Dissociating Behavior and Spatial Working Memory Demands Using an H Maze

[Abstract] The development of mazes for animal experiments has allowed for the investigation of cognitive maps and place cells, spatial working memory, naturalistic navigation, perseverance, exploration, and choice and motivated behavior. However, many mazes, such as the T maze, currently developed to test learning and memory, do not distinguish temporally and spatially between the encoding and recall periods, which makes it difficult to study these stages separately when analyzing animal behavior and electrophysiology. Other mazes, such as the radial maze, rely on single visits to portions of the maze, making maze coverage sparse for place cell and electrophysiology experiments. In this protocol, we present instructions for building and training an animal on a spatial appetitive choice task on a low-cost double-sided T (or H) maze. This maze has several advantages over the traditional T maze and radial mazes. This maze is unique in that it temporally and directionally dissociates the memory encoding and retrieval periods, while requiring the same behaviors of the animal during both periods. This design allows for independent investigation of brain mechanisms, such as cross-region theta coordination, during memory encoding and retrieval, while at least partially dissociating these stages from behavior. This maze has been previously used in our laboratory to investigate cell firing, single-region local field potential (LFP) patterns, and cross region LFP coherence in the hippocampus, lateral septum, prefrontal cortex, and ventral tegmental area, as well as to investigate the effects of hippocampal theta perturbations on task performance.

Half a century later, Olton developed a radial arm maze to investigate choice behavior and spatial working memory beyond simple T maze alternation (Olton and Samuelson, 1976;Olton et al., 1979). In the radial maze, animals are expected to visit each arm of the maze once, with no repetitions or revisits.
Although this maze requires learning, a trained animal normally visits an arm one time per session (Foreman and Ermakova, 1998), making maze coverage sparse for place cell and electrophysiology experiments. Working memory in the radial maze also incorporates all previous choices during a session, which means memory errors can compound and a single choice is dependent on all previous choices.
Our lab has developed a double-sided T (or H) maze which has several advantages over the traditional T maze and radial mazes. In this task, an animal is forced to an arm on one side T of the maze, and then must choose the same side of the maze on the opposing T in order to get rewarded.
While this maze task requires a learning period and thus can be used for studying spatial learning, a trained rat can run dozens of trials a day, each of which requires coverage of at least half of the track.
The maze is therefore ideal for use in place cell and other experiments which require robust tract coverage. Furthermore, each choice in the maze requires knowledge of only the immediately preceding navigation, so errors are non-compounding. Finally, this maze is unique in that it temporally and directionally disassociates the memory encoding and retrieval periods, while requiring the same behaviors of the animal during both periods (i.e., both periods require a run down a center stem and then a single turn, which is mirrored across encoding and retrieval). This design allows for independent investigation of brain mechanisms, such as cross-region theta coordination, during memory encoding and retrieval, while at least partially dissociating these stages from behavior.
In this protocol, we will discuss the steps involved in building a low-cost version of this maze for use in rat, as well as the steps necessary for training an animal to become proficient at the task. This maze has been previously used in our laboratory to investigate cell firing, single-region LFP patterns, and cross region LFP coherence in the hippocampus, lateral septum, prefrontal cortex, and ventral tegmental area (

Animals
Animals used were Long Evans rats. All animals tested were males at about 350-500 g, but females and animals of varying weights could be used. A scaled version of this maze has also been successfully used with mice (Siegle and Wilson, 2014 4. Wood or corrugated plastic to create a block to force the animal to one side at the forced portion of the maze (for an automated version servo controlled doors can be used [Gomperts et al., 2015]). 5. Risers to elevate the maze, we used blue traffic cones.

