Combined change of behavioral traits for domestication and gene‐networks in mice selectively bred for active tameness

Abstract Tameness is a major element of animal domestication and involves two components: motivation to approach humans (active tameness) and reluctance to avoid humans (passive tameness). To understand the behavioral and genetic mechanisms of active tameness in mice, we had previously conducted selective breeding for long durations of contact and heading toward human hands in an active tameness test using a wild‐derived heterogeneous stock. Although the study showed a significant increase in contacting and heading with the 12th generation of breeding, the effect on other behavioral indices related to tameness and change of gene expression levels underlying selective breeding was unclear. Here, we analyzed nine tameness‐related traits at a later stage of selective breeding and analyzed how gene expression levels were changed by the selective breeding. We found that five traits, including contacting and heading, showed behavioral change in the selective groups comparing to the control through the generations. Furthermore, we conducted cluster analyses to evaluate the relationships among the nine traits and found that contacting and heading combined in an independent cluster in the selected groups, but not in the control groups. RNA‐Seq of hippocampal tissue revealed differential expression of 136 genes between the selection and control groups, while the pathway analysis identified the networks associated with these genes. These results suggest that active tameness was hidden in the control groups but became apparent in the selected populations by selective breeding, potentially driven by changes in gene expression networks.


| INTRODUCTION
Tameness is one of the major elements of animal domestication 1 with a critical role in its early stages. 2 Tameness has been evaluated using behavioral tests and genetic and/or gene expression analysis in several animal species, such as foxes (Vulpes vulpes), 3,4 rats (Rattus norvegicus), 5-8 mice (Mus musculus), [9][10][11] and red junglefowl (Gallus gallus). 12,13 These studies demonstrated that genetic factors influence tameness in animals.
Tameness may be divided into two components: the motivation to approach humans (active tameness) and the reluctance to avoid them (passive tameness). 9,14 In mice, active and passive tameness have been evaluated by using three behavioral tests (active tameness, passive tameness, and stay-on-hand tests). 9 Using these tests, we previously measured nine indices: heading, contacting, locomotion, and jumping in the active tameness test; heading, accepting, locomotion, and jumping in the passive tameness test; and staying time in the stay-on-hand test. A difference between wild and laboratory mouse strains was observed in passive tameness but not in active tameness. 9 These results implied that selective pressure for passive tameness, but not for active tameness, has occurred during mouse domestication. To entirely understand the mechanism of tameness, we need to analyze active tameness in mice.
Based on these results, we previously bred a novel wild-derived heterogeneous stock (WHS) 15 that was established by crossing eight wild strains, followed by selective breeding for active tameness. 10 At the 12th generation of breeding, we obtained selected groups that exhibited a higher level of active tameness than that of the nonselected groups. Using the selected and non-selected groups, we conducted genetic analysis and found two genetic loci on chromosome 11 that are associated with active tameness in the selected group.
However, it is not clear how the behavioral indices of active and passive tameness relate to each other in terms of behavioral and genetic aspects. In the previous study using data obtained from inbred strains, a relatively higher correlation coefficient was observed between contacting (an index quantifying active tameness) and accepting and staying (indexes quantifying passive tameness), where Pearson's correlation coefficients were 0.55 and 0.60, respectively, but did not achieve significance with the Bonferroni correction. 9 The study evaluated the correlation between the traits, but more comprehensive analyses were needed to clarify the relationships among traits related to tameness.
The link between the brain and domestication also poses a major question. The size of hippocampal regions in domesticated animals, such as sheep and pigs, is smaller compared with those in their wild species counterparts. 16,17 In foxes, animals selected for higher tameness showed up-regulation of adult neurogenesis in the hippocampus. 17 A 2-fold increase in gene expression for epigenetic regulation has also been reported in domesticated foxes compared with non-selected foxes. 18 The hippocampus is a brain region known to be affected by domestication and stress in several species, such as rats and foxes, 17,19 and varied gene expression levels were observed after stress in mice 20 and chicken. 21 Therefore, it is of interest to compare hippocampal gene expression levels and behavior in selected and non-selected animals.
In the course of selective breeding, the three tests (active tameness, passive tameness, and stay-on-hand) were conducted in each generation. Our previous paper reported on the selective breeding, 10 where we demonstrated heading and contacting in active tameness between generation 3 (G3) and G12, as selective breeding was conducted for these two indices; the first index was the contacting and the second was the heading. Given that further generations of selective breeding up to G16 increased the difference between selected and non-selected groups, conducting comprehensive analyses of behavioral indices for tameness as well as gene expression in the hippocampus would be valuable.
In this study we first analyzed the sex and generation effect for each behavioral index, then conducted analysis of the change of behavioral indices as well as cluster analyses based on correlation between all tameness-related indices. Finally, we performed RNA-seq analysis from mice brains to uncover the differences in gene expression and to identify the associated gene networks between the mouse populations that potentially affect tameness.

