Foxp2-R552H mutant mice were bred from strains previously described (Groszer et al., 2008); heterozygous fertile males were paired with C57BL6/J females. The pairs were housed in regular plastic home cages at average temperature of 25°C and a 12-h light-dark cycle. Wood shaving served as bedding, water and food were available ad libitum. To avoid post mating in the postpartum estrus female and ensure a calm raising of the pups by the female, fathers were removed at the day of birth of the pups. Pups were sexed, tagged at weaning (postnatal day 22). A tail sample was also taken for genotyping purposes. Young males were then group housed blind to genotype.

Mice were genotyped by polymerase chain reaction (PCR) using Titanium Taq polymerase and restriction digestion of genomic DNA from tail samples. The following primers were used: Foxp2_Forward 5′-GTTCCTCTGGACATTTCAAC-3′ and Foxp2_Reverse 5′-TGTGAGCATGCCTTTAGCTG-3′. PCR conditions were as follows: 95°C for 3 min (1 cycle), 95°C for 30 s (13 cycles), 65°C (−0.5°C/cycle) for 30 s (13 cycles), 68°C for 45 s (13 cycles), 95°C for 30 s (25 cycles), 58°C for 30 s (25 cycles), 68°C for 40 s (25 cycles), 68°C for 7 min (1 cycle). The 603 bp PCR products were digested overnight at 37°C with HgaI which yields fragments of 372 and 231 bp for the wildtype allele, while the mutant R552H allele remains undigested.

Vocalizations were recorded from males when they were around 8–9 weeks old (young adults). For this first set of animals, a total of 19 adult male mice were recorded, 10 heterozygous, and 9 wildtypes. One wildtype animal did not sing at all in any condition and was thus removed from the study (10 het/8 WT). After an overnight experience with a sexually mature wildtype female, male mice were placed back in the same social housed cages (4/5 mice per cage) until the test day. The males were then removed from their cages, placed in a new cage and then singly habituated in the sound recording environment (15″ × 24″ × 12″ beach cooler with a tube for pumped air circulation input, no light and a hanging microphone, as a soundproof compartment (Arriaga et al., 2012; Chabout et al., in press) for 15 min. Although no recordings were made for this period, overall observation of the live audio recording on the computer monitor by Avisoft Recorder USG software showed no songs were emitted during the habituation. We then exposed the males to one of the four different social contexts to stimulate singing (Chabout et al., 2015): (1) Fresh female urine collected from at least two different wildtype females from two distinct grouped housed cages within minutes of exposure on a urine-dipped cotton tip placed inside the male's cage (UF); (2) awake and behaving adult wildtype female placed inside the cage (LF); (3) an anesthetized wildtype female placed on the metal lid of the cage (AF); and (4) an anesthetized adult wildtype male placed on the metal lid of the cage (AM). We modified our original abbreviations for these context descriptions (Chabout et al., 2015) to have a more consistent systematic naming: first characteristic of the context (U-urine; L-Live; A-Anesthetized) followed by sex of the stimulus animal (F-female; M-male). Exposure and recordings lasted for 5 min. The same mouse was exposed on 3 consecutive days to the same social context (either UF, LF, AF, or AM), but the identity of the stimulus (specific animal) was changed every day to ensure against a familiarity effect. Then the next week, the same mouse was exposed to a different stimulus following the same procedure. We repeated this for 4 consecutive weeks, where the order of stimulus was shuffled between weeks and genotypes such that each animal received a different stimulus, in order to normalize against any possible order effect. We tried to use females in pro-estrus or estrus (wide vaginal opening and pink surround) for the female stimuli when possible with the scheduled recordings. The anesthetized animals were anesthetized with ketamine/xylazine (100 and 10 mg/kg, respectively, intraperitoneally) and put on a heat pad outside of the test cage between recording sessions for at least 5 min. Between trials, the recording cages were cleaned with 1% Trifectant and water.

