Skip to main content
eLife logoLink to eLife
. 2020 Dec 29;9:e63493. doi: 10.7554/eLife.63493

Circuit and synaptic organization of forebrain-to-midbrain pathways that promote and suppress vocalization

Valerie Michael 1, Jack Goffinet 1, John Pearson 1,2, Fan Wang 1, Katherine Tschida 3,, Richard Mooney 1,
Editors: Catherine Emily Carr4, Catherine Dulac5
PMCID: PMC7793624  PMID: 33372655

Abstract

Animals vocalize only in certain behavioral contexts, but the circuits and synapses through which forebrain neurons trigger or suppress vocalization remain unknown. Here, we used transsynaptic tracing to identify two populations of inhibitory neurons that lie upstream of neurons in the periaqueductal gray (PAG) that gate the production of ultrasonic vocalizations (USVs) in mice (i.e. PAG-USV neurons). Activating PAG-projecting neurons in the preoptic area of the hypothalamus (POAPAG neurons) elicited USV production in the absence of social cues. In contrast, activating PAG-projecting neurons in the central-medial boundary zone of the amygdala (AmgC/M-PAG neurons) transiently suppressed USV production without disrupting non-vocal social behavior. Optogenetics-assisted circuit mapping in brain slices revealed that POAPAG neurons directly inhibit PAG interneurons, which in turn inhibit PAG-USV neurons, whereas AmgC/M-PAG neurons directly inhibit PAG-USV neurons. These experiments identify two major forebrain inputs to the PAG that trigger and suppress vocalization, respectively, while also establishing the synaptic mechanisms through which these neurons exert opposing behavioral effects.

Research organism: Mouse

Introduction

The decision to vocalize is often a matter of life and death, as vocalizations are an important medium for sexual and social signaling between conspecifics but may also inadvertently advertise the caller’s location to eavesdropping predators. Consequently, many factors influence the decision to vocalize, including the presence of external sensory and social cues, as well as the animal’s own internal state and past experience. Work from the last five decades has established the midbrain periaqueductal gray (PAG) as an obligatory gate for the production of vocalizations in all mammals (Fenzl and Schuller, 2002; Jürgens, 1994; Jürgens, 2002; Jürgens, 2009; Subramanian, et al., 2020; Sugiyama et al., 2010; Tschida et al., 2019), and it is thought that forebrain inputs to the PAG regulate the production of vocalizations in a context-dependent fashion. In line with this idea, forebrain regions including the cortex, amygdala, and hypothalamus have been implicated in regulating vocalization as a function of social context (Bennett et al., 2019; Dujardin and Jürgens, 2006; Gao et al., 2019; Green et al., 2018; Jürgens, 1982; Jürgens, 2002; Kyuhou and Gemba, 1998; Ma and Kanwal, 2014; Manteuffel et al., 2007). Notably, although electrical or pharmacological activation of various forebrain regions can elicit vocalizations (Jürgens, 2009; Jürgens and Ploog, 1970; Jürgens and Richter, 1986), these effects depend on an intact PAG (Jürgens and Pratt, 1979; Lu and Jürgens, 1993; Siebert and Jürgens, 2003), suggesting that the PAG acts as an essential hub for descending forebrain control of vocalization. Despite the centrality of the PAG to vocalization, the synaptic mechanisms through which forebrain neurons interact with the PAG vocal gating circuit to either promote or suppress vocalization remain unknown.

