Computational Neurobiology of the Hippocampus 
Terrence J. Sejnowski, Ph.D. — Investigator 
Dr. Sejnowski is also Professor at the Salk Institute and Professor of Biology and Neuroscience at the Uni- 
versity of California, San Diego. He received his B.S. degree in physics from Case Western University and 
his M.A. and Ph.D. degrees in physics from Princeton University. He was a postdoctoral fellow with Alan 
Gelperin in the Biology Department at Princeton and with Stephen Kuffler at the Harvard Medical School, 
where he studied mechanisms of synaptic transmission. Dr. Sejnowski was a member of the faculty of the 
Biophysics Department at the Johns Hopkins University before moving to San Diego. He was chosen to 
give this year's Messenger Lectures at Cornell University. 
ALTHOUGH there has been an explosion of 
discoveries over the last several decades 
concerning the structure of the brain at the cellu- 
lar and molecular levels, v^^e do not yet under- 
stand how the nervous system enables us to see 
and hear, to learn skills and remember events, to 
plan actions and make decisions. The long-range 
goal of this laboratory is to explain how neural 
systems effect these complex behaviors. Our gen- 
eral approach is to use what we know about the 
structure and function of identified neurons and 
neuronal networks to construct computational 
models at several levels of investigation. 
At the biophysical level, the computational 
mechanisms are based on chemical and electrical 
signals within and between neurons. Among the 
more important of these are the signals at syn- 
apses, which carry information between neurons. 
On cortical pyramidal neurons, most of the syn- 
apses occur on thorn-like structures protruding 
from the dendritic shaft, called dendritic spines. 
The function of these spines is unknown, but is 
likely to be important because they are found on 
neurons in most species. 
We have recently completed a study of the ef- 
fectiveness of inhibitory synapses on dendritic 
spines. About 10 percent of the spines on cortical 
pyramidal neurons have both excitatory and in- 
hibitory inputs. It had been thought that this ar- 
rangement allowed the inhibitory synapse to 
"veto" the excitatory one when they were simul- 
taneously activated. In our computer simulations 
of these events, however, we have found that the 
inhibitory synapses were only effective at reduc- 
ing the magnitude of the excitatory postsyn- 
aptic potential (EPSP) under very restricted 
circumstances. 
The reason is that the volume of a spine is so 
small that even tiny conductance changes in the 
membrane allow enough ions to enter and leave 
the spine to change the intracellular concentra- 
tions significantly. This would tend to reduce the 
forces driving these ions across the membrane 
and thus reduce the dependent synaptic currents. 
For our simulations, we infer that, to be effec- 
tive, the inhibitory synapses found on cortical 
spines must be mediated by through GABAg 
receptors. 
At the neuronal level, we are exploring with 
computational models the effects of mechanisms 
that couple the synaptic signals in the dendrites 
to the spike-triggering region of the neuron, 
which is located near the soma. According to the 
traditional view of dendritic processing, current 
injected into dendrites from synaptic activity is 
passively conveyed to the cell body by the den- 
drites' cable properties. In recent years voltage- 
dependent calcium currents have been identified 
in dendrites that could boost the coupling be- 
tween synaptic activity and spike generation. 
This is particularly interesting because an in- 
crease in the excitability of the neuron often ac- 
companies the long-term potentiation (LTP) of 
excitatory synapses, a rapid and persistent eleva- 
tion of the EPSP. 
Could the increase in excitability seen during 
LTP result from a change in voltage-dependent 
calcium currents? To test this hypothesis, we 
used simulations of intradendritic LTP experi- 
ments. The shape of our simulated neuron was 
based on the detailed branching pattern of a hip- 
pocampal neuron that had been injected with a 
dye and reconstructed in a computer. We found 
that a small addition of voltage-sensitive calcium 
channels to the dendritic membrane made a previ- 
ously subthreshold input suprathreshold, with no 
significant change in the EPSP. 
Our most recent work concerns the effect of 
these changes in dendritic excitability on LTP 
specificity. We simulated two sets of synaptic in- 
puts, one stimulated and the other control. Po- 
tentiation at the stimulated input had little effect 
on the control input if the latter was electrically 
closer to the cell body, or if the two inputs 
were segregated on different primary dendritic 
branches. Specificity is degraded if the control 
input is farther away than the stimulated. These 
results show that modulation of dendritic excit- 
ability is a plausible mechanism for EPSP-spike 
potentiation. 
We are testing this computational model of 
399 
