Western University Western University
Scholarship@Western Scholarship@Western
Brain and Mind Institute Researchers'
Publications
Brain and Mind Institute
10-1-2013
The touchscreen operant platform for testing learning and The touchscreen operant platform for testing learning and
memory in rats and mice. memory in rats and mice.
Alexa E Horner
Christopher J Heath
Martha Hvoslef-Eide
Brianne A Kent
Chi Hun Kim
See next page for additional authors
Follow this and additional works at: https://ir.lib.uwo.ca/brainpub
Part of the Neurosciences Commons, and the Psychology Commons
Citation of this paper: Citation of this paper:
Horner, Alexa E; Heath, Christopher J; Hvoslef-Eide, Martha; Kent, Brianne A; Kim, Chi Hun; Nilsson, Simon
R O; Alsiö, Johan; Oomen, Charlotte A; Holmes, Andrew; Saksida, Lisa M; and Bussey, Timothy J, "The
touchscreen operant platform for testing learning and memory in rats and mice." (2013).
Brain and Mind
Institute Researchers' Publications
. 99.
https://ir.lib.uwo.ca/brainpub/99
© 2013 Nature America, Inc. All rights reserved.
PROTOCOL
NATURE PROTOCOLS
|
VOL.8 NO.10
|
2013
|
1961
INTRODUCTION
This protocol describes an automated touchscreen platform with
which a remarkable diversity of cognitive functions may be tested
in rodents. During more than 2 decades of research, a number
of tasks have been designed and validated for the platform, each
allowing the researcher to probe a unique set of functions
1–5
.
Together these form a comprehensive battery of tasks, several
of which may be used in concert by the researcher to elucidate
a cognitive profile for a given rodent model; alternatively, they
can be used more selectively to examine specific aspects of the
cognitive repertoire in a hypothesis-driven manner.
The touchscreen platform has been used in a number of stud-
ies, in a variety of ways. First, putative rodent models of human
conditions including Alzheimer’s disease
6,7
, schizophrenia
8–10
,
Huntington’s disease
11
, frontotemporal dementia (A.E.H., B.A.K.,
T.J.B. and L.M.S., unpublished data), aging
12
, exposure to stress
13
and substance abuse
14
have been studied. Notably, we recently
demonstrated the utility of this platform for parallel cognitive
testing of humans with schizophrenia and a putative mouse model
of the disease (discs, large homolog 2 (Dlg2) knockout) sharing a
similar genetic basis
8
. Second, these tasks have been used to inves-
tigate the neural underpinnings of a number of different cogni-
tive functions, targeting the rhinal
15–18
, medial and ventromedial
prefrontal
13,19–21
, anterior and posterior cingulate
22–26
, medial
frontal
22,23
, orbitofrontal
13,27,28
, infralimbic
28
and prelimbic
29
cortices. In the striatum, studies of dorsolateral and dorsome-
dial areas
13,20,21
and the nucleus accumbens
22,25,26,30,31
have been
performed. Roles for a number of other brain regions, including
the amygdala
25,32
, distinct thalamic nuclei
29
, the subthalamic
nucleus
33
, the fornix
17,34
, subiculum
32
, hippocampus
27,35–41
,
pedunculopontine tegmental nucleus
42
, medial septal/vertical
limb of diagonal band (cholinergic neurons)
43
and nucleus basalis
magnocellularis (cholinergic neurons)
44
have also been identified
in a number of tasks. Third, the efficacy of systemic pharma-
cological agents has been studied, using compounds active on
the cholinergic
7,45–48
, dopaminergic
14,31,48–50
, glutamatergic
9,48
and serotonergic
50–53
systems. Fourth, the function of specific
genes
8,51,54,55
, receptors
56
, receptor subunits
57–59
and structural
plasticity processes, such as adult hippocampal neurogenesis
12,60
,
have been assessed.
