Novel Object Recognition and Object Location Behavioral Testing in Mice on a Budget

This protocol provides a cost-effective method to conduct object location and novel object recognition behavioral testing in mice. These tests enable the evaluation of hippocampal function as well as function of other cortical regions, such as the prefrontal cortex, involved in object recognition¹⁰. The OLT and NORT have the advantage of avoiding stimuli with strong emotional valence that are required for the Morris water maze, contextual fear conditioning, Barnes maze or radial arm maze. They also avoid the need for food deprivation as required for the radial arm maze. Furthermore, this protocol describes a simple two-day testing procedure that does not require extensive or complicated equipment for execution or analysis. One disadvantage of these tasks is that they do not allow for measures of learning or acquisition. A difference in novelty investigation could be due to poorer learning about objects during training, poorer memory for what was learned or both. Total time spent investigating objects is an important measure for ruling out any inherent differences in exploration drive but is not a measure of learning. If measures of learning are important for an experimental question, a water maze, Barnes maze, or radial arm maze would likely be preferable.

Custom building of behavioral arenas has the potential to save hundreds of dollars and bring object testing within the financial reach of a wide variety of labs. This protocol eliminates many obstacles and streamlines the process of fabrication to make in-house arena construction more accessible to scientists with no specialized training in acrylics. It is important to note that purchasing colored acrylic sheets that will contrast with the mice, such as white acrylic for black mice and black acrylic for white mice, will facilitate data acquisition and analysis, especially when using commercially available analysis software. Ordering cut-to-size sheets with "routed edges" eliminates the need for a table saw (Table 1), and the use of acrylic cement removes the need to drill and countersink pilot holes. Screwing in fasteners, drilling, and cutting acrylic often causes it break, chip and crack due to its brittle nature. Because the cement is a solvent, it will flow into the area being joined, dissolving and softening any acrylic it encounters. Thus, it should not be applied to each piece separately as if it were a traditional glue. Unlike glue, the cement will not fill negative spaces or adhere to surfaces. This is the primary reason to order "routed edges" as this will ensure a smooth and flat edge, creating a much better bond. When the cement dries, it will have fused the two acrylic sheets into a single piece in a process called "solvent welding". Much like metal welding, the finished product is a single piece, but the welded area will always remain the weakest location. As such, once the arenas are in use, care should be taken to avoid direct impact or extreme stress at those junctures.

This protocol also shows how to set up 4 arenas for simultaneous testing of up to 4 mice (Figure 1 and Figure 2). The opaque walls of the arenas prevent mice from seeing one another during testing, but there is still a possibility that having other mice in the room causes odor or noise distractions that can impair testing. Habituation trials as detailed here can help mitigate this concern, as mice are exposed to the multi-animal room conditions before behavioral testing. However, if distractions from other mice or experimenter noise associated with handling other mice is a strong concern, one arena with one mouse can be used, as well, though it will increase the time required to complete the OLT and NORT with multiple mice. More than 4 arenas could theoretically be used as well, but most cameras do not have a wide enough field of view to show that many arenas with good resolution.

The dimensions and distances provided here are general guidelines for mouse behavioral testing in a typical testing room that is 16 x 16 x 16 m³ in dimensions (Figure 1B and 1C). Set up of appropriate environmental cues, arenas, and video recording equipment must be optimized for each environment. Cues can consist of large shapes or patterns (typically in black and white) that enable mice to spatially orient themselves during the OLT. Instead of placing cues at different locations across from each other, cues can also be mounted to the walls of the testing area. This protocol recommends dividing the testing room with a curtain to hide the researcher and computer during behavioral testing. During all habituation, inter-trial intervals, and active trials, researchers should close the curtain to separate themselves from the testing area. If this is not practical, the computer can remain in view of the mice, but the researcher must move out of view during the task. If the researcher is present, the mice may try to rely on her or him as a spatial cue.