For reward delivery
A remote reward delivery system, we used two BrandTech TM BRAND TM seripettor TM Bottletop Dispensers (Fisher Scientific, catalog number: 03-840-010) attached to two common reagent bottles.
We ran tubing from the bottles to the reward site in the maze, as described in 'procedure'. Previous procedures have also used 45 mg food pellets (Gomperts et al., 2015). animal is forced to the right forced arm, they must also visit the right choice arm to receive a reward.
The correct choice must be made on the first try in order to be rewarded. The side the animal is forced to is randomly chosen, with the caveat that the animal can be forced to the same side of the maze for no more than three trials in a row (to avoid the animal thinking only one side of the maze is ever correct). A reward is delivered remotely by an experimenter who is also responsible for moving the blockade that forces the animal to a forced arm. A new trial begins when the animal returns to the forced arm.
Of note, if one seeks to compare neurophysiology associated with choice-associated reward to forced reward, the animal can also be rewarded at the forced side of the maze (Gomperts et al., 2015). 5 www.bio-protocol.org/e3947   does not receive a reward. The animal then checks the right arm for a reward, which is not administered as the initial choice was incorrect.

Animal housing
Animals were single housed with enrichment on a 12/12 light dark cycle. Animals were trained during the light cycle as to capture post-training sleep, but could easily be run on the dark cycle. Animals were food deprived to 85-90% of starting weight, with free access to water.

Maze construction
1. Construct the maze to the parameters in Figure 1. Ends of forced and choice arms can be made slightly larger to allow for turnaround room.
2. If planning on doing electrophysiology and the maze is constructed of metal, use a conductive wire to ground the maze to a ground source in the room.
3. We advise covering the maze in black contact paper, which makes for easier cleaning and provides high contrast for animal tracking with overhead cameras. 4. Construct two blocks the width of the maze (~2.5 in/6.3 cm), by about 6 inches deep and 10 inches tall. These will be used to block off the forced arm during the learning and test phase, as well as the incorrect choice arm during the learning phase. These blocks should be deep and tall enough so that the rat cannot crawl around or over them. 7 www.bio-protocol.org/e3947

Reward delivery
We recommend using a liquid food reward so the animal cannot travel with the reward; we used 20% chocolate milk powder and 10% sucrose in water. We inexpensively created a remote reward delivery system as follows: 1. Drill a hole in the wall of the maze at the two reward delivery sites.
2. Glue or cement a small dish, such as a screw cover cap, beneath the drilled hole.
3. Feed a pipette tip through the hole so it is over the dish, and connect the pipette tip to tubing.
The tubing can then be run to the Seripettor TM Bottletop Dispensers for remote delivery. It is important to fill the length of the tubing with the reward liquid before putting the animal in the maze.

Animal tracking
Install overhead cameras above the maze making sure the entire maze is in the field of view. Adjust tracking software to your desired parameters, which will vary depending on if you are tracking the animal using an LED or not. We adjusted camera parameters using a trial animal to avoid unintended maze exposure with a test animal. Some tracking programs, such as Bonsai, allow you to set the parameters post recording.

Operating the maze
Because the maze is not automated, it requires manual operation. Set up an area of the room where the maze operator can be hidden behind two semi-sheer curtains. The two curtains should part at the forced point of the maze. This setup allows the maze operator to remain hidden from view, while within reach of the object blocking the non-forced arm of the maze to adjust when the animal is at the maze choice arms. This setup also allows the operator the observe the choices made by the animal for manual reward delivery. Please refer to Video 1 for a demonstration on operating the maze.
A reward is dispensed when the animal's entire body, including tail tip, is entirely in the correct choice arm. While all four feet in the correct arm could be used as a criterion, animals often change or correct their choices with four feet (but not the entire length of their tail) in the maze. Using tail tip as criterion helps eliminate incorrect administration of reward.

C. Animal training
For an overview of the animal training timeline, see Figure 3.

Handling phase (minimum of 10 days)
Handle the animals for a minimum of 15 min a day for 10 days before placing on the maze. Allow the rat to familiarize themselves to you by holding them in your lap, stroking them lightly, and playing with them, over gradually increasing lengths of time. As the version of the maze we use is not automated, the animal must be familiar with you as to avoid being distracted when you operate the maze. Begin food depriving the animal about a week prior to the start of training the animal on the maze. 8 www.bio-protocol.org/e3947  unrewarded exploration the maze, followed by two days of reward exploration, and 3 days of trials where the animal is double forced (i.e., forced at both the forced and choice sides of the maze to make the correct turns). Following these three training stages, the animal is ready to begin learning the maze task.