To replicate our key findings using a different population of Foxp2-R552H mice from the same founder line, a total of 31 mice were recorded, 16 heterozygous and 15 wildtype. Males were treated the same as above, except that, for litter delivery reasons all males were treated sequentially in the three contexts in the following order: UF, AF, and LF. In this second experiment, the timing of litter deliveries of different males on different days did not allow us to randomize the study with groups of the same age or a maximum of 1 week apart. We still managed to test equal numbers of heterozygous and wildtype on test days. The first experiment above was conducted in October and the replicate experiment in March.

Sounds were recorded with UltraSoundGate CM16/CMPA ultrasound microphones that were suspended over the center of each cage in the recording box, high enough so that the receiving angle of the microphone covered the whole area of the test cage. The microphones were connected to a multichannel ultrasound recording interface Ultrasound Gate 416H, which was plugged into a computer running Avisoft Recorder USG software v4.2.18 (Sampling frequency: 250 kHz, FFT-length: 1024 points; 16-bits). All recording hardware and software were from Avisoft Bioacoustics® (Berlin, Germany). Further, detail of the recording method is described in Chabout et al. (in press).

Following standard definitions as described in Arriaga and Jarvis (2013), we considered a sound note as the most basic acoustic unit, formed by a single continuous sound with or without variations in fundamental frequency. One or more notes can be combined to form a “call” or a “syllable,” as a single acoustic unit not separated by silence. We distinguish “calls” and “syllables” by the pattern of usage. Calls are typically produced in isolation or in short bursts, and are usually repeated single acoustic unit types. Syllables, however, derive their classification from being included in a longer series of rapidly produced vocalizations of varying types. We define a sequence as a succession of syllables spaced by short intervals, with each sequence separated by a longer interval (250 ms or more) of silence as described in Chabout et al. (2015) and the main text. Thus, a song is a sequence of syllables, often elaborate, delivered periodically and sometimes with rhythm. When pitched to the human hearing range, male USV sequences in the four social contexts are strikingly reminiscent of the songs of certain songbirds (Holy and Guo, 2005; Arriaga et al., 2012).

Acoustic waveforms were processed using a custom MATLAB program (Arriaga et al., 2012), originally modified from code written by Timothy E. Holy (Holy and Guo, 2005) that we call “Mouse Song Analyzer v1.3” and is available on our website (http://jarvislab.net/research/mouse-vocal-communication/). Briefly, the software computed the sonograms from each waveform (256 samples/block, half overlap), thresholded to eliminate the white noise component of the signal, and truncated for frequencies outside the USV song range (35–125 kHz). We used a criterion of 10 ms minimum of silence to separate two syllables and 3 ms as the minimum duration of a syllable. The identified syllables were then classified by presence or absence of instantaneous “pitch jumps” separating notes within a syllable into four categories: (1) simple syllables without any pitch jumps (type “s”); (2) complex syllables containing two notes separated by a single upward (type “u”) or (3) downward (type “d”) pitch jump; and (4) more complex syllables containing two or more multiple pitch jumps (type “m”; Figure 1B). Any sounds that the software could not classify were put into an “unclassified” category and removed from the analysis. Manual visual inspection of the sonograms of the unclassified sounds revealed that most of them were either syllables that overlapped with mechanical, non-vocal noises the mouse made, such as scratching, walking on the plastic cage, chewing on the cage lid etc., or non-vocal mechanical sounds that included frequencies that reached above our 25 kHz cut off. All analyses were conducted on a total of 10,720 classified syllables in the urine condition (UF), 19,193 syllables in the anesthetized female condition (AF), 41,209 in the live female condition (LF), and 1,293 in the anesthetized male condition (AM). Sonograms were analyzed and the following spectral features were calculated automatically by the Mouse Song Analyzer MATLAB code from the sonograms of each of the classified syllable types: Syllable duration, inter-syllable interval (ISI), mean (pitch), minimum, maximum, start, and end frequencies, frequency modulation, spectral purity, amplitude, and bandwidth. Spectral purity was calculated as the instantaneous maximum power at the peak frequency normalized by the instantaneous total power in the spectrum, averaged across the entire syllable; a pure tone has a spectral purity of 1, and white noise approaches 0. In the main text, we only report on five main features (e.g. mean frequency), as more minor features (e.g. end or start mean frequencies) did not reveal new information relative to the main features.

