Biology Reference
In-Depth Information
was associated with a “remapping” of the hippocampal representation of the environment (Fig
5), including creation of new place cells and abolition of preexisting place fields (Dragoi
et
al.
, 2003), supporting the view that place features of pyramidal cells emerge within
hippocampal circuits (McNaughton
et al.
, 1996; Lever
et al.
, 2002). LTP-induced effects
have been shown to be context-dependent since place fields associated with one direction of
movement were often selectively modified without affecting the neuron's place representation
when moving in the opposite direction. Therefore single neurons can be part of several
representations (O'Keefe & Nadel, 1978; Markus
et al.
, 1995; Wood
et al.
, 1999) and inputs
from these representations can be modified selectively.
Changes in place field representation
did not affect the size and shape of place fields. No effect was observed on theta power and
theta cycle compression of distances between place fields. Importantly the global firing rate
of the network was preserved after the LTP protocol, suggesting that LTP rearranges the
place representation in the hippocampus without altering the functional properties of the
network (Dragoi
et al.
, 2003).
Interventional synaptic plasticity studies demonstrate that the persistence of the fields
over a period of time greater than a few minutes or hours seems to depend on a mechanisms
common with these underlying LTP-models. This suggests that spatial mnemonic functions
recruit place fields in a non-NMDA-dependent process and then associates them, via an
NMDAR- and protein synthesis-dependent processes, so that upon re-entry, the same map can
be recalled. Experience-dependent plasticity of place field can be explained on a cellular
level, but how these changes are mediated by the network connectivity remains difficult to
reveal. The following section analyses the up-to-date computational models and their
experimental substrates of how network dynamics define the functional properties of
hippocamapl system.
5.
H
IPPPOCAMPAL
N
ETWORK AS A
M
EMORY
S
YSTEM
The memory stored in any neuronal network can be represented by the firing rates of the
population of neurons that are stored by the associative synaptic modification, and can be
correctly recalled later (Treves & Rolls, 1991, 1992; Rolls & Kesner, 2006). Computational
models suggest that autoassociation networks that undergo Hebbian modification can store
the number of different memories, each one expressed as a stable attractor. An attractor
network is one in which a stable pattern of firing is maintained once it has been started. In
hippocampal region the CA3 neurons are proposed to operate as an attractor network (Treves
& Rolls, 1991, 1992; Rolls
et al.
, 1997; Kesner & Rolls, 2001). Associative modification is
mediated by long-term potentiation, and this synaptic modification appears to be involved in
learning (Morris, 2003; Morris
et al.
, 2003; Lynch, 2004).
In order for most associative networks to store information efficiently, heterosynaptic
long-term depression is required (Rolls & Treves, 1990; Treves & Rolls, 1991; Fazeli &
Collingridge, 1996; Rolls & Treves, 1998; Rolls & Deco, 2002). Without heterosynaptic
LTD, there would otherwise always be a correlation between any set of positively firing
inputs acting as the input pattern vector to a neuron. LTD effectively enables the average
firing of each input axon to be subtracted from its input at any one time (Rolls, 1996).