Familiarizing the animal with the reward (3 days)
Three days before beginning maze exploration, during the animal's dark/feeding period, leave a small dish in the animal's home cage with some prepared liquid reward. This is to familiarize the animal with the reward food and to avoid any neophobia on the maze.

Unrewarded exploration phase (2 days)
This phase is designed to get the animal familiar with the track and the surrounding landmarks. This phase should last for two consecutive days and no arms on the maze should be blocked off.
1. Place the animal on the maze (location was not kept consistent).
2. Let the animal explore, uninterrupted, for 30 min or until the animal is still for 5 min, whichever comes first.
3. Remove the animal from the maze. 4. On the second day, if the entirety of the track was not covered on the first exploration day, be sure to place the animal in the unexplored area when placing them on the track.

Rewarded exploration phase (2 days)
This phase is designed to familiarize the animal with the reward locations on the maze. This phase should last for two consecutive days and begin the day after the end of the unrewarded exploration phase. 9 www.bio-protocol.org/e3947 2. Place the animal on the track, at either one of the forced arms or the middle stem.
3. Allow the animal to explore and find the filled reward wells. 4. When the animal has finished a reward, remotely dispense 0.5 ml reward into the now empty well after the animal has exited that arm of the maze. 5. Continue doing this for 30 min or until the animal is still for 5 min, whichever comes first. There may be some initial hesitancy to take the reward on the maze; this is normal.

Double forced learning phase (3 days)
This phase is designed to familiarize the animal with the idea that they will be rewarded on the choice side after visiting the same side of the maze they were at on the maze's forced side. This phase is easiest to complete with two people; one to move the blockade on each side of the maze.
1. Before putting the animal on the track, set up the remote reward dispensers to dispense 0.4 ml of reward.
2. Prepare and make note of the direction the blockade will be on for 30 trials (you will likely only get through 5-12 trials on these days). As stated, the animal will be forced to variable sides of the maze with no more than three consecutive trials forced to the same side. To prepare, either run prepared code that will output trials in this pattern (we used https://github.com/hsw28/behavior/blob/master/maze/leftright.py) or you can assign heads and tails of a coin to left and right and flip a coin. If the coin comes up the same way three times (for example, instructs three times forced right), do not flip for the next trial and assign it to the opposite direction (i.e., left) then resume flipping to designate the next trial.
3. Block off the designated side of the maze for the first trial on both the reward/choice arms and the forced arms. You will therefore be forcing the animal to the correct side of the maze on the choice side.
4. Place the animal on the unblocked forced arm 5. Allow the animal to run to the rewarded arm and receive the reward.
6. While the animal is taking the reward, set up for the next trial. If the animal is forced the same way, this will require no set up. If the animal is to be forced to the opposite direction, both blockades will need to be moved.
7. The next trial is considered initiated when the animal has returned to the forced side of the maze (the unblocked arm for the next trial)-you will not be moving the animal yourself. Tail tip in the arm was considered a trial initiation.
8. Allow the animal to run the next trial with the appropriate sides blocked off according to your plan from step two. Although, at this stage, there is no choice at the choice arm, it is still a good idea to get in the habit of not dispensing a reward until the animal has committed to a choice/reward arm with tail tip in. 10 www.bio-protocol.org/e3947 9. Repeat steps 6-8 for 30 trials, 30 min, or until the animal is still for 10 min, whichever comes first.
It is common to only get 5-12 trials in on these training days.
10. Repeat the above steps for a total of three consecutive days.

Test phase (variable length)
This is the final phase of learning and testing. The animal should acquire task proficiency (>75% correct) during this phase, although the time it takes to gain proficiency varies greatly (Figure 5B).
We believe a three-day moving average over 75% correct would be an acceptable criteria. 7. If the animal chooses the incorrect arm, they are not rewarded even if they subsequently visit the correct arm. The animal must initiate a new trial for the chance to be rewarded. 8. As soon as the animal makes their choice, whether correct or incorrect, setup for the next trial by moving the block on the forced side to the side needed for the next trial, as determined in step 2.

