Synaptic plasticity in simple systems Flashcards
What is learning and memory?
· Learning: ”the process of acquiring new information” – through experience, teaching or study
Memory: ”the persistence of learning in a state that can be revealed at a later time” – information that is learnt to be stored, encoded and retrieved at a later time
What is nativism and empiricism?
· Nativism: Plato - ’All knowledge is innate’ – hardwired into our brains at the time of birth
- Empiricism: Aristotle - Learning is the process of drawing information into the mind. The mind is a ‘tabula rasa’ on which experience is subsequently recorded. The mind is moulded from the environment and sensory experience
What is the physical basis of memory?
· Stephen Rose (1995): “Something, somewhere has to change”.
· Engrams are the physical manifestation of memory in the brain. Permanent physical change
· Plasticity is the physical process of experience-dependent change in the brain. Leads to memory formation.
Are neurons static?
Neurons are not static; their dendrites are continuously active.
How do we test memory?
· Simulated memory
· Memory in a petri dish – artificial networks from real neurons
Biological memory – most realistic, nervous system of whole organisms
What theories are used for experimentation?
· William James (1890) – when two elementary brain processes
· Donald Hebb (1949) – when axon of site A is close to site B
· McClelland & Rumelhart (1986) – when unit a and unit b are simultaneously excited
· These views convey very similar ideas: memory can be encoded by experience-dependent changes in excitability.
Connectionism: networks of neurons – memory is not stored in single neurons, instead in dense networks of neurons
What is synaptic plasticity and how can it be measured?
· Synaptic plasticity – how effectively neurons can communicate with each other
· Before learning: space between the neurons is important (called the synapse), neuron 1 is getting activity propagating across the synapse to the second neuron
· During learning (trial 1):
· Trial two: activity profile grows in the postsynaptic neuron
After learning: increase in activity in the postsynaptic neuron with less energy in the pre-synaptic neuron, can excite his neighbours so neurons can communicate more effectively
What is LTP and LTD?
· Long-term potentiation (LTP) is the long-lasting enhancement in efficacy of the synapse between two neurons.
· Long-term depression (LTD) is an alternative form of plasticity in which there is a decrease in efficacy between two neurons. (opposite to LTP)
· Occurs in many areas in the central nervous system
· These forms of Hebbian learning might form the physiological basis of memory in the brain.
· Presynaptic neuron, synapse and postsynaptic neuron
· Slow stimulation of the presynaptic neuron, take a recording of the activity in the postsynaptic neuron
Tetanic stimulation, large stimulation fired within one or two seconds.
What is associative LTP?
· A tetanus will elicit LTP if applied to a strong input.
· It will not do so if applied to a weak input alone.
However, when a tetanus is applied to both inputs simultaneously, LTP occurs in both. This is known as associative LTP.
What are characteristics of LTP?
· Synapses in LTP behave according to Hebbian rules. These are:
1. The tetanus induces repeated contiguous pre- and post-synaptic activity.
2. This results in increased efficacy between pre- and post-synaptic neurons.
· These synapses have four properties:
1. Rapidity – LTP can be induced
2. Cooperativity
3. Associativity – weak stimulation of a single pathway is insufficient for LTP if one pathway is strongly activated the neighbouring neuron experience LTP
Input specificity
What is connectionism?
· Memory is distributed in networks of neurons.
· Parallel distributed processing (PDP) McLelland & Rumelhart (1986) – inspired the parallel nature of brain networks
· Input and output with a dense network in-between, activity is propagated across the layers
Architecture consists of a neural network made up of idealised neurons connected with Hebb-like synapses. – artificial synapses
What is backprogation?
· Backpropogation: The networks are presented with training sets. Containing several examples of inputs with corresponding outputs
· Actual output – desired output = error. Connections have values reflecting their strengths.
· errors
The error signal is propagated backwards and the initially random weightings are adjusted with learning until errors are minimal.
What are the advantages of connectionism?
· Biological realism. – don’t need to be biologically realistic, easy to learn and test
· Networks learn through experience. – trial and error, how we operate in the real world, using mistakes to adjust behaviour
· Graceful degradation. – lesions to the units degrades memory but doesn’t abolish, distributed across the networks
Analytical solutions not required. – parallel networks meaning errors can be resolved easily
What are the disadvantages of connectionism?
