, hippocampus, CA1, dynamics.

It is widely accepted that the precise time of occurrence of individual pre- and post-synaptic action potentials (APs) plays an important role in the modification of synaptic efficacy and that the back propagation of APs is crucial for the induction of long-term potentiation (LTP) and long-term depression (LTD) (Magee and Johnston, 1997; Markram et al., 1997; Bi and Poo, 2001; Stuart and Hausser, 2001). The findings of Stuart and Sakmann (1994), that suprathreshold presynaptic inputs create back propagating APs, are consistent with this notion. Bi and Poo (1998) later showed that spike timing dependent synaptic plasticity *Corresponding author. Tel: ⫹81-42-739-8432; fax: ⫹81-42-739-8858. E-mail address: [email protected] (T. Aihara). Abbreviations: AHP, afterhyperpolarization; ANOVA, analysis of variance; AP, action potential; CS, conditioning stimulus; DD, distal dendrite; ISI, inter-stimulus interval; LTD, long-term depression; LTP, long-term potentiation; NMDA, N-methyl-D-aspartate; PD, proximal dendrite; P1, first period; P2, second period; P3, third period; S.E., standard error of the mean; SR, stratum radiatum; STDP, spike timing dependent synaptic plasticity.

0306-4522/07$30.00⫹0.00 © 2006 IBRO. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.neuroscience.2006.11.025

80

T. Aihara et al. / Neuroscience 145 (2007) 80 – 87

81

In this study, to investigate how the magnitude of STDP was related to the rate of Ca2⫹ increase (“velocity”) during induction-stimuli application, location dependency of STDP profiles was analyzed in terms of Ca2⫹ influx. Furthermore, the dynamics of Ca2⫹ mobility during the application of STDP-inducing stimuli were analyzed. The relation between the velocity and the dynamics of Ca2⫹ mobility was discussed in terms of promotive and demotive factors of STDP induction.

EXPERIMENTAL PROCEDURES General method The experiments were performed on hippocampal slices, with a thickness of 400 ␮m, taken from 4 week-old male Wister rats. The tissue was sliced at an angle of 30 – 45° to the long axis of the hippocampus. This angle was selected because the plane was parallel to the alveus fibers that were on the surface of the tissue. The slices were maintained at room temperature in an experimental submerge chamber with a normal medium of artificial cerebrospinal fluid (ACSF) (142 mM NaCl, 5.0 mM KCl, 2.6 mM NaH2PO4, 2.0 mM MgSO4, 2.0 mM CaCl2, 26 mM NaHCO3, 10 mM glucose). The positions of the stimulus electrodes are shown in Fig. 1A. One stimulating bipolar tungsten electrode was placed at a fixed position in the stratum radiatum (SR) to stimulate the Shaffer commissural collaterals (s.c.) of the CA3 (Stim. A). The stimulation is spread and delivered to both PD and DD, then we can simultaneously observe STDP at PD and DD by optical imaging as described below. The other bipolar electrode was placed at a fixed position in the stratum oriens bordering the alveus (Stim. B) to induce back-propagated APs. The optical recording area covered the target network in CA1 area.

Stimulation The intensity of the electric pulse used to stimulate Stim. A and Stim. B was fixed at a constant value. This value was set to be exactly half the intensity necessary to produce a maximum population response and antidromic response, respectively, in the CA1 region (0.1– 0.5 mA). The duration of the stimulus pulse, the total number of pulses and the inter-stimulus interval (ISI) were fixed at 0.2 ms, 200 pulses and 2 s, respectively. The stimulus condition, ISI⫽2 s, was known not to induce LTP or LTD (Aihara et al., 1997). A pair of pulse stimuli was used to stimulate the Stim. A and Stim. B with various sets of relative timing (␶⫽tB⫺tA), where Stim. A (tA) was the timing reference for Stim. B (tB). These sets with ␶⫽0, ⫾10, ⫾20, ⫾50 ms, are shown in Fig. 1B and were used to induce STDP. We used ⫹20 ms and ⫺20 ms as the representative peak values of LTD time window because Nishiyama et al. (2000) showed that the timing for the two peaks in LTD window were close to ⫹20 ms and ⫺20 ms.

