An Integrated Platform for In Vivo Electrophysiology in Spatial Cognition Experiments

Results

To demonstrate the potential of our platform we will describe a number of applications for these software packages across three laboratories and four institutions. The first will be of behavioral data collected from a “spatial reorientation” task that requires mice to use distal cues to get reoriented in space while running down a linear track (Stimmell et al., 2019). Furthermore, we would like to highlight additional publications that have used this platform. These include three publications employing a sequence task in rats (Bower et al., 2005; Euston and McNaughton, 2006; Euston et al., 2007), one publication using a task that requires rats to navigate toward a randomly selected cue light out of 32 possible options (Wilber et al., 2014), and one using both (Wilber et al., 2017). It is worth noting that all of these publications, involving both the sequence task and the random lights task, incorporated a pre-training phase training rats to shuttle between the ends of a linear track. Additionally, we present an instance where the “Stim Trigger” software was combined with a virtual maze to conduct experiments with mice (Cushing et al., 2020). We also include an example with the object-place paired associate (OPPA) task (Sanchez et al., 2019). Finally, we describe its implementation on an unpublished task, the map-to-action transformation (MAT) task.

First, we will elaborate on the application of this software platform for the spatial reorientation task, which was previously described, and has been used for both rats and mice (Rosenzweig et al., 2003; Stimmell et al., 2019). Briefly, the animal must travel back and forth over a linear track with an unmarked rewarded location. The starting point within the linear track is randomized, thus the distance to the rewarded location changes across trials. The animal is moved to a new starting location after each trial while in an enclosed start box where the animal receives a water reward for running to the end of the track and back into the start box.

Before running the actual task, and after a week of recovery after medial forebrain bundle electrode implantation, mice were placed in a custom box with a nose poke port. They were trained to approach and poke their nose into port, which triggered a brain stimulation reinforcement. “Stim Trigger” controlled the characteristics of the brain stimulation like duration and frequency for each beam break. The other parameters like current and electrode wire combinations were configured on the stimulus isolator. Adjustments to these were made to achieve the highest response rate over a week. The optimal configuration was used as reinforcer for the entire length of the experiment.

To run the task, the researchers used three zone markers. The configuration of the zones was the following: three tracked zones, one to start the trial (i.e., exit from the start box), one rewarded zone and one to mark the arrival at the end of the track. As the actual starting point varied in each trial by design, a manual zone skip feature was employed, triggered by a handheld device specifically designed for advancing presentation slides, to mark the beginning of each trial. The virtual location of the “Start zone” was set so the animal could not reach it, and the trial start was marked by pressing the “Next Cue” button using a handheld presentation clicker. Then, the “Rewarded Zone” became active and the animal could trigger the reward, which in this case was electrical stimulation of the medial forebrain bundle. For this application, the delay feature was used for the “Rewarded Zone” so that the animal must remain in the reward zone for a delay period to or obtained a brain stimulation reward. If the animal advances through the reward zone, or obtained a reward in the zone, then the next nonrewarded zone, called the “End Zone,” becomes active. After reaching the “End Zone,” the animal returns to the start box and consumes a water reward while the track is moved to the next randomly selected start location. Thus, at the end of the session the data for the start of the outward trajectory, entry into and exit from the reward zone, reward timestamp if any, and the time the rodent arrived at the end of the track can be extracted.

A postprocessing script extracts the animal location from the video file and then calculates the velocity for each trial fixed with respect to the position relative to the reward zone. The velocity profile is visually displayed, enabling the user to identify and exclude trials with issues, such as instances of lost tracking that are too extensive to allow for accurate position estimation by the software. This exclusion process can be performed during visualization of the current session. This analysis code is available on request.

The results showed that 3xTg-AD female mice performed significantly worse than non-Tg age-matched controls for the 1.5- and 2-s delays, but not for other delays (Stimmell et al., 2019). The six-month non-Tg female mice were able to identify the location of the reward zone and slowed down in preparation for stopping in the zone, whereas the six-month 3xTg-AD females did not. Overall, the study showed that six-month 3xTg-AD female mice were impaired at spatial reorientation compared with non-Tg mice.

