Transcript Hippocampus

Declarative Memory
Declarative or episodic memory is for arbitrary and non-repeated associations. For
example, when birds or squirrels store seeds for the winter they must
remember the arbitrary locations and their contents (episode) even if they have
just been to the particular places just once and several months earlier. In humans
it would mean remembering a particular person and what they said (an episode)
as opposed to learning a motor skill or improving one’s perception.
There is extensive evidence that the hippocampus if essential for episodic memory in
humans and animals.
So the next section will cover:
1. Hippocampal anatomy.
2. Evidence that the hippocampus is essential for human memory.
3. What does the hippocampus do in other animals?
4. Spatial maps
5. Hippocampal and entorhinal cortex physiology- rhythms.
6. Hippocampal and entorhinal cortex physiology- place cells.
7. Non-spatial hippocampal memory.
8. Tentative conclusions.
9. Hippocampal cellular physiology.
10. Consolidation.
11. Adult neurogenesis in the hippocampus
The hippocampus: anatomy 1
The hippocampus is medial pallium- that is, it is arises from the dorsal medial part of
the neural tube (bounded by the ventricle). This is seen in reptiles, birds and
mammals like the opossum. In primates especially the growth of the brain drags the
hippocampus into the temporal lobe.
In all cases the hippocampus is a 3 layer cortex. It has two main subdivisions: the
dentate gyrus (DG) and the cornu ammonis (CA). The CA is itself subdivided into 2
regions: CA1 and CA3.
Connecting the 3 layer hippocampus to the standard 6 layer neocortex are 2 regions
intimately involved in hippocampal function: the subiculum and entorhinal cortex.
The hippocampus: anatomy 2
Here is a classic image from Ramon y Cajal illustrating the principle cells (excitatory
projection cells) of the hippocampus (we won’t deal with the numerous GABA
interneurons). The DG has modified pyramidal cells (called granule cells) that have only
apical dendrites. The CA fields contain fairly typical pyramidal cells with both basal and
apical dendrites, both covered in spines.
The hippocampus: main input
The hippocampus gets most of its input from neocortex. Not directly but via the entorhinal
cortex. These entorhinal to hippocampal projections go to both the CA, CA1,3 fields and
to DG. The entire cortex can influence the hippocampus this way.
In addition the hippocampus receives input from septum and hypothalamus via the
fornix/fimbria.
The hippocampus: schematic connections
Plasticity:
1. Perforant pathway:
postsynaptic LTP and LTD
(NMDA-R dependent)..
2. Mossy fibers: presynaptic
LTP (NOT NMDA-R).
3. Schaffer collaterals:
postsynaptic LTP and LTP
(NMDA-R dependent).
This diagram illustrates the famous trisynaptic pathway. Fibers from the entorhinal cortex
reach the DG via the perforant path and contact the DG granule cell dendritic spines. The
DG granule cells then project mossy fibers to the proximal apical dendritic spines of CA3
pyramidal cells. The CA3 pyramidal cells then project (Schaffer collaterals) to the CA3
apical dendritic spines. CA3 pyramids the project out of the hippocampus.
Also shown is the projection of CA3 pyramidal cells down to the hypothalamus (mamillary
body) via the fornix (this pathway is both input and output).
The hippocampus: detailed connections
Deficits after hippocampus damage in humans
A classic measure of human declarative memory- remembering faces- is completely
wiped out in patients with bilateral hippocampal damage.
This and a great deal of related work initiated by the neurosurgeons, Drs. Penfield
and Scoville and the neuropsychologist, Dr. Brenda Milner, clearly demonstrated the
importance of the hippocampus for memory.
Patients with frontal lobe damage are fine on this declarative memory task.
In contrast frontal lobe, but not hippocampal damage produces deficits in the ability
to change strategies in a card sorting task.
Memory remaining after hippocampus damage
A motor task such as mirror drawing is normally learned by the famous hippocampal
damaged patient, H.M. This is despite the fact that he was unable to remember
doing this task from day to day.
So sensory-motor learning is not affected by hippocampal damage.
Neither is emotional learning.
