Subject, surgery, and behavioral task
Male C57BL6/J mice (n = 2) weighing 20 ~ 25 g were used in this study. Mice were housed individually in a Plexiglas cage under 12 h/12 h light–dark cycle. Water was limited but food was available ad libitum and body weights were maintained about 80 % of free-feeding weight. All procedures regarding animal care and handling were approved by the Institution of Animal Care and Use Committee of Kyungpook national university (South Korea) and all experimental protocols were performed in accordance with the guidelines for the Care and Use of Laboratory Animals of the National Institutes of Health (USA).
The mice were anesthetized with an intraperitoneal injection of tribromoethanol (Avertin, 0.0125 mg/g of body weight). Deep anesthesia was confirmed by tail and paw pinches, which resulted in no withdrawal behavior. Then mouse head was fixed on a stereotaxic apparatus. Ophthalmic ointment was applied to prevent from drying eyes. After shaving and midline incision, skull surface was cleared using saline. Periosteum was scrapped off and holes were drilled and screws were implanted. A microdrive having 8 tetrodes was implanted above the M1 cortex targeting striatum, and dental cement was applied. After 1 week of recovery period, mouse was introduced to behavioral task.
The schematic drawing of behavioral chamber, which is a modified version used for 5-choice serial reaction time task, is depicted in Fig. 2a . At the start position, mouse was required to nose-poke to initiate a trial. Upon initiation visual cues on the opposite end of the maze were illuminated and the mouse was trained to move toward the cue position and nose-poke to the cued hole to receive reward. Correct nose pokes resulted in water delivery as a reward, but incorrect nose poke resulted in sudden black out of 5 to 8 s as a penalty. Since the water nozzle was located near the start position, mouse had to turn and move back to the start position to consume reward.
Development of automatic commutator control system
First step: refinement of the video tracking to improve the accuracy to detect behavioral parameters such as X, Y coordinates or head-directions
Second step: mechanical assembly of commutator system
Third step: development of algorithm to compute and judge whether or not the animal has made a turn
Final step: calibration of commutator with open loop tracking
First step: video tracking refinement
LED manipulation for video noise reduction
For video tracking, we used digital amplifier (Digital Lynx SX, Neuralynx) with a headstage with two LEDs, red and green mounted alongside (HS-36, Neuralynx). Since it is impossible to control the intensity of the commercial LEDs, we covered the LEDs physically with layers of paraffin films to attenuate the center points’ intensities of LEDs to prevent saturation. Although we used a 2 mm diameter tether (TETH-HS-36-Litz, Neuralynx), sometimes the tether positioned between LEDs and camera, obstructing LED detection. We therefore added the curved aluminum reflectors behind LEDs to widen the spatial coverages of LEDs (Fig. 1a inset).
The position of LEDs were adjusted to be perpendicular to animal’s rostro-caudal axis, that is, red and green LEDs were positioned left and right side of the animal’s head, respectively (Fig. 1a inset). This allowed continuous tracking even when an animal changes its head direction vertically, for example, during grooming or rearing against wall.
Improving SNR by image smoothing
Since the behavioral chamber had glossy walls which could reflect the LEDs, causing errors in detection, the walls were matted via abrasion. In the altered chamber the reflected LED light was diffuse and its intensity was attenuated. To improve the SNR of video image, we smoothed incoming images at the initial sensing stage (Fig. 1a.h). The camera provided a dial to control the focusing level manually, so we adjusted the level slightly out of focus to get pre-smoothing filtered images. For general behavioral monitoring purposes, we used an additional video camera.
Second step: mechanical assembly
We used 56 circuits slip-ring (PSR-C56, Pan-Link technology; Fig. 1b.e). The wires of both ends were soldered to connectors (MDR connector, 50 pin, 3 M Korea) which matches to the adapter of existing amplifier system (ADPT-HS-36-DRS, Neuralynx).