Following a method described in a previous study of ours (Chabout et al., 2015), we used our custom script generated in Microsoft Excel (2013) that detects silences (gap > 250 ms), and letter-coded sequences of syllables and silence (Chabout et al., 2015). These data were used to calculate the “conditional probabilities” of different syllable transition types for each mouse:

We then averaged the probabilities from all males within a group and contexts, to obtain conditional probabilities for the entire group. We graphed these group-context conditional probabilities into syntax diagrams using Graphviz v2.36 (http://www.graphviz.org/), with nodes designating different syllable categories or silence, and arrows the transitions between the syllables and silence. Arrow thickness in pixel size was made proportional to the conditional probability values. In the diagrams, we only include transitions that were produced by the mice with a probability higher than 0.05 to show the common transitions, and not rare events. The statistical analyses of syntax described below include all transitions recorded, even if rare, such as produced by one animal.

Using these conditional probabilities, we then investigated whether the transition dynamics, as characterized by these transition probabilities, varied significantly between the two genotypes, using two different modes of a novel statistical approach. The first allowed us to test for statistical differences in transition dynamics in the animals from the same genotype between two different contexts. The second allowed us to test the differences in transitions between two independent groups of animals from two different genotypes within the same context. This approach allowed us to test differences in transitions to and from different syllables, and provided additional insights into differences in individual transition types that made up these sets of probabilities. The procedure comprised two stages. In the first stage, we focused separately on each of the 24 transition types and tested whether the corresponding context-specific or genotype-specific distributions are different. We used non-parametric rank based tests, avoiding restrictive parametric assumptions on the transition probabilities. For the within genotype comparisons between contexts, we used paired sample Wilcoxon signed rank sum tests (WSR). For the between genotype comparisons within the same context, we used two sample Wilcoxon-Mann-Whitney (WMW) tests. In the second stage, we combined the p-values returned by these “local” tests to obtain test statistics and p-values for testing the differences in the transition probabilities to and from different syllables as well as for testing the differences in the global dynamics. We used the method of Zaykin et al. (2002), which is robust to the presence of a few outlying local p-values. The null distributions of the combined test statistics were determined using a permutation based Monte Carlo method that accounts for the correlation among the local p-values. The p-values for the combined tests were finally corrected for multiple combined tests corresponding to different syllables. We include in the supplement our R program generated scripts (which we called Syntax Decoder) for the syntax analyses. See Data Sheet 1 in Supplementary Material for additional details.

At ~20 weeks of age Oxygen (O₂) consumption and carbon dioxide (CO₂) production were measured in 8 C57 wildtype and 9 Foxp2-R552H heterozygous mice individually using CLAMS chambers from Columbus Instruments system (Columbus, OH). Measurements were recorded every 20 min over ~48 h. The respiratory exchange ratio (RER) was calculated as the ratio of CO₂ production (V_CO2) to O₂ consumption (V_O2) at any given time (Thupari et al., 2002). V_CO2 is the expired CO₂ volume at ml/kg·h and V_O2 is the expired O₂ volume in the same units.

After all their recording sessions, six heterozygous and six wildtype males were used to trace the connections between the laryngeal muscles and the primary motor cortex M1 following a procedure described in Arriaga et al. (2012, 2015). We used a recombinant strain of pseudorabies Bartha (PRV-Bartha) expressing enhanced Green Fluorescent Protein (eGFP) under the control of the histomegalovirus immediate-early gene promoter (Smith et al., 2000; Card and Enquist, 2001). Live virus was received from Dr. Lynn Enquist's laboratory at Princeton University at a titer of 9.55 × 10⁸ pfu/ml (Virus center grant #P40RR018604), aliquoted at 4 μl per tube, then stored at −80°C, and thawed immediately before injection. General anesthesia was induced with 1% isoflurane. A midline incision of ~1.5 cm was made under the hyoid bone. The skin, fat tissue and membranes were carefully separated to allow access to the deep muscles. We gently pulled back the sternohyoid muscle to expose the larynx and its muscles. A total of 1 μl was injected into the cricothyroid muscle at a rate of 0.05 μl per min using a Nanofil microsyringe system with a 34-gauge stainless steel needle (World Precision Instrument, Sarasota, FL). After 5 min, the micromanipulator was retracted, and the same injection was repeated for the cricoarythenoid lateralis muscle. Injections were made bilaterally in both muscles. A single puncture point was made for the injection to avoid any leakage outside the muscles and spreading to other tissues.

