Automated, Quantitative Cognitive/Behavioral Screening of Mice: For Genetics, Pharmacology, Animal Cognition and Undergraduate Instruction

Summary

Fully automated system for measuring physiologically meaningful properties of the mechanisms mediating spatial localization, temporal localization, duration, rate and probability estimation, risk assessment, impulsivity, and the accuracy and precision of memory, in order to assess the effects of genetic and pharmacological manipulations on foundational mechanisms of cognition in mice.

Abstract

We describe a high-throughput, high-volume, fully automated, live-in 24/7 behavioral testing system for assessing the effects of genetic and pharmacological manipulations on basic mechanisms of cognition and learning in mice. A standard polypropylene mouse housing tub is connected through an acrylic tube to a standard commercial mouse test box. The test box has 3 hoppers, 2 of which are connected to pellet feeders. All are internally illuminable with an LED and monitored for head entries by infrared (IR) beams. Mice live in the environment, which eliminates handling during screening. They obtain their food during two or more daily feeding periods by performing in operant (instrumental) and Pavlovian (classical) protocols, for which we have written protocol-control software and quasi-real-time data analysis and graphing software. The data analysis and graphing routines are written in a MATLAB-based language created to simplify greatly the analysis of large time-stamped behavioral and physiological event records and to preserve a full data trail from raw data through all intermediate analyses to the published graphs and statistics within a single data structure. The data-analysis code harvests the data several times a day and subjects it to statistical and graphical analyses, which are automatically stored in the "cloud" and on in-lab computers. Thus, the progress of individual mice is visualized and quantified daily. The data-analysis code talks to the protocol-control code, permitting the automated advance from protocol to protocol of individual subjects. The behavioral protocols implemented are matching, autoshaping, timed hopper-switching, risk assessment in timed hopper-switching, impulsivity measurement, and the circadian anticipation of food availability. Open-source protocol-control and data-analysis code makes the addition of new protocols simple. Eight test environments fit in a 48 in x 24 in x 78 in cabinet; two such cabinets (16 environments) may be controlled by one computer.

Introduction

To bring the powerful techniques of genetics, molecular genetics, molecular biology, and neuropharmacology to bear on elucidating the cellular and molecular mechanisms that mediate basic mechanisms of cognition, we need high-volume, high-through-put psychophysical screening methods that quantify physiologically meaningful properties of cognitive mechanisms. A psychophysically measurable, physiological meaningful quantitative property of a mechanism is a property that can be measured by behavioral means and also by electrophysiological or biochemical means. Examples are the absorption spectrum of rhodopsin, the free-running period of the circadian clock, and the refractory period of reward axons in the medial forebrain bundle^1,2. Psychophysical measurements that can be compared to cellular and molecular measurements lay a foundation for linking cellular and molecular mechanisms to psychological mechanisms through quantitative correspondence. For example, the fact that the in situ absorption spectrum of the rhodopsin in the outer segments of rods superimposes on the human scotopic spectral sensitivity function is strong evidence that the photon-triggered isomerization of rhodopsin is the first step in scotopic vision. The quantitative aspects of complex behavior patterns are also central to the use of QTL methods in behavioral genetics^3,4.

The performance of mice (and rats) on well-established instrumental and Pavlovian learning protocols depends on brain mechanisms that measure abstract quantities like time, number, duration, rate, probability, risk, and spatial location. For example, the speed of acquisition of Pavlovian conditioned responses depends on the ratio between the average interval between the reinforcing events (typically, food deliveries) and the average latency to reinforcement following the onset of the signal for impending reinforcement^5-7. For a second example, the ratio of the average duration of the visits to two feeding hoppers in a matching protocol approximately equals the ratio of the rates of reinforcement at those two hoppers^8-10.

The behavioral testing methods currently in wide use by neuroscientists interested in underlying mechanisms are, for the most part, low volume, low through-put, and labor intensive²⁶. Moreover, they do not measure quantities that can be compared with quantities measured by electrophysiological and biochemical methods, as, for example, the behaviorally measured periods and phases of circadian oscillators may be compared to electrophysiological and biochemical measures of circadian period and phase. Current behavioral testing methods focus on categories of learning, such as spatial learning, temporal learning, or fear learning, rather than on underlying mechanisms. The widely used water maze test of spatial learning^11-15 is an example of these shortcomings. Spatial learning is a category. Learning in that category depends on many mechanisms, one of which is the mechanism of dead reckoning^16,17. Dead reckoning depends in turn on the odometer, the mechanism that measures distance run¹⁸. Similarly, temporal learning is a category. A circadian clock is among the mechanisms on which learning in that category depends, because an oscillator with an approximately 24 hr period is required for animals to learn the time of day at which events occur^17,19. The clock that enables food anticipation has yet to be discovered¹⁹.

