Exp Neurobiol 2019; 28(2): 146-157
Published online April 8, 2019
© The Korean Society for Brain and Neural Sciences
Dong-Hee Kim1, Yoon-Sun Jang1, Won Kyung Jeon2,3, and Jung-Soo Han1*
1Department of Biological Sciences, Konkuk University, Seoul 05029, Korea.
2Herbal Medicine Research Division, Korea Institute of Oriental Medicine, Daejeon 34054, Korea.
3Convergence Research Center for Diagnosis, Treatment and Care System of Dementia, Korea Institute of Science and Technology, Seoul 02792, Korea.
Correspondence to: *To whom correspondence should be addressed.
TEL: 82-2-450-3293, FAX: 82-2-3436-5432
Genetically modified mouse models are being used predominantly to understand brain functions and diseases. Well-designed and controlled behavioral analyses of genetically modified mice have successfully led to the identification of gene functions, understanding of brain diseases, and development of treatments. Recently, complex and higher cognitive functions have been examined in mice with genetic mutations. Therefore, research strategies for cognitive phenotyping should be sophisticated and evolve to convey the exact meaning of the findings and provide robust translational tools for testing hypotheses and developing treatments. This review addresses issues of experimental design and discusses studies that have examined cognitive function using mouse strain differences, genetically modified mice, and transgenic mice for Alzheimer's disease.
Keywords: Hippocampus, Mice, Water maze, Fear conditioning, Object recognition, Prefrontal cortex
Genetically modified mouse models allow us to better understand brain functions, causes of brain diseases, and develop therapeutic treatments. Anatomical, biochemical, physiological, and behavioral analyses of genetically modified mice provide insights into the influences of a target gene on brain function and behavior. Because the final output of the nervous system is behavior, behavioral measures are essential for analyzing these mice. Therefore, the success of genetically modified mouse models depends on a robust, well-replicated phenotype [1,2,3].
The behavior of rats is well characterized and there is a long and illustrious literature regarding the measuring of cognitive functions, including learning and memory, with well-validated and controlled experimental design and methods . However, environmental factors and the animal handling of the experimenter may confound behavioral data [4,5]. Therefore, to ensure accurate interpretation of cognitive phenotypes, experimental designs, including appropriate control groups and experimental procedures, have been elaborated [3,6,7,8] and should be implemented in future studies. In addition, control measures of general health, sensory abilities, and motor function are examined to avoid over-interpretation and misinterpretation of experimental results [1,2].
What is more important than these points mentioned above is that a researcher understands mouse behavior, along with a deep understanding of the research field. In particular, these comments are further emphasized when studying cognitive function and diseases with cognitive impairments such as Alzheimer's disease (AD). Therefore, we first describe differences between mouse strains with respect to cognitive function and then introduce a number of studies that have revealed the neurobiological basis of cognitive functions using genetically modified mice. Finally, cognitive behavioral assays for AD transgenic mice are reviewed and comments regarding future directions are presented.
Inbred strains of mice are useful in the study of the association between cognitive traits and neurobiological functions [9,10,11,12,13]. Because each mouse in an inbred strain can be virtually identical in genotype, the cognitive characteristics of inbred mice can be related to neurobiological characteristics of the same strain of mice. Hence, numerous studies that show behavioral differences in inbred strains of mice can provide criteria in selecting a background strain for constructing genetically modified mice and in evaluating these behavioral results. For example, these studies can be informative in analyses of a hippocampal phenotype that differs between the comparison strains. Specifically, because DBA/2 mice show a specific behavioral profile, together with a deficiency in synaptic plasticity and a signaling pathway in the hippocampus [14,15,16,17,18,19,20,21], this strain may be a suitable model for the study of mutant mice with the goal of enhancing hippocampal-dependent learning and memory [22,23]. On the other hand, these studies indicate that, compared with other inbred strains including DBA/2 mice, C57BL/6 mice may be the strain with a superior background for the genetic analysis of the molecular/cellular mechanisms underlying hippocampus-dependent learning and memory [8,11,12,19,20,24,25]. Therefore, in this section of the present review, behavioral differences between C57BL/6 and DBA/2 mice, which have been used for developing transgenic or knock-out mouse models, are presented focusing on behavioral tasks that measure hippocampus-dependent learning and memory.
The C57BL/6 and DBA/2 strains have been shown to have relatively similar performances in some behavioral assessments using the water maze [26,27,28,29]. However, other studies have reported that C57BL/6 and DBA/2 mice differed in their performance on tasks that required hippocampal integrity. C57BL/6 mice performed significantly better than DBA/2 mice on a stationary hidden platform version of the water maze task (place training)  and context fear conditioning .