| Rearing condition and animal experiments
Mice were maintained in accordance with guidelines of the National Institute of Genetics (NIG) in Japan. All procedures were carried out with permission from the Committee for Animal Care and Use of the NIG (No. . All mice were bred and kept under specific-pathogenfree conditions at the NIG with food and water available ad libitum. The mice were kept in a temperature-controlled room (23 ± 2 C) under 12/12 hours light/dark cycle. All mice were weaned from their parents at 3 weeks of age and housed in same-sex groups with littermates in standard-size plastic cages containing wood chips until the tameness tests. When there was only one same-sex littermate, the mouse was individually housed after weaning. During cage exchange and behavioral tests, each mouse was gently caught by its tail using large tweezers covered with silicon tubing to reduce pain.

| Behavioral assay
The active tameness, passive tameness, and stay-on-hand tests were conducted to quantify nine indices (active tameness test: heading, locomotion, contacting, jumping; passive tameness test: heading, locomotion, accepting, jumping; stay-on hand test: staying). 9,11 The three tests were conducted during the light period using a test apparatus made from gray colored polyvinylchloride of 40 × 40 × 40 cm (O'Hara & Co. Ltd., Tokyo, Japan). The field was illuminated with 100 lux at the center, and video recorded using a digital camera, CX5 (Ricoh Company, Ltd., Tokyo, Japan). Each mouse was tested at 6 weeks of age, with the exception of some animals in generations 14-15, which were tested when they were up to 10 weeks old. The ages of all the animals tested for tameness are indicated in Table S1.
An active tameness test was established to measure an animal's active responses to a human's hand. After placing the mouse in the center of the test field, an experimenter then put their hand at the bottom of the field, with their fingers slightly moving, and measured the duration of heading, locomotion, contacting, and jumping for 1 minute. The passive tameness test was established to measure an animal's passive responses to a human's hand and was conducted immediately after the active tameness test. An experimenter put their hand at the bottom of the test field without moving the fingers during the 1 minute trial. The hand then chased the mouse slowly until touching the body of the mouse. If the mouse accepted being touched, the experimenter attempted to touch the mouse for as long as possible during the 1 minute trial. The durations of heading, accepting, and jumping were counted. The stay-on-hand test quantified behavioral responses to forced stimulation by a human hand and was conducted immediately after the passive tameness test. An experimenter picked up the mouse by the tail with large tweezers and placed the mouse on their hand; while the mouse stayed on the hand, the experiment's thumb stroked the back of the mouse softly at a frequency of once per second. The maximum time for staying was 10 seconds in each trial. If the mouse left the hand, the trial was stopped and repeated three times. The staying times were counted for each of three trials and their median was used as the staying time data. The number of mice used in this study are summarized in Table 1.
Behavioral assays in the different generations (generations 3-7, generations 8-12, and generations [13][14][15][16] were conducted by three independent experimenters who were trained to perform similar hand movements. To do this, a video observer checked the procedure of the tameness test until the experimenters showed similar movements.
Video observations were conducted by a single, third party individual.

| Wild-derived heterogeneous stocks
WHS of mice were previously established 10 and derived from eight wild mouse strains: BFM/2Ms, PGN2/Ms, HMI/Ms, BLG2/Ms, NJL/Ms, KJR/Ms, CHD/Ms, and MSM/Ms. As a result of large genetic differences among these strains, the WHS has higher genetic diversity compared with other existing heterogeneous stocks of mice. Details on the genetic characteristics of WHS are reported in a previous publication. 10 The eight wild strains were genetically mixed by crosses using the circular rotation rule followed by the random rotation rule from G3 to avoid intercross mating (Figure 1). At G2, pairs in a group were expanded from 8 to 16 and the group was also expanded into two groups, each of which consisted of 16 pairs. At the G3 generation, the genomes of all eight strains were mixed randomly in each mouse, so genetic diversity was expanded more extensively than in the eight founder strains. After G3, 16 pairs of crosses were made by randomly choosing the pairs from the 16 families to avoid intercrossing in each generation.

| Selective breeding
Using the WHS, selected groups and control groups without any selection were established and bred as described in the previous study. 10 Briefly, the selective breeding experiment was conducted using WHS mice to increase active tameness. The mice that exhibited the highest score for contacting in the active tameness test among five mice were used for breeding. In the case where the scores for contacting were identical in the two mice with the highest score, the mouse that showed a higher score for heading in the active tameness test was used for the breeding. These mice were then mated between different families.
We expanded the WHS stock into two groups, selected group named Ms:WHS-S1 (S1) and control group named Ms:WHS-C1 (C1) at G3. These two groups, S1 and C1, were each expanded into a further two groups, control group named Ms:WHS-C2 (C2) and selected group named Ms:WHS-S2 (S2), respectively, from G5 ( Figure 1). For the WHS, up to 80 mice of each sex were tested from each group (C1, C2, S1, and S2) at each generation.
In this study, we used the data for two behavioral indices (heading and contacting in the active tameness test, from G3 to G12), which was published in the previous study. 10 Data from the other seven indices for G3 to G12 and all nine indices for the following four generations (G12 to G16) are unique to this study.

| Evaluation for the effect of sex
The effect of sex in the nine tameness-related behavioral indices was evaluated using two-way analysis of variance (ANOVA). The effects of sex, generation, and/or sex × generation interaction were assessed in four groups of WHS mice. Multiple comparisons were corrected using the Bonferroni correction (nine behavioral traits, P < .006). We used the lm and ANOVA functions implemented in R version 3.3.2 for the analyses.

| Clustering analysis
The four groups of WHS mice were separated and the individual scores analyzed. In total, data collected from 437 mice at G16 were used for the analysis. The distance between clusters was calculated using the Ward method as implemented by the hclust function in R version 3.3.2. The significance was set at 5% for each clustering analysis using pvclust function in the R package pvclust v2.2.0.

| RNA-seq experiments
Ten mice from each of the four WHS groups at G16-G17 were evalu- Ingenuity pathway analysis (IPA; version 57662101, Ingenuity Systems, Redwood City, CA, USA) was used for the network analysis on 136 genes in the IPA database that were differently expressed between the selected and control groups, based on q-value <0.05 computed via sleuth analysis as described above. The dataset includes data on mice nervous system tissues whose confidence levels were "Experimentally Observed" or "predicted with high confidence." The network analysis was conducted using canonical pathway information.

| Functional characterization of differentially expressed genes in the human GWAS catalog database
We used the GWAS catalog database to identify the functions of the differentially expressed genes that were found in both mice and humans. 25 The GWAS catalog file (v1.0.2 -with added ontology annotations, GWAS Catalog study accession numbers and genotyping technology) was downloaded from https://www.ebi.ac.uk/gwas/ (1) C2 S1 S1 S1 S1 S1 3 | RESULTS

| Generation and sex effects in tamenessrelated indices
To examine how selective breeding affects the behavioral phenotype in mice, we analyzed changes in behavioral indices related to tameness with increasing generations ( Figure S1-S4, Table 1, Table S1).
There was no effect of sex in either the control or selection groups, except for locomotion in the active tameness test, accepting in the passive tameness test, and staying in the stay-on-hand test in the C1 group, and heading in active tameness test in the C2 group (Tables 2   and 3, Figure S1-S4). A generational effect was detected in five out of nine indices in the S1 group and six indices in the S2 group ( Figure 2, Table 3). There were significant generational effects on the six indices in each of the C1 and C2 groups, even though no index was selected. (Figure 2, Table 2). Interestingly, contacting in the active tameness test and heading in the passive tameness test decreased through the generations in the control groups, C1 and C2 ( Figure 2) as reported in the previous report. 10

| Behavioral differences between selected and control groups
Behavioral differences between selected and control groups were evaluated over the generations (Figures 2 and 3, Table 1). Given that the control groups were bred without any selection for tameness but showed changes of tameness-related indices as the generations proceeded, the difference between the selected and control groups may show a true effect of selection for active tameness (Figure 3). For the majority of the five indices (heading and contacting in the active tameness test, heading and accepting in the passive tameness test, and staying in the stay-on hand test), durations in the control groups were lower than those in the selection groups after G9 ( Figure 3). However, the control groups were higher than the selection groups in two indices (locomotion and jumping) after G10 in the passive tameness test (Figure 3).

| Identification of components of behavior in tameness tests
We hypothesized that the phenotypic association between each index related to tameness is high when a significant cluster is observed T A B L E 2 (Continued) among the indices. A behavioral/genetic base for the behavioral components may then be shared among the indices. To identify common behavioral factors associated with tameness, we conducted clustering analyses of nine behavioral indices for the two selected groups and two control groups. Two large clusters were found in the WHS control groups, C1 and C2 (Figure 4). The first cluster includes accepting and locomotion in the passive tameness test and locomotion in the active tameness test and the second cluster contains other indices. In the two selection groups, S1 and S2, in addition to the first cluster still present, another cluster including two major selection indices, contacting and heading in the active tameness test, was prominent (P < .05). The difference in the clusters between the control and selection groups may be because of more contacting and heading behavior in the active tameness test in the selected groups, so these two behaviors are more notable in the selected groups than in the control groups.

| Gene expression profiles
To analyze gene expression levels in the hippocampus under selective breeding, we conducted RNA-seq analysis using samples obtained from 10 animals from each group of C1, C2, S1, and S2.
The heatmap of all the 28,675 genes analyzed is shown in Figure S5.
No obvious difference in expression levels between the selection and control groups were observed in the heatmap. We found differential expression in 136 genes among the 15,982 genes assessed in C1, C2, S1, and S2 groups (Table S2, Figure S6). Because S1 was expanded from C2 and S2 from C1 (Figure 1), C1 is more genetically similar to S2 than S1, while C2 is more genetically similar to S1 than S2. Therefore, we also compared expression differences in two comparative pairs, C1 vs S2 ( Figure 5A) and C2 vs S1 ( Figure 5B). We found differential expression in 184 of 16,038 genes assessed in the C1 vs S2 groups (Table S3) and 494 of 16,322 genes assessed in the C2 vs S1 groups (Table S4). Compared with the control groups, 107 (C1 vs S2) and 194 (C2 vs S1) genes had lower expression in the selected groups, although expression of 77 (C1 vs S2) and 300 (C2 vs S1) genes was higher in the selected groups. We found that seven genes, Gm20498, Pmvk, Dusp18, Zfp738, Kcnt2, Slc8a3, and Gm16165, exhibited expression change in the same direction between groups in two pairs, C1 vs S2 and C2 vs S1 (Figures 5 and   6, Table 4).
We then conducted IPA analysis to identify the molecular pathways that were different between selected and control groups. The IPA analysis identified 14 networks that are associated with the differentially expressed genes (

| Searching GWAS catalog data
Searching through 197,711 records in the GWAS catalog data, we identified 26 records that were associated with the seven candidate genes (Table 6). SLC8A3 and PMVK have been associated with bipolar disorder and Parkinson disease, respectively. Multiple GWAS studies reported that KCNT2 and DUSP18 are associated with Alzheimer disease, and KCNT2 is also associated with chronotype (Table 6). These results indicate the important roles of these genes in brain function.