A clock is a time-measuring mechanism. Endogenous oscillators with a wide range of periods allow the brain to locate events in time by recording the phases of those clocks^16,17. The ability to record locations in time enables the measurement of durations, that is, distances between locations in time. Associative learning depends on the brain's measurements of durations^5,6,20,21. Counters are number-measuring mechanisms. Number measuring enables probability estimation, because a probability is the proportion between the numerosity of a subset and the numerosity of the superset. Number measuring and duration measuring enable rate estimation, because a rate is the number of events divided by the duration of the interval over which that number was measured. Measurements of duration, number, rate, and probability enable behavioral adjustments to changing risks.^22,23 Our method focuses on measuring the accuracy and precision of these foundational mechanisms. Accuracy is the extent to which the brain's measure corresponds to an objective measure. Precision is the variation or uncertainty in the brain's measure of a fixed objective value, for example, a fixed duration. Weber's Law is the oldest and most securely established result in psychophysics. It asserts that the precision of the brain's measure of a quantity is a fixed fraction of that quantity. The Weber Fraction, which is the statistician's coefficient of variation in a distribution (σ/μ), measures precision. The ratio of the psychophysical mean (e.g. mean judged duration) to the objective mean (mean objective duration) is the measure of accuracy.

The method presented here maximizes volume (number of animals being screened at any one time in a given amount of lab space) and throughput (amount of information obtained divided by the average duration of the screening of a single animal) while minimizing the amount of human labor required to make the measurements and maximizing the immediacy with which the results of the screening become known.

The data-analysis software architecture presented here automatically puts the raw data and all the summary results and statistics derived from the data together in a single data structure, with field headings that render intelligible the vast seas of numbers therein contained. The analytic software only operates on data in that structure, and always stores the results of its operations in fields within that same structure. This insures an intact trail from raw data to published summaries and graphs.

The software automatically writes into the structure the experiment-control programs that governed the fully automated testing, and it automatically indicates which raw data came from which program. Thus, it preserves an impeccable data trail, with no doubt as to which experimental conditions were in force for each animal at each point in the testing and no doubt about how the summary statistics were derived from the raw data. This method of data preservation greatly facilitates the development of standardized behavioral screening data bases, making it possible for other laboratories to further analyze these rich data sets.

This method minimizes the risk of loss of support for the firmware and software on which it depends. The testing apparatus is trivially modified from a long-established commercial source. The programming languages are the custom language provided by the hardware manufacturer, for protocol control, and, for data analysis and graphing, a purpose-built, noncommercial, open-source toolbox (TSsystem) written in a very widely supported commercial scientific programming, data analysis and graphing language. The toolbox contains high-level commands for extracting structural information and summary statistics from lengthy time-stamped event records. The protocol-implementing programs and the data-analyzing programs are open source and thoroughly documented.

The screening system is schematized in Figure 1. Ten cabinets, each containing 8 test environments may be set up in a 10 ft x 15 ft laboratory room, enabling 80 mice to be run at one time. Cables passing through a port in a party wall should connect the environments to the electronic/electrical interface cards and PCs in another room. The PCs run the protocol-control programs. One computer is required for every 2 cabinets (16 test environments). The PCs must be connected via a Local Area Network to a server running the data-analysis and graphing software.

Protocol

The three fully automated protocols in the TSsystem (matching, appetitive instrumental and classical conditioning) and the switch protocol have been approved by the Animal Care and Facilities Committee at Rutgers New Brunswick.

1. Setting Up the Physical System

1. Set up the test environments in the cabinets (see Figure 1).
2. Install the experiment-control software provided with the test environments on the protocol-control computers.
   Note: Do not use these computers for any other purpose!

2. Setting Up the Software System

1. Set up the LAN (local area network) so that the server on which data-analysis software is installed can access the hard disks of the computer(s) controlling the test environments (see Figure 1).
2. Establish a file-synchronizing account for data storage in the "cloud".
3. Put the TSsystem folder and its subfolders on the search path of the commercial programming language in a cloud-synchronized folder.
   Note: The TSsystem is a software toolbox, that is, a library of over 30 high-level functions that facilitate the creation of complex data-analysis and data-graphing code that automatically processes the data whenever it is harvested from the output files generated by the experiment-control program. All of the commands operate on data in fields of the Experiment structure and put the results in other fields in the same structure (see Figure 2). These open-source commands are written in one of the most widely used commercial scientific programming and graphing languages. It has many other "toolboxes", including most usefully a statistics toolbox.

3. Starting an Experiment

1. Call TSbegin (see Figure 3).
   Note: TSbegin is an interactive GUI (Graphic User Interface) in the TSsystem toolbox. It leads the user through the process of creating the hierarchical data structure into which the raw data and all results derived from it will be placed by the other functions in the TSsystem toolbox.
2. Call TSaddprotocol (see Figure 4).
   Note: TSaddprotocol is a GUI in the TSsystem toolbox. It leads the user through the process of specifying control parameters for an experimental protocol, specifying decision code that will automate the decision to terminate the protocol and go on to the next one, and specifying the decision criteria to be used.
3. Place mice in the 24/7 live-in test environments, one mouse per environment.
   Notes: Take care to note the ID number of the mouse that goes into each of the numbered experimental environments (Box 1, Box 2, etc.). Also, note the letter that identifies the experiment-control computer on the Local Area Network (LAN) and its IP address.
4. Call TSstartsession (Figure 5).
   Note: TSstartsession is a GUI in the TSsystem toolbox . It leads the user through the process of starting an experimental session. Experimental sessions last one or two weeks, during which several different behavioral testing protocols are run. TSstartsession stores the information that goes into the macro that the protocol-control software reads when a session is started. Included is the path to and name of the code file that the protocol-control software reads. TSsystem's analytic software reads this code into the hierarchical data structure, so there is never doubt as to the exact protocol in force at any time.
5. Go to the control computers and call the macros written to the MedPC folder, in order to start the session for the boxes controlled by that computer.