A major challenge to understanding the synaptic mechanisms through which descending forebrain neurons influence vocalization was that, until recently, the identity of the PAG neurons that play an obligatory role in vocal gating remained unknown. The PAG is a functionally heterogeneous structure important to many survival behaviors (Bandler and Shipley, 1994; Carrive, 1993; Evans et al., 2018; Holstege, 2014; Tovote et al., 2016), thus hindering the identification of vocalization-related PAG neurons and forebrain inputs to these neurons that might influence vocalization. To overcome this challenge, we recently used an intersectional activity-dependent genetic tagging technique to identify neurons in the PAG of the mouse that gate the production of ultrasonic vocalizations (USVs; i.e. PAG-USV neurons Tschida et al., 2019), which mice produce in a variety of social contexts (Chabout et al., 2015; Holy and Guo, 2005; Maggio and Whitney, 1985; Neunuebel et al., 2015; Nyby, 1979; Portfors and Perkel, 2014; Whitney et al., 1974). The identification of PAG-USV neurons opens the door to identifying their monosynaptic inputs and to understanding how these afferent synapses modulate neural activity within the PAG vocal gating circuit to influence vocal behavior.

Here, we combined intersectional methods and transsynaptic tracing to identify neurons that provide monosynaptic input to PAG-USV neurons and to local PAG inhibitory interneurons. Using this transsynaptic tracing as an entry point, we then identified inhibitory neurons in both the hypothalamus and the amygdala that provide synaptic input to the PAG vocal gating circuit. In male and female mice, we found that optogenetic stimulation of hypothalamic afferents to the PAG-USV circuit promoted USV production in the absence of any social cues, whereas similar stimulation of amygdalar afferents to the PAG-USV circuit in males suppressed spontaneous USV production elicited by social encounters with females. Lastly, we used optogenetic-assisted circuit mapping in brain slices to deduce the synaptic mechanisms through which these forebrain afferents act on PAG-USV neurons and PAG interneurons to exert their opposing effects on USV production. This study provides the first functional description of the synaptic logic that governs the decision to vocalize, a behavior fundamental to communication and survival.

Results

Inhibitory neurons in the hypothalamus and amygdala provide input to the PAG vocal gating circuit

To identify forebrain neurons that provide input to the PAG vocal gating circuit, we performed transsynaptic tracing from PAG-USV neurons, which are primarily glutamatergic and reside in the caudolateral PAG (Tschida et al., 2019). Briefly, to label inputs to PAG-USV neurons, we used an activity-dependent labeling strategy (CANE; see Materials and methods) to express Cre-dependent helper viruses in PAG-USV neurons (Rodriguez et al., 2017; Sakurai et al., 2016; Tschida et al., 2019). A pseudotyped replication-deficient rabies virus was subsequently injected into the caudolateral PAG, allowing selective transsynaptic labeling of direct inputs to PAG-USV neurons (Figure 1A, see Materials and methods). Due to the difficulty in eliciting robust and reliable USVs from female mice, transsynaptic tracing from CANE-tagged PAG-USV neurons was performed only in male mice. Because the activity of many glutamatergic PAG neurons is shaped by potent inhibition from GABAergic PAG neurons (Tovote et al., 2016), we also performed transsynaptic tracing from local GABAergic neurons in the caudolateral and ventrolateral PAG that likely provide inhibition onto PAG-USV neurons, an idea we confirmed in a later section of the Results. To label direct inputs to local GABAergic PAG interneurons, Cre-dependent helper viruses were injected into the caudo/ventrolateral PAG of a VGAT-Cre mouse, and a pseudotyped replication-deficient rabies virus was subsequently injected at the same site to enable transsynaptic tracing from these cells (Figure 1B; N = 4 males, N = 2 females). These rabies tracing experiments revealed monosynaptic inputs to PAG-USV neurons as well as to local GABAergic PAG neurons from a variety of forebrain areas (Figure 1—source data 1, Figure 1—figure supplements 14). We subsequently focused on thoroughly characterizing forebrain afferents from the hypothalamus and amygdala, two brain regions important for the regulation and production of emotional and social behaviors, including vocalization (Chen and Hong, 2018; Duvarci and Pare, 2014; Ehrlich et al., 2009; Gothard, 2020; Janak and Tye, 2015; LeDoux, 2007; Sternson, 2013).

Figure 1. Inhibitory neurons in the hypothalamus and amygdala provide input to the periaqueductal gray (PAG) vocal gating circuit.