About 120 h after infection, when the virus is expected to infect 2nd order LMC neurons (Arriaga et al., 2012, 2015), mice were given an overdose of pentobarbital sodium and perfused transcardially with 0.1M PBS followed by 4% paraformaldehyde (PFA) in 0.1M PBS. Brains were removed, post fixed in 4% PFA overnight, then cryoprotected in 30% sucrose in 0.1M PBS until they sank at the bottom of the tube. Brains were then frozen in TissueTek® O.C.T. compound. Forty micrometer coronal sections were cut on a cryostat and put into 0.1M PBS. Forebrain (from +0.50 mm to −0.46 mm) sections were mounted directly on SuperFrost® Plus slides with Vectastain with DAPI (Vector Labs) to observe eGFP expression pattern. Pictures of the slides where taken either with Olympus DVX10 or Olympus BX61 (for high magnification). Then the number of positive layer V neurons in M1 per section was quantified manually and graphed in an Excel (2013) file.

To measure the expression profile of Foxp2 protein in these and adjacent neurons, we unmounted the sections with PRV positive cells in 0.1M PBS and stained them with a FOXP2 antibody (abcam 160046). Sections were washed 3 times in 0.1M PBS, then incubated in 0.1M PBS + 0.3% Triton (X100) + 10% NGS for 1 h at room temperature. Section were incubated overnight at 4°C with anti-FOXP2 at a 1:5000 dilution in 0.1M PBS + 3% BSA + 0.3% Triton (X100) + 10% NGS. After three washes in 0.1M PBS, a fluorescent secondary anti-Rabbit Cy3 anti-body was used at a 1:500 dilution in PBS 1X + 3% BSA + 0.3% Triton (X100) +10% NGS for 1.5 h at room temperature. Washed sections were mounted with DAPI medium (Vectashield) and coverslipped.

Statistical analyses were conducted using either IBM SPSS Statistic software (v.22.0) or R (R Development Core Team, 2011). Two-way repeated measures ANOVA or MANOVA were used to compare male subject performances across genotypes, stimuli, or across syllable types. For the Repeated-measure ANOVA, when the assumption of sphericity was violated (Mauchly's test) we reported the corrected degrees of freedom using Greenhouse-Geisser correction. Post-hoc analyses were performed using WMW tests for independent variables. Student's paired t-test comparisons were used for dependent variables when appropriate.

Since the Foxp2-R552H mutation was backcrossed on a different wildtype background (C57BL6/J, hereafter called C57) than the strain used for our previous study (B6D2F1/J, hereafter called B6; Chabout et al., 2015), we first checked whether the C57 wildtype also showed social context differences. Although we did not find social context differences in acoustic features of C57 adult male USV song syllables (Figure S1), they produced higher rates of syllables in the presence of a live female (LF; Figure 2A [unlike B6 males which had highest rates for fresh female urine UF (Chabout et al., 2015)]. Like B6 males (Chabout et al., 2015), C57 males produced intermediate rates in the presence of an anesthetized female (AF) and very few or no syllables in the presence of an anesthetized male (AM; Figure 2A).