Advantages and disadvantages of the touchscreen platform
The advantages of the touchscreen platform have been discussed
in detail elsewhere
1–3
. Briefly, this platform offers the potential for
a high degree of standardization, minimal experimenter involve-
ment and high translational potential (e.g., similarity to human
CANTAB (Cambridge neuropsychological test automated battery)
tests). It includes assays of various neuropsychological constructs,
including attention and cognitive flexibility, and it uses appeti-
tive rather than aversive motivation. One obvious advantage of
using computer-generated visual stimuli is that the perceptual
features (size, shape, contrast, luminance and so on) and similari-
ties of the stimuli can be easily manipulated
3,61
. Furthermore, in
object-based tasks in which the objects are displayed in different
locations on the touchscreen, there is no potential for the use
of odor cues, unlike some (dry) maze tasks, which can modify
results. The platform also lends itself to applications that allow
for the measurement of brain functions in vivo as animals per-
form a task (for example, via single-unit neuronal recordings
62
).
There is potential for the incorporation of other powerful metho-
dologies (e.g., optogenetics) into the touchscreen platform.
Although we focus on rodents in this article, touchscreens have
been used with pigeons and nonhuman primates, as well as with
mice and rats
2,63–68
.
It is worth noting that although automated methods such as the
touchscreen platform reduce experimenter effort, the tasks can
take many more sessions to run than equivalent tests using, e.g.,
odors. However, because tests on large numbers of animals (>20)
can be run in parallel, experiments can often be completed in the
same number of days (or fewer) as they can with ‘hand-testing’
methods in which an experimenter tests one animal at a time.
The touchscreen operant platform for testing
learning and memory in rats and mice
Alexa E Horner
1–3
, Christopher J Heath
2,3
, Martha Hvoslef-Eide
2,3
, Brianne A Kent
2,3
, Chi Hun Kim
2,3
,
Simon R O Nilsson
2,3
, Johan Alsiö
2,3
, Charlotte A Oomen
2,3
, Andrew Holmes
4
, Lisa M Saksida
2,3
& Timothy J Bussey
2,3
1
Synome Ltd., Babraham Research Campus, Cambridge, UK.
2
Department of Psychology, University of Cambridge, Cambridge, UK.
3
Behavioural and Clinical
Neuroscience Institute, University of Cambridge, Cambridge, UK.
4
Laboratory of Behavioral and Genomic Neuroscience, National Institute on Alcohol Abuse and
Alcoholism (NIAAA), US National Institutes of Health, Bethesda, Maryland, USA. Correspondence should be addressed to A.E.H. (alexa.horner@cantab.net).
Published online 19 September 2013; doi:10.1038/nprot.2013.122
An increasingly popular method of assessing cognitive functions in rodents is the automated touchscreen platform, on which a
number of different cognitive tests can be run in a manner very similar to touchscreen methods currently used to test human
subjects. This methodology is low stress (using appetitive rather than aversive reinforcement), has high translational potential
and lends itself to a high degree of standardization and throughput. Applications include the study of cognition in rodent models
of psychiatric and neurodegenerative diseases (e.g., Alzheimer’s disease, schizophrenia, Huntington’s disease, frontotemporal
dementia), as well as the characterization of the role of select brain regions, neurotransmitter systems and genes in rodents.
This protocol describes how to perform four touchscreen assays of learning and memory: visual discrimination, object-location
paired-associates learning, visuomotor conditional learning and autoshaping. It is accompanied by two further protocols (also
published in this issue) that use the touchscreen platform to assess executive function, working memory and pattern separation.
© 2013 Nature America, Inc. All rights reserved.
PROTOCOL
1962
|
VOL.8 NO.10
|
2013
|
NATURE PROTOCOLS
Furthermore, although the hand tester is working on one experi-
ment all day, the experimenter with a number of automated units
can work on several experiments. Of course, to achieve this high
throughput, one needs the apparatus, which means a larger initial
financial outlay than is required for most hand-testing methods.
Again, however, if one considers all factors, such as salaries and
person-hours spent on experiments, and the fact that such appa-
ratus can be used for many years before needing to be replaced,
in the long run automation may actually be less expensive than
hand-testing alternatives.