Optical imaging The method of optical imaging using voltage sensitive dye is following the technique of Tsukada et al. (2005). The relative fluorescence change (⌬F/F) was measured, where F is the initial fluorescent intensity and ⌬F is the change in the intensity of fluorescence. A decrease in fluorescence corresponds to a positive value and an increase in fluorescence corresponds to a negative value in the membrane potential change. STDP (LTP and LTD) was calculated by the magnitude of the population response measured from these fluorescence changes. For optical imaging using Ca2⫹ sensitive dye, slices were stained for 30 min with 10 ␮M fura-2/AM in normal medium and were then washed and recovered for an additional 10 min. Slices were viewed with a

Fig. 1. (A) Stimulation sites and recording area. Stim. A was placed at a fixed position in the SR of the CA1 area to stimulate the Schaffer collateral commissural of the CA3 region. Stim. B was set on the stratum oriens bordering the alveolar, the output layer of the CA1, to initiate back propagating APs. The optical recording area (120⫻120 pixels) was 1.25 mm⫻1.25 mm. DG, dentate gyrus; F, fimbria. (B) The sets of stimulus timing between Stim. A and Stim. B. Paired stimulus patterns with spike timing (␶) were applied. The duration, total number and ISI of stimulus pulses were fixed at 0.2 ms, 200 pulses and 2 s, respectively.

10⫻ objective illustrated in Fig. 1A. The Ca2⫹ measurements were made by ratio imaging of fura-2 (Grynkiewicz et al., 1985), with the use of 340/380 nm excitation, an inverted microscope, and a cooled frame-transfer CCD-camera system (Argus50, Hamamatsu Photonics Co., Hamamatsu, Japan). The transmitted light was detected by a 120⫻120 square pixels, each with a receptive area of 1.25⫻1.25 mm2. This had an adequate resolution in space (40⫻40 ␮m2 using 4⫻4 binning with 10⫻10 ␮m2/single photopixel) and in time (117 ms sampling, 29.1 ms/single frame) to analyze the spatio-temporal activities of the CA1 neural network. In this experiment, two representative points of the CA1 area, one distal and one proximal from the cell body, were used to compare the characteristics (location dependency) of Ca2⫹ transients on inducing STDP.

Measurement of LTP and LTD At the beginning of each experiment, the “test stimulus,” (TS, at Stim. A site, single pulse) was applied once every 20 s (0.05 Hz) for more than 20 min to ensure that the amplitude of the population spike was stable. Then, one of the paired stimuli with ␶ was delivered as a “conditioning stimulus (CS).” as shown in Fig. 1B.

82

T. Aihara et al. / Neuroscience 145 (2007) 80 – 87

Fig. 2. (A) STDP at PD. ; The results from naive slice. ␱; Those with bicuculline (25 ␮M). All values were expressed as the mean⫾S.E. (%) in 5.5 pixels of seven slices and results were analyzed for significance (P⬍0.05) by ANOVA. A significant difference appeared at ␶⫽20 ms. (B) Three measurements of Ca2⫹ transient during stimulus. (C) Results of velocity for relative timing (D) Results of peak ratio. (E) Results of integration.

After applying the CS of 200 paired pulses, the same TS was applied every 20 s to estimate the change in responsiveness induced by the CS. Sixteen responses were averaged in the experiment to improve the signal to noise ratio for both unconditioned and conditioned (15–25 min after the CS) TS responses. The magnitudes of LTP and LTD were estimated using mean percentage changes in the EPSP peak in hippocampal CA1 slices (conditioned TS-response/unconditioned TS-response). The detail is shown in the paper (Tsukada et al., 2005). A new slice was used for stimulus sequence of TS, CS, and TS. Seven slices were used for each paired stimulus with different t (⫺20 –20 ms).

