also block the spatial learning in rodents that lesion 
studies have associated with hippocampal function. 
An additional reason LTP has been so widely ac- 
cepted as a memory mechanism is that there are no 
alternatives. 
Despite the wide acceptance of LTP as a cellular 
substrate for memory, the actual evidence in favor of 
this mechanism is inconclusive. The most direct evi- 
dence comes from the experiments, noted above, in 
which NMDA receptor function is blocked with 
drugs like APV. But APV and similar agents also 
block the slow component of synaptic currents, and 
studies in many brain areas have demonstrated that 
this slow component is often required for computa- 
tions. An alternative explanation of the NMDA- 
blocking results, then, would be that APV prevents 
proper functioning of hippocampal circuits so they 
cannot perform some computation required for 
learning to be established by a mechanism entirely 
independent of LTP. How, then, might LTP be used 
in the hippocampus? Perhaps LTP modifies hippo- 
campal circuits so they can compute more effi- 
ciently. LTP would thus be producing activity- 
dependent circuit modifications like those that 
result in ocular dominance columns in area 17, but 
modifications that are constantly updated. 
On this view, LTP would be used to tune up hip- 
pocampal circuits continually so they could carry 
out some computation associated with forming 
memories, but would not be the storage mechanism 
itself. 
Because blocking NMDA receptors simulta- 
neously prevents the triggering of LTP and interferes 
with synaptic transmission, the only way to test the 
hypothesis that LTP is a memory storage mechanism 
is to find ways of preventing LTP that do not involve 
modifications of synaptic transmission. Dr. Alcino 
Silva, in the laboratory of Dr. Susumu Tonegawa 
(HHMI, Massachusetts Institute of Technology) , has 
produced a mutant mouse that lacks the a subunit of 
calciura/calmodulin kinase type II (CaM kinase 11), 
an enzyme that pharmacological studies have impli- 
cated in some step of the production of LTP. The 
brains of these mutant mice, and their behavior, are 
grossly normal. The Stevens laboratory, in collabora- 
tion with Dr. Tonegawa, has found, however, that 
the mice are deficient in LTP production but that 
their synaptic transmission is intact. Behavioral 
work, done in collaboration by the laboratories of 
Dr. Tonegawa and Dr. Jeanne Wehner, finds that the 
mutant mice are also deficient in spatial learning. 
These studies thus strengthen the hypothesis that 
LTP is a memory mechanism. 
Field potential recordings survey a very large pop- 
ulation of neurons and thus are useful in assessing 
the extent to which some particular manipulation 
interferes with the triggering of LTP. Such field po- 
tential recording revealed, using a standard proce- 
dure that almost always produces LTP in the litter- 
mate controls, that the mutant mice only rarely 
exhibit LTP. 
The advantage of field potential recordings is that 
the properties of very large populations of neurons 
are surveyed. The difficulty with the technique is 
that the triggering of LTP is known to depend on the 
neuron's depolarization, which in turn results from 
the net effect of the synaptic input. If the synaptic 
input is insufficient for any reason at all, LTP would 
be diminished simply because the stimulus did not 
produce an adequate depolarization. A more sensi- 
tive test for LTP, then, is to use whole-cell recording 
and control the neuron's membrane potential by volt- 
age clamp. This method examines one by one (rather 
than revealing population properties) but is very 
sensitive in that stimulation of just one or a few bou- 
tons can produce LTP. The Stevens laboratory thus 
studied neurons from mutant mice and from normal 
littermates; again, they found that LTP is greatly di- 
minished. From both field potentials and whole-cell 
recordings, then, the mutant animals appear to be 
deficient in LTP, but some fraction of the neurons do 
show what appears to be normal, if small, LTP. 
LTP is triggered by the calcium influx through 
NMDA receptor channels. The mutant mice might, 
through either a developmental or regulatory mech- 
anism that depends on the activity of CaM kinase II, 
lack effective NMDA receptor function. To examine 
this question. Dr. Stevens and his colleagues com- 
pared the behavior of NMDA receptor channels in 
hippocampal neurons from mutant and normal ani- 
mals. They find that the amplitude and time course 
of the NMDA component of synaptic currents ap- 
pears not to differ between normal and mutant 
mice. NMDA receptor channels have a special prop- 
erty: the fraction of the time the channel-ligand 
complex spends in a conducting state depends on 
the neuron's voltage. The voltage dependence is 
known, in some circumstances, to be modified by 
phosphorylation. The Stevens laboratory compared 
this voltage dependence in mutant and controlled 
mice and found this property of NMDA receptors to 
be unaffected in the mutant animals. 
In summary, postsynaptic mechanisms are intact 
in the mutant mice, so hippocampal computations 
should be performed normally. Nevertheless, both 
LTP and spatial learning are deficient. These obser- 
vations strengthen the hypothesis that LTP is indeed 
a memory mechanism, but a final interpretation of 
the results must await studies that identify the pre- 
cise mechanism underlying the defects. 
NEUROSCIENCE 44 1 