All behavioral testing should be completed in a temperature- and humidity-controlled environment with dim, but even illumination at around 310 lux and minimal extraneous sound or strong environmental odor cues, such as perfumes on experimenters. Between each trial and testing day, all arenas and objects should be cleaned with animal facility recommended methods of sterilization such as wiping with 70% ethanol or unscented bleach wipes to minimize olfactory cues. If chemical disinfectants are used, a final rinse with 70% ethanol is recommended because many chemical disinfectants can be irritating to the animals’ feet.  As with any behavioral task, handling mice for several days before testing is necessary to familiarize them with the individuals who will be performing the OLT and NORT and reduce stress during testing^20,22. As mice can experience acute stress due to unfamiliar individuals in the vicinity of the testing area, it is also recommended that all behavioral testing should be completed by the same individual(s). The testing parameters and conditions detailed in this protocol have been optimized for 6 to 9 week-old adult C57Bl/6 mice and would be most useful in revealing memory impairments due to injury in this age group or memory impairments due age itself in older mice. If the aim is to test for memory improvements in young mice, a longer ITI ranging from 1 hour to 1 day would be more appropriate to avoid ceiling effects on performance in the easier 20 min ITI version. Indeed, ITIs can range from 5 min (for immediate recall) to several hours or days (for remote memories), depending on the specific needs of the experiment and the strains and ages of the mice. Importantly, regardless of the length, all ITIs should be consistent between sessions in an experiment. As different strains and ages of mice exhibit differences in behavior and learning, the timing for each trial and interval, testing area arrangement, and objects used can be modified according to the particular strain of mice, their age, and the specific injury/disease/intervention model being tested^9,21,23,25.

While the hippocampal dependence of the spatial memory functions tested in the OLT are well-established, the NORT may or may not rely on the hippocampus. Interpretation of data from the NORT should take this caveat in to account. The determining variables, for whether the hippocampus is involved, in object recognition memory are not agreed upon yet, but could include ITI length or saliency of spatial cues¹⁸. Notably, the presented protocol uses spatial OLT before the NORT, which may bias mice towards using hippocampal processes in the NORT. Thus, it is important to note that the order can be reversed, or each task can be run independently, depending on the experimenters' questions and needs.

Object selection is a critical aspect of both the OLT and NORT^22,24. Ideal objects are heavy enough not to be easily displaced by a mouse and made of material, like glass or metal, that a mouse cannot damage by chewing or scratching. Wooden, foam, or soft plastic objects are not appropriate as they are easily deformed and are difficult to keep odor-free. Furthermore, the objects used in the trials should all be relatively similar in size, texture, odor and material. Figure 1D gives an example of an appropriate object that could be used. This orange plastic figurine of a chick is filled with sand to give it enough weight and sealed to prevent leakage of cleaning reagents or other odor-causing agents. Because of the shape of the top, mice are unable to climb on top of or sit on this object. For the OLT or NORT, this object is best paired with another object of similar size, weight, material, color and complexity, such as a similar plastic figurine of a rabbit. To ensure that object investigation truly reflects preference for novelty, all objects must be validated for equivalent intrinsic value with a minimum of 8 mice in the same strain, sex, and age of the experimental group as described in section 2.9 of the protocol. Additionally, the objects should be randomized in terms of which object is the novel or moved object between mice in the same study to further ensure that inherent characteristics of the objects are not affecting preference. Object placement can also greatly affect the success or failure of the OLT and NORT. Objects must be counterbalanced in the arenas and not be too close to the walls. A corner crowded by an object is an attractive place for mice to hide and this will confound measures of investigation.

An important prerequisite to data collection is defining "active investigation", which is when a mouse engages an object with its nose pointed at the object no more than 2 cm away. A mouse moving over the top of the object or looking past the object does not qualify as active investigation. Furthermore, since the OLT and NORT depend on a mouse remembering either the spatial location or actual features of the object, there must be sufficient study of these during the training trials. Thus, the researcher must define a minimum investigation time and exclude any subjects that do not meet that baseline level of investigation, traditionally set at 20 seconds²¹.

Quantification of object investigation time can be accomplished in various ways either with or without expensive analysis software. If manual scoring is used, as described here, the most comprehensive data collection can be achieved by recording the time stamps on the video when the mouse is investigating the object. Recording the start and stop times of object investigation during the three individual trials creates a permanent log of unbiased time stamps and facilitates accurate manual calculation of data as opposed to alternative methods such as using a stop watch to additively record the total time a mouse investigates each object.

Commercial software packages are also available for scoring object investigation. Commercially available software can provide a wealth of data beyond the hand scoring data described here, including total distance traveled, amount of time spent in certain areas of the arena, and speed of movement^16,23. Software, once properly calibrated and confirmed to reliably detect object investigation, can also yield data much faster than manual scoring. This faster data yield may, in the long run, yield net savings compared to the human-hours required for manual scoring. However, most software packages for behavioral analysis have high up-front costs that can be prohibitive for many labs and manual scoring can often be accomplished by student researchers, making the hourly costs of behavioral scoring minimal. Commercial software also frequently comes with restrictions on how many users can access the software concurrently, limiting the data throughput and time savings. Though there are many advantages to these software packages, they are not necessary to gather the essential information of time investigating objects from the OLT and NORT. The ability to manually acquire this data makes novel object tasks more financially accessible to a wider variety of researchers.

