Investigating Motor Skill Learning Processes with a Robotic Manipulandum

Summary

A paradigm is presented for training and analysis of an automated skilled reaching task in rats. Analysis of pulling attempts reveals distinct subprocesses of motor learning.

Abstract

Skilled reaching tasks are commonly used in studies of motor skill learning and motor function under healthy and pathological conditions, but can be time-intensive and ambiguous to quantify beyond simple success rates. Here, we describe the training procedure for reach-and-pull tasks with ETH Pattus, a robotic platform for automated forelimb reaching training that records pulling and hand rotation movements in rats. Kinematic quantification of the performed pulling attempts reveals the presence of distinct temporal profiles of movement parameters such as pulling velocity, spatial variability of the pulling trajectory, deviation from midline, as well as pulling success. We show how minor adjustments in the training paradigm result in alterations in these parameters, revealing their relation to task difficulty, general motor function or skilled task execution. Combined with electrophysiological, pharmacological and optogenetic techniques, this paradigm can be used to explore the mechanisms underlying motor learning and memory formation, as well as loss and recovery of function (e.g. after stroke).

Introduction

Motor tasks are widely used to assess behavioral and neural changes related to motor learning or to alterations in motor function in neurological or pharmacological animal models. Fine motor function can be difficult to quantify in rodents, however. Tasks requiring manual dexterity, such as manipulation of cereal¹, pasta², or sunflower seeds³ are sensitive and do not require extensive training of the animal. Their main drawback is that these tasks yield mostly qualitative results and can be difficult to score unambiguously.

Skilled reaching tasks, such as variations of the single pellet reaching task are more straightforward to quantify^4,5. However, kinematic factors that underlie successful execution of these tasks can only be inferred to a limited extent and require labor-intensive frame-by-frame video analysis.

Robotic devices have gained popularity as means of quantifying aspects of forelimb function and motor skill. Several automated reaching tasks are available. The majority focus on a single aspect of a forelimb movement, such as pulling of a handle along a linear guide^6,7, simple distal limb movements⁸, or pronation and supination of the paw⁹. While these devices show promise for the analysis of motor function in, they only reflect the complex motor actions executed during single pellet reaching to a limited extend.

Here, we demonstrate the use of a three-degree-of-freedom robotic device, ETH Pattus, developed for training and assessment of various motor tasks in rats^10,11. It records planar and rotational movement of rat forelimb movements in reach, grasp, and pulling tasks carried out in the horizontal plane. Rats interact with the robot via a 6 mm-diameter spherical handle that can be reached through a window in the testing cage (width: 15 cm, length: 40 cm, height: 45 cm) and moved in the horizontal plane (pushing and pulling movements) and rotated (pronation-supination movements). Thus, it enables the rat to carry out movements that approximate those executed during conventional single pellet reaching tasks. The window is 10 mm wide and located 50 mm above the cage floor. The handle is located 55 mm above the floor. A sliding door blocks access to the handle between reaching trials and opens when the robot reaches its start position and closes after a trial is completed. After a correctly executed movement, rats receive a food reward on the opposite side of the testing cage.

The robot is controlled via software and records output from 3 rotary encoders at 1000 Hz, resulting in information about the position of the handle in the horizontal plane, as well as its rotation angle (for details, see reference¹¹). The conditions required for successful task execution are defined in the software prior to each training session (e.g. minimum required pulling distance and maximum deviation from midline in a reach-and-pull task). An initial standardized reference position of the handle is recorded with a fixed holder at the start of each training session. This reference is used for all trials within a session, assuring a constant start position of the handle for each trial. Constant positioning of the handle relative to the cage window is assured by alignment of marks on the cage and robot (Figure 1).

Video recordings of the reaching movements are recorded using a small high speed camera (120 frames/s, 640 x 480 resolution). A small display in the camera's view shows the rat's identification number, training session, trial number and trial result (success or failed). These videos are used to verify recorded results and to assess the effects of reaching movements that precede the touching, pulling or rotation of the handle.

Here, we demonstrate the use of this robotic platform in variations of a reach-and-pull task. This task can be trained within a period of time that is comparable to other skilled reaching paradigms and yields reproducible results. We describe a typical training protocol, as well as some of the main output parameters. Moreover, we show how minor changes in the used training protocol can result in altered time courses of behavioral outcomes that may represent independent subprocesses within the motor skill learning process.

Protocol

The experiments presented here were approved by the Veterinary Office of the Canton of Zurich, Switzerland and were carried out according to national and institutional regulations.

1. Feeding Conditions

NOTE: All training sessions are performed under a scheduled feeding protocol.

1. Feed the rats 50 g/kg of standard chow once per day, after training is completed. This amount of food is sufficient to prevent major weight loss (body weight is >90% of free-feeding weight), but small enough to ensure reproducible behavioral conditioning. Weigh the rats daily to ensure their body weight remains stable.
   NOTE: Additional overnight (10-12 h) food deprivation may be useful prior to the first reward-touch session (step 2.3).