The virtual version of the task is identical except that the animal’s virtual location is restarted automatically to the next starting location from the end of the track (Cushing et al., 2020). During the virtual task, the animal is placed on a tablet surface coated with mineral oil to facilitate smooth movement of its paws. The animal’s head is fixed to ensure that as it walks, the tablet detects the movement of its paws, thereby navigating the animal through the virtual environment, which is projected on the floor and three surrounding wall tablets. Tablet holders are 3D printed with a low-cost resin-based 3D printer and the design files for these tablet holders are available on request. The East, North and West tablets display the rest of the virtual room, which moves along with the animal locomotion. The virtual maze software (Molina et al., 2016) was paired with “Stim Trigger” which controlled the delivery of brain stimulation as described above.

Our next detailed example includes in vivo electrophysiological data. Rats were first trained on an alternation training task, where they learned to shuttle back and forth along a linear track created by walls that connect two opposite zones on the circular platform. Rats receive a brain stimulation that is delivered in at both ends of the linear track. Following the alternation training, rats underwent training on a “random lights task,” where sequences of up to 900 elements were drawn randomly with replacement from the 32 light/reward zones.

Subsequently, rats were trained on the complex spatial sequence task, which involves rats learning to navigate to unmarked locations fixed in space in a specific sequence. Landmarks are distributed around the room for spatial orientation.

We used a sequence (1-2-3-4-1-2-3-5-; Fig. 5A; Movie 1) that had a repeating path segment (1-2-3) followed by one of two distinct actions. Specifically, the rat learned in context 5-1-2-3, to go to 4 for reward, while in context 4-1-2-3 the rat must go to 5.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

A, Complex Spatial Sequence Task. Schematic for the complex spatial sequence task. The rat always starts at zone 5 and continues to zones 1-2-3-4-1-2-3-5-. The Zones and Rewarded Zones are set in the “Maze” GUI. As no reward delay was used for this task, as soon as the animal enters the currently active zone, an automatic reward is delivered and the next “Rewarded zone” becomes active following the input order. B, Capture of the behavior during a “cued run.” C, Two plots showing the activity of two hippocampal place cells as a heat plot of occupancy-adjusted firing rate, using an evenly spaced color-map with max rate indicated in red. The firing frequency is also shown. Only areas with a high enough occupancy during the task are represented in the figures.

Thus, each action belongs to two spatial contexts, so navigation to zone 4 or 5 requires a map-to-action transformation. This emulates the common spatial memory problem one encounters when driving through an intersection and remembering the appropriate action given the current route and goals (e.g., turn left to a favorite restaurant vs right to home). Sets of three unguided (“memory”) runs through the complete 1-2-3-4-1-2-3-5 sequence were interleaved with sets of three “cued” runs in which a light at each goal led the rat through the sequence (Fig. 5B). During memory runs, following an error, a light cue directs the rat to the next zone in the sequence. This complex spatial task, while likely engaging the hippocampus, hippocampal activity alone (e.g., splitter cells) is insufficient to predict the rat’s action (Bower et al., 2005). Note, because the sequence makes use of five zones selected from 32 evenly spaced zones distributed around the perimeter of the platform, alternate sequences which match the distance traveled for various elements of the sequence can be created or new sequences can be created by flipping and rotating the sequence to make a novel sequence (we have made use of both of these options in our experiments). The data presented here are the activity of two place cells recorded from the rat’s dorsal hippocampus while it navigated the 8-item sequence on circular platform. The cells have distinct firing patterns focused in a specific location on the platform (Fig. 5C). The animal can be trained by reinforcing their behavior with electrical medial forebrain bundle stimulation or food reward. Rats have been trained under both conditions on the three tasks described in this section. Additionally, this method allowed the researchers to detect the animal’s precise location during behavior, and record the neuronal activity with high temporal resolution, using the recording system.