9.
A new trial is considered initiated when the animal returns to the forced side of the maze, with tail tip in the unblocked arm. 10. Repeat steps 5-9 for 30 trials, 30 min, or until the animal is still for 10 min.
11. An animal is considered to have learned the task when they complete the task at 75% correct for two consecutive days.
For an example of an animal running two trials on the maze see Video 1.

Data analysis
A tail-tip in the forced arm was the criteria used for the initiation of new trials. A tail tip in the choice arm was used to determine when an animal made a choice on the choice side of the maze. For a representative example, see Figure 5.  The animal has not yet learned they must return to the forced arm to initiate a new trial. Middle: Animal has begun to learn that they must return to the forced side of the maze, but still meanders and is not very focused on running and initiating trials. Right: Animal is well trained. They consistently return to the forced arms and are focused on the task with little wandering or turning mid-arm.
2. During the initial stages of training, the animal will tend to run back and forth between the reward arms. This is normal; it takes time for the animal to learn they have to return to the forced side of the maze. This will likely greatly decrease the number of trials initiated (Figure 5A). For an example, see Video 1.
3. Animals at the beginning of training will also tend to run halfway up or down the middle arm. A new trial is not considered initiated until the tail tip is in a forced arm, and a new trial should also NOT be started if the animal has not yet visited a choice arm. This will also greatly decrease the number of total trials (Figure 5A). 12 www.bio-protocol.org/e3947   4. Even a well-trained animal will often check the other reward arm before returning to the forced side of the maze to initiate a new trial (Note than in some automated variations of this maze, after the animal chooses at the choice side, a door is lowered to prevent the animal from checking the other side of the maze [Gomperts et al., 2015]). For an example, see Video 1.

5.
If an animal perseverates on one side of the maze (incorrectly turns the same way on > 80% of trials) for several days it can be useful to return to the double-forced stage of training for two days. A return to the double-forced stage can also be used if 10 days have passed and the animal is not improving.
6. At the start of training, many animals are very curious about the blockade on the forced side of the maze; this is normal. However, your animal should NOT be able to climb around or over the blockade. If they are able to, build a taller and/or deeper blockade. 7. At the start of training, it is normal to get very few trials done in 30 min. This occurs because animals run slowly and spend a lot of time going back and forth between reward arms, rather than initiating new trials by returning to the forced arms. As the animals learn the task and how to initiate new trials, they will become much faster ( Figure 5A).
8. There can be great fluctuations in percent trials correct in one animal. This can be due to natural behavior patterns mirroring trial demands, without the animal actually learning the task. For instance, if your animal has a strong left turn bias and a number of trials require a left turn, it may appear the animal has learned the task when they do not actually understand the task demands. Similarly, if the randomly chosen blocked sides alternate, it can mirror animals' natural tendency to alternate, but the animals have not actually learned the task (Figure 5B). 9. We found there to be extensive variability between animals and their ability to learn this task.
The fastest animal learned within 5 days and consistently performed at over 95% correct. Most animals took 2-4 weeks and were generally around 80% correct. A small proportion of animals were never able to learn to proficiency ( Figure 5B). 13 www.bio-protocol.org/e3947  11. Because this non-automated version of the task requires the input of the experimenter, it is important to stereotype your own behavior to avoid serving as a cue in the task. The experimenter who trained the animal should perform the task and avoid wearing any scented products. As noted above, the experimenter should sit behind a curtain when running the animal to avoid any visual cues, and the blockade should be moved only when the animal is at the other side of the track, ideally when they are receiving a reward and distracted. If multiple animals are being run, the track should be cleaned between each animal. 14 www.bio-protocol.org/e3947

Competing interests
There are no competing interests.

Ethics
All procedures were performed within MIT Committee on Animal Care and NIH guidelines under Wilson protocol 0417-037-20, valid from 2017-2020.