Chaotic stimulus dependent long-term potentiation in the
hippocampal CA1 area
Hironori Matsuda
a, Minoru Tsukada
a,b,*, Takeshi Aihara
b,
Masami Tatsuno
c, Kazuyuki Aihara
d,eaBrain Science Research Center,Tamagawa Uni6ersity Research Institute,6-1-1Tamagawagakuen,Machida,
Tokyo194-8610, Japan
bFaculty of Engineering,Tamagawa Uni6ersity,6-1-1Tamagawagakuen,Machida,Tokyo194-8610,Japan
cBrain Science Institute,Institute for Physical and Chemical Research(RIKEN),2-1Hirosawa,Wako,Saitama351-0198,Japan dGraduate School of Engineering,The Uni6ersity of Tokyo,7-3-1Hongo,Bunkyou,Tokyo113-8656,Japan
eCREST,Japan Science and Technology Corporation(JST),4-1-8,Hon-cho,Kawaguchi,Saitama332-0012,Japan
Abstract
In our previous report [Tsukada, M., Aihara, T., Saito, H., Kato, H., 1996. Neural Netw. 9, 1357 – 1365], the temporal pattern sensitivity of long-term potentiation (LTP) in hippocampal CA1 neurons was estimated by using Markov chain stimuli (MS) with different values of the serial correlation coefficientr1between successive
interstim-ulus-intervals. In this paper, the effect of chaotic stimuli (CS) on induction of LTP in the hippocampal CA1 area was investigated in comparison with that of MS and periodic pattern stimuli (PS). The CS were produced by a modified Bernoulli map, so that interstimulus sequences with various values ofr1can be generated by changing the parameter
B. These stimuli had an identical first order statistics (mean interstimulus-interval), but their higher order statistics such as the serial correlation coefficients were different. The LTP induced by CS at B=2 was significantly larger in magnitude than that of PS and MS, and also depended on the initial value of CS at B=2 and 3. These results suggest that chaotic signals play an important role for memory coding in the hippocampal CA1 network. © 2000 Elsevier Science Ireland Ltd. All rights reserved.
Keywords:Long-term potentiation (LTP); Hippocampal CA1 neurons; Chaotic stimulus (CS)
www.elsevier.com/locate/biosystems
1. Introduction
Long-term potentiation (LTP) of synaptic transmission has been demonstrated in many
bio-logical neural networks, especially in the hippocampus, which is a region critical for mem-ory function. LTP is generally considered to be the cellular basis of learning and memory in the brain (Bliss and Lømo, 1973; McNaughton and Morris, 1987; Bliss and Collingridge, 1993). Alter-natively, a popular hypothesis is that specific properties of the Ca2+ signal, including its
magni-tude and perhaps its temporal structure, might
* Corresponding author. Tel.:+81-42-739-8430; fax:+ 81-42-739-8858.
E-mail address: [email protected] (M. Tsukada).
alone completely determine the direction of synaptic change caused synaptic activity (Artola and Singer, 1993).
LTP is a persistent synaptic enhancement in-duced by a brief period of high-frequency electri-cal stimulation (100 – 500 Hz, 3 – 10 pulses) of the afferents (Bliss and Gardner-Medwin, 1973; Bliss and Lømo, 1973; Dunwiddie and Lynch, 1978; Yamamoto and Sawada, 1981). Furthermore, it is reported that LTP is induced by low frequency stimulation (5 – 10 Hz, 100 – 200 pulses) with some specific time structure (Rose and Dunwiddie, 1986; Tsukada et al., 1990, 1991, 1994; Aihara et al., 1997). Tsukada et al. (1996) also demon-strated that LTP was induced by two different time scales: One is the short interstimulus-interval like bursting spikes and the other is the long interstimulus-interval of a few hundreds millisec-ond representing theta wave. In addition, the effect of chaotic stimuli (CS) on LTP magnitude in CA1 was investigated by a simple neuron model (Tatsuno and Aizawa, 1997), and it is suggested that there is the chaotic time series dependent LTP.
To investigate the effect of CS on the LTP in the hippocampal CA1 area, three different types of CS produced by modified Bernoulli Maps were applied. This effect on LTP was compared with that of periodic pattern stimuli (PS) and Markov chain stimuli (MS). We discuss three focal points of the effect of chaotic stimulation in this study. First is the difference between CS and PS with the same mean rate and first order serial correlation. Second is the difference between CS and MS with the same mean rate and first order serial correla-tion. The last is the initial value dependence of LTP in the case of CS.
