(C) 1999 Dr Roger J Morris, UK

The probe test: is it always a measure of spatial memory for transgenic mice?

Our aim, in the watermaze, is to test the capacity of rodents to learn, and
retrieve, spatially-encoded information; instead we are restricted to
asking the mice to solve a problem for which the ability to learn and
remember a spatially-cued task will significantly enhance performance.

But is the capability for spatial learning the main variable in this test?
Experience suggests that two problems in particular are worth considering,
especially when dealing with transgenic mice.

The first is the ability of the animal to concentrate upon the task in
hand. If the mice are too nervous, or too weak to be at their ease
swimming, their performance is compromised independently of spatial
learning (both effects have misled studies of spatial learning in
transgenic KO mice). It is important that mice be tested in appropriate
assays to determine whether they differ significantly from the control
group in levels of anxiety, willingness to explore new territory, and
general locomotory activity. The latter is monitored in the watermaze as
swimming speed, the former given by such assays as the elevated plus maze,
holeboard and open field. Our experience suggests that problems in the
watermaze such as thigmotaxis and passive floating are minimised if mice
first get accustomed to being removed from their home cage to solve
relatively unthreatening problems, before being placed in a large tank of
water with no visible means of escape. Thus both practical and theoretical
considerations suggest testing the mice for exploration/anxiety prior to
the watermaze, and the results may advise against using the watermaze as a
test of spatial memory.

The second concerns the strategy used by the mice to locate the platform.
Have they solved the problem purely by reference to the spatial cues, or
has the physical process of "swimming in the general area until bumping
into a submerged object" been a significant part of the strategy? In the
initial phase of the watermaze in which rodents search for the hidden
platform, in each trial the rodent has only to bump into the platform once
to locate it and escape; is this single encounter the result of remembering
how to triangulate the location from external cues, or just good luck?

Of the methods devised to eliminate a 'lucky bump' from the analysis, the
probe test is considerably the least expensive and most convenient, and
works very well for rats. A rat, knowing that the platform was in a
certain position, will trawl repeatedly over that position looking for it.
The rat is thus indicating that it knows the position independently of such
tactile cues as hitting the platform, and this is real knowledge rather
than a chance passage through the target area. This simple test is
conducted when the animals have learnt the straight forward task of
locating the hidden platform; i.e. when their elapsed time to find the
platform has dropped from an initial maximum (e.g. 90 sec for the trial) to
a base level where no significant improvement is occurring each day
(generally around 10-25 sec, depending on the size of the pool/platform).

This test, however, depends as much upon the rodent's strategy in solving a
problem as do the initial searches for the hidden platform; but in the
probe test the problem is one of spatial memory if, and only if, the rodent
does not confront the fact that the platform is no longer there. If the
rodent realises that the platform really is not there, that the hitherto
reliable spatial analysis will not enable it to escape the water, then its
subsequent behaviour cannot be interpreted as driven by spatial analysis.
It will resort to alternative survival-dependent problem solving. If the
mutants differ from the controls in levels of anxiety, or in other less
readily identifiable variables (depression, quick-wittedness, etc), their
response to finding that the platform has been withdrawn will almost
certainly vary from that of the controls.

We find that mice tend to be much more adaptable than rats upon finding
that the platform has gone. Our mice swim immediately to the former
position of the platform. A significant number then swim back to where
they were put into the pool (which is randomly varied trial by trial, but
they relocate this point accurately) and swim back to the former position
of the platform. Some even repeat this once more. Most spend some time
swimming around in the central 70% of the pool trying to locate the
platform (and will again cross the former position). At some stage, many
head for the side of the pool from which they will be retrieved at the end
of the trial, where some float. After 90 sec, they escape the pool -
confirming the non-spatial experience of the previous trials.

Since most mice return more than once to the former platform position, the
groups under test will often spend significantly more time in the former
quadrant, and pass over the former position of the platform, than they will
in control quadrants/counters. The probe test thus works in mice as well
as rats. Where a mutant group of mice differ from the control group in this
test, I suspect that in some cases the source is not a defect in spatial
memory, but differences in survival strategy by rodents that realise
spatial memory will not help them escape from the water. The frequency with
which defects are only found in the probe test, but not in the initial
searching, concerns me since in very few cases are potentially confounding
properties such as anxiety investigated.

The adaptability of the mice to variation of the test can be assessed if
one reverses the position of the platform. On the day following the probe
test, return the platform to the pool, but in a new position. Start the
mouse off by placing it for 15 sec on the platform in its new position, so
it knows there is again a platform to find, but (if it learns quickly) in a
new position. Then run a series of trials for the mouse to find the
platform in its new position, comparing not only time/path, but passages
through counters and time in quadrants as simple indicators of strategy.
We have found that mice of one knockout strain, on the first day of
platform reversal, return immediately to the former site of the platform,
and then begin a circular search pattern for the new location (which is
usually successful within 90 sec); their wild type littermates tend to
restrict their search mainly to the former position of the platform, and
only after two futile trials do they begin to search more widely to find
the platform. By trials 3/4, both groups are equally successful in finding
the platform in its new position. The genotype-specific difference seen on
day 1 of reversal is, I suggest, one of survival strategy, not spatial
learning.

So, in summary, how the rodent solves any of the problems confronting it in
the watermaze depends not just on its capacity for spatial learning, but
also its strategy for solving problems; as these problems are varied
(probe vs hidden platform), it may be a difference in strategy, rather than
in learning or memory, that dominates the response of knockout, compared to
wild-type, mice. I must also state that this is not a common view in the
field - the probe test, after all, has proved to be one of the best ways
for producing a significant difference in behaviour between control and
mutant groups, and the temptation to attribute this difference to defects
in spatial learning is rarely resisted.

It is my opinion that, given the unknown effect of most mutations upon the
emotional, sensory and cognitive functions of the mice under test, the
prime test of spatial memory is given by the initial search for the hidden
platform, provided this is set up appropriately. That is, the initial
search must be set up so that the animal realises early on that the problem
is difficult but solvable (avoiding problems of thigmotaxis and floating);
the learning curve must show progressive improvements in performance over
many trials (thus cancelling out the "lucky bump"); and the mice should
not be overtrained. And above all, the mice must be physically and
emotionally able to perform the task.