2. Training Procedure for a Reach-and-pull Task

1. Preparation: Allow the rats to habituate to their new home cages for at least a week after arrival in the animal facility. Handle the rats regularly during this time and give dustless precision pellets in the home cage to habituate the rats to the new food. These pellets will be used as rewards throughout the training protocol.
2. Habituation: Place the rats in the testing cage for 30-45 min and provide 30-50 pellets in the feeding bowl, mixed with powdered chow. Open and close the cage window and run the pellet dispenser occasionally to habituate the rats to their sound.
   1. Repeat this for 2-3 days.
3. Reward-touch: Train the rats to touch the spherical handle through the cage window and to then move to the opposite side of the cage to retrieve a food reward.
   1. Adjust the software settings so that the handle is located just outside the testing cage window at the start of each trial and align the handle with the center of the cage window. When the trials are successful, i.e., as soon as a light touch on the handle (0.25 mm displacement in any direction) has been detected, a tone sounds and a reward is dispensed. Classify trials as failed when no touch has been detected for 180 s after the window opens.
   2. Put the rat in the training cage. Prompt the rat to reach out by letting it grab at a pellet held near the handle. Direct the rat's attention to the handle and food bowl by tapping on the cage.
   3. Stop prompting when the rat independently reaches through the cage window and retrieves the food pellet.
   4. Continue until 100 trials (touches) are completed or until 60 min have passed, whichever comes first.
   5. Continue training for 3-4 days and begin the next stage of training (step 2.4) when rats achieve 100 trials within 30 min on 2 consecutive days.
      NOTE: Do not over-train this step. The goal of reward-touch is to achieve reliable interaction between the rat and robot, so that this behavior can be shaped in subsequent training.
4. Free pull (FP): Train the rats to reach out and pull the robot's handle.
   1. Adjust the software settings so that the handle is located 18 mm from the window at the start of each trial, and must be pulled for at least 10 mm without interruption for a successful trial. There are no lateral restrictions on the pulling movement in this stage.
      1. Classify a trial as failed when the handle has not been moved for 180 s after the window opens, when the handle is moved outside of the reachable workspace (more than 12 mm from midline), or when the rat has pulled less than 10 mm within 5 s after the first touch has been detected.
   2. Take note of the number of times the left and right paw are used during the first 20 trials of the first FP session. The paw that is used in at least 80% of the trials is considered the preferred paw.
      NOTE: Paw preference may already be clear in reward-touch sessions.
   3. Move the handle laterally until it is aligned with the edge of the window to facilitate pulling with the preferred paw (i.e. move the robot 5 mm to the left side of the window for right-handed rats and vice-versa).
      NOTE: Place the handle in this exact same position relative to the cage for all following training sessions for this rat. Assure exact placement by marks on the cage wall and on the robot.
   4. Put the rat in the training cage and train until 100 trials are completed or until 60 min have passed, whichever comes first.
      NOTE: If the rat does not reach out far enough, prompt it by letting it grab at a pellet held near the handle. Rats may stop trying to pull after repeated failed attempts. Tap on the cage, let them grab for pellets held with a pair of forceps or dispense a pellet to restore their motivation.
   5. For experiments involving only FP training, continue training as described in 2.4.
      NOTE: Typically, 1-2 FP sessions are needed to help transition from reward touch to SP(Straight Pull) training. The goal of these FP sessions is to habituate rats to reach out, grab and pull the handle, rather than to only touch it. As with reward-touch training, it is important not to over-train if the goal is to transition to a next training step.
5. Straight pull (SP): Train the rats to pull the handle without deviating more than 2 mm from midline.
   NOTE: The midline is defined relative to the start position of the robot, not to the midpoint of the cage window. Thus, a pulling attempt ending at the midpoint of the cage window will result in a pulling trajectory that deviates more than 2 mm from midline.
   1. Adjust the software settings so that only trials where the pulling movement does not deviate more than 2 mm from midline on either side are rewarded by a tone and a pellet. Keep all other parameters as described in step 2.4.
   2. Put the rat in the training cage and train until 100 trials are completed or until 60 min have passed, whichever comes first.
      NOTE: Rats may become extremely agitated and stop trying to pull after repeated failed attempts. Tap on the cage to redirect their attention to the reaching task, let them grab for pellets held with a pair of forceps or dispense a pellet to restore their motivation.
   3. Continue training until the rats reach plateau performance, or adapt the training period according to the goal of an experiment.

Results

Here, we show 3 variations of a reach-and pull task using male Long-Evans rats (10-12 weeks old). In the free-pull (FP) group (N = 6), rats were trained to pull the robot's handle for a 22 day period without lateral restrictions. Animals in the straight-pull 1 (SP1) group (N = 12) were trained to pull the handle without deviating more than 2 mm from midline. These animals transitioned directly from reward-touch (step 2.3) to straight-pull training (step 2.5). For both FP and SP1 anima...

Discussion

Skilled reaching tasks are commonly used to study motor skill acquisition as well as impairment of motor function under pathological conditions⁶. Reliable and unambiguous analysis of reaching behavior is essential for the study of cellular mechanisms underlying motor skill acquisition, as well as neurophysiological processes involved in loss and subsequent recovery of function in animal models of neurological disease. The results presented here show how spatial and temporal aspects of the pulling ...

Materials

Name	Company	Catalog Number	Comments
ETH Pattus			ETH Pattus was made by the Rehabilitation Engineering Laboratory of Prof. Gassert at ETH Zurich.
Training cage			The plexiglass training cage was made in-house.
Pellet dispenser	Campden Instruments	80209
45-mg dustless precision pellets	Bio-Serv	F0021-J
GoPro Hero 3+ Silver Edition	digitec.ch	284528	Small highspeed camera
Small display	Adafruit Industries	#50, #661	128 x 32 SPI OLED display controlled via an Arduino Uno microcontroller and Labview software
LabVIEW 2012	National Instruments	776678-3513	ETH Pattus is compatible with more recent Labview versions.
Matlab 2014b	The Mathworks	MLALL