Relative to C57 wildtypes, the Foxp2 heterozygous C57 male littermates did not differ in any acoustic features measured (Figure S1 white vs. black bars). However, in female-associated contexts (UF, LF, and AF), Foxp2 heterozygotes had a non-significant trend for lower syllable production rates (Figure 2A), which was related to an interaction with sequence length, described later in this study. These adult findings are consistent with a previous study on pup calls (Gaub et al., 2010), which found no differences in syllable acoustic structure or production rate in Foxp2-R552H heterozygotes.

Relative to B6 males in our previous study (Chabout et al., 2015), C57 wildtype males in the current study produced fewer differences in syllable repertoire composition across context, where only the down “d” pitch jump syllable type was produced proportionally more in the presence of female-associated stimuli (UF, LF, and AF) compared to anesthetized males (AM; Figure 2B). Foxp2 heterozygous males lost the “d” syllable context-dependent difference, and also produced proportionally less complex multiple “m” pitch-jump syllables in the anesthetized male context (Figure 2B). Despite these within-genotype effects, differences were not detected when comparing between genotypes. These findings suggest subtle differences in context-dependent syllable repertoire composition in Foxp2 heterozygotes, which appear to affect production of more complex syllables.

To analyze syllable syntax (i.e., sequencing), we used our previous approach of defining a song-bout sequence based on automated quantification of Inter-Syllable Intervals (ISI; Chabout et al., 2015). Similar to B6 males (Chabout et al., 2015), C57 males had several peaks in ISI distribution, with the shortest two [short interval (SI) and medium interval (MI)] corresponding to silences between syllables within a bout, and a longer interval (LI) of 250 ms or more (2 times the S.D. of the central peak) corresponding to separating sequences (Figures 3A,B). There were no overt differences in ISI peak timing between wildtypes and Foxp2 heterozygotes within or across contexts (Figure 3A).

We next measured the ratio of complex sequences (containing at least two occurrences of the complex syllable type “m”) vs. simple sequences (containing one or no “m”) in the different contexts, and found that in contrast to B6 males in the female urine context (Chabout et al., 2015), wildtype C57 males produced a >3-fold increase in sequences with complex “m” syllables specifically in the live female context (LF; Figures 4A,C,D). Foxp2 heterozygotes lost this context-dependent increase (Figures 4A,C,D). We know that females (at least B6) prefer to listen to these more complex pitch jump syllable sequences (Chabout et al., 2015).

Foxp2 heterozygous males also produced shorter sequences (i.e., a lower number of syllables per sequence) than their wildtype littermates in the female associated contexts (UF, LF, AF; Figure 4B). Additionally, there was a positive correlation between syllable sequence length (Figure 4B) and production rate (Figure 2A) in all contexts involving the presence of another animal (LF, AF, and AM; Figure 5). However, only in the live female context was there a difference in the correlations (slopes) between genotypes, where the heterozygotes produced both proportionally shorter sequences and lower syllable rates (Figure 5C).

The above findings led us to investigate whether there were differences in internal song sequence structure of heterozygous animals. We calculated the conditional probabilities of different transition types (i.e., with fixed starting syllables; Figure S2) and generated graphical syntax diagrams (Chabout et al., 2015) (Figures 6A–C; for common transitions with an occurrence >0.05 probability; red lines in Figure S2). Similar to B6 male mice (Chabout et al., 2015), C57 wildtypes in all contexts typically started a sequence with the “s” syllable type, followed by either looping with the “s” type or transitioning to the “d” and then to other syllable types (Figures 6A–C). At this probability cut-off level, the “s,” “d,” and “m” types were repeated in loops, whereas the “u” type was not. However, instead of producing greater syntax diversity in the female urine context as previously found in B6 males (Chabout et al., 2015), C57 males produced greater syntax diversity in the live female context, also involving transitions with “m” type syllables (Figures 6A–C). The Foxp2 heterozygotes produced all the same syllable transition types as the C57 wildtype in the urine and anesthetized female contexts, but they did not switch to the more diverse syntax in the live female context (Figure 6B). Instead, syntax of heterozygous animals in the presence of live females was more similar to the socially-reduced contexts (urine only or anesthetized females). There also appeared to be differences in relative proportions of transition types between wildtypes and heterozygotes under different social contexts (Figures 6A–C; differences in syllable transition probabilities, represented by arrow thickness).

