Long-term depression (LTD), in neurophysiology, is the weakening of a neuronal synapse that lasts from hours to days. It results from either strong synaptic stimulation (as occurs in the cerebellar Purkinje cells) or persistent weak synaptic stimulation (as in the hippocampus). LTD is thought to result from changes in postsynaptic receptor density, although changes in presynaptic release may also play a role. Cerebellar LTD has been hypothesized to be important for motor learning. However, it is likely that other plasticity mechanisms play a role as well. Hippocampal LTD may be important for the clearing of old memory traces. Hippocampal/cortical LTD can be dependent of NMDA receptors, metabotrophic glutamate receptors (mGluR) or endocannabinoids. LTD is distinct from synaptic depotentiation, which is the reversal of long-term potentiation (Barrionuevo 1980). LTD is a novel reduction in synaptic strength - specifically, an activity-dependant reduction in the excitatory post-synaptic potential compared to the baseline level.
How can a similar calcium influx cause both LTP and LTD? It depends on the timing and frequency of the input (Dudek & Bear 1992) The Bienenstock, Cooper and Munro model (BCM model - 1982) explains how the type of Ca2+ signal leads to both LTP and LTD. They postulate that when there is a low calcium influx, it leads to LTD, and a Ca2+ entry above threshold leads to LTP. The threshold level is on a sliding scale, and depends on the history of the synapse. If the synapse has already been subject to LTP, the threshold is raised, increasing the probability that a calcium influx will yield LTD. In this way, it provides a "negative feedback" system to maintain synaptic plasticity. Frequency of the stimulus is another factor for the LTP/LTD switch. Using single spike timing dependent plasticity methods (STDP), it was shown that LTP is induced when the neurotransmitter release occurs 5-15ms before a back-propagating action potential, and LTD is induced when the stimulus occurs 5-15ms after the bAP. (Markram et al 1997, Kennedy 2005) There is a plasticity window: if the pre-synaptic and post-synaptic spikes are too far apart (i.e. more than 15ms apart), there is little chance of plasticity (Bi & Poo 1998). The possible window for LTD is wider than that for LTP (Feldmann 2000) - although note that this threshold depends on synaptic history.
According to the Lisman model, protein phosphatase 1 and 2B play the same role in LTD as CaMKII does in LTP, but is activated at lower calcium levels and thus provide a mechanism to induce LTD at low calcium levels and LTP at high calcium levels.