An additional feature of this protocol is that the first session of habituation is essentially a trial in an open field, which can yield data on activity and anxiety levels. Activity levels can be quantified using total distance traveled or average speed, if tracking software is available. Similarly, with tracking software, anxiety measures can be derived from time spent in the center of the arena or distance traveled in the center. These data can also be acquired manually by overlaying the arena video with a grid and quantifying crossings. This open-field data can help rule out gross motor deficits or excessive anxiety that might interfere with object investigation later.

The statistical tests used in this protocol are representative of traditional methods of analysis. When comparing 2 groups of mice, a two-sample, two-tailed t-test is recommended to test the significance of the difference in percent time investigating the moved or novel object between groups. If the sample sizes are uneven, have exceptionally high variability, or show uneven variability between groups, a non-parametric test is recommended instead, as many of the underlying assumptions of the t-test will be violated in these conditions. When comparing 3 or more groups of mice, ANOVAs are recommended to test whether group has a significant effect on percent time investigating the moved or novel object. Error-corrected post-hoc tests, such as Tukey's test or Bonferroni-corrected pairwise comparisons, can then be applied to test the difference in percent investigation time between each pair of groups. With the OLT, an additional way to analyze data is to test for a significant change in percent time investigating the moved object from training to OLT trials. With only one group of mice, this test would take the form of a paired, two-tailed t-test, testing for significant changes in percent time with the moved object from training to OLT in each mouse (keeping in mind that these are paired analyses and not independent measurements, since each mouse yields two data points). With two groups of mice, a repeated measures two-way ANOVA would be used, with the repeated measure being the trial (training versus OLT) and group assignment being the second factor. Post-hoc tests for differences between treatment groups should compare the groups to each other within trials.

Alternatively, if evidence of memory in the OLT or NORT is all that is being tested, a one-sample t-test of percent time investigating an object versus the fixed value of 50% can be used. A percent time significantly higher than 50% suggests memory for the object. A percent time significantly below 50% suggests aversion (for some reason), and object choices should be re-evaluated. However, the one-sample t-test cannot reveal whether two or more groups differ from each other. For example, in an experiment with two groups of mice, if Group A spent 60% of the time with the novel object for p = 0.049 by one-sample t-test comparison to 50%, and Group B only spent 59% of the time with the novel object for p = 0.051, Group A is significantly different from 50% (and group B is not). However, it is erroneous to conclude that these two groups are significantly different from each other. A two-sample t-test comparing A and B can easily reveal that these two groups are statistically indistinguishable. If the end goal is to compare the memory performance of two or more groups, those groups must be statistically compared to each other, not only compared to an external standard. A similar guideline applies for a comparison of time spent with a moved object between training and OLT. In this case, finding a significant difference in percent time with a moved object between training and OLT in one group and not in another does NOT show that these groups are statistically different. Groups must be post-hoc compared against each other within each trial, not just across trials within group.

It is important to keep in mind that the investigation time of any individual mouse cannot be used as evidence for or against memory. Rather, memory for object location or identity in these tasks can only be concluded based on aggregate data that is statistically compared to either another group's aggregate data or the fixed chance levels (50% for time, 0 for discrimination ratio). The sample size needed will depend heavily on the effect size of a particular manipulation and variability in behavior, both of which in turn will depend on the mice being used. Age, sex, and manipulations all impact variability. In the example data presented in Figure 3A and 3B, n = 14 subjects were used, yielding an effect size of 0.68 and power of 0.65 for a paired t-test with α = 0.05. If a power of 0.8 were desired for this comparison, a sample size of 18 would be required.

This discussion is framed around p-values and significance cutoffs because these are the measures and analyses most typically reported for OLT and NORT data, and therefore are likely to familiar to both experimenters and reviewers. This reliance on p-values has been heavily criticized as statistically invalid²⁶. However, though alternative analysis methods exist and are endorsed in some journals²⁷, none have been broadly adopted by the behavioral and biomedical fields as standard²⁶.

In summary, this protocol effectively tests memory in mice at minimal costs. Recommendations for appropriate modifications to the protocol are included to ensure successful implementation with any small rodent model. Application of this protocol to specific injury or therapeutic intervention models can reveal valuable functional relevance that complements the cellular and molecular mechanisms being studied.