2. Methods
2.1. General methods
Experiments were conducted on hippocampal slices (thickness; 500 mm) prepared from
guinea-pig. The slices were maintained at about 30°C in an experimental chamber with artificial cere-brospinal fluid. A detailed description of the
tech-nique is given by Ito et al. (1989). Positions of electrodes for the stimulation and the extracellular recording are schematically shown in Fig. 1. A bipolar tungsten electrode was placed at a fixed position (Fig. 1, Stim) in the stratum radiatum of the specific region to stimulate the Schaffer collat-eral commissural fibers (SC) of CA3. An electrode for extracellular recordings was placed in the pyramidal cell body layer (Fig. 1, Rec).
2.2. LTP induced stimulus
2.2.1. Chaotic stimulus
The time series of CS was produced from the modified Bernoulli Map xnxn+1 (mod.1),
We transformed the value of xn to the spike
interval tin the following way:
t(x
n)=
!
50ms05xnB1/2
950ms1/2BxnB1
and which were named CS (Fig. 3). These stimuli all had the same mean rate (2 Hz), and they were
Fig. 2. Experimental protocol and estimated method of LTP. The amplitude of the population spikes (hc and ht) were computed as the difference between the negative peak and the averaged value of the leading and trailing positive peaks. The average of the amplitudes of population spikes (hc) to test stim. obtained before each type of stimuli is taken as the control (unconditioned TS-response, 100%). For the prepara-tion of each slice, the amplitudes of the 30 populaprepara-tion spikes evoked by the test stimuli (TS) were given 20 – 30min after each type of stimuli (six types of chaotic stimuli/five types periodic pattern stimuli). The magnitude of LTP was estimated by the ratio (%) between this response (conditioned TS-re-sponse;ht) and the unconditioned TS-response (hc).
value (B=1, 2, 3). Fig. 3(a) shows sample trains of CS patterns, and serial correlation coefficients (r1 – 15) of these stimulus patterns for different
parameter B is shown in Fig. 4.
2.2.2. Periodic pattern stimulus
We produced PS, which are composed from short and long interstimulus intervals. That is, a temporal pattern of successive short intervals (50 ms) and successive long intervals (950 ms) were repeated periodically. PS had the same mean rate as CS. To compare the induced LTP between CS and PS, we used the stimuli which have the same first-order serial correlation. Fig. 3(b) shows the ‘Imax’ and ‘Imin’ of PS corresponding to ‘Imax’ and
‘Imin’ of CS. In the case of B=1, only one
corre-sponding PS stimulus pattern was produced.
Fig. 3. Examples of (a) chaotic stimuli (CS); (b) periodic pattern stimuli (PS); and (c) Markov chain stimuli (MS).
Fig. 4. Serial correlation coefficientsr1 – 15of chaotic stimulus
(CS) and Markov chain stimulus (MS).
shows the comparison of the correlation coeffi-cients from first to fifteenth (r1 – 15) of the CS Imax,
CS Iminand MS. For B=2 and 3, the correlation
coefficients of CS Imax were very different from
those of MS, especially for the higher order r’s.
2.3. Experimental procedures
The stimulus intensity was adjusted to half of that producing the maximum amplitude of the population spike. At the beginning of each exper-iment, the test stimulus (TS) was applied once every 20 s to ensure that the amplitude of the population spike was stable. Then, a LTP induc-ing potentiation stimulus (PoS) was delivered, and it was again followed by TS. The amplitude of the population spike was computed as the difference between the negative peak and the averaged value of the leading and trailing positive peaks. The magnitude of LTP was estimated by mean per-centage changes of the amplitudes of the popula-tion spikes between before and 20 – 30 min end of LTP induced stimulus (Fig. 2). A naive slice was used for each stimulus sequence of TS/PoS/TS. To keep the same stimulus and recording condi-tions, we placed the stimulus and recording elec-trode described in general methods in the fixed position each slice.
2.4. Statistical analysis
All values were expressed as the mean9S.E.M. (%). To analyze the data statistically, we com-pared the LTP magnitude between CS (n=8), PS (n=5) and MS (n=5) within similar first order serial correlation by the Kruskal – Wallis test. And also Mann – Whitney U-test was used to compare the magnitude of LTP induced by CS and PS with similar first order serial correlation. We consid-ered differences significant at PB0.05.
3. Results
3.1. The comparison of LTP induced by chaotic and periodic pattern stimuli
Table 1 summarizes the changes of the popula-2.2.3. Marko6chain stimulus
The third type of stimulus sequence, which we refer as MS consist of mixed short (50 ms) and long (950 ms) interstimulus intervals. The impor-tant property of MS is that the occurrence of the next interstimulus interval is dependent only on the previous interstimulus interval, and the transi-tion probability matrix determines this stochastic process. These Markov sequences were used by Tsukada et al. (1990, 1994) to examine whether the induced LTP was dependent on the stimulus temporal structure. For a given pair of short and long intervals, we used three different stimulus set, which is r1=0.0, 0.4 and 0.8, respectively.