Episodic memory in animals
Here is the famous delayed non-match to sample task
(DNMS); a variant of this task is delayed match to
sample. The animal has to recognize, after a delay, an
object different from the one it has seen previously.
After an intertrial interval this task is repeated but with
different objects.
Humans, monkeys are very good at this task.
Hippocampal damage completely wipes out this form
of memory.
Equivalent tests also work for rodents.
This is a test for episodic memory and believed to be
the animal version of declarative memory.
Spatial learning in animals
(b) In the radial spatial maze all the arms are
initially baited with food. Once the rat finds
and eats the food in a maze arm, it is not
replaced.
The rat quickly learns to NOT visit the arms it
has already visited (they no longer contain
food).
This is an example of an episodic memory
that involves the rat remembering specific
spatial locations.
Hippocampal lesions completely wipe out this
form of memory.
Note: Olton originally called this working
memory but that term is used in a different
way now.
(c) In this case some arms are never baited.
Again the rat figures this out and never visits
the unbaited arms. This type of learning is not
permanently prevented by the same
hippocampal lesions.
So spatial learning is quite subtle.
Hippocampal activation during spatial navigation:
Landmark Learning
Subjects (human) navigated through a virtual town on a computer screen. Their brain
activity was monitored with positron emission tomography (PET).
The regions with the greatest increases in activity were the right hippocampus and
the left caudate (part of the striatum).
This clearly implicates the hippocampus in the processing of spatial information.
The only spatial information available in this task is the spatial relations of objects in
the environment. These are known as landmarks and it is thought that learning the
relative position of landmarks is critical to spatial learning.
Spatial navigation: Path Integration in insects
(b). Detours are imposed in the ant’s path.
The animal takes into account the distance
and direction it has traveled during the
detour and still finds its way home.
(a). An ant leaves its nest (N) and forages
with a very complicated path until it finds
some food (F). It then carries the food back to
the nest via a direct path. The ant must
somehow take into account all the twists and
turns it has made as well as the linear
distance it has traveled in any direction.
In mathematical terms this is easy to do. For
example, lets imagine that you know the
velocity of an object and its final position.
Velocity is the derivative of distance with
respect to time; so if you integrate velocity
you get back the distance traveled. This
information + the final position allows you to
compute the initial position. The same applies
to rotations. Hence the term: path
integration.
Spatial navigation: Path Integration in mammals
These experiments were done with hamsters
and in total darkness- no landmarks were
visible.
The arena is rotated at the beginning of
each experiment. The animal is guided to
food somewhere near the middle of the
arena and must then find its way back.
It succeeds in finding its way back
presumably using only path integration.
There have been many experiments demonstrating path integration in mammals. The
neural basis of this ability is not known precisely. It presumably requires internal
(idiothetic) cues as to the direction and distance of movements. This might come from
several sources including: vestibular input, proprioceptive input and corollary
discharge. At least vestibular input appears to be required for this ability and this
information can reach the hippocampus.
Head direction cells discharge when the animal’s head is oriented in a specific
direction. So they do have the directional signal required. These cells are found in
brain areas projecting to the hippocampus and so are likely involved in path
integration.
The consensus model is that landmark and path integration information are both
required for spatial navigation and that these kinds of information are present in brain
structures projecting to the hippocampus.
Spatial learning: the Morris water maze
A rat is released into a tank full of opaque water. There is usually a landmark visible- a highly
visible cue card at one position at the edge of the tank. The rat has to find its way to a
platform that is under the opaque water and therefore invisible (or it will eventually drown). It
quickly learns to do so. The rat is presumably using some combination of landmarks (the cue
card) and path integration to solve this spatial problem. This is now the standard test for
spatial learning in rodents.
Lesioning the hippocampus prevents this spatial learning.
A variety of interventions in the hippocampus also degrade this spatial learning. These
include: blocking NMDA receptors (pharmacology or genetically), blocking CaMKII etc.
Spatial learning and the hippocampus
Hippocampal lesions prevent rats learning the water maze.