4. Data Analysis

1. If you have created a new protocol, write appropriate data-analysis and graphing code using the commands in the TSsystem toolbox, which greatly simplify the creation of complex data analyses.
   Note: Data-analysis and graphing code for the three protocols whose results are described below are included in the TSsystem toolbox. Because, they are open source, they may be modified at will. The code for these analyses is extensively commented, which makes it easier to create code for analyzing the results from user-specified protocols.
2. For the duration of the experiment (24 hr to many weeks), monitor email for alerts from the server indicating possible equipment malfunctions (power failures, spontaneous, control-computer reboots, pellet feeder malfunctions, etc.), which the TSsystem data-analyzing program detects.
3. Study the plots of performance that the data-analyzing code written in TSsystem produces every time it is called by the analysis timer (typically 2-4 times/day).
   Note: The analysis timer calls the data-analysis and graphing program at user-specified intervals. The called program is written with functions in TSsystem. It reads into the hierarchical data structure the raw data harvested from the file to which the protocol-control software writes. Then, it analyzes the data and graphs the results of the analyses. The file containing the hierarchical data structure is stored in a file-synchronization folder in the cloud. This provides automatic off-site backup. The automatic file-synchronization stores copies of the structure file on the computers of all the personnel and collaborators who have been granted access. Specified graphs are automatically emailed to specified personnel and collaborators. A principal investigator can monitor the progress of the testing from anywhere in the world at any time, and, if necessary, revise the experimental protocol, on line, remote from the site where the mice are being tested.
4. Use TSbrowser to study the data and summary statistics in the hierarchical data structure as they become available, in quasi real time (see Figure 2).

Results

The system can and should be used to run protocols tailored to the aims of the individual investigator or classroom teacher. However, we have developed a suite of 3 protocols that should prove useful in large-scale screening of genetically manipulated mice and large-scale pharmacological testing: the matching protocol, the 2-hopper autoshaping protocol, and the switch protocol. The matching protocol measures the mouse's capacity to estimate incomes (food pellets per unit time) at two different locations, to remember whic...

Discussion

Our method yields a wide range of physiological meaningful, quantitative results on the functioning of several different mechanisms of cognition, learning and memory, for many mice at once, in a minimum amount of time, with a minimum of human labor, and with no handling of the experimental subjects during days, weeks, or months of testing. These attributes suit it for genetic and pharmacological screening programs. It uses minimally modified off-the-shelf hardware (test boxes and nest tubs). It produces more data from mo...

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
SmartCtrl Connection Panel	Med Associates	SG-716B (115)	control panel for inputs/outputs
SmartCtrl Interface Module	Med Associates	DIG-716B (114)	smart card for each chamber
Universal Cable	Med Associates	SG-210CB (115)	cable from smart card to control panel
Tabletop Interface Cabinet	Med Associates	SG-6080C (109)	cabinet to hold smart cards
Rack Mount Power Supply	Med Associates	SG-500 (112)	28 volt power
Wide Mouse Test Chamber	Med Associates	ENV-307W (31)	test chamber
Filler Panel Package	Med Associates	ENV-307W-FP (32)	various-size panels for test chamber
Wide Mouse Modular Grid Floor	Med Associates	ENV-307W-GF (31)	test chamber floor grid
Head Entry Detector	Med Associates	ENV-303HDW (62)	head entry/pellet entry into hopper
Pellet Dispenser	Med Associates	ENV-203-20 (73)	feeder
Pellet Receptacle	Med Associates	ENV-303W (61)	hopper
Pellet Receptacle Light	Med Associates	ENV-303RL (62)	hopper light
House Light	Med Associates	ENV-315W (43)	house light
IR Controller	Med Associates	ENV-253B (77)	entry detector for tube between nest and test
Fan	Med Associates	ENV-025F28 (42)	exhaust fan for each chamber
Polypropylene Nest Tub			nest box
Acrylic Connection Tube			connection between nest and test areas
Steel Cabinet			cabinet to hold test chambers (78"H, 48"W, 24"D)
Windows computer			running MedPC experiment-control software
Server			running Matlab, linked to exper-control computer by LAN
Software
MedPC software	Med Associates		proprietary process-control programming language
Matlab w Statistics Toolbox	Matlab		proprietary data analysis and graphing programing system
TSsystem		in Supplementary Material w updates from senior author	Open-source Matlab Toolbox
Note: This is the euipment needed for one cabinet, containing 8 test environments. Hardware must be replicated for each such cabinet. However one computer can control 2 cabinets (16 test environments)