Fig. 3(c) shows sample trains of MS. For small values ofr1 (=0.0), the short and long intervals
have no correlation. On the other hand, for large values ofr1(=0.8), long trains of short and long
Table 1
Comparison among LTPs induced by chaotic stimuli (CS), periodic pattern stimuli (PS) and Markov chain stimuli (MS)a
B=2 B=3
aValues are means9S.E.M.
bIndicatesPB0.05 versus the magnitude of LTP induced by the corresponded to PS. cIndicatesPB0.05 versus the magnitude of LTP induced by I
maxof same B vale dIndicatesPB0.05 versus the magnitude of LTP induced by the corresponded to MS. tion spike amplitude induced by CS, PS and MS.
‘B=1, 2 and 3’ show parameter of the modified Bernoulli Map. ‘Imax’ and ‘Imin’ represent CS
which induced the highest and the lowest LTP in numerical simulation, respectively. We compared the magnitude of LTP induced by CS and PS which have the similar first order serial correla-tion (Table 1). At B=2, CS Imax (r1=0.495),
induces significantly larger LTP than that of PS (PB0.05). Similarly, in the Imin (r1=0.605) case,
CS induced significantly higher LTP than that of PS. However, there were no significant differences between CS and PS in the case of B=1 and 3. Furthermore, LTP induced by CS Imaxat B=2 is
the highest among our CS experimental data, and is significantly higher than that of PS and MS.
Therefore it seems that CS at B=2 induced larger LTP than that of corresponding PS, but CS with B=1 and B=3 didn’t. Consequently, it seems that LTP of the hippocampal CA1 area depend on specific chaotic temporal structure.
3.2. The comparison of LTP induced by chaotic and Marko6chain stimuli
We compared the magnitude of LTP induced by CS and MS, which have the similar first order serial correlation (Table 1). In B=1, there was no significant difference between CS and MS. In B=2, the magnitude of LTP induced by CS Imax
was significantly larger than that of corresponding MS (PB0.05). However in B=3, the magnitude of LTP induced by CS Imax was significantly
smaller than that of corresponding MS (PB0.05).
The higher order serial correlation (r2 – 15) was
different between these two stimuli, though the first order serial correlation is similar (Fig. 4). The largest LTP at B=2 may be interpreted as that it was caused by the higher order serial correlation of CS Imax. On the other hand, in B=3, the
induced LTP by CS Imaxwas smaller than that of
MS, and this might be because that too many successive long interstimulus intervals (82 inter-vals) caused depotentiation of LTP (Fujii et al., 1996).
3.3. The initial 6alue dependence of chaotic time series
At B=1, there were no significant differences in the LTP magnitude by an initial value of CS. However, in the case of B=2 and 3, it was found that the magnitude of LTP was strongly depen-dent on initial value (PB0.05).
4. Discussion
Chaotic response has been observed experimen-tally in rat CA3 neurons (Hayashi and Ishizuka, 1995). On the other hand, Tsukada et al. (1991, 1994, 1996) reported that the hippocampal LTP is highly sensitive to the temporal structure of the stimuli.
effect of the CS on CA1 LTP/LTD is the first step toward this end. In this study, we used the modified Bernoulli Map to produce CS, because the serial correlation can be easily controlled by changing the value of B, and also because the results can be compared with the previous results by Tsukada et al. (1996). The important point here is that all of three stimulus pattern have an almost identical first-order serial correlation, but the higher-order serial correlation coefficients are clearly different.
In comparison with CS and PS which has no fluctuation, the magnitude of LTP was signifi-cantly different at B=2. This result suggests that the fluctuation of CS may have an important role to induce large LTP. Aihara et al. (1997) investigated the relation between the number of short inter-stimulus intervals (S-ISI) and the in-duced LTP, and reported that the magnitude of LTP is proportional to the number of S-ISI, at least up to 200 S-ISI’s. However, our results (CS, B=3) showed that the induced LTP didn’t de-pend on the number of S-ISI, that is, the in-duced LTP magnitude is significantly different for different initial conditions, even though the number of S-ISI is fixed as 100. Therefore, our results indicate that the temporal structure of short inter-stimulus interval and long inter-stimu-lus interval plays an important role in LTP in-duction.
In the case for CS and MS, the magnitude of LTP induced by CS B=2 Imax was significantly
larger than that of MS, but this relation was reversed for CS B=3 Imax and corresponding
MS (Table 1). In spite of the same number of pulses, same mean rate, and similar first-order serial correlation, there is a big difference be-tween the induced LTP. This result suggests that the higher-order serial correlation coefficients (r2 – 15) have an effect on the LTP amplitude. Fig.