Another potential limitation is that the use of visual stimuli
precludes the use of certain subjects, such as mice with genetic
alterations that cause rapid retinal degeneration. (Albino rats,
however, seem to have sufficient acuity to perform as well in the
touchscreen as pigmented rats
3
.) In addition, as with most appe-
titive, operant paradigms, the use of food reward may introduce
possible problems; for example, an experimental treatment may
affect appetite or interact with the physiological effects of food
restriction. These limitations should be kept in mind, although we
do believe that all things considered, the advantages conferred by
avoiding aversive stimuli far outweigh the disadvantages of using
appetitive stimuli. Touchscreen tasks require intact motoric func-
tion such that subjects are able to traverse the testing chamber,
respond to the screen and collect and consume the food reward.
Again, however, these demands are much lower than many cur-
rently used behavioral paradigms. Importantly, the impact of
most of these potential changes can be assessed by taking a battery
approach, by running appropriate control experiments and/or
inspecting relevant dependent variables such as trial omissions
and/or reaction times to respond or to collect the reward. If one
takes a battery approach, testing the effect of a given experimental
manipulation on several tasks, then the tasks can act as mutual
controls by virtue of the fact that they involve the same types of
apparatus, stimuli, responses and reinforcement
1
; comparisons
can be made confidently between tasks in the battery because such
variations are minimal. For example, if an animal performs poorly
in object-location paired-associates learning (which theoretically
requires cognitive functions including visual discrimination and
learning of object-location associations; discussed further below),
but well in visual discrimination (which requires visual discrimi-
nation learning; discussed further below), it would be reasonable
to conclude that the former impairment is not due to a general
problem in perceptually discriminating images. Similarly, we have
found that muscarinic M2 receptor–knockout mice are impaired
in object-location paired-associates learning, but they actually
demonstrate improved attention in the five-choice serial reac-
tion time (5-CSRT) task (Romberg, C. et al., unpublished data)
making it very unlikely that the former impairment is due to an
attentional deficit.
Finally, for researchers for whom the ethological validity of
a method is important, rodents using touchscreens may not be
the method of choice. However, we note that the behavior in
the touchscreen is built on the natural tendency of rodents to
approach and explore novelty in the environment; the explora-
tion is detected by the touchscreen, and the animal learns, again
quite naturally, the consequences of exploring certain stimuli.
In this sense, the method is no less ethologically valid than hav-
ing rodents swim in an artificial pool in a laboratory setting, or
other commonly used laboratory methods. In any case, we see
the touchscreen method as complementing rather than replacing
other methods such as foraging paradigms.
Assessing learning and memory
This protocol describes four tasks that may be used to assess
aspects of learning and memory. The first three of these rely pri-
marily on appetitively motivated instrumental learning, and are
preceded by ‘pretraining’, in which subjects must learn to make
instrumental responses in the touchscreen apparatus. Visual dis-
crimination is a relatively simple task, in which subjects must
learn to consistently respond to one of two visual stimuli. In
object-location paired-associates learning, the correct stimu-
lus is identified by the conjunction of a visual stimulus and its
location on the touchscreen. In visuomotor conditional learning
(VMCL), the correct response (left or right) depends on which
conditional visual stimulus is presented. Autoshaping is unique in
the battery, primarily testing Pavlovian stimulus-reinforcer learn-
ing. Two accompanying protocols discuss additional tasks that
may be used to assess working memory and pattern separation
4
(trial-unique non–matching-to-location (TUNL) and location
discrimination (LD)) and executive function
5
(reversal, extinc-
tion and the 5-CSRT task). Other tasks that will further expand
the range of the battery are constantly in development.
Visual discrimination (Step 10A)
Learning to discriminate between environmental stimuli is essen-
tial in order to successfully shape decisions and adaptively guide
behavior. Understanding the neural mechanisms supporting dis-
crimination learning is of major interest to cognitive neuroscience,
and it may have implications for delineating the pathophysiology
of cognitive impairments in neuropsychiatric disorders such as
schizophrenia and Parkinson’s disease. Basic preclinical research in
animals is key to this work, and various methods for testing discrim-
ination learning have been developed, including touchscreen-based
systems in nonhuman primates
69
. In addition to the basic pairwise
discrimination procedure, certain variations have also been devel-
oped, including multidimensional
70
(to test attentional set-shifting),
concurrent
23
and conditional (see VMCL) discriminations, as well
as transverse patterning
34
(to test configural learning).
