Two key questions emerge by the above considerations: (a) what is the characteristic of response (EPSP or EPSC) which better represent the code of the single bit of the
synaptic information? and (b) how does the single bit of synaptic information produces a
synaptic code at the postsynaptic level? The two questions are not independent to each
other. The best candidate to code for the single bit of information seems to be the
EPSP (or EPSC) “amplitude.” The amplitude depends on the characteristic of the
synapse (number of receptors, PSD input impedance, spine neck resistance, etc.)
and on the activity of the dendrite on which the synapse is hosted.
The amplitude of an EPSP occurring when the postsynaptic membrane is close to
the reverse potential 0 mV can approach 0 mV. This means that the postsynaptic
mechanism of tuning can, depending on its state, nullify the information. Alternatively, an EPSP starting when the postsynaptic membrane potential is close to the
resting potential (or even in a hyperpolarized state), the amplitude is maximized
[65, 66].
Interestingly, if the single bit of information is coded by the EPSP (EPSC)
amplitude, while a diffuse excitation depolarizing the membrane reduces the
amount of information passed by the synapse, the inhibition works in the opposite
direction. Driving the membrane potential far from the reverse potential, in fact,
the inhibitory inputs play a favor of increasing the amplitude [30].
Assuming that the single spike represents the single bit of information of a
neuron, a sequence of spikes emitted by a presynaptic neuron represent a “word”
that is the full representation of a stimulus in that neuron. The synaptic codification
of this “word” should be an equivalent sequence of EPSP. This does not always hold.
As we have said, the probability that an EPSP is generated when a spike arrives is
less than 1. Moreover, EPSPs sum non linearly at the postsynaptic side and the
amplitude and shape of the resulting sum depend on the time between the EPSPs. In
addition, the different EPSPs are modulated postsynaptically each differently
depending on the coincidence of their start and the phase of the wave produced by
the dendritic activity. The same presynaptic “word” can then have different postsynaptic representations since formed by different number of EPSPs coded with
different amplitudes and presenting different shapes and duration because of the
different NMDA contributions. In short, rarely the same repeated stimulus
represented by a sequence of spikes will have a fixed clearly identifiable representation at the postsynaptic side. This variability of synaptic representation of a
“word” is probably the main cause of the variability of the postsynaptic neuronal
“word” (different sequences of the postsynaptic spikes). This means that the single
presynaptic “word” almost never determines the postsynaptic spike sequence
(postsynaptic “word”) which is always the results of the cooperation of all the
inputs arriving in a given time window. Although in many experimental results it is
possible to identify a sort of relationship between a stimulus and some characteristics of the spikes sequence it induces in a given neuron, probably in the real brain,
the situation is much more complex.
A last comment on how the mechanisms of postsynaptic regulation play a role in
the information processing by considering the different information arriving from
many neurons on a single one. If we consider the inputs on a single neuron coming
from two areas of the brain and located in the close proximity on the dendritic tree,
the area which sends early the information can inhibit the information of the other
area. A massive input arriving from many excited neurons of a firstly activated area
will produce a strong depolarization of the dendritic area which will inhibit (if not
nullify) the information arriving from the other area. This can be probably a mechanism which regulates, at the single neuron level, the competition between two
antagonist inputs involving different areas of the brain but also a mechanism of
“decision-making.” The priority for the response, in this case, is time dependent
102
Advances in Neural Signal Processing
synaptic information? and (b) how does the single bit of synaptic information produces a
synaptic code at the postsynaptic level? The two questions are not independent to each
other. The best candidate to code for the single bit of information seems to be the
EPSP (or EPSC) “amplitude.” The amplitude depends on the characteristic of the
synapse (number of receptors, PSD input impedance, spine neck resistance, etc.)
and on the activity of the dendrite on which the synapse is hosted.
The amplitude of an EPSP occurring when the postsynaptic membrane is close to
the reverse potential 0 mV can approach 0 mV. This means that the postsynaptic
mechanism of tuning can, depending on its state, nullify the information. Alternatively, an EPSP starting when the postsynaptic membrane potential is close to the
resting potential (or even in a hyperpolarized state), the amplitude is maximized
[65, 66].
Interestingly, if the single bit of information is coded by the EPSP (EPSC)
amplitude, while a diffuse excitation depolarizing the membrane reduces the
amount of information passed by the synapse, the inhibition works in the opposite
direction. Driving the membrane potential far from the reverse potential, in fact,
the inhibitory inputs play a favor of increasing the amplitude [30].
Assuming that the single spike represents the single bit of information of a
neuron, a sequence of spikes emitted by a presynaptic neuron represent a “word”
that is the full representation of a stimulus in that neuron. The synaptic codification
of this “word” should be an equivalent sequence of EPSP. This does not always hold.
As we have said, the probability that an EPSP is generated when a spike arrives is
less than 1. Moreover, EPSPs sum non linearly at the postsynaptic side and the
amplitude and shape of the resulting sum depend on the time between the EPSPs. In
addition, the different EPSPs are modulated postsynaptically each differently
depending on the coincidence of their start and the phase of the wave produced by
the dendritic activity. The same presynaptic “word” can then have different postsynaptic representations since formed by different number of EPSPs coded with
different amplitudes and presenting different shapes and duration because of the
different NMDA contributions. In short, rarely the same repeated stimulus
represented by a sequence of spikes will have a fixed clearly identifiable representation at the postsynaptic side. This variability of synaptic representation of a
“word” is probably the main cause of the variability of the postsynaptic neuronal
“word” (different sequences of the postsynaptic spikes). This means that the single
presynaptic “word” almost never determines the postsynaptic spike sequence
(postsynaptic “word”) which is always the results of the cooperation of all the
inputs arriving in a given time window. Although in many experimental results it is
possible to identify a sort of relationship between a stimulus and some characteristics of the spikes sequence it induces in a given neuron, probably in the real brain,
the situation is much more complex.
A last comment on how the mechanisms of postsynaptic regulation play a role in
the information processing by considering the different information arriving from
many neurons on a single one. If we consider the inputs on a single neuron coming
from two areas of the brain and located in the close proximity on the dendritic tree,
the area which sends early the information can inhibit the information of the other
area. A massive input arriving from many excited neurons of a firstly activated area
will produce a strong depolarization of the dendritic area which will inhibit (if not
nullify) the information arriving from the other area. This can be probably a mechanism which regulates, at the single neuron level, the competition between two
antagonist inputs involving different areas of the brain but also a mechanism of
“decision-making.” The priority for the response, in this case, is time dependent
102
Advances in Neural Signal Processing