USV in the male–female interaction test
In this study, we recorded USVs emitted by male mice in response to MCH females (see Materials and Methods
). A single female mouse was introduced into the cage in which a male mouse was formerly placed, and the emitted USVs recorded. It has been reported that female mice do not emit USVs during interaction with males. However, we needed to confirm this finding under our experimental conditions. To examine whether the female mice emitted USVs, male B6 mice, which emit USVs frequently, were devocalized and then introduced into the same cage with an intact female mouse. We found that no USVs were detected from the females during interaction with devocalized male mice (). In contrast, when sham-operated male mice were introduced to the female, frequent USV was detected (). These results confirmed that male mice emitted USVs during male–female interaction but females did not emit USVs under our experimental conditions. Therefore, it can be assumed that the following USV data obtained in this study were emitted by males.
Effect of male devocalization on USV during male–female interaction.
Characterization of USV patterns in 13 inbred mouse strains
To study USVs emitted by males during male–female interaction, we used males from 13 inbred strains: 10 wild-derived strains, one strain derived from fancy mice, and two laboratory strains. For the female counterparts, we used MCH mice, which were produced as hybrids of four different inbred strains that originated in the ICR outbred colony. As a consequence, the MCH females displayed genetic heterogeneity among individuals, and the USVs emitted from male mice were assumed not to be specific to a certain female strain. We found that most of the wild-derived mice did not emit USVs. In particular, for PGN2, CAST/Ei, HMI, and NJL mice, USVs were emitted in fewer than 40 % of the pairs of mice we examined (). Even if these strains emitted USVs, the number of calls was very small. In contrast, the laboratory mouse strains (B6 and BALB/c) emitted USVs in all trials. The frequency of emission in the trials showed a strain effect (chi-square test, p<0.05). In addition, call latency, i.e. the time from an encounter with a female to emission of the first call, showed a significant effect of strain (p<0.001) on one-way analysis of variance (ANOVA), and was generally longer in wild-derived mouse strains than in laboratory mouse strains ().
A number of mouse strains showed a characteristic pattern of USVs (). To characterize the pattern of USVs in more detail, we used sound data from mice who emitted more than 10 calls. Given that NJL, HMI, PGN2, and CAST/Ei mice did not emit a sufficient number of USVs for subsequent analyses, these strains were not characterized further. For the remaining nine strains, we calculated the following parameters: frequency (start point, mid point, end point, maximum point, and minimum point) and call duration. The results are summarized in . The frequency and duration of the USVs showed a significant effect of strain on ANOVA, but call latency and number of calls were not significantly different among strains. Among the nine inbred mouse strains, BALB/c mice displayed the lowest frequency and longest duration for the USVs, whereas BLG2 mice showed the shortest duration and highest frequency.
Analysis of each category of waveform.
Differences in ultrasonic vocalization among male mice of nine inbred strains.
Strain differences in the waveform composition of USVs
It has been reported that the USVs of mice are complexes of different sonographic components (waveforms) 
. To characterize the structures of the USVs that were emitted by the strains investigated, we categorized the waveforms into nine types (Table S1
), which is consistent with the results described in a recent report 
, with minor modification. The frequency of each waveform was calculated for each animal and analyzed in detail to investigate the characteristics of each strain. The percentage compositions of the waveforms are shown in . The waveform categories showed a significant effect of strain on MANOVA (p<0.0001). Moreover, the strain effect for the frequency with which each waveform category was emitted was tested by ANOVA. Upward, Downward, Jump, Short, and A-type waveforms all showed a significant effect of strain (p<0.01, 0.01, 0.05, 0.001, 0.0001, respectively). BALB/c mice showed a high percentage of A-type waveforms 
, whereas B6 and BLG2 mice showed a high percentage of Short-type calls. The main characteristics of the waveform compositions in CHD, JF1, KJR, and MSM mice appeared to be similar. To evaluate the similarity among strains, we performed cluster analysis on the basis of the waveform patterns (). On this basis, the strains were clustered into three groups by the Ward method. The classification of the strains into these three groups did not reflect their genetic relationships. The pattern of USVs was as different among closely related strains as among genetically remote strains. Strain differences in these waveform patterns have been reported previously 
. In addition to the strain differences with respect to waveform patterns, the duration and frequency of USVs showed significant strain effects (Table S2
Principal component analysis (PCA) of ultrasonic vocalization
We performed PCA to simplify the differences among strains, using variables that had shown a significant strain effect in the ANOVA. In the analysis, highly correlated variables (R>0.9) were combined into one variable. More than 90% of the variance in the data was explained by principal components (PC) 1 to 5 (PC1–PC5) (). We interpreted the component that showed more than 5% proportion of variance. For PC1, the frequency at each point and duration of each waveform showed high factor loadings. Frequency and duration were negatively correlated; thus, a high score for PC1 indicated USV of high frequency with short duration. For PC2, call duration and the maximum frequency of the Flat waveform were positively correlated, but the minimum frequency was negatively correlated. For PC3, the percentage composition of Jump and the maximum frequency of U-type waveforms displayed high factor loadings. PC4 indicated duration until the maximum or minimum peak for waveforms of the Short, A-type, and Jump type. For PC5, the slope of the Downward waveform showed high factor loadings. In summary, major differences among the mouse strains occurred with respect to frequency and duration, as well as for Flat, U-type, and Downward waveforms. To clarify the differences among the nine inbred strains, we have presented the standardized scores for the five components in each strain on radar charts (). The charts represent the character of the USV pattern in each strain. The BALB/c mice displayed a high score for PC1, which indicated calls of lower frequency and longer duration. The KJR mice displayed high scores for PC2–4 as compared with other strains. CHD and JF1 mice, and BLG2 and SWN mice, displayed similar USV patterns to each other, but the USV patterns of the BFM/2 and MSM mice were unique among the nine strains.
Proportion of variance for the principal components.
Radar charts representing the characters of the USV pattern in each strain.
Analysis of behavior during male–female interactions
In the second stage of this study, we investigated the role of the differences in USV patterns with respect to male–female social interaction. ANOVA revealed no significant effect of strain on any of the behavioral components including the positive behavior of females toward males (genital sniffing and grooming; Figure S1
). Thus, these social behavioral components were not significantly different among strains. ANOVA only revealed a significant strain effect (p<0.01) for clicks made by female mice (). A click made by a female is an audible call and is thought to be an expression of a negative reaction of the female to a male mouse 
. Indeed, we analyzed the correlation between clicks from females and kicking of males by females in all the video data, and found a positive correlation (R
0.6868, p<0.001) (). This finding supports the idea that the click indicates a negative reaction of the female to a male. Thus, we used click, a negative reaction, as an opposite index of female preference for males. The KJR strain, which displayed high scores in PC2–4, triggered the fewest female clicks. Also, the PC scores PC2, PC3, and PC4 showed a weak trend for correlation with the number of female clicks (R
−0.316, −0.272, −0.278, respectively). Therefore, the PC2–4-related waveforms emitted by KJR males may have a positive effect on male–female interactions.
Figure 5 A. Number of female clicks during male–female interaction. Female clicks showed a strain effect employing ANOVA (p<0.01). Data indicate mean ± standard error (n=3). B. Correlation between females that kicked and (more ...)
Response of female mice to the playback of USVs of selected waveforms
To examine whether the pattern of USVs could be correlated with any sign of preference or aversion in females, playback experiments using two USV files that had been created by cutting and pasting were conducted. We extracted waveforms on the basis of the factor loadings for PC2–4 from USVs emitted by KJR mice and created a sound file that was named "HIGH2-4". As a negative control, we produced another sound file, "LOW2-4", which was created by extracting waveforms in PC2–4 from the strains with the lowest scores for these components (see Materials and Methods
). As a consequence, HIGH2-4 included Flat waveforms with gradually increasing frequency, U-type waveforms that were similar to the shape of a reversed letter ”J„, and Jump waveforms that jumped downwards (). The LOW2-4 file included Flat waveforms with gradually decreasing frequency, U-type waveforms that were similar to the shape of the letter ”J„, and Jump waveforms that jumped upwards (). In the playback experiment, we used two speakers (
, ) to play the different files to evaluate whether the female mice preferred HIGH2-4 and avoided LOW2-4.
Preference of female mice for male USVs as investigated by the two-speaker method.
In the first experiment, HIGH2-4 and LOW2-4 were played at the same time, one in each of the two speaker zones, to allow the mouse to choose one of the USV patterns (). We measured two parameters: number of entries into the sound zones () and duration of contact with the mesh of the speaker (). The number of entries did not show a significant difference between HIGH2-4 and LOW2-4 (). However, the duration of contact with the mesh of the speaker did display a significant difference, and females clearly preferred HIGH2-4 ().
In the second experiment, we tested the preference of female mice for white noise against HIGH2-4, or white noise against LOW2-4 (). In the case of white noise and HIGH2-4, female mice showed a significantly longer duration of contact with the HIGH2-4 speaker than with the white noise speaker (). In contrast, a significant difference was not observed between LOW2-4 and white noise (). These results showed that female mice prefer HIGH2-4 to white noise, but not LOW2-4. The LOW2-4 USV had a similar effect to white noise and therefore might not have a strong aversive effect on females. These results highlight the importance of differences in the waveforms of USVs with respect to the preference of females for males of certain strains.