Measurement of Ca2ⴙ signals The Ca2⫹ transient was measured during TS, and CS which induced STDP. The Ca2⫹ increase was measured as the difference from the resting Ca2⫹ level during TS before CS. All values were expressed as the mean (%)⫾standard error of the mean (S.E.) and the results were evaluated statistically (P⬍0.05) by analysis of variance (ANOVA).

RESULTS The relation between the magnitude of STDP and the Ca2ⴙ dynamics The magnitudes of STDP for relative timing of pre- and post-synaptic activation were measured at PD at a dis-

tance of 60 ␮m from the cell body layer shown in Fig. 2A. The profile shows an LTP window at the timing of coincidence of EPSP and back-propagated AP, which is at 0 ms of relative timing, and two LTD time-windows at around ⫺20 and 20 ms (Nishiyama et al., 2000; Tsukada et al., 2005). A representation of the time courses of optical signals of Ca2⫹ influx showing ratio imaging data (as examples, ␶⫽5, 20 ms) during the experimental sequence was shown in Fig. 2B. S and E mark the start and end points of the paired stimuli for induction of STDP, respectively. We estimated Ca2⫹ transient during stimulus application using three methods of analysis. The first was “velocity” of Ca2⫹ increase. The second one was the “peak ratio” during stimulus application. The last method was the “integration of ratios” during stimulation; the gray area in Fig. 2B. The results were shown in Fig. 2 (C; velocity, D; peak ratio, E; integration). The shape of velocity seen using Ca2⫹ imaging in Fig. 2C most closely fitted the shape of the STDP profile using voltage imaging (Fig. 2A). Profiles of peak ratio and integration were significantly different from the STDP profile at 50 ms of relative timing. To measure the correlation between the three Ca⫹2 dynamics profiles and STDP profile, the correlation coefficient, r, was calculated. As a result, r⫽0.987, 0.879, 0.873 in the case

T. Aihara et al. / Neuroscience 145 (2007) 80 – 87

of velocity, peak ratio, and integration, respectively. Therefore, velocity was used as the estimation of Ca2⫹ transient to compare with the magnitude of LTP and LTD in this study. The location dependency of STDP and Ca2ⴙ dynamics In our experiment, the dendritic area was divided into two parts (Fig. 3A). PD represents the proximal region of the SR, 60 ␮m close to the soma. The DD is the distal region of the SR, 280 mm further from the soma. Profiles of STDP for relative timing of pre- and postsynaptic activity were also classified into two types depending on their layer specific location along the dendrite, as seen in Fig. 3B and 3D. One was characterized by a symmetric time window observed at PD (Fig. 3B) and the other by an asymmetric time window at DD (Fig. 3D). The bath-application of bicuculline (25 ␮M) to hippocampal slices, shown by open circles in Fig. 3B, revealed that GABAergic interneuron projections were responsible for the symmetry of a time window. LTD on one side at ␶⫽20 ms was clearly blocked by bicuculline, while on the other side for ␶⫽⫺20 ms, LTD

83

was only slightly affected by bicuculline. In following, the profile of LTP and LTD in the slice with bicuculline showed an asymmetric window for spike timing. The LTD for ␶⫽20 ms is closely related to the GABAergic inhibitory connections in the hippocampal CA1 network, while the asymmetric window, in which LTD appeared at only ␶⫽⫺20 ms, was found at the distal region of the SR where projections of GABAergic interneurons are not dense. Bicuculline application, however, showed LTP at 20 ms. To investigate how Ca2⫹ transient during the induction of STDP affects the magnitude depending on the location in the CA1 network, the velocity of Ca2⫹ transient during induction of STDP for relative timings was measured at PD and DD. The results at PD and DD are shown in Fig. 3C and 3E, respectively. There were significant difference between the relative timing at ␶⫽⫺50, 50 ms and ␶⫽⫺20, 20 ms in the case of PD in Fig. 3C. However, there was no significant difference between ␶⫽50 ms and ␶⫽20 ms in the case of DD in Fig. 3E. It was found that the profiles of the velocity of Ca2⫹ transient showed symmetry at PD and asymmetry at DD corresponding to the magnitudes of

Fig. 3. (A) Illustration of a neuron. (B) STDP and (C) Ca2⫹ velocity at PD for relative timing. (D) STDP and (E) Ca2⫹ velocity at DD for relative timing. Closed circles and squares show the results from naive slice. Open circles and squares show those with bicuculline (25 ␮M). Statistical difference (P⬍0.05) by ANOVA is shown by a * on bold lines from results of naïve slice and by a * on dashed lines from those of bicuculline application.

84

T. Aihara et al. / Neuroscience 145 (2007) 80 – 87

STDP in Fig. 3B and 3D. Furthermore, under bicuculline application shown by open squares in Fig. 3C and 3E, the profile at PD changed from symmetrical to asymmetrical and the low velocity of Ca2⫹ transient at ␶⫽20 ms was changed to a medium level similar to ␶⫽⫺50, 50 ms which showed little LTP and LTD. The result also corresponded to the change of profiles of STDP shown by open circles under 25 ␮M bicuculline application in Fig. 3B and 3D. The velocity of Ca2⫹ transient was closely related to the magnitude of STDP.

The dynamics of Ca2ⴙ mobility To investigate the causes for differences in the velocities for relative timing, the dynamics of the decay time constant and peak height of the time course of Ca2⫹ transient during STDP induction were measured. Measurements of the average of five pairings in three periods during stimulation are shown in Fig. 4A. The first is the average in the first period (P1) from the first pairing to the fifth paring of Stim. A and B. The second is the average in the second

Fig. 4. (A) The time course of Ca2⫹ transients. Blue and red lines show ratios at PD and DD, respectively. P1, P2, P3 show the period of five pair-stimulations at the start and at graduation of Ca2⫹ increase, and at stable state of Ca2⫹ dynamics, respectively. (B) Measurements of a peak height and a decay time constant of Ca2⫹ transient. (C) Peak heights at P1, P2, P3 for relative timings (⫺20, 0, 20 ms) (D) Decay time constants of Ca2⫹ transient at P1, P2, P3 for relative timings (⫺20, 0, 20 ms). Blue and red lines show results at PD and DD, respectively in (C), (D). Statistical difference (P⬍0.05) by ANOVA is shown by a * on bold lines. For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.

T. Aihara et al. / Neuroscience 145 (2007) 80 – 87

85

period (P2) for which maximum Ca2⫹ velocities were shown in the stimulus. The last was the average in the third period (P3) which was a stable state of Ca2⫹ transient. Fig. 4B shows an example of the measurement of peak height and time constant of time course of Ca2⫹ transient. Fig. 4C shows averages of the peak height for each pairing at P1, P2, P3 during STDP inducing stimuli whose relative timings were ␶⫽⫺20, 0, 20 ms. The peak height of DD was smaller than that of PD for any relative stimulus timing and there was no significant difference among the three periods. The order of the peak heights of PD was ␶⫽0⬎␶⫽20⬎␶⫽⫺20 ms, and that of DD was also ␶⫽0⬎␶⫽20⬎␶⫽⫺20 ms. The decay time constants are shown in Fig. 4D. Though the time constant of PD was a little larger than that of DD, at ⫺20 ms of relative timing there was no difference between the three periods. However, both time constants of PD and DD were strongly increased in P2 at 0 ms of relative timing. The level of time constant of PD was gradually decreased and that of DD was further increased. The time constant of PD at 20 ms was smaller than that of DD, contrary to the case at ⫺20 ms, though there was no difference between the three periods. The effect of GABAergic interneurons on STDP induction was investigated by analyzing the time constant and peak height during STDP inducing stimuli with a relative timing of ␶⫽20 ms. This relative timing was of particular interest since bicuculline application showed a drastic change in STDP induction, marked by the disappearance of LTD induction, as shown in Fig. 3C. The results were graphed in Fig. 5. There was no significant difference in both the peak height of PD and DD, though an increase was observed after bicuculline application in each period in Fig. 5A. On the other hand, there was a significant difference between the decay time constant of PD in the naive slice and with the bicuculline application at P2 and P3. There was no significant difference of the decay constants of DD at each period.

DISCUSSION In order to investigate what measurements for the estimation of Ca2⫹ dynamics correspond most closely to the magnitude of LTP and LTD in local areas, three modes of measurement were fitted to STDP profile. As a result, the velocity method was shown to be the most suitable measurement (Fig. 2). It was found that the most important factor in calculating the time course of Ca2⫹ transient was not the value of the stabilized peak, but the velocity of Ca2⫹ elevation. The results suggest that the magnitude of STDP was strongly related to the period of a large increase of Ca2⫹ transient. Measurements of the velocity of Ca2⫹ transient showed that the profiles correspond closely to those of the magnitude of STDP at both the PD and DD. GABAergic interneurons usually have an inhibitory effect on the amplification of the back-propagated action spike from the soma by shunting the membrane resistance (Hausser et al., 2001; Williams and Stuart, 2003). However, since the application of bicuculline resulted in a reduction of LTD (Fig. 3B) and an increase in Ca2⫹ velocity (Fig. 3C) at ␶⫽20 ms at PD, it was considered that the feed-forward

Fig. 5. (A) Peak heights at P1, P2, P3 for the relative timing, ␶⫽20 ms, with and without bicuculline at PD and DD. (B) Decay time constants of Ca2⫹ transient at P1, P2, P3 for the relative timing, ␶⫽20 ms. A significant statistical difference (P⬍0.05) by ANOVA is denoted by a * on bold lines.

projection of GABAergic interneurons was blocked and that the level of Ca2⫹ velocity recovered to a neutral level and, thus, did not induce LTP and LTD. On the other hand, since the projection of GABAergic interneurons to the DD area is not dense, the change of Ca2⫹ velocity between with and without bicuculline trials may not be significant (Fig. 3E). The involvement of a local spike in inducing LTP at 20 ms (Fig. 3D) is discussed later. These results indicate the high velocity of Ca2⫹ during stimulation induced a significant amount of LTP and that the low velocity of Ca2⫹ induced LTD. Interestingly, between Ca2⫹ velocity for LTP

86

T. Aihara et al. / Neuroscience 145 (2007) 80 – 87

and that for LTD, there was a neutral level which did not induce LTP and LTD. These results suggest that the estimation for the relation between STDP and Ca2⫹ dynamics by our measurement of Ca2⫹ transient (“velocity”) was similar to the extended BCM rule (Biemenstock et al., 1982; Artla et al., 1990; Tamura et al., 1992). Bender et al. (2006) reported that the LTP inducing component of STDP was dependent on postsynaptic NMDA receptors, and LTD associated with STDP is attributed to the activation of metabotropic glutamate receptors. Therefore, the high and low velocity of Ca2⫹ dynamics might be related to the promotion of the above two systems, respectively. One report on the dendrite locationdependent STDP window in cortical pyramidal neurons (Froemke et al., 2005) showed asymmetrical profiles of STDP without LTD window at positive timings in both PD and DD. The window of LTD at DD is broader than that at PD, which caused NMDA receptor suppression. Extracellular electrodes were placed ⬍10 ␮s away from the apical dendrite in the experiment, which is a much shorter distance than that of our experiments (Stim. A in Fig. 1, 500 ␮m). It means that the input to GABAergic interneuron to induce IPSP in our experiments was larger than that of the experiments by Froemke et al. (2005). Therefore, our profile of STDP might be a symmetrical STDP window with added LTD time window at the positive timings. On the other hand, the extracellular stimulation (Stim. B in Fig. 1) in our experiment activates some pyramidal neurons. Then, the feedback inhibition by inhibitory inter-neurons in our experiments is larger than that from stimulation to one neuron in the experiment by Froemke et al. (2005). It suggests that the effect of NMDA suppression on LTD might be mixed or submerged in the effect of feedback inhibition in our experiment. Lastly, to investigate the incidence for differences of the relative timing in velocities of the Ca2⫹ cumulative effect, the dynamics of peak height for each pairing of Ca2⫹ mobility during STDP induction were measured. As a result, the peak height of PD was larger than that of DD for all relative stimulus timing and there was no significant difference of peak height among the three periods (P1, P2, P3) as shown in Fig. 4C. It is believed that the peak height of PD was larger than that of DD because the density of synaptic input form stimulus site to PD was larger than at DD. While there was no significant difference among the three periods for each relative timings, there are differences between the peak height for relative timing whose order was ␶⫽0⬎␶⫽20⬎␶⫽⫺20 ms. Yuste and Denk (1995) showed that supralinear Ca2⫹ transient was induced by AP–EPSP pairing and sublinear Ca2⫹ transient was induced by EPSP–AP pairing, corresponding respectively to ⫺20 ms and 0 ms of relative timing in this study. At 0 ms, supralinear Ca2⫹ transient was induced by the suppression of A-type K⫹ channels by EPSP, which plays a promotive factor in the induction of LTP (Hoffman et al., 1997; Migliore et al., 1999). On the other hand, at ⫺20 ms of relative timing, EPSP at the local area was depressed by afterhyperpolarization (AHP) caused by K⫹ channel activation following back-propagated APs and also by feed-