Initial studies using a touchscreen discrimination procedure
were published almost 20 years ago; they used a configuration
that included a monitor, off-the-shelf operant hardware and
customized software
2
(see also ref. 71). Briefly, the procedure
entails simultaneous presentation of two stimuli, and the meas-
urement, over multitrial sessions, of the animal’s ability to reliably
touch the stimulus designated the conditioned stimulus (CS)
+
(rewarded) in favor of the other stimulus (CS
−
, nonrewarded).
Discrimination learning requires at least two processes: learning
to perceptually discriminate the stimuli, and learning which of
the two stimuli is associated with reward. It also provides the
basis for testing reversal learning
5
, in which the stimulus-reward
contingencies acquired during discrimination are reversed.
The task has been used to investigate a variety of questions con-
cerning the neural basis and pharmacological modulation of visual
discrimination learning. These include testing the effects of drug
treatments including psychotomimetics and putative cognitive
enhancers
9,13,18,47
; gene mutations, particularly of glutamate sig-
naling molecules
8,10,51,55,57–59
; discrete brain lesions
13,23,28,72–74
;
and environmental manipulations such as exposure to stress
13
.
© 2013 Nature America, Inc. All rights reserved.
PROTOCOL
NATURE PROTOCOLS
|
VOL.8 NO.10
|
2013
|
1963
Object-location paired-associates learning (Step 10B)
The formation of an association between two individually neu-
tral stimuli, named paired-associate learning (PAL), has been
extensively studied in humans using a variety of modalities
(verbal, visual, locations). Although PAL has traditionally been
assessed using pairings of words tested by cued recall, the human
CANTAB PAL task
75
does not rely on verbal stimuli and thus
provides a version of PAL that is more amenable to modeling
in animals. The computerized PAL task requires the subject to
form an association between a visual stimulus and its location
on a screen, demonstrated under cued-recall conditions. Over
more than 2 decades, CANTAB PAL has been validated as sensi-
tive to detecting deficits in a range of conditions such as schizo-
phrenia
76–78
, Huntington’s disease
79
, Parkinson’s disease
75
, major
depressive disorder
80
, unipolar and bipolar mood disorders
81
and
Alzheimer’s disease
75,79,82–86
.
Given the profile of neuropsychiatric disorders to which
object-location learning is sensitive, it is not surprising to find
that encoding and retrieval of object-location associations has
been linked to hippocampal and prefrontal cortical function
87–89
.
Importantly, the same areas have been implicated in the rodent
touchscreen object-location paired-associates learning task devel-
oped by Talpos and colleagues
41,90
, in which the animal is required
to learn three individual object-location associations. Each visual
stimulus (object) is correct in a unique location, which stays sta-
ble throughout training. On each trial, two different objects are
presented, one in its correct location and the other in an incorrect
location. The third location remains blank. The rodent task differs
from that of CANTAB PAL in that the stimuli are not trial unique,
and the task does not feature a delay. Importantly, however, the
requirement to use both object and location information to solve
the task is maintained. Indeed, assessment of paired-associates
learning using CANTAB PAL in people with DLG2 mutations
produced a similar phenotypic profile to that observed by using
the rodent object-location paired-associates learning task
8
with
Dlg2-knockout mice, indicating the translational potential of the
paradigm. We note that, in this task, the animal can approach
locations on the screen from many different angles, which is in
contrast to the behavior that we see in, e.g., the VMCL task.
Pharmacological manipulation of the rodent object-location
paired-associates learning task indicates that both facilitation and
disruption of performance is possible. Antagonism of NMDA or
AMPA receptors in the hippocampus impairs performance in rats,
but leaves accuracy unaltered for a similar control task, which
may be solved by visuomotor conditional learning as opposed
to the formation of object-location associations
41
. Systemic
pharmacological manipulations in mice have further implicated
cholinergic muscarinic receptors in performance of the task, with
a facilitation observed in wild-type animals using donepezil
45
.
