Do different threshold levels of predict the direction of synaptic modification?

V

could be brought about in these experiments either by antidromic Na^ spikes with some soma depolarization due to summation of DAPs, or mainly by soma depolarization with a lower frequency of action potentials following depolarizing current pulses. In the measurements of cytosolic [Ca^^], it was found that it increased as a result of action potentials or (in the presence of tetrodotoxin) induced by depolarizing pulses alone, above a threshold value (unpublished observations; see also Ross et al., 1991; Jaffe et a l , 1992).

The intriguing question that remains is how ^/s^jthat a high [Ca^^] in dendrites can give rise to LTP and a lower [Ca^'^j to LTD? The work of Artola et al. (1990), Kimura et al. (1990), Yoshimura et a i (1991), Brocher et al. (1992) and Hirsch and Crepel (1992) in neocortex, points to the crucial role played by Ca^"^ in the induction of LTP when it is at a high level due to a high level of postsynaptic activity, and of LTD when at a lower level, but above a certain threshold. In this study, measurements of cytosolic Ca^'*' using Fura-2 showed that less Ca^"^ entered the cell soma during

postsynaptic depolarization and firing in a medium containing 25 mM Mg^"'’ and 2 mM Ca^^ than in normal acsf. How do the results of Artola et al. (1990) correlate with those of this study? If we are to accept the theory of Artola et al. (1990), how can we explain the apparent lack of LTP or LTD observed in acsf, although the somatic Ca^^ entry measured during conditioning in acsf was greater than in the presence of a raised [Mg^^] which did induce LTD? If one assumes that soma Ca^^ is a reasonable

representation of the dendritic Ca^"^ level, then one would expect to observe LTD or LTP following conditioning in acsf. Alternatively, the results of this study suggest that the Ca^"^ entry that occurred in normal acsf was above the threshold for producing LTD, but below that for inducing LTP. Indeed, in this study it was shown that the rise of [Ca^‘^1 per action potential in the bathing medium containing 15 mM Mg^"^ and no added Ca^"*" was 17% less than in that containing 25 mM Mg^"^ and 2 mM Ca^"^, which

suggests that there is a narrow divide between the conditions for LTP, LTD and zero effect. Furthermore, the data of other workers obtained using Ca^^ chelators in visual cortex also point to there being a narrow divide (Artola et al. , 1990; Kimura et al.,

1990; Yoshimura et a l , 1991; Brocher et al .,1992; Hirsch and Crepel., 1992).

Hypothesis 3

(i.e. [Ca^'*’] range) above that for LTD induction and below that required for the induction of LTP. If this level were reached in conditioning, it could result in neither LTD nor LTP induction. Alternatively, a [Ca^^] threshold might exist that if reached following conditioning, favoured the induction of both LTP and LTD of different synaptic inputs. In this case both LTD and LTP might occlude each other, and appear to the experimenter as an unsuccessful attempt to induce either phenomenon. In the presence of a high [Mg^^] solution during conditioning, postsynaptic Ca^^ entry could be reduced to a dendritic level that favoured LTD induction.

Thus by blocking the dendritic depolarizing effect of GABA released from interneurones activated during conditioning, CNQX could also restrict Ca^^ entry postsynaptically to a level that favoured the induction of LTD.

Figure 6.1 summarizes this hypothetical situation, which may explain why LTD did not occur following conditioning in acsf, although a greater level of Ca^"^ entry occurred in acsf than in the high [Mg^‘*‘] solutions.

Lisman (1989) has proposed a model for the induction of Hebbian LTP and anti- Hebbian LTD that depends on the quantitative level of the postsynaptic rise in [Ca^^]. Lisman (1989) suggested that the synaptic weight of each synapse could be stored locally by the group of Ca^"^/calmodulin kinase II (Cam-kinase II) molecules contained in the postsynaptic density (Kennedy et al., 1983; Kelly et at., 1984), a cytoplasmic structure directly abutting the postsynaptic receptors within each dendritic spine

(Siekevitz, 1985). The bidirectional control of synaptic weight could be brought about by the control of different enzymes by the postsynaptic [Ca^^]. A high elevation of Ca^'^ postsynaptically could lead to phosphorylation of CaM-kinase II, an increase in synaptic weight, and LTP, while a moderate level of Ca^'*' could activate protein phosphatase 1, an enzyme directly controlled by Ca^^ through reactions with

phosphatase inhibitor 1, cAMP-dependent kinase, calcineurin, and adenylate cyclase. Protein phosphatase 1 dephosphorylates CaM-kinase II, and leads to a reduction of synaptic weight, i.e., to LTD. Conclusive evidence for a common mechanism for LTP and LTD is still lacking, however, because the complete or partial blockade of LTD may reflect mechanisms with a higher sensitivity to intracellular [Ca^^] than those underlying LTP. The test of these hypotheses will depend on the ability to measure EPSPs, together with localized changes in [Ca^^] (Muller and Connor, 1991; Guthrie et at., 1991) and related biochemical events at the region of the activated synapses and the

i

LTP

LTP/LTD cancel out

(0

Ü

LTD

No change

Figure 6.1. A model of how postsynaptic dendritic levels of intracellular Ca^^ may determine changes in the efficacy of synaptic transmission. The Ca^^ levels are

arbitrary, and each box is drawn the same dimensions for clarity, although this may not be the case in reality.

examination of whether the postsynaptic cell could behave like an analogue computer that can store a synaptic weight and modify it in accordance with the Hebbian

(associative) and anti-Hebbian (non-associative) learning rules.