back inhibition caused by GABAergic neurons after postsynaptic firing, both of which are both demotive factors. Consequently, there may be sublinear Ca2⫹ transient elicited by both AHP and IPSP. Also, at 20 ms, sublinear Ca2⫹ transient might be controlled by the shunting effect of feed-foward inhibition of GABAergic neurons, IPSP (Hausser et al., 2001; Williams and Stuart, 2003), which is also a demotive factor. This presence of two demotive factors of EPSP (AHP and feedback IPSP) at ⫺20 ms and only one demotive factor (feed-foward IPSP) at 20 ms may be the physiological mechanism that determines the order of the peak heights of relative timing. Furthermore, the results of the decay time constants are shown in Fig. 4D. Although no significant difference was seen between the three periods of the decay time constant of both PD and DD at ⫺20 ms and 20 ms of relative timing, both decay time constants of PD and DD were strongly increased in P2 and P3 at only 0 ms of relative timing. Interestingly, the level of the decay time constant of PD was gradually decreased and that of DD was further increased. These results suggest that the large induction of LTP was dependent on the rapid increase of Ca2⫹ in period P2. Furthermore, since the inter-pair interval of stimulus was 2 s in this experiment, the cumulative increase of Ca2⫹ through NMDA channels and voltage dependent Ca2⫹ channels may be very small. This suggests that large amounts of LTP would not be induced unless another factor was at work. The increase in the decay time constant might be due to many factors such as altered Ca2⫹ clearance. Saturation or regulation of Ca2⫹ intensity might be apparent in results showing that the time constant of P3 which has a higher Ca2⫹ concentration was smaller than that of P2 at 0 ms in DD. This result suggests that the magnitude of LTP was strongly affected by Ca2⫹ transients of period P2, which corresponds to the velocity measurement used to estimate Ca2⫹ transient for the induction of LTP and provided the validity of the analysis in this experiment. Our results showed the dynamic feature is more important for STDP than levels of calcium increase. It might be considered that this function caused the nonlinear Ca2⫹ dependence of postsynaptic downstream mechanisms (Dan and Poo, 2004). Furthermore, the decay time constant of DD was larger than that of PD at 20 ms. This supports the possibility of Ca2⫹ increase by a local spike (Golding et al., 2002; Mehta 2004). To investigate the effect of GABAergic interneurons on inducing STDP, dynamics of the decay time constant during stimuli with and without bicuculline were measured (Fig. 5). There was a significant difference among decay time constants but not among peak heights at the three periods. These results suggest that GABAergic interneurons affect dynamics (i.e. decay constant) of Ca2⫹ influx and thereby modify STDP induction. It is considered that this effect may be the cause underlying the change in STDP profile from symmetric to asymmetric, seen in Fig. 3C. Our optical method was able to address the analysis of spatio-temporal coding in CA1 neural networks by removing the spatial limitation of data acquisition using an electrophysiological technique. Our results suggest that the location