(A) (Left) Viral strategy shown for transsynaptic labeling of direct inputs to PAG-USV neurons (performed in N = 4 males). (Right) Confocal images are shown of starter PAG-USV neurons, upstream neurons labeled within the preoptic area of the hypothalamus (POA), and upstream neurons labeled within the AmgC/M and CeA. (B) Same, for transsynaptic labeling of direct inputs to GABAergic PAG neurons (performed in N = 4 males, N = 2 females). (C) Representative confocal image of in situ hybridization performed on transsynaptically labeled POA neurons that provide direct input to GABAergic PAG neurons (labeled with GFP, shown in white), showing overlap with expression of VGAT (green) and VGlut2 (red). DAPI shown in blue. (D) Quantification of overlap of transsynaptically labeled POA and amygdala neurons (CeA and AmgC/M combined) with VGAT and VGlut2 (N = 2 male mice per condition). Total numbers of neurons scored for overlap in each condition are indicated by the numbers over the bars. See also Figure 1—figure supplements 14, and Figure 1—source datas 12.

Figure 1—source data 1. Source data for Figure 1A–C.
Figure 1—source data 2. Source data for Figure 1D.

Figure 1.

Figure 1—figure supplement 1. Monosynaptic rabies-based tracing reveals preoptic and amygdala inputs to the midbrain vocalization circuit.

Figure 1—figure supplement 1.

(A) Example confocal images showing transsynaptic labeling of neurons in the preoptic area of the hypothalamus (preoptic area [POA], green) that provide monosynaptic input to GABAergic periaqueductal gray (PAG) neurons (left-most confocal image, VGAT-Cre female; middle confocal image, VGAT-Cre male) or to PAG-USV neurons (right confocal image, FosTVA male; blue, Neurotrace). (B–C) Same, for transsynaptically labeled neurons in (B) the central-medial boundary region of the amygdala (AmgC/M) and (C) the central amygdala (CeA). Coronal brain atlas images show the approximate plane of section of the confocal images to the right. Red rectangles superimposed on the atlas images indicate the approximate location of the brain regions shown in the confocal images. See also Figure 1—figure supplements 24. Scale bars, 250 µm. All brain atlas images were obtained from the Allen Mouse Brain Reference Atlas, (https://mouse.brain-map.org/experiment/thumbnails/100048576?Image_type=atlas).
Figure 1—figure supplement 2. Monosynaptic rabies-based tracing reveals cortical inputs to the midbrain vocalization circuit.

Figure 1—figure supplement 2.

(A) Example confocal images showing transsynaptic labeling of neurons in the infralimbic cortex (green) that provide monosynaptic input to GABAergic periaqueductal gray (PAG) neurons (left-most confocal image, VGAT-Cre female; middle confocal image, VGAT-Cre male) or to PAG-USV neurons (right confocal image, FosTVA male; blue, Neurotrace). (B–E) Same, for transsynaptically labeled neurons in (B) insular cortex, (C) motor (M1, M2), and cingulate (Cg) cortices, (D) M1, M2, and somatosensory cortex (S1), and (E) auditory cortex. Coronal brain atlas images show the approximate plane of section of the confocal images to the right. Red rectangles superimposed on the atlas images indicate the approximate location of the brain regions shown in the confocal images. See also Figure 1—figure supplements 1, 3 and 4. Scale bars, 250 µm. All brain atlas images were obtained from the Allen Mouse Brain Reference Atlas, (https://mouse.brain-map.org/experiment/thumbnails/100048576?Image_type=atlas).
Figure 1—figure supplement 3. Monosynaptic rabies-based tracing reveals additional hypothalamic inputs to the midbrain vocalization circuit.

Figure 1—figure supplement 3.