4 shows the comparison of serial correlation co-efficients between CS and MS. The higher order serial correlation coefficients(r2 – 15) are different
at B=2 Imax and B=3 Imax of CS, and these
differences may cause the observed LTP differ-ence. And especially at B=3 Imax CS contains
very long successive intervals (82) and very long short intervals (90), and it is known that
succes-sive long interstimulus intervals induce depotenti-ation of LTP. Fujii et al. (1996) also reported that long-term suppression of LTP was produced by successive long interstimulus interval (1 Hz, more than 80 pulses). This successive long inter-stimulus intervals may cause a decrease in mag-nitude of LTP at B=3 Imax of CS.
We found that the magnitude of LTP was dependent on the initial value of CS at B=2 and 3. As chaotic time series have the sensitivity to an initial condition, there is a possibility that information can be encoded in the initial values of CS.
This result suggests that a special chaotic dy-namics with higher order serial correlations plays an important role to induce even larger LTP in the hippocampal CA1 area. It is likely that the induction of LTP in CA1 is sensitive to CS, and chaotic time series may play an important role from the view point of memory coding. The fur-ther study to elucidate the relation between de-tailed time structure of chaotic time series and mechanism of LTP induction is necessary to un-derstand memory information representation in the hippocampus.
Acknowledgements
This study was supported in part by a Japanese Society for the Promotion of Science Research for the Future Program (JSPS-RFTF 96I00104).
References
Aihara, T., Tsukada, M., Crair, M.C., Shinomoto, S., 1997. Stimulus-dependent induction of long-term potentiation in CA1 area of the hippocampus: experiment and model. Hippocampus 7, 416 – 426.
Artola, A., Singer, W., 1993. Long-term depression of excita-tory synaptic transmission and its relationship to long-term potentiation. Trends Neurosci. 16, 480 – 487.
Bliss, T.V.P., Collingridge, G.L.A., 1993. Synptic model of memory: long-term potentiation in the hippocampus. Na-ture 361, 31 – 39.
Bliss, T.V.P., Lømo, T., 1973. Long-lasting potentiation of synaptic transmission in the dentate area of the anes-thetized rabbit following stimulation of the perforant path. J. Physiol. (Lond.) 232, 331 – 356.
Dunwiddie, T., Lynch, G., 1978. Long-term potentiation and depression of synaptic responses in the rat hippocampus localization and frequency dependency. J. Physiol. 276, 353 – 367.
Fujii, S., Kuroda, Y., Miura, M., Furuse, H., Sasaki, H., Kaneko, K., Ito, K., Chen, Z., Kato, H., 1996. The long-term suppressive effect of prior activation of synaptic inputs by low-frequency stimulation on induction of long-term potentiation in CA1 neurons of guinea pig hippocam-pal slices. Exp. Brain Res. 111, 305 – 312.
Hayashi, H., Ishizuka, S., 1995. Chaotic responses of the hippocaampal CA3 region to a mossy fiber stimulation in vitro. Brain Res. 686, 194 – 206.
Ito, K., Miyakawa, H., Igumi, Y., Kato, H., 1989. Long-term potentiation induced by repeated paired stimulation of the striatum radiatum and alvens in CA1 pyramidal neurons of hippocampal slices. Biomed. Res. 10, 111 – 124. McNaughton, B.L., Morris, R.C., 1987. Hippocampal
synap-tic enhancement and information storage within a dis-tributed memory system. Trends Neurosci. 10, 408 – 415.
Rose, G.M., Dunwiddie, T.V., 1986. Induction of hippocam-pal long-term potentiation using physiologically patterned stimulation. Neurosci. Lett. 69, 244 – 248.
Tatsuno, M., Aizawa, Y., 1997. Theory for characterization of hippocampal long-term potentiation induced by time-struc-tured stimuli. J. Phys. Soc. Jpn. 66, 572 – 575.
Tsukada, M., Aihara, T., Mizuno, M., Kato, H., Ito, K., 1990. Long-term potentiation to temporal pattern stimuli in hippocampal slices. In: Iwai, E., Mishikin, M. (Eds.), Vision, Memory, and the Temporal Lobe. Elsevier, New York, pp. 251 – 254.
Tsukada, M., Aihara, T., Mizuno, M., Kato, H., Ito, K., 1991. An induction algorithm for LTP in hippocampal CA1 neurons studied by temporal pattern stimulation. Proc. IJCNN-91 Int. Joint Conference Neurosci. 3, 213 – 224. Tsukada, M., Aihara, T., Mizuno, M., Kato, H., Ito, K., 1994.
Temporal pattern sensitivity of long-term potentiation in hippocampal CA1 neurons. Biol. Cybern. 70, 495 – 503. Tsukada, M., Aihara, T., Saito, H., Kato, H., 1996.
Hippocampal LTP depends on spatial and temporal corre-lation of inputs. Neural Netw. 9, 1357 – 1365.
Yamamoto, C., Sawada, S., 1981. Important factors in induc-tion of long-term potentiainduc-tion in thin hippocampal sec-tions. Exp. Neurol. 74, 122 – 130.