Despite these discrepancies in behavioral reports, two strain differences in the neurobiological mechanisms support the behavioral superiority of C57BL/6 mice in hippocampus-dependent tasks over DBA/2 mice. For example, C57BL/6 mice show greater long-term potentiation and higher expression levels of protein kinase C in the hippocampus than DBA/2 mice [16,18,30]. Interestingly, even though both C57BL/6 and DBA/2 mice performed with similar accuracy in either the place or visible platform training (cued training), hippocampal phosphorylated cAMP response element-binding protein (pCREB) levels were higher in C57BL/6 than in DBA/2 mice after place training, while no differences in hippocampal pCREB levels were reported between the two inbred mice strains after cued training [8,31]. These findings indicate that the hippocampus is better engaged in place training in C57BL/6 mice than in DBA/2 mice.
Additional evidence that C57BL/6 mice are better place learners or use better place strategy compared to DBA/2 mice comes from studies using a different maze task or a modified water maze task. First, C57BL/6 mice have been shown to choose a better place strategy in a plus maze task compared to DBA/2 mice . Furthermore, strategy preferences of C57BL/6 and DBA/2 mice were assessed in a redundant place/cued version of the water maze task, which was developed in a study by McDonald and White . In the modified task, mice received training using a stationary visible platform for 2 days and then using a stationary submerged platform for 2 days. The 4-day sequence was repeated twice. On the final (competition) day, a visible platform was positioned in the quadrant opposite to where it had been located throughout training (see Fig. 1A). Mice were released at starting points equidistant from the training platform and the newly located, visible platform to assess place/cued strategies. As shown in Fig. 1B, mice that visited the training platform location prior to escaping to the visible platform were classified as “place responders,” whereas those swimming directly to the visible (cued) platform were classified as “cued responders.” The number of place responders was higher in C57BL/6 than in DBA/2 mice .
It is notable that in the strategy preference task, as the cued and subsequent place training progressed, C57BL/6 mice performed better than DBA/2 mice on the third place learning task . Therefore, a subsequent experiment was conducted to examine how prior learning affected subsequent learning. Training in the water maze consisted either of cued training followed by place training or the reverse order (i.e., place training followed by cued training) (Fig. 2). Both strains of mice showed equivalent performance in the initial cued or place training (Fig. 2B and 2C). However, C57BL/6 mice performed significantly better than DBA/2 mice, both in place training followed by cued training and in cued training followed by place training (Fig. 2B) [6,33]. These findings indicate not only differences between the two inbred strains in the flexibility of C57BL/6 and DBA/2 mice to switch their learning strategies, but also suggest the importance of designs and training procedures to demonstrate experimental hypotheses.
Pavlovian fear conditioning is a form of associative learning in which an initially neutral stimulus, such as a tone (conditioned stimulus; CS), is paired with the presentation of a stimulus with aversive properties, such as a foot-shock (unconditioned stimulus; US). Freezing responses, an expression of associative fear learning, are elicited by CS presentation or by environment/context cues where the pairing occurred. Several types of fear conditioning have been used in animal studies (Fig. 3). Delay conditioning, in which discrete CS and US co-terminate, is dependent upon the amygdala [34,35], while contextual conditioning, in which the environment serves as the CS, is dependent upon the hippocampus [36,37]. Another hippocampus-dependent type of fear conditioning is trace conditioning, in which the CS and US are separated by a stimulus-free trace interval , and which has been used in numerous studies with genetically modified and inbred mice, along with context conditioning [14,15,19,29,30,39,40,41,42,43,44,45].
While there are many training procedures in the literature that have examined trace fear, a training procedure to minimize nonassociative effects was developed and performed by Smith et al. (2007) to demonstrate strain differences in performance in trace fear conditioning in mice. Subsequent studies using the same training procedure of trace fear conditioning compared differences in freezing responses between C57BL/6 and DBA/2 mice. The C57BL/6 mice displayed higher freezing responses to the tone-CS during testing than DBA/2 mice. Hippocampal levels of pCREB measured after conditioning were higher in C57BL/6 than in DBA/2 mice . These findings provide additional evidence for the superiority of C57BL/6 mice over DBA/2 mice in hippocampus-dependent tasks.
The novel object recognition task has emerged as one of most used tasks for examining cognitive status in genetically modified mice, including AD mice [45,46]. The task was used for the first time to measure recognition memory in the rat, based on the rat's behavior to explore a novel object more than a familiar one . Although there is considerable variation in novel object recognition task procedures, this task is typically conducted in a familiar square or rectangular high-walled arena lacking evident spatial cues (Fig. 4). After habituation to the arena, the rats are exposed to two identical objects. After a delay, rats are exposed to a familiar and a new object. Rats typically spend more time exploring the new object than the familiar one.
Rats with hippocampal lesions exhibited impairments in the novel object recognition task [48,49]. The status of recognition memory with different delays has been examined in several inbred mice. Inbred strain differences have been observed in some studies but not others [27,50]. It was reported that DBA/2 mice performed worse than C57BL/6 mice in this task . However, it is difficult to compare between studies because there are numerous potential variations of this task, such as the delay interval, size of arena, and scoring methods used.