To rotate it using a DC motor (5 V FM1502, D&J WITH, Korea; Fig. 1b.f), a pulley of diameter 60 mm was glued to rotary joint which connected to tether (Fig. 1b.d). A small sized pulley (diameter of 6 mm) was attached to the DC motor shaft. Those two pulleys were linked using a rubber belt so that motor power was transferred through the pulley. The distance between center points of each pulley was approximately 85 mm.
We used a commercially available DC motor driver which was designed based on TB6612FNG chip (Toshiba; Fig. 1b.g). It provides bidirectional control (clock- and counter clock-wise) using two TTL channels. To control the DC motor, the output terminal of the digital IO device (NI PCI-6601, National Instruments) was connected to the motor controller through cables (Fig. 1a).
The frame of commutator was constructed using two acrylic boards (100 × 100 mm) and steel posts (PCB supports, height of 130 mm) as shown in Fig. 1b. The boards were used for top and bottom of the frame and posts were served as vertical supports for the frame and embedded firmly at each edge of the board. The slip-ring, DC motor, and motor driver were then tightly secured on the frame using either glues or screws. Several holes for slip-ring and posts were drilled on the bottom board.
Third step: development of algorithm
The head position is defined by the mid-point between the LEDs. The head direction is defined by the angle composed of a reference line (e. g., y axis of an image) and the midline of animal (rostral to caudal), which is orthogonal to the axis passing both LEDs from the video image. We took advantage of the data acquisition software, Cheetah (Neuralynx). Tracking parameters were then obtained at 29.97 fps (frames per second) and streamed to processing computer through a NetCom library (Neuralynx) for online processing. Customized software was designed using Labview (National Instruments) to apply the algorithm described below.
The behavioral task consisted of a single trial repeated many times during a daily session. Across the entire session the mice spent the majority of their time near the starting position, therefore we set a virtual rectangle near the starting position in the video image (Fig. 2a). The rectangle was set empirically so that mouse could not rotate within it and could be detected when coming back to the position for reward. Under the condition, only the choice phase was needed to examine whether a mouse rotated 360°. The head direction, the relative angle from the positive Y axis of the video image, increases along the clockwise angular direction. When the mouse is near the start point, its head direction is about 180°. A single rotation of 360° through both direction (clock-wise; CW or counter-clock-wise; CCW) has a trajectory of sudden change of the angle of which absolute value is larger than a threshold value, say, 330°. Consequently, the total rotation number (TRN) is rotations of CW – CCW. For example, the TRN of +1 and −2 indicate a single rotation of CW, and two rotations of CCW direction, respectively. Zero indicates no rotation.
In summary, algorithm to detect whether a mouse rotated is as follows (Fig. 2b).
When the mouse occupies the starting position, wait.
As soon as the mouse departs the starting position, stores a trajectory of head direction until the mouse comes into the position back.
Compute rotation number (RN) from the trajectory.
Count CW rotation (RNCW)
Count CCW rotation (RNCCW)
Total rotation number (TRN) is then TRN = RNCW - RNCCW.
Generate motor command for the given TRN.
To drive the DC motor, we used pulse modulated control. We set single pulse duration as 100 ms, and changed its duty cycle. Then we generated a series of pulses, a pulse train, to drive the motor. As described above, actual disentanglement of tether is carried out when the mouse is in the start position, rewarding and/or initiating the next trial. Therefore, it is important to rotate the DC motor exactly 360°. Detailed calibration procedures are described in the next section.
Final step: calibration and application
There are two parameters to control the motor angles, the pulse width and the number of pulses. Therefore, by inspecting the combination of both parameters, it is possible to rotate the tether close to 360°. Initially we set the pulse width as 20 ms, that is, duty cycle of 20 %. Then pulse trains of having 20 to 30 pulses were transmitted to motor controller. We found that 26 pulses made one rotation nearly perfectly. The given pulse trains were further validated by sending it repeatedly, under simultaneous online recording of its angles. We checked the angle difference before and after motor rotation.
Finally, the designed system was applied to mouse performing behavioral tasks which lasted approximately 50 min and composed of more than 100 trials.