(A) Example confocal images showing transsynaptic labeling of neurons in the anterior hypothalamus (green) that provide monosynaptic input to GABAergic periaqueductal gray (PAG) neurons (left-most confocal image, VGAT-Cre female; middle confocal image, VGAT-Cre male) or to PAG-USV neurons (right confocal image, FosTVA male; blue, Neurotrace). (B–D) Same, for transsynaptically labeled neurons in (B) the paraventricular nucleus (PVN), lateral hypothalamus (LH), and zone incerta (ZI), (C) ventromedial hypothalamus (VMH) and LH, (D) posterior hypothalamus, and (E) premammillary nucleus. Coronal brain atlas images show the approximate plane of section of the confocal images to the right. Red rectangles superimposed on the atlas images indicate the location of the brain regions shown in the confocal images. See also Figure 1—figure supplements 1, 2 and 4. Scale bars, 250 µm. All brain atlas images were obtained from the Allen Mouse Brain Reference Atlas, (https://mouse.brain-map.org/experiment/thumbnails/100048576?Image_type=atlas).
Figure 1—figure supplement 4. Monosynaptic rabies-based tracing reveals additional subcortical inputs to the midbrain vocalization circuit.

Figure 1—figure supplement 4.

(A) Example confocal images showing transsynaptic labeling of neurons in the ventral pallidum (green) that provide monosynaptic input to GABAergic periaqueductal gray (PAG) neurons (left-most confocal image, VGAT-Cre female; middle confocal image, VGAT-Cre male) or to PAG-USV neurons (right confocal image, FosTVA male; blue, Neurotrace). (B–D) Same, for transsynaptically labeling of neurons in (B) lateral septum, (C) bed nucleus of the stria terminals (BNST), (D) lateral habenula, and (E) paraventricular nucleus of the thalamus (PVT). Coronal brain atlas images show the approximate plane of section of the confocal images to the right. Red rectangles superimposed on the atlas images indicates the location of the brain regions shown in the confocal images. See also Figure 1—figure supplements 13. Scale bars, 250 µm. All brain atlas images were obtained from the Allen Mouse Brain Reference Atlas, (https://mouse.brain-map.org/experiment/thumbnails/100048576?Image_type=atlas).

Within the hypothalamus, we observed labeling of neurons in the medial preoptic area (POA, Figure 1A–B), a region that plays a crucial role in sexual behavior (Balthazart and Ball, 2007; McKinsey et al., 2018; Newman, 1999; Wei et al., 2018) and more specifically in the production of courtship vocalizations in rodents and in songbirds (Alger and Riters, 2006; Bean et al., 1981; Floody, 1989; Floody, 2009; Floody et al., 1998; Fu and Brudzynski, 1994; Gao et al., 2019; Riters and Ball, 1999; Vandries et al., 2019). Within the amygdala, we observed inputs to both PAG cell types from neurons spanning the rostral portion of the boundary between the central and medial amygdala (referred to here as the central-medial boundary zone (AmgC/M), see below) continuing caudally to the central amygdala (CeA) (Figure 1A–B). Although the amygdala contributes to the sensory processing of and behavioral responses to social and emotional vocalizations (Fecteau et al., 2007; Gadziola et al., 2016; Hall et al., 2013; Schönfeld et al., 2020), whether and how the amygdala contributes to the production of vocalizations remains understudied (see Hall et al., 2013; Ma and Kanwal, 2014; Matsumoto et al., 2012).

To characterize the neurotransmitter phenotypes of these upstream hypothalamic and amygdala neurons, we performed two-color in situ hybridization on transsynaptically labeled neurons for mRNA transcripts expressed in glutamatergic and GABAergic cells (vesicular glutamate transporter (vGluT2) and vesicular GABA transporter (VGAT); Figure 1C, see Materials and methods). This experiment revealed that the majority (~84%, Figure 1C–D) of PAG-projecting POA neurons (i.e. POAPAG neurons) and almost all (~98%, Figure 1D) of PAG-projecting AmgC/M and CeA neurons (i.e. AmgC/M-PAG and CeAPAG neurons) are GABAergic. In summary, the PAG vocal gating circuit receives input from inhibitory neurons in both the preoptic hypothalamus and the amygdala.

Activating PAG-projecting POA neurons elicits USVs in the absence of social cues

The POA plays a crucial role in courtship, raising the possibility that POAPAG neurons are important to promoting USV production. To test this idea, we selectively expressed channelrhodopsin (ChR2) in POAPAG neurons by injecting a Cre-dependent AAV driving ChR2 expression into the POA and injecting a retrogradely infecting AAV that drives Cre expression into the caudolateral PAG, the region in which PAG-USV neurons are concentrated (Figure 2A). Optogenetic activation of POAPAG cell bodies was sufficient to elicit USVs in male and female mice that were singly tested in the absence of social partners or social cues (USVs elicited in N = 6 of 8 males, N = 3 of 4 females; 10 Hz trains or tonic pulses of 1–2 s duration; Figure 2A–B, Video 1). Although optogenetic activation of POAPAG neurons often elicited robust USV production, the efficacy of optogenetic stimulation (number of USVs elicited per trial, number of successful trials) as well as the latency from stimulation to USV onset were variable both within and across individual mice (Figure 2B and F, and Figure 2—figure supplement 1). This vocal effect was specifically attributable to optogenetic activation of the POA, as delivery of blue light to the POA of GFP-expressing mice failed to elicit USVs (AAV-FLEX-GFP injected into the POA of Esr1-Cre males, see below for additional Esr1-Cre data, N = 5, Figure 2F). In summary, optogenetic activation of POAPAG neurons is capable of promoting USV production in both male and female mice, consistent with the known role of the POA in promoting appetitive courtship behaviors.

Figure 2. Activating periaqueductal gray (PAG)-projecting POA neurons elicits ultrasonic vocalizations (USVs) in the absence of social cues.

(A) (Left) Viral strategy to express ChR2 in POAPAG neurons. (Right) Example trial showing that optogenetic activation of POAPAG neurons elicits USV production in an isolated animal. (B) (Left) Raster plot shows USVs elicited in many trials in a representative mouse following optogenetic activation of POAPAG neurons. (Middle) Mean USV rate aligned to delivery of blue light pulses plotted for that same mouse. (Right) Mean USV rate plotted for N = 8 mice following optogenetic activation of POAPAG neurons. Please note that one mouse in which USVs were elicited by optogenetic stimuli that did not include the 2s-long, 10 Hz stimulus is excluded from the summary analysis shown in the right-most panel. Gray shading above and below the mean in the middle and right panels represents S.E.M. See also Figure 2—figure supplement 1. (C) Representative confocal image and quantification of in situ hybridization performed on POAPAG neurons (tdTomato, red), showing overlap with Esr1 (white) and VGAT (green). DAPI is blue, N = 2 mice. (D) (Left) Viral strategy used to express ChR2 in Esr1+ POA neurons. (Middle) Raster plot shows USVs elicited in many trials in a representative mouse following optogenetic activation of Esr1+ POA neurons. (Right) Mean USV rate plotted for N = 7 mice following optogenetic activation of Esr1+ POA neurons. Gray shading above and below the mean represents S.E.M. (E) Same as (D), for experiments in which the axon terminals of Esr1+ POA neurons were optogenetically activated within the PAG. Data shown for stimulation with10s-long, 20 Hz blue light pulses. Please note that one mouse in which USVs were elicited by optogenetic stimuli that did not include the 10s-long, 20 Hz stimulus is excluded from the summary analysis shown in the right-most panel. (F) Summary plots show mean number of USVs per second of optogenetic stimulation (left, p=0.0013, one-way ANOVA between all groups, with post-hoc t-tests showing that each experimental condition was significantly different from control conditions at p<0.01), mean number of optogenetic trials with USVs (middle, p=1.8E-6, one-way ANOVA between all groups, with post-hoc t-tests showing that each experimental condition was significantly different from control conditions at p<0.01), and mean latency from onset of optogenetic stimulus to onset of first USV (right) for mice in which optogenetic stimulation was applied to POAPAG neurons (N = 9 mice), Esr1+ POA neurons (N = 7 mice), Esr1+ POA axon terminals within the PAG (N = 5 mice), GFP-expressing Esr1+ POA neurons (N = 5 mice), VGlut+ POA neurons (N = 3 mice), VMHPAG neurons (N = 3), and Esr1+ POA axon terminals within the ventral tegmental area (VTA) (N = 4 mice). See also Figure 2—figure supplements 13 and Figure 2—source data 1.

Figure 2—source data 1. Source data for Figure 2C and F.

Figure 2.

Figure 2—figure supplement 1. Additional characterization of ultrasonic vocalizations (USVs) elicited by optogenetic activation of preoptic area (POA) neurons.

Figure 2—figure supplement 1.

(A) Mean USV rate across trials aligned to delivery of blue light pulses plotted for N = 8 individual mice following optogenetic activation of POAPAG neurons. Plots represent the individual animals summarized in the right-most panel of Figure 2B. The example animal represented in the left panels of 2B is mouse number four here. Gray shading above and below the mean in the middle and right panel represents S.E.M. (B) Top: the minimum latency from the onset of optogenetic stimulation to the onset of the first optogenetically elicited USV is plotted for three groups of mice: ChR2 expressed in POAPAG neurons (N = 9 mice, blue points), ChR2 expressed in Esr1+ POA neurons (N = 7 mice, black points), and ChR2 expressed in axon terminals within the PAG of Esr1+ POA neurons (N = 5 mice, red points). Bottom: same plot as above, but with the y range restricted from 0 to 0.5 s. (C) Quantification of the duration of USV bouts elicited by 2 s, 10–20 Hz optogenetic stimulation of POA neurons. Each column represents a different mouse, and each point represents the duration of an USV bout elicited by a 2s-long optogenetic stimulus. Colors as in (B). Please note that five animals shown in (B) had only a small number of 2s-long optogenetic activation trials and were therefore not included in the analysis shown in (C). See also Figure 2—figure supplement 1—source data 1.
Figure 2—figure supplement 1—source data 1. Source data for panels B and C of Figure 2—figure supplement 1.
Figure 2—figure supplement 2. Additional information related to the optogenetic activation of POAPAG, POA-Esr1+ neurons, and AmgC/M-PAG neurons.

Figure 2—figure supplement 2.

(A) Real-time place preference tests were performed in which either POA or AmgC/M neurons were optogenetically activated when mice were in one of two sides of a test chamber (see Materials and methods). (B) The mean speeds of mice were calculated and aligned to the onset of optogenetic activation of POAPAG neurons (top left), POA-Esr1+ neurons (top right), and AmgC/M-PAG neurons (bottom left). (Bottom, right) Distances between male AmgC/M-PAG-ChR2 mice and their female social partners were calculated and aligned to the onset of optogenetic activation of AmgC/M-PAG neurons. Gray shading represents S.E.M. See also Figure 2—figure supplement 2—source data 1.
Figure 2—figure supplement 2—source data 1. Source data for panel A of Figure 2—figure supplement 2.
Figure 2—figure supplement 3. Dual tracing of the axonal projections of POAPAG and AmgC/M-PAG neurons.

Figure 2—figure supplement 3.

POAPAG neurons are labeled with GFP, and AmgC/M-PAG neurons are labeled with tdTomato. (Top left) Plane of section through the preoptic area (POA) shows the beginning of POAPAG cell body labeling in green and axonal projections to the lateral POA from AmgC/M-PAG neurons in red. (Middle left) Plane of section through the AmgC/M-PAG shows axonal projections from POAPAG neurons that overlap with AmgC/M-PAG cell bodies. (Bottom left) Plane of section through the ventral tegmental area (VTA)/SNr. (Top right) Plane of section including the rostral PAG and retrorubral area. (Bottom right) Place of section through the caudal PAG, showing the overlapping terminal fields of POAPAG and AmgC/M-PAG neurons. These representative images are from a female mouse, and we found that POAPAG and AmgC/M-PAG neurons have similar axonal projections in males and females (data not shown).

Video 1. Optogenetic activation of POAPAG neurons elicits ultrasonic vocalizations (USVs).

Download video file (917.3KB, mp4)

An isolated male mouse is shown which has ChR2 is expressed in POAPAG neurons. Optogenetic activation of these neurons with pulses of blue light elicits USV production. Video is shown at the top, a spectrogram (bottom) showing the optogenetically elicited USVs is synchronized to the video, and pitch-shifted audio (80 kHz to 5 kHz transformation) is included to place the ultrasonic vocalizations (USVs) within the human hearing range.

To begin to describe the molecular phenotype of POAPAG neurons, we used situ hybridization to establish that these cells express VGAT (319/319 neurons were VGAT+; Figure 2C), similar to the POA neurons that we labeled via transsynaptic tracing from the PAG vocal gating circuit. We also noted that the majority of POAPAG neurons co-express Estrogen Receptor 1 (Esr1), a prominent marker for neurons in the POA (278/319 Figure 2C; Fang et al., 2018; Moffitt et al., 2018; Wei et al., 2018). Given that POAPAG neurons express Esr1, we next tested whether optogenetic activation of Esr1+ POA neurons was sufficient to elicit USV production, by injecting a Cre-dependent AAV driving the expression of ChR2 into the POA of Esr1-Cre mice. We observed that optogenetically activating Esr1+ POA neurons was sufficient to elicit USV production in male (N = 4 of 5) and female mice (N = 3 of 4) that were tested in the absence of any social partners or social cues (Figure 2D). In contrast, optogenetic activation of VGlut2+ neurons within the POA failed to elicit USV production (Figure 2F, N = 3 males, POA of VGlut-Cre mice injected with AAV-FLEX-ChR2). Our findings confirm and extend the recent finding that optogenetic activation of GABAergic POA neurons elicits USV production in male and female mice (Gao et al., 2019).

To test whether activation of the Esr1+ POA neurons that project to the PAG is sufficient to elicit USVs, we optogenetically activated the axon terminals of Esr1+ POA neurons within the PAG (Figure 2E). Bilateral Esr1+ POAPAG terminal activation within the PAG was sufficient to elicit USV production (N = 1 of 2 males; N = 4 of 6 females, 20 Hz trains of 2–10 s duration). This treatment also evoked escape behavior in four of eight of the tested animals, which was not observed following optogenetic activation of Esr1+ POA cell bodies, suggesting that viral spread to PAG-projecting neurons nearby to the POA may account for these effects. Finally, we sought to test the idea that optogenetic activation of Esr1+ POA neurons promotes USV production through their projections to the PAG rather than through other regions that they also innervate, and we also tested whether USV production could be elicited by activating non-POA hypothalamic inputs to the PAG. In fact, USVs were not elicited by optogenetically activating either Esr1+ POA axon terminals in the ventral tegmental area (VTA) (Figure 2F, 0/2 females, 0/2 males) or PAG-projecting neurons within the ventromedial hypothalamus (VMH) (Figure 2F, 0/3 males, AAV-retro-Cre injected in the PAG, AAV-FLEX-ChR2 injected in the VMH). Therefore, GABAergic POA neurons, including Esr1+ cells, act via their synapses in the caudolateral PAG to promote USV production.

Although these control experiments are consistent with the idea that POAPAG and Esr1+ POA neurons act directly on the vocal gating mechanism in the PAG, a remaining possibility is that they promote USV production through hedonic reinforcement. To control for this possibility, we performed real-time place preference tests in which optogenetic stimulation of either POAPAG or Esr1+ POA neurons was applied when mice were in only one of two sides of the test chamber. We observed that optogenetic activation of POAPAG neurons drove a slightly negative place preference on average (mean PP = 0.39 +/- 0.07 for N = 7 mice; Figure 2—figure supplement 2, panel A) and that optogenetic activation of Esr1+ POA cell bodies did not positively reinforce place preference (mean PP = 0.46 +/- 0.02 for N = 5 mice). In contrast, optogenetic activation of Esr1+ POA axon terminals within the VTA positively reinforced place preference (mean PP = 0.59 +/- 0.06, N = 4 mice, Figure 2—figure supplement 2, panel A). We also note that when using the same stimulation parameters that were sufficient to elicit USVs, optogenetic activation of either POAPAG or Esr1+ POA neurons did not drive mounting of other mice (N = 7 POAPAG-ChR2 mice tested; N = 4 POA-Esr1-ChR2 mice tested) nor did it induce overt locomotion (Figure 2—figure supplement 2, panel B). These experiments indicate that activation of POAPAG neurons can elicit USVs in a manner that does not depend on positive reinforcement and without recruiting other courtship behaviors.

Because the POA lies upstream of the PAG, we anticipated that optogenetic activation of the POA would elicit USV production at longer latencies than observed for optogenetic activation of PAG-USV neurons. Indeed, we found that the minimum and mean latencies to elicit USVs by optogenetic stimulation of POA neurons were 664.5 +/- 320.9 ms and 1782.6 +/- 407.6 ms, respectively (Figure 2—figure supplement 1, calculated from N = 9 POAPAG-ChR2 and N = 7 POA-Esr1-ChR2 mice). These latencies are longer than those observed when optogenetically activating PAG-USV neurons (PAG-USV activation: min. latency from laser onset to first USV was 23.4 ± 8.6 ms, mean latency was 406.6 ± 0.5 ms) (Tschida et al., 2019) but are comparable to the latencies from optogenetic activation of the hypothalamus to observed effects on behavior that have been reported in other studies (Lin et al., 2011; Wei et al., 2018). We also found that USV bouts elicited by optogenetic activation of the POA often outlasted the duration of the optogenetic stimulation, sometimes by many seconds (Figure 2, Figure 2—figure supplement 1). This contrasts with what is observed following optogenetic activation of PAG-USV neurons, in which USV bout durations map on tightly to the duration of optogenetic stimulation (Tschida et al., 2019), and suggests that brief optogenetic stimulation in the POA can be transformed into longer lasting changes in neural activity within the POA or across POA-to-PAG synapses.

Acoustic characterization of USVs elicited by activation of POA neurons

Given that optogenetic stimulation of the POA elicited USVs in the absence of any social cues, we wondered whether such optogenetically evoked USVs were acoustically similar to the USVs that mice produce during social interactions. To compare the acoustic features of optogenetically elicited USVs to those of USVs produced spontaneously to a nearby female, we employed a recently described method using variational autoencoders (VAEs) (Goffinet et al., 2019; Sainburg et al., 2019). Briefly, the VAE is an unsupervised modeling approach that uses spectrograms of vocalizations as inputs and from these data learns a pair of probabilistic maps, an ‘encoder’ and a ‘decoder,’ capable of compressing vocalizations into a small number of latent features while attempting to preserve as much information as possible (Figure 3A–B). Notably, this method does not rely on user-defined acoustic features, nor does it require clustering of vocalizations into categories. We applied this approach to spectrograms of USVs to compare the acoustic features of female-directed and optogenetically elicited USVs from the same mice and found that the VAE converged on a concise latent representation of only five dimensions. We then employed a dimensionality reduction method (UMAP) (McInnes et al., 2018) to visualize the latent features of these USVs in 2D space (Figure 3C). This analysis revealed that for some mice, female-directed and optogenetically elicited USVs were acoustically similar (Figure 3C, left), while for other mice, a subset of optogenetically elicited USVs were acoustically distