Automated Discrimination Learning and Video-based Analysis of Decision Making in Zebrafish (Danio rerio)

Bishen Singh1, Luciano Zu3, Jacqueline Summers1, Saman Asdjodi1, Jared Giordano1, Eric Glasgow2 and Jagmeet S. Kanwal11 1Dept. of Neurology, 2 Dept. of Tumor Biology, Georgetown University Med. Ctr., Washington, D.C.; 3Univ. degli Studi di Roma 'La Sapienza', Roma, Italy

INTRODUCTIONDirected swimming towards a target requires spatial memory, sensory inputs and motivation. An animal may be trained to perform this task via associative (classical or operant) conditioning.

In a stimulus-driven, operant conditioning paradigm, we present a pulse of light via LED’s and/or sounds via an underwater transducer. A webcam placed below a glass tank records fish swimming behavior. During operant conditioning, a fish must interrupt a light beam at one location to obtain a small food reward at the same or different location. The timing-gated interrupt activates robotic arm and feeder stepper motors via custom software controlling a microprocessor (Arduino). In this way, full automation of stimulus-triggered place-sensitive conditioning is achieved. Precise multi day scheduling of training, including timing, location and intensity of stimulus parameters, and feeder control is accomplished via a user-friendly interface.

Our training paradigm permits tracking of learning by monitoring swimming, turning, location and response times of individual fish. This facilitates comparison of performance within and across a cohort of animals. We demonstrate the ability of zebrafish to discriminate complex sounds using the newly developed methodology.

Current methods used for associative conditioning often involve human intervention, which is labor intensive, stressful to animals, and introduces noise in the data. Our relatively simple yet flexible paradigm requires a simple apparatus and minimal human intervention. Our scheduling and control software and apparatus (NEMOTRAINER) can quickly and efficiently screen drugs and test the effects of of CRISPR based and optogenetic modification of neural circuits on sensation, locomotion, learning and memory.

Fig. 1. Diagrammatic representation of the audio-visual training apparatus. Fish training/testing trials on stimulus-directed swimming were conducted individually and in groups within the apparatus.

METHODS: Materials

Fig. 4. Flow chart depicting the algorithm for automated training. The training procedure assigns user-defined delays for turning “on” and “off” light and sound or any other type of stimulus as part of the setup (green). Stimulus repetition (blue) provides multiple opportunities in close succession within each of multiple daily runs (pink) for the animal to learn the task.

REFERENCES & SUPPORT

Fig. 5. Screen captures of single video frames. location of free-swimming fish a) pre sound presentation; b) and c) post sound presentation. Collective decision making [5] likely determines final location of all fish towards upper for upward FM (UFM) vs. lower for downward FM (DFM) partitions.

1 Associative Conditioning

SUMMARY

2 Interface Design

Fig. 3. Screen captures of user interface (above) and training schedule (below) for monitoring sensors and control of associative conditioning. “Ardulink” (vers. 0.4.2; Zu, 2013), a JAVA facility, allows simultaneous implementation of communication protocols with Arduino. User-definable settings enable either classical or operant conditioning via customized multi-day scheduling and precise control of stimulus parameters for training.

Fig. 6. Box plots and jittered scattergrams showing distances of 6 fish from correct (target) side after training in response to the presentation of sounds within a single trial (4). Data were obtained from 1 s before (PRE) and during the second (POST2) and third (POST3) second post stimulus. Shorter distances from target indicate better learning. On average, fish were closer to target during POST2 and started to wander away during POST3. Data from 2 fish trapped behind dividers are not included in this plot.

4 Video-based Analysis3 Conditioning Paradigm

1. We demonstrate fully automated training and testing of freely swimming zebrafish within a stimulus-dependent directional memory task.

2. NEMOTRAINER can be used to test the ability of zebrafish to discriminate between different colored lights and frequency modulation (FM) in sounds.

3. Zebrafish are attracted to LEDs and appear to learn upward FMs better than downward FMs.

RESULTS

• Zebrafish maintenance• 14:10 light:dark cycle• Fed daily with brine & dried flake food• Habituated to housing 3 days prior to training

• Hardware• Arduino microprocessor, stepper motors and LEDs. • Desktop computer, webcam (Logitech d90), Debut -video

recording software.• Circular glass tank, aerator, temperature probe, amplifier, plastic

tubing and BNC cables. • Underwater sound transducer: digital sound files.

Training Paradigm:Fish are trained via a reward-based conditioning paradigm to associate a sound with a particular location (side) of the tank. For operant conditioning, zebrafish trigger motion sensors to obtain a small reward. Triggering of the sensor is accomplished by swimming to the correct side of the tank, which activates release of food into that chamber. For classical conditioning, a small food reward is delivered after a user-specified delay. Each trial extended over 5 days with 6 possible repetitions (reps.) or chances to obtain food for each of 6 runs per day.

Day 1: Free run: intermittent reward paradigmDay 2: Delayed LEDs, alternatingDay 3: Delayed LEDs, randomizedDay 4: Delayed LEDs, randomizedDay 5: Sounds Only

• Zebrafish are an excellent model organism for neurological studies. • Commonly used in genetics, oncology, and developmental

biology• Genome completely sequenced• Rich expression of innate behavior (e.g. dominant and

submissive behavior, food searching behavior, shoaling)• Easily maintained and bred under laboratory conditions

• Hypothesis: Using an automated system of LEDs and auditory cues, zebrafish can be trained via classical and operant conditioning to trigger motion sensors and earn a food reward

• Goal: To create a fully programmable, user-friendly, low-cost system that can be easily replicated and expanded to allow for training multiple animals in parallel.

BACKGROUND

METHODS: Setup, Training and Tracking

METHODS: Software development & Data Analysis

1. Blaser RE, Vira DG. Experiments on learning in zebrafish (Danio rerio): a promising model of neurocognitive function. Neurosci Biobehav Rev. 2014 May;42:224–31.

2. Higgs DM, Souza MJ, Wilkins HR, Presson JC, Popper AN. Age-and size-related changes in the inner ear and hearing ability of the adult zebrafish (Danio rerio). JARO. 2002;3(2):174–84.

3. Manabe K, Dooling RJ, Takaku S. An automated device for appetitive conditioning in zebrafish (Danio rerio). Zebrafish. 2013 Dec;10(4):518–23.

4. Miller N, Garnier S, Hartnett AT, Couzin ID. Both information and social cohesion determine collective decisions in animal groups. Proc Natl Acad Sci U S A. 2013 Mar 26;110(13):5263–8.

5. Mueller KP, Neuhauss SC. Automated visual choice discrimination learning in zebrafish (Danio rerio). J Integr Neurosci. 2012;11(01):73–85.

6. Pérez-Escudero A, Vicente-Page J, Hinz RC, Arganda S, de Polavieja GG. idTracker: tracking individuals in a group by automatic identification of unmarked animals. Nat Methods. 2014 Jul;11(7):743–8.

ACKNOWLEDGEMENTS: Supported in part by BGRO, Georgetown University.

Fig. 2. Tracking individual fish behavior. Tracks created using iDTracker (Pérez-Escudero et. al., 2014) from video recordings a) before, and b) during presentation of upward FM. Tracks begin at locations 1s pre and terminate 4s post sound onset, showing fish moving towards lower partition.

Fig. 7. Line plots showing individual variation in learned performance to the presentation of upward and downward FMs as reflected in proximity to the correct side (target) on which they were trained. Averaged performance for all 8 fish across 5 trials (n=40) is shown in plots and averaged behavior plots are superimposed. UFMs elicited better performance.

SoundTransducer

a

b

c

UFM

DFM