To determine whether these syntax findings are statistically different, we could not use our previous approach (Chabout et al., 2015) as it was only sufficient for comparing differences within the same animals from one condition (e.g., context) to another. Thus, we developed a new approach based on Markov chain frameworks, Wilcoxon-Mann-Whitney rank sum tests, and Monte Carlo permutations, to test whether the syllable transition dynamics varied significantly between two groups of animals (two sample test), i.e., wildtypes and heterozygotes, within contexts, as well as between different contexts (paired test) within genotypes (see Section Methods and Data Sheet 1 in Supplementary Material). We tested for statistical differences at three levels: globally for the entire syntax network; for all transitions to and from a particular syllable type; and for individual transitions between two specific syllable types. In the to syllable case, we asked: when starting with different syllable types (“d,” “m,” “s,” “u,” or silence), do the probabilities of transitioning to a particular specified syllable type (say “d”) differ between the two groups of animals? In the from syllable case, we asked: when starting with a particular specified syllable type (say “m”), do the probabilities of transitioning to different syllables (“d,” “m,” “s,” “u,” and silence) differ between the two groups? These analyses included all transitions, whether they were produced at < 0.05 occurrence.

In the pairwise analyses with genotypes fixed, consistent with the syntax graphs, C57 wildtypes had global statistically significant syntax differences between contexts (e.g., AF vs. UF and AF vs. FE), whereas Foxp2 heterozygotes did not (Table 1A). When examining transitions to (top row) and from (last column) different syllable types, relative to C57 wildtypes, Foxp2 heterozygotes had weaker differences in transition probabilities (greener colors/higher p-values), particularly in the anesthetized female context (Figure 7). These global to and from transition differences were due to differences in specific syllable transition types in the heterozygotes across contexts compared to wild type (Figure 7; greener colors in inner cells of heatmaps for heterozygotes).

In the two sample analyses directly comparing genotypes, consistent with the pairwise analyses, the heterozygotes had global statistically significant syntax differences with wildtype, in the urine and live female contexts (Table 1B; Figures 6D,E). This was in part due to relatively stronger differences in transition probabilities to (top row) silence in the urine and live female contexts (Figures 6D,E); these relative differences survive a Benjamini and Hochberg false-discovery-rate correction for multiple testing (Table S1; Figure S3). When examining the specific transitions that contributed to these differences (inner cells of the heat maps in Figures 6D–F; Figure S2 for direction of the changes), the strongest differences between heterozygous and wildtype animals were in the transitions from all syllables to silence in the live female context (Figure 6F and Figure S2B), from all syllables except “u” to silence in the female urine context (Figure 6D and Figure S2A), and mainly from “d” to silence in the anesthetized female context (Figure 6E and Figure S2C). These differences can be explained by two types of global transition changes: (1) heterozygous mice producing more transitions to silence in all contexts (thicker arrows for heterozygotes in Figures 6A–C), consistent with heterozygotes producing shorter sequences; and (2) heterozygous mice producing decreased transitions from other syllables to “d” (Figures 6A,D) and from “m” syllables to other syllables (Figures 6C,F), consistent with the analyses within genotype.

Taken together, the above findings indicate that compared to wildtypes, heterozygous males produced shorter sequences in most contexts (due mainly to transitioning to silence more often from specific syllable types), had reduced internal sequencing with more acoustically complex syllable types, and did not increase syntax diversity with live females.

The strain and genotype differences in the results above, subtle in some cases and large in others, along with the variable conclusions in different studies on pup calls (Shu et al., 2005; French et al., 2007; Groszer et al., 2008; Gaub et al., 2010; French and Fisher, 2014) led us to seek replication of the key findings in an independent set of 31 animals (16 heterozygous and 15 wildtype males) and at a different time of the year (Fall/October instead of Spring/March). The only other difference from the first set of experiments was that we performed analyses across contexts in a sequential fashion (UF, AF, and FE order) instead of a randomized design (see Section Methods).