Knockout of muscarinic M2 but not muscarinic M1 receptors
impairs acquisition of the task
56
(Romberg, C. et al., unpublished
results). Task performance is sensitive to amphetamine but not
to PCP, ketamine or LSD
91
. Thus, the task offers an automated
and sensitive measure of rodent object-location paired-associates
learning and performance, which has translational potential.
VMCL (Step 10C)
In VMCL, animals learn a conditional rule of the type ‘If vis-
ual stimulus A is presented, make motor response X; if visual
stimulus B is presented, make motor response Y’. There has been
considerable interest in such visuomotor mapping in primates
92
.
Generally, it appears that across monkeys and rodents, hippocam-
pal damage does not consistently produce impairments in such
tasks, although the hippocampal system can become involved
when mappings are acquired rapidly or involve object-location
rather than visuomotor associations
93,94
. Rodent VMCL in oper-
ant chambers requires discrete left-right responses and thus prob-
ably involves visuomotor associations, which are likely to require
stimulus-response habit learning; as would therefore be expected,
the task is more sensitive to damage in the striatum than in the
hippocampus
95,96
. The VMCL task in the touchscreen is indeed
designed to maximize stimulus-response learning and minimize
other cognitive demands. Thus, the discrimination is chosen to be
an easy one (in practice probably solved via light-dark discrimi-
nation), to reduce perceptual demands. Furthermore, a ‘limited
hold’ (time limit) for responding promotes the same rapid head-
turn-and-nose-poke motor response on each trial, encouraging
a visuomotor strategy and limiting the extent to which subjects
can move away from the screen and reapproach the choice stimuli
from different angles, which might promote alternative learning
strategies. Touchscreen VMCL does not require medial prefrontal
cortex, anterior cingulate cortex, hippocampus, perirhinal cortex,
anterior thalamus or mediodorsal thalamus, but it does depend
on posterior cingulate cortex (late in learning only)
23,74,97
, thus
conferring the specificity needed to dissociate function as part of
a touchscreen test battery
23
. As in pairwise visual discrimination
learning, the task can also be reversed to engage a different set
of brain regions
22,43,74
. The VMCL task may be particularly rel-
evant to Huntington’s and Parkinson’s disease, in which cognitive
impairments include deficits in habit learning
98,99
.
Autoshaping (Step 10D)
The autoshaping task assesses Pavlovian approach learning.
It capitalizes on the process of ‘autoshaping’, which was first
observed in experiments in which pigeons came to reliably peck
at an illuminated key (CS) presented immediately before delivery
of grain at a separate location
100
, and has been reported in many
species
8,101–106
. It is considered to rely on Pavlovian, as opposed
to instrumental, associations
22
. Although a behavioral chamber
equipped with levers can be used to assess rodent autoshaping
107
,
this protocol details the use of a touchscreen system as originally
described by Bussey et al.
22
.
Autoshaping is a discriminative conditioning procedure, in
which a stimulus is presented on either the left or right side of
the touchscreen, with one side defined as CS
+
(rewarded CS) and
the other as CS
−
(nonrewarded CS). Reward is delivered upon ter-
mination of the CS
+
but not the CS
−
. With repeated presentations,
rodents increase CS
+
approaches and decrease CS
−
approaches,
indicating that the predictive relationship between CS
+
presen-
tation and reward delivery has been learned
22
. To demonstrate
the Pavlovian nature of the association, a reward omission
procedure
22,108
can be implemented, in which CS
+
approaches
cause reward to be withheld. Under this altered contingency, ani-
mals continue to respond to the CS
+
, which is consistent with a
Pavlovian CS-UR association
22,108
.
This task requires minimal pretraining and animals quickly
develop the necessary stimulus discrimination, making it rela-
tively rapid to complete. Therefore, it has been used extensively