- Notion that they can learn anything given enough training. – unlike our real brains, we cannot be good at everything
- Information is sub-symbolic. – abstract connection views
- Training is time consuming. – have to go through the process of input and output
- Networks forget learned material fast. – when they’re trained on other data sets, unlike our real brain
Learning in simple networks?
· Is there any evidence that networks of real neurons use Hebbian principles?
· It is possible to manipulate and observe the activity of real networks.
· LTP and LTD can be shown:
· In artificially grown neurons – in lab conditions
· In slices of brain tissue – taken from animals
In the nervous system
How are Real neurons used in artificial networks?
· Networks of real neurons can be grown in vitro (in cell culture in the lab) and studied. Plasticity can be studied in these networks.
· Tao et al (2000): Neurons taken from embryonic rats spontaneously branch out and form synaptic connections with each other. LTP can spread to other synapses, wanted to investigate how it happened
Found the spread of LTP were highly selected
Can LTP be measured in real neural networks?
· LTP can also be studied in real neuronal networks in the brain. Brain slices can be maintained in culture and investigated.
· The hippocampus has an important role in memory formation. LTP has been demonstrated in several areas within its circuitry.
· Bliss & Lomo (1974): Stimulated perforant pathway in hippocampus of rabbits.
Recorded from dentate gyrus.
What are the mechanisms of LTP?
· LTP relies on a cascade of molecular events that are becoming well-understood.
· LTP increases the efficacy of synapses in three ways:
1. More neurotransmitter is released from the pre- synaptic terminal.
2. More receptors available for binding at the post- synaptic terminal.
Retrograde messengers – pass backwards from the post synaptic neuron to the presynaptic neuron, stimulating the pre to post more neurotransmitters into the cleft
What does LTP depend on?
· LTP depends on NMDA receptors.
· NMDA receptors contain ion channels that can admit calcium Ca2+ (Ca2+) or sodium (Na+) ions.
The channel is normally blocked by the magnesium (Mg2+) ion. – preventing any neurotransmitters from going into the receptor
What to presynaptic neurons trigger?
· Presynaptic action potentials trigger glutamate release.
· Glutamate binds to the receptor. – not activated by glutamate as it’s blocked
· No Ca2+ can enter.
· Mg2+ ion dislodged when post- synaptic terminal is depolarized.
The pre- and post-synaptic terminal must be active for LTP to work.
Is LTP the basis of learning?
· Martin & Morris (2002): there are 4 criteria for testing this hypothesis:
1. Detectability. Changes should be detectable in the nervous system.
2. Mimicry. Artificially induced synaptic weight changes. – planting artificial memory in animals
3. Anterograde alteration. Preventing induction of synaptic weight changes during a learning experience.
Retrograde alteration. Altering synaptic weight changes induced by a prior learning experience.
What is Detectability?
· Do neurons shows changes in synaptic strength after behavioural learning? Can this be measured after learning has occurred?
· Aplysia are commonly used as an experimental model.
· Few neurons arranges in simple networks
· Behaviour and neural activity easy to measure
· Reflex: touch results in the withdrawal of the siphon, mantle or gill.
Light touch causes the gill and siphon to withdraw, a strong stimulus causes the aplysia to squirt ink through the siphon.
Classical conditioning in aplysia
· Electrical stimulation of tail (US) = gill withdrawal (UR)
· Electrical stimulation of tail (US) + Mild stimulus to mantle (CS+) = gill withdrawal (UR)
· Mild stimulus to mantle (CS+) = gill withdrawal (CR).
· Mild stimulus to siphon (CS-) = never paired with electrical stimulation of tail (US).
Classical conditioning in aplysia - how does it happen?
· Murphy + Glanzmann (1999):
· Assessed the strength of a specific sensory-to-motor synapse before, during, and after classical conditioning. – received stimulation through an electrode
· Learning-related changes in synaptic efficacy.
· Changes were NMDA-dependent.
Conclusion: this form of learning causes excitability changes and these are dependent on the NMDA receptor. – supports the hypothesis that it plays a role in classical conditioning