T. Aihara et al. / Neuroscience 145 (2007) 80 – 87

dependence of STDP is caused by Ca2⫹ dynamics modulated by the behavior of IPSP and AHP for the relative timing of input in the network. As suggested by the synaptic modification related to Ca2⫹ dynamics in the network observed here, the timing of inputs (spatio-temporal information) is might be an algorithm which manipulates factors (inhibitory connection so on) in the network for inducing LTP and LTD. Acknowledgments—We thank Dr. H. Kato at Yamagata University and Mr. N. Yoshida at NYU for valuable discussions and advice on physiological experiments. This study was supported by COE program of JSPS and GASRPA (Integrative Brain Research) form MEXT (70192838).

REFERENCES Aihara T, Tsukada M, Michael CC, Shinomoto S (1997) Stimulus dependent induction of long-term potentiation in CA1 area of the hippocampus: experiment and model. Hippocampus 7:416 – 426. Altola A, Singer W (1993) Long term depression of excitatory synaptic transmission and its relationship to long-term potentiation. Trends Neurosci 16:480 – 487. Artla A, Brocher S, Singer W (1990) Different voltage-dependent thresholds for inducing long-term depression and long-term potentiation in slices of rat visual cortex. Nature 347:69 –72. Bear CC, Malenka RC (1994) Synaptic plasticity: LTP and LTD. Curr Opin Neurobiol 4:389 –399. Bender VA, Bender KJ, Brasier DJ, Feldman DE (2006) Two coincidence detectors for spike timing-dependent plasticity in somatosensory cortex. J Neurosci 26(16):4166 – 4177. Bi G-Q, Poo M-M (2001) Synaptic modification by correlated activity: Hebb’s postulate revisited. Annu Rev Neurosci 24:139 –166. Bi G-Q, Poo M-M (1998) Synaptic modifications in cultured hippocampal neurons: Dependence on spike timing, synaptic strength, and postsynaptic cell type. J Neurosci 18:10464 –10472. Biemenstock EL, Cooper LN, Munro PW (1982) Theory for the development of neuron selectivity: orientation specificity and binocular interaction in visual cortex. J Neurosci 2:32– 48. Buzsaki G, Penttonen M, Nadasdy Z, Bragin A (1996) Pattern and inhibition-dependent invasion of pyramidal cell dendrites by fast spikes in the hippocampus in vivo. Proc Natl Acad Sci U S A 93:9921–9925. Connor JA, Cormier RJ (2000) Cumulative effects of glutamate microstimulation on Ca2⫹ responses of CA1 hippocampal pyramidal neurons in slice. J Neurophysiol 83:90 –98. Dan Y, Poo M-M (2004) Spike timing-dependent plasticity of neural circuits. Neuron 44:23–30. Debance D, Tompson SM (1994) Calcium: A trigger for long-term depression and potentiation in the hippocampus. News Physiol Sci 9:256 –260. Feldman DE, Nicoll RA, Malenka RC, Isaac JTR (1998) Long-term depression at thalamo-cortical synapses in developing rat somatosensory cortex. Neuron 21:347–357. Froemke RC, Poo M-M, Dan Y (2005) Spike-timing-dependent synaptic plasticity depends on dendritic location. Nature 434:221–225. Goda Y, Stevens CF (1996) Long-term depression properties in a simple system. Neuron 16:103–111. Golding NL, Staff NP, Spruston N (2002) Dendritic spikes as a mechanism for cooperative long-term potentiation. Nature 418:326–331. Hansel C, Artola A, Singer W (1997) Relation between dendritic Ca2⫹ levels and the polarity of synaptic long-term modifications in rat visual cortex neurons. Eur J Neurosci 9:2309 –2322.