Rodents, including mice, tend to avoid the ingestion of novel foods. This initial neophobia is decreased with repeated exposure to the novel food. Our laboratory examined differences in retention of exposure to novel olfactory/gustatory information in two inbred mouse strains. Consumption levels of C57BL/6 mice were significantly increased compared to those of DBA/2 mice (Fig. 5). It could be interpreted that C57BL/6 mice remembered and recognized the novel food that they were previously exposed to better than DBA/2 mice.
Pavlovian conditioning is one of the most systematically studied subjects by psychologists . In order to reveal the cognitive phenotype of genetically modified mice, Pavlovian aversive conditioning, an association of CS with aversive US, has been extensively used, while Pavlovian appetitive conditioning with an appetitive US, such as food or water, is not prevalent. The principle of associative learning and its neural mechanisms have been studied and revealed mostly by Pavlovian appetitive conditioning using rats as subjects. A number of studies that have identified a specific cognitive function with Pavlovian appetitive conditioning have been introduced.
The traditional view of conditioning is that it is a passive and automatic process, but it is now clear that conditioning is affected by predictive and informational relationships between CS and US . The establishment of a predictive relationship between a CS and a motivationally significant US through Pavlovian conditioning can endow the CS with motivational and emotional power. For example, a neutral CS (e.g., light or tone) paired with food delivery acquires incentive motivation. Therefore, this CS reinforces later the Pavlovian conditioning or instrumental conditioning and modulates the performance of other learned or unlearned responses (e.g., lever pressing for food; eating food) [54,55]. Although the neural mechanisms and circuitry for cognitive function, including attention and reward expectancy, have been studied in rats and monkeys, mouse models are valuable for understanding the cellular and molecular mechanism of these cognitive functions. A mouse-based paradigm to examine Pavlovian influences on instrumental conditioning has been developed (Fig. 6) .
Genetically modified mice targeting molecular and cellular mechanisms underlying synaptic plasticity and learning and memory development have been used to reveal the involvement of genes in appetitive incentive learning. For example, mice with knock-in mutations in the phosphorylation sites of alpha-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA) receptor subunit GluR1 exhibited impairments of synaptic plasticity and memory retention . These phosphomutants also exhibited impairment in appetitive incentive learning [56,58]. Deletion of pentraxin, which is regulated by neuronal activity and co-clustered with AMPA receptors, affected the processing of sensory-specific incentive value . It has been reported that the BDNF receptor, TrkB, known to play a key role in neural development and plasticity, is critical for the acquisition of appetitive incentive learning . Motivational function, as measured using appetitive incentive learning, was impoverished in transgenic mice expressing a putative dominant-negative disrupted in schizophrenia 1 .
AD is a neurodegenerative disorder characterized by progressive decline of cognitive function . Mouse AD models are essential for understanding the basic underlying biology of AD and developing therapeutic drugs. Three mouse models of AD and three cognitive assays that are frequently used are summarized in Table 1 (also reviewed in [63,64]). Some AD studies using animals have used only female mice. This is due to reports showing that accumulated levels of amyloid beta and plaque in female mice were higher than those in male mice in some AD mouse models [65,66,67,68]. However, sex differences in AD pathology in AD mouse models is still controversial (reviewed in ). Furthermore, because the frequency and length of the estrous cycle changes with age , and neuroplasticity is affected by levels of estrogen , a measurement of cognitive status of AD mice should be carefully conducted [72,73].
The present review introduces interesting findings regarding the decline of cognitive flexibility in aged mice or AD mice using a behavioral task based on the modified behavioral protocol of a redundant place/cue version of the water maze task described earlier (Fig. 1). Twenty-three-month-old aged mice and 12-month-old adult mice were serially trained in cued and place versions of the water maze task and then underwent a strategy preference test (Fig. 2). Aged and adult mice showed no differences in cued and place training, but, on the preference test, the aged mice chose a cued strategy more frequently than the adult mice . We also examined the learning strategy preferences of 5XFAD mice using this same training protocol. No differences between 5XFAD and non-transgenic control mice were observed in cued and place training. However, in the strategy preference test, the 5XFAD mice preferred more often a cued strategy than control mice .
Numerous researchers have raised the question of how many behavioral tests are necessary and what tests are key in revealing the phenotype of their genetically modified mice. There is no single answer to this question. When using an experimental design with appropriate control groups and comparison conditions, the choice of a cognitive test should be optimized to address the investigator's hypotheses based on anatomical and neurophysiological results.
Behavioral and neurobiological findings regarding cognitive function have largely come from studies of mice, rats, and primates. However, researchers have tried to apply parameters and behavioral protocols used in studies with rats to mouse studies, without considering behavioral and physiological differences between rats and mice. As a result, although it takes a lot of time and effort, a study is sometimes not successful. Therefore, parameters and behavioral protocols appropriate for a mouse study might be useful for studying cognitive functions in AD transgenic mice, or in mice with conditional gene targeting using Cre-LoxP recombination.
In studies where the sex has not been specified under Age, both male and female mice were used or no information on sex was provided.