A strategy of numerous studies has been to use pharmacological or, more recently,
molecular genetic means to perturb NMDA receptors (in the hippocampus) or their
downstream targets and then measure synaptic plasticity (LTP), water maze learning and
place cells (next section). The conclusion of most (not all) of these studies is that
perturbing NMDA receptors blocks LTP and spatial learning and perturbs place cells.
Theta rhythms in hippocampus
Neurons in the hippocampus (and elsewhere)
often fire in a rhythmic manner at around 4
Hz. This is known as the theta rhythm. Theta
will occur during behaviors such as sniffing
and during spatial navigation.
The firing of place cells (next section) is often
phase locked to the theta rhythm.
Synaptic plasticity (LTP) in hippocampus is
maximized when the inputs are stimulated at
the theta rhythm: theta burst stimulation.
The relationship between oscillatory
discharge of hippocampal pyramidal cells,
synaptic plasticity, spatial navigation and
other learned discriminations is a subject
under intense investigation and beyond the
scope of this course.
The cellular and network basis of theta and
other rhythms is not understood and also the
subject of intense study.
There is other oscillatory activity in the
hippocampus: gamma oscillations (30-50 Hz)
and high frequency “sharp-wave ripples”.
The interactions and roles of these rhythmic
activities is unknown.
Hippocampal Place Cells 1
A subset of hippocampal neurons discharge whenever the
animal (typically a rodent) is in a specific location within the test
chamber. Cell 1 in the figure discharges in the lower left region
and continues to do so even after a partition is removed to
create a larger chamber.
Cell 2 does not discharge in the small arena but develops a
place field after the barrier is removed.
This figure begins to give you and idea of the complexity of place
cells. It is known that, while place cell discharge can be
controlled by visual landmarks, they can also occur in total
darkness. This has led to the hypothesis that they are controlled
by both landmark and path integration information; although
there are theories about this relationship, they are not proven.
Place fields are highly labile and many environmental
manipulations will change place fields; the basis of this effect is
not known.
In most cases molecular genetic and other disruptions that
interfere with LTP also reduce the stability of place fields and
new spatial learning, but the causal relationships are not clear.
There are also many reports of hippocampal neurons that
respond to non-spatial signals related to learning; for example, in
association with eyeblink conditioning and odor discrimination.
Entorhinal Cortex Place Cells 1
Recent studies have shown that the entorhinal
cortex contains very well defined place cells.
These have a regular triangular pattern of
place fields and the population of these cells
can exactly code for the location of the
animals. Further evidence demonstrates that
place cells are first generated in entorhinal
cortex and then passed on to hippocampus.
How place cells are generated is not known
exactly.
Entorhinal Cortex Place Cells 2
Entorhinal cortex place cell’s place fields are
controlled by the location of a cue card. This
implies that landmarks control the place
cells. This makes sense since the rat must
orient itself to structures in its environment.
Entorhinal cortex place cell’s fields remain stable
after long periods in the dark. This implies that
path integration operates to maintain the place
fields in the absence of landmark cues.
So entorhinal place cells are generated by a combination of landmark and path integration
inputs. Presumably these are linked together by plastic synapses but this is not
understood. It is known that entorhinal place fields are rapidly generated and then remain
stable but the cellular mechanisms are not known.
Non-spatial information in hippocampal place cells
Here are examples of a place cells that discharge
differently depending on the final destination. The
cell in A discharged when the animal was going
to turn to the left. The cells in B and C discharged
when the animal was going to turn to the right.
This is despite the fact that they traverse the
same location no matter whether they will (later
on) turn to the left or right.
So hippocampal place cells can vary their
discharge for complicated non-spatial reasons.
There are numerous other examples of this type
of behaviour.
Hippocampal Place and Episodic memory
Place cell discharge rate is controlled by
a non-spatial cue.
Place cell spatial discharge pattern is
controlled by differing locations with a
constant non-spatial cue.
Hippocampus: memory for what and where: episodic
and spatial?
The figure to the left (a) represents the theory that the hippocampus contains a map of
space.