87

Hausser M, Major G, Stuart GJ (2001) Differential shunting of EPSPs by action potentials. Science 291(5501):138 –141. Hoffman DA, Magee JC, Colbert CM, Johnston D (1997) K⫹ channel regulation of signal propagation in dendrites of hippocampal pyramidal neurons. Nature 387:869 – 875. Koester HJ, Sakmann B (1998) Calcium dynamics in single spine during coincident pre- and postsynaptic activity depend on relative timing of back-propagating action potentials and subthreshold excitatory postsynaptic potentials. Proc Natl Acad Sci U S A 95:9596–9601. Larkum ME, Zhu JJ, Sakmann B (1999) A new cellular mechanism for coupling inputs arriving at different cortical layers. Nature 398: 338 –341. Lisman J (1989) A mechanism for the Hebb and the anti-Hebb processes underlying learning and memory. Proc Natl Acad Sci U S A 86:9574 –9578. Grynkiewicz G, Poenie M, Tsien RY (1985) A new generation of Ca2⫹ indicators with greatly improved fluorescent properties. J Biol Chem 260:3440 –3450. Magee JC, Johnston D (1997) A synaptically controlled, associative signal for Hebbian plasticity in hippocampal neurons. Science 275:209 –213. Markram H, Lubke L, Frotscher M, Sakmann B (1997) Regulation of synaptic efficacy by coincidence of postsynaptic APs and EPSPs. Science 275:213–215. Mehta MR (2004) Cooperative LTP can map memory sequences on dendritic branches. Trends Neurosci 27(2):69 –72. Migliore M, Hoffman DA, Magee JC, Johnston D (1999) Role of an A-type K⫹ conductance in the back-propagation of action potentials in the dendrites of hippocampal pyramidal neurons. J Comput Neurosci 7(1):5–15. Nishiyama M, Hong K, Mikoshiba K, Poo M-M, Kato K (2000) Calcium stores regulate the polarity and input specificity of synaptic modification. Nature 408:584 –588. Schiller J, Schiller Y, Clapham DE (1998) NMDA receptors amplify calcium influx into dendritic spines during associative pre- and postsynaptic activation. Nat Neurosci 1(2):114 –118. Selig DK, Lee HK, Bear MF, Malenka RC (1995) Reexamination of the effects of MCPG on hippocampal LTP, LTD, and depotentiation. J Neurophysiol 74:1075–1082. Stuart G, Sakmann B (1994) Active propagation of somatic action potentials into neocortical pyramidal cell dendrites. Nature 367:67–72. Stuart G, Hausser M (2001) Dendritic coincidence detection of EPSPs and action potentials. Nat Neurosci 4(1):63–71. Tamura H, Tsumoto T, Hata H (1992) Activity-dependent potentiation and depression of visual cortical responses to optic nerve stimulation in kittens. J Neurophysiol 68:1603–1612. Tsubokawa H, Ross WN (1996) IPSPs modulate spike back propagation and associated [Ca2⫹]i changes in the dendrites of hippocampal CA1 pyramidal neurons. J Neurophysiol 76:2896 –2906. Tsukada M, Aihara T, Kobayashi Y, Shimazaki H (2005) Spatial analysis of spike-timing dependent LTP and LTD in the CA1 area of hippocampal slices using optical imaging. Hippocampus 15:104–109. Williams SR, Stuart GJ (2003) Role of dendritic synapse location in the control of action potential output. Trends Neurosci 26(3):147–154. Yang SN, Tang YG, Zucker RS (1999) Selective induction of LTP and LTD by postsynaptic [Ca2⫹]i. J Neurophysiol 81:781–787. Yuste R, Denk W (1995) Dendritic spines as basic functional units of neuronal integration. Nature 375:682– 684. Yuste R, Majewska A, Cash SS, Denk W (1999) Mechanisms of calcium influx into hippocampal spines: heterogeneity among spines, coincidence detection by NMDA receptors, and optical quantal analysis. J Neurosci 19(6):1976 –1987.

(Accepted 13 November 2006) (Available online 16 January 2007)