| DISCUSSION
We conducted selective breeding for active tameness in mice and successfully selected groups that exhibited a high level of active tameness, and thus a high motivation to approach a human's hand. In addition to the active tameness test, we also conducted two more behavioral tests, passive tameness and stay-on-hand tests, at each generation for each group. From these behavioral tests, we obtained data for nine behavioral indices. Sexual difference in several mouse behaviors has been reported previously. [26][27][28] In the present study, however, we found no clear sex effect in nine behavioral indices of tameness in the four WHS groups. Given that mice are externally not highly sexually dimorphic and tameness is not a behavior directly associated with sex, we believe these tameness-related behaviors did not exhibit a sex difference.
To clarify the relationship among the nine behavioral indices, we In addition, the groups that were selected for contacting showed higher heading values compared with control groups (Figures 2 and 3).
Furthermore, two other behavioral indices (heading in passive tameness and staying in the stay-on-hand test), which are located in the next clusters within the same branch, also showed higher values in the selected groups than the control groups. These results illustrated that contacting and heading in the active tameness test and the other two indices could partially share behavioral and/or genetic components. It is interesting that heading and contacting in the active tameness test are not clearly separated in the control groups but classified into independent clusters in the selected groups, S1 and S2. It is possible that selection for contacting makes the behavioral indices for active tameness more apparent after the selective breeding.
In the previous study, when we established the three behavioral tests, we intended to measure the active and passive tameness separately. 10 Given that two indices, contacting and heading, in the active tameness test and two indices, accepting and heading, in the passive tameness test, as well as staying in the stay-on-hand test, are separated in different clusters, our strategy of using three tameness tests that can measure active and passive tameness separately was validated.
As we previously reported, contacting in the active tameness test decreased in control groups as the generations proceeded. 11 32 Another report of RNA-Seq analysis from rats revealed several candidate genes including Gltscr2, Lgi4, Zfp40, and Slc17a7 associated with tameness and aggressiveness. 8 The present study uncovered differential gene expression in 136 genes and in F I G U R E 3 Phenotypic differences in nine behavioral indices between two groups. Differences in the durations of nine indices between all combinations of groups were evaluated over 16 generations. Blue cells indicate lower values in the group on the right side of the comparison.
Red cells indicate higher values in the group on the right side of the comparison expression networks between populations. We found differential expression in Dusp18 ( Figure 6, Table 4), which affects response to external stimuli. Hyperactivity and abnormal behavioral response to light was observed in mice mutated with a Dusp18 intragenic deletion (IMPC Database release 2014). For the expression network analysis, we found two gene networks that were clearly different between selected and non-selected groups. In the first network, we found SOX2 to be the fourth most connected molecule. It has been proposed that domestication of animals is strongly associated with diverse changes in behavioral, morphological, and physiological traits. This is called "domestication syndrome" and the mild deficit of neural crest cells might play a key role in the phenotypic change. 2 Sox2 is one of the genes having a critical role in the development of neural crest cells during developmental stage. 33 Although our RNA-seq analysis targeted gene expression in the hippocampus, it will be valuable to analyze gene expression in neural crest cells in the future. It is interesting that the first gene network consisted of neuropeptides B/W receptor 1, neuropeptide Y, and POMC (Table 5). POMC is expressed as a precursor of the adrenocorticotropic hormone, expressed in response to stress stimuli and also demonstrated to be less expressed in tamed foxes 32 ; therefore, a similar mechanism may be expected for tameness in both mice and foxes. In the second network, we identified oxytocin receptor, which plays a key role in social preference and recognition, suggesting an association between social behavior and tameness. [34][35][36] In addition, FOS, a product of immediate  T A B L E 5 Expression networks for differential gene expression  Note: Genes with significantly higher and lower expression in the selection group are indicated by upward and downward arrows, respectively, in parentheses.
sensitivity. Thus, further studies are needed to reveal how stress sensitivity affects the difference between tameness and non-tameness.
We conducted comparative analysis of seven DEGs commonly found in the two sets of RNA-seq analyses with the results in the previously reported GWAS. 25 Four of seven differentially expressed genes identified in human GWAS studies, were tentatively reported as loci associated with human neurological diseases: Slc8a3 for bipolar disorder, 41 Pmvk for Parkinson's disease, 42 Kcnt2 and Dusp18 for Alzheimer's disease, 43 and Kcnt2 for insomnia and chronotype traits. 44,45 These diseases are not directly related to tameness or domestication in animals; however, combined studies on human GWAS and DEGs between tamed and non-tamed mice indicate that these genes are playing important roles in maintaining a variety of normal functions in the brains.
We previously reported that two closely located genomic regions on chromosome 11, ATR1 and ATR2, were selected in the group of mouse that are selectively bred for tameness using the WHS 10 . We found nine functionally annotated genes, Wscd1, Xaf1, Atp2a3, Sez6, Sdf2, Aldoc, Vtn, Atp5g1, and Sp6k, to be differentially expressed between C1 and S2 groups in the ATR1 and ATR2 regions (Table S3). Also, two genes, Srr and Ramp2, were differentially expressed between C2 and S1 groups in these regions (Table S4). Thus, it will be important to understand how these genes affect gene expression levels and gene networks in future studies.
In summary, we conducted selective breeding for active tameness and successfully bred tame mice. At the same time, we identified a combined change in other behavioral traits, such as accepting, heading, locomotion, and jumping in passive tameness, and staying in the stay-onhand test. Our cluster analysis demonstrated that the behavioral component of active tameness was notable in the selected groups where contacting and heading in the active tameness test were selected. The results suggested that the behavioral component of active tameness was F I G U R E 7 The top two networks of expressed genes identified by IPA analysis. Information on network shapes, path designer shapes, and network relationship are shown in the box. Diagrams shown with green and red colors indicate down-and up-regulated genes with significant differences between groups, respectively. A, ID1 network that scored 13.
Thirty-five genes are involved and associate with behavior, neurological disease, and organismal injury and abnormality. B, ID2 network that scored 11. Thirty-five genes are involved and associated with cellular development, cellular growth and proliferation, and nervous system development and function hidden in other behaviors in the non-selected group of mice but became apparent in the selected groups. Behavioral components of active tameness may be influenced by other behaviors, such as anxiety, fear, sociability, and novelty-seeking behaviors. 9,10 In addition, we found changes in the gene expression pattern between the selected and non-selected groups. Further analyses will be required to quantify behaviors with the generally used behavioral tests to analyze the relationships between tameness and other behavioral aspects. Furthermore, it is necessary to characterize the neural cells that are activated during contact with human hands to understand the neurobiological mechanism underlying active tameness in mice.

ACKNOWLEDGMENTS
We thank Motoko Nihei for analyzing the video-recordings and Junko Kitayama for helping with RNA-seq analysis. This work was supported by