The figure to the right (b) represents the theory that the hippocampus codes for
relationships: primarily between episodes (events- have I encountered the cylinder
before and what does it mean to me?) and where they occur (place fields- where do I
expect the cylinder and near what else is it located?). The evidence is more in favour
of the “relational” hypothesis to the right although this subject is still at the center of
intense research.
Hippocampal Cellular Physiology: LTP and LTD
Different patterns of stimulation can induce
LTP (theta burst) or LTD (low rates) in the
same hippocampal pyramidal cell. The
plasticity is specific to the pathway
stimulated. The molecular basis involves
NMDA receptors and the different affinities
of CaMKII (kinase) and calcineurin
(phosphatase) for Ca2+.
Very long lasting LTP (late LTP) also
required transcriptional regulation and new
protein synthesis.
The theory (well supported) is that
associating places with events requires
LTP and LTD.
Consolidation 1
H. M. and other patients with hippocampal damage have anterograde amnesia; that
is they are unable to retain long term memories for events (declarative or episodic
memory) that occur after their hippocampal damage.
In these and other cases the patients (or experimental animal) may be unable to
remember events before the trauma (or experimental manipulation). This is
retrograde amnesia. Typically retrograde amnesia is itself graded: it is most severe for
events just before the trauma while events much further away in time may be
remembered. This appears to be a paradoxical results: better memories for distant
versus recent events.
An quantitative example has been noted for electroconvulsive shock therapy (ECTfor depression); in this case memories for the 3 years prior to ECT were lost while
earlier memories were intact.
Hippocampal lesions in animals produce retrograde amnesias that range from a few
days to 12 weeks.
Consolidation 2
These results (and many others) have led to the theory that, after initial learning has
taken place, there is a lengthy period during which the new memories become
independent of the hippocampus- the initial site of storage. They now become more
permanently stored in the neocortex. This process is referred to as consolidation.
The term consolidation is also used to refer to the cascade of molecular interactions
that convert LTP into very long lasting LTP (transcriptional regulation and new
protein synthesis). I shall indicate the usage as appropriate.
There is extensive evidence that cellular/molecular basis of behavioural
consolidation includes transcriptional regulation, protein synthesis, the change
(strengthening and weakening) of existing synaptic connections and the formation of
new connections in hippocampus and cortex. There may also be an involvement of
neurogenesis (next section).
Hippocampus
Temporary memory
storage
consolidation
Molecular
Cellular
Network
changes
Neocortex
Long term memory
storage
Consolidation 3
Consolidation Theory
According to one widely held theory episodic learning first involves rapid changes in
hippocampal synaptic connections; the changes include both phosphorylation of
synaptic proteins and new protein synthesis and last for hours to days.
Subsequently the hippocampus transfers the new memories to neocortex over a
period of days to weeks.
The latter process might involve memory replay during sleep (in hippocampus)
perhaps during sharp wave ripples.
Cellular Bases of Consolidation 1
Mice with heterozygous null mutation of CaMKII have impaired LTP in neocortex and do not
consolidate memories. Memories are intact for a few days and then detiorate.
Mice with homozygous null mutations do not have hippocampal LTP and do not form
memories in the first place.
Cellular Bases of Consolidation 2
Radioactive 2-deoxyglucose is used to
probe for the distribution of high levels
of activity during recall of recent versus
remote memories.
Cellular Bases of Consolidation 3
An immediate early gene (c-fos) is used to
probe for the storage sites of remote and
recent memory traces in parietal cortex.
Adult Neurogenesis in Hippocampus
Neural stem cells located in the
subgranular zone (SGZ) of the
hippocampus (on the ventricle)
differentiate into immature
neurons and then migrate into
the dentate gyrus where they
mature into DG granule cells.
These cells then make
appropriate synaptic
connections.
Environmental conditions promote survival of newly
generated hippocampal neurons
Enriched environment for rats
Greater number of
neurons in DG of rats
from enriched
environments.
Development of new DG neurons
Current evidence now suggests that new DG neurons may also be important in the storage of
new memories- they appear to be more “plastic” than mature neurons. How this type of
memory storage relates to consolidation and the transfer of memories from hippocampus to
neocortex is not understood at all.