Overall, the results were consistent with the first set of experiments: no significant differences between genotypes in the acoustic features measured (Figure S4); switching to a higher ratio of complex-vs.-simple sequences in wildtype and absence of such switching in heterozygotes in the live female context (Figure S5A); and shorter sequences in heterozygotes (Figure S5B). However, we noted some differences compared to the first experiments: a higher rate of singing in the anesthetized instead of live female context for both genotypes (Figure S5C vs. Figure 2); a higher rate of singing for the heterozygous animals in the urine context (Figure S5C), even though they had shorter sequences (Figure S5B); increased use of “m” syllables in the repertoires of both genotypes (Figure S5D vs. Figure 4), and differences in the exact transitions that differed between heterozygotes and wildtypes in each context (Figure 6 vs. Figure S6). Such variability between experiments could be due either to the random vs. sequential experimental design, a motivation to sing more complex courtship songs in the fall vs. the spring, or some other unmeasured variable. The findings, however, remain consistent with our main conclusions that FoxP2 heterozygotes produce less complex and shorter syllable sequences relative to wildtypes under the same conditions.

The production of shorter USV sequences in heterozygotes led us to wonder if this could be due to shortness of breath compared to wildtype. We examined the consumption/production rates of oxygen and carbon dioxide (V_O2/V_CO2) in all mice in a 48-h period using isolated CLAMS (Comprehensive Lab Animal Monitoring System) chambers. Although, we found some large differences among some animals, there was there was no significant difference between genotypes [repeated measures ANOVA: V_O2: F_{(1, 13)} = 1.279, p = 0.27; V_CO2: F_{(1, 13)} = 0.544, p = 0.47; RER: F_{(1, 13)} = 3.83, p = 0.072; Figure S7]. These findings indicate no detectable deficits in respiration in heterozygotes that could explain their production of shorter sequences.

It has been proposed that human FOXP2 may contribute to speech acquisition and production through effects on vocal motor pathways of the cortex and basal ganglia, as the human LMC region and parts of the anterior striatum both show altered activation in human KE family FOXP2 heterozygotes during speech/language-related tasks (Liégeois et al., 2003). The recently discovered mouse rudimentary LMC region that projects to the anterior striatum and to nucleus ambiguous (Amb) brainstem vocal motor neurons (Figure 1A; Arriaga et al., 2012) is within the same coordinate region where Foxp2 is conspicuously expressed in layer-5 neurons of M1 compared to other parts of M1 (Hisaoka et al., 2010; Pfenning et al., 2014). This prompted us to ask whether these LMC layer-5 neurons have any change in connectivity or other properties in heterozygous mice.

Using our previous approach (Arriaga et al., 2012, 2015), we injected laryngeal muscles with a pseudorabies virus that travels retrogradely and transynaptically through functional synapses, and confirmed the presence of M1 LMC layer-5 neurons in C57 male mice (Figure 8A). Double-labeling experiments confirmed that these cells were located in the same region of M1 that has Foxp2-expressing layer-5 neurons, but that specific laryngeal connected layer-5 neurons expressed less Foxp2 (Figure 8B); this difference of less Foxp2 expression could be due to real differences in Foxp2 expression in laryngeal connected layer-5 neurons or toxicity to the neurons from the pseudorabies virus. The Foxp2 heterozygotes had these same laryngeal connected layer-5 cells, with no significant difference in the total number of labeled cells (Figure 8C). However, heterozygous mice showed a significant posterior shift and a more shallow peak in the distribution of LMC layer-5 neurons compared to wildtypes, resulting in the heterozygous LMC layer-5 neurons being more spread out in the cortex (Figure 8D). We therefore conclude that the heterozygous Foxp2 mutation did not change the presence, number, or gross connectivity of these laryngeal premotor neurons, but did alter their relative localization in the cortex. Future studies will be required to determine if there is a causal relationship between the change in distribution of these cells and the alterations in USV sequencing in the heterozygous animals.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.