The next example of the use of our platform is with the object-place paired associate (OPPA) task. (previously described by Sanchez et al., 2019). In brief, rats are trained to travel to the end of a two-arm maze where they have to displace a specific object to obtain a reward out of two possible options. The animal must then navigate back to the center of the maze and go to the other arm. The goal of the current project is to analyze neural activity in the central platform, on each maze arm, and the choice platform where the objects are located. To do this, the Maze software collects accurate timestamps of the location of the animal, as well as entry in zones of interest. The maze software is set-up to detect movements in seven zones, one for each platform at the end of the arms, two for the maze arms, one for the central platform of the maze and two zones outside of the maze to timestamp the manual reward delivery (Fig. 6). To obtain the location of the animal, Bonsai can send location data via UDP, as well as head direction (angle) to the maze software. Each zone entry is sent to the recording system as a TTL to obtain the precise timestamps of the maze events for post processing.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Maze GUI for OPPA task. A, Control of electronic hardware including reward type: none as manual food reward is used. B, Five zones of interest. C, Counters and configuration controls. D, Allocated coordinates for zone location. E, LED test and auto calibration controls. F, Animal ID and comment boxes. G, Real time position tracking with zones (*) in green, unused in black. Unused zones are listed above and below the zones that are positioned on the maze for visualization.

The final example is the map-to-action transformation (MAT) task. Rats first learn to navigate to the reward location (R in Fig. 7, left) from a randomly selected set of seven start locations (S1–S7) using distal cues. The maze is surrounded by landmarks for orientation. Accurate navigation results in food or brain stimulation reward delivered at the reward location. After achieving allocentric condition criterion (over 80% correct trials for at least three out of four sessions), the transformation condition begins. In this condition, the rat is held at each start location in a translucent box for 10 s. This is done for Allocentric and Egocentric conditions to produce parallel datasets for each task. Next, an opaque box is placed over the rat while curtains are pulled obscuring distal cues (Fig. 7, middle). The box is then removed, and the rat must now navigate to R while the distal cues are obscured. The hypothesis is that in this condition the animals must transform the allocentric location of R into an egocentric action sequence to reach the reward.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Map-to-action (MAT) task. Layout for the allocentric (left), transformation (middle), and egocentric conditions (right) of the MAT task. For allocentric and transformation, using “Maze” software, one “Rewarded zone” is used. All the arms S1–S7 and R are marked as “Zones” as well as a center zone. The coordinates as well as the size of the zones can be set in the “Maze” GUI (Fig. 2B,D). No reward delay was used for this task. For egocentric the experimenter determines zone entries manually and triggers the reward using “Send Manually” in “Stim Trigger” (Fig. 3).

After achieving the same criterion for the map-to-action transformation condition, the egocentric paradigm begins where the egocentric relationship between the start location and the rewarded location is maintained (e.g., 4 to 6). However, these locations are randomized for each trial while the distal cues are obscured by the curtains. There are two variations, right turn (Fig. 7), right red and left turn, green. Each rat runs both the right turn and left turn variation separately, and the order is counterbalanced across rats. In this condition, the rat needs to develop a self-centered motor strategy to find the reward and cannot rely on any external cues which are unavailable (distal) or changed every trial (proximal).

Spatial zones are positioned in “Maze” software in order to timestamp zone entry for each of the 8 arms as well as the center of the maze. Some animals are trained with food reward, while others with electrical stimulation of the medial forebrain bundle as the reinforcer. For the animals that receive medial forebrain bundle stimulation reinforcement, the delivery of the brain stimulation pulse train via TTL to the recording system is also timestamped. For all training phases, a nonrewarded zone is positioned in the center of the maze to timestamp entry into the center of the maze. In the case of the allocentric and transformation tasks, the rewarded location remains consistent throughout the session. Two rewarded zones are designated, one located within the correct zone and the other positioned outside of the maze. This ensures that on zone entry, the reward is delivered only once until the experimenter utilizes the “Next Cue” button to reactivate the rewarded zone on the maze. The zone that is positioned off the maze is also used to signal the end of an incorrect trial.

Because of the continuously changing rewarded location for the egocentric task, the “Stim Trigger” software is used to deliver and timestamp the reinforcement delivery for correct trials. Furthermore, manually triggered TTLs can mark events that are not location-dependent, such as the animal placement into boxes. To do this, a wireless relay switch activated by a wireless radio frequency remote control can be used. This relay system is connected to the Neuralynx TTL acquisition board. Ten control rats have successfully completed all three paradigms of the task. Training for each phase takes a similar number of sessions, between 19 and 23.