Anatomy of

the Hippocampus

Computational Models of Neural Systems
Lecture 3.2

David S. Touretzky
September, 2015
 Human Hippocampus

10/06/15 Computational Models of Neural Systems 2
 Human Hippocampus

10/06/15 Computational Models of Neural Systems 3
 Hippocampus Means “Seahorse”

Dissected human hippocampus next to a specimen of hippocampus leria, one 
of several dozen species.
10/06/15 Computational Models of Neural Systems 4
 Rat Hippocampal Formation

septal

temporal

From (Amaral & Witter, 1989)

10/06/15 Computational Models of Neural Systems 5
Ammon: Egyptian god with ram's horns

10/06/15 Computational Models of Neural Systems 6
 Rat Hippocampal Formation

septal

temporal

10/06/15 Computational Models of Neural Systems 7
Components of Hippocampal Formation
● The “hippocampus”
contains CA1 - CA3.

CA = Cornu Amonis
Ammon's Horn
(Ammon's horn)
● The “hippocampal
formation” includes
CA, dentate gyrus
(“tooth-like bump”),
entorhinal cortex,
subiculum, pre- and
parasubiculum.
● Hilus: reciprocally
connected to DG.

10/06/15 Computational Models of Neural Systems 8
 Early Anatomy: Cajal's Drawings

10/06/15 Computational Models of Neural Systems 9
 Rat Hippocampus

10/06/15 Computational Models of Neural Systems 10
 Monkey Hippocampus

10/06/15 Computational Models of Neural Systems 11
 Human Hippocampus

The volume of the human hippocampus is about 100 
times that of the rat, and 10 times that of the monkey.

10/06/15 Computational Models of Neural Systems 12
 Basic Circuit

Layer III perforant path

EC           Layer II perf. path

Schaffer
Mossy fibers collaterals
DG CA3 CA1
Recurrent
The mossy fiber synapse is one of the 
collaterals
largest and most powerful synapses in 
the brain.

10/06/15 Computational Models of Neural Systems 13
 Connectivity Is Mostly Uni-Directional

10/06/15 Computational Models of Neural Systems 14
 What About CA2?
● Small relative to CA3
and CA1
● Originally dismissed as
a “transitional area”
● Driven by Schaffer
collaterals from CA3
● Projects back to CA3
and to two layers of CA1
● Does not show activity-
dependent LTP as found
in CA3 and CA1
● In mice, contributes to
social memory
Caruana, Alexander, and Dudek (2012), Fig. 1

10/06/15 Computational Models of Neural Systems 15
 More of the Picture

● EC: entorhinal cortex Other cortical &

EC
– Layer II projects to DG/CA3 subcortical areas
– Layer III projects to CA1/Sub
pp
– CA1 & Sub project back DG Hilus
to EC layer V mf

● DG: dentate gyrus CA3

– mossy fibers project to CA Sch

CA1
● CA3
– Schaffer collaterals to CA1
Sub
● Sub: subiculum
10/06/15 Computational Models of Neural Systems 16
 Even More

PR: perirhinal cortex; POR: postrhinal cortex; EC: entorhinal cortex; PrS: presubiculum; PaS: 
parasubiculum; DG: dentate gyrus; CA: Cornu amonis; S: subiculum; RSP: retrosplenial cortex; 
Par/Oc: parietal/occipital cortex

10/06/15 Computational Models of Neural Systems 17
 Three Major Fiber Systems
● Angular bundle
from EC: perforant
path (and more)

● Fimbria/fornix to
subcortical areas
– fimbria: red
– fornix: yellow

● Dorsal and ventral

commissures link
hippocampi
– dorsal: purple
– ventral: green

10/06/15 Computational Models of Neural Systems 18
 Alveus/Fimbria/Fornix

● Second input/output
pathway for
hippocampus
(first is via angular
bundle from EC)

● Communicates with
subcortical structures:
septal nuclei and
mammillary bodies

10/06/15 Computational Models of Neural Systems 19
(commissural
 pathway) (perforant path)

10/06/15 Computational Models of Neural Systems 20
Dendritic Arborization of Principal Cells

suprapyramidal blade

infrapyramidal blade

10/06/15 Computational Models of Neural Systems 21
 Stucture of the Dentate Gyrus
– Granule cell layer holds principal cell bodies
● their axons form the mossy fiber pathway
– Molecular layer: gc dendritic tree; afferent connections
– Polymorphic cell layer (hilus): interneurons, mf collaterals

10/06/15 Computational Models of Neural Systems 22
 Mossy Fiber Synapse Onto CA3 Cells
● The dentate gyrus projection
to CA3 terminates in
large mossy fiber synapses.

● CA3 dendrites have “thorny

excresences” with complex
spine shapes. A mossy fiber
can make 30-40 synapses
within one excresence.

● Each granule cell contacts

only about 15 CA3 pyramidal
cells. Each pyramidal cell
receives input from only
about 72 granule cells.
10/06/15 Computational Models of Neural Systems 23
Structure of CA3

● stratum radiatum:
entorhinal afferents;
mossy fibers enter
from DG, make
synapses in s. lucidum

thorny excresences
● stratum lucidum
● stratum pyramidale

● stratum oriens:
recurrent collaterals
10/06/15 Computational Models of Neural Systems 24
Structure of CA1
● stratum lacunosum-
moleculare:
entorhinal afferents

● stratum radiatum:
Schaeffer collaterals,
commissural fibers

● stratum pyramidale

● stratum oriens

10/06/15 Computational Models of Neural Systems 25
 Interneurons in CA3/CA1

10/06/15 Computational Models of Neural Systems 26
Structured Projections: EC Layer II vs. III
EC layer II
EC layer III

10/06/15 Computational Models of Neural Systems 27
 Layer II Projections

MPP/LPP = 
medial/lateral
perforant path

10/06/15 Computational Models of Neural Systems 28
 Layer III Projections

10/06/15 Computational Models of Neural Systems 29
 Temporal-Lobe.com

10/06/15 Computational Models of Neural Systems 30
10/06/15 Computational Models of Neural Systems 31
 Some Numbers for the Rat
● Dentate Gyrus
– 1.2 million granule cells
– 4K basket cells
– 32K hilar interneurons (20K mossy cells)

● CA3/CA1
– 330K /420K pyramidal cells
– various interneurons

● Entorhinal cortex layer II

– Around 200K cells (20% interneurons?)

● Subiculum
– Around 180K cells
10/06/15 Computational Models of Neural Systems 32
 Rat Connectivity
● Perforant path projection to DG
– Around 4500 spines per granule cell (75% from EC)
– One EC cells makes about 18,000 synapses with granule cells

● CA3: three distinct inputs

– 50-80 mossy fibers from DG
– 3,500 perforant path synapses from EC II
– 12,000 recurrent collaterals from other CA3 cells
● 8,000 to basilar dendrites (stratum oriens)
● 4,000 to apical dendrites (stratum radiatum)

● CA1: inputs from CA3 and EC

– From CA3 Schaffer collaterals: 4,500 basilar, 6500 apical synapses
– From EC layer III: 2,500 synapses
10/06/15 Computational Models of Neural Systems 33
 Why Does the Circuitry Look Like This?
● Why so many kinds of interneurons?
● Why have recurrent connections in
some regions and not others?
● What do these regions compute?

CA3
10/06/15 Computational Models of Neural Systems 34
 The Hippocampus Is Very Well
Connected to Other Brain Areas
Connected via EC and fimbria-fornix pathway to:
– Prefrontal / orbitorfrontal cortices
– Cingulate cortex
– Piriform cortex
– Perirhinal and Postrhinal cortices (sensory)
– Striatum
– Amygdala
– Septum
– Mammillary bodies
– Thalamus
– Hypothalamus

10/06/15 Computational Models of Neural Systems 35
 LEA/MEA = lateral/medial 
entorhinal area

PER
`

10/06/15 Computational Models of Neural Systems 36
 Neuromodulatory Projections to HC

NA: Attention
5HT: Mood
DA: Reward
ACh: Novelty

acetylcholine
locus coeruleus raphe nuclei substantia nigra medial septum
10/06/15 Computational Models of Neural Systems 37
 Hippocampal EEG
● Awake
– Active exploration:
● theta (8-12 Hz) plus gamma (40 Hz)
– Resting:
● LIA (Large-amplitude Irregular Activity) w/transient sharp waves

● Asleep
– REM sleep: theta
– Slow wave sleep: LIA
– S-SIA: small-amplitude irregular activity (Jarosiewicz & Skaggs)

10/06/15 Computational Models of Neural Systems 38
 Hippocampal EEG

Awake (theta) REM sleep
`

Sharp waves

Slow wave sleep (LIA) REM    SIA

LIA            S­SIA LIA                      REM

10/06/15 Computational Models of Neural Systems 39
 What Does the Hippocampus Do?
● Formation of new episodic memories?
– Anterograde amnesia (H.M. and others)

● Cognitive map?
– Place cells in rats; spatial attention cells in monkeys

● Configural association theory?

– Lesioned rats are impaired on tasks requiring them
to recognize cue configurations

10/06/15 Computational Models of Neural Systems 40
 A Hippocampal Prosthesis?
● Hampson et al (2013) Journal of Neural Engineering, 10(6).

10/06/15 Computational Models of Neural Systems 41
 Delayed Match to Sample Task

10/06/15 Computational Models of Neural Systems 42
 Effects of Stimulation

10/06/15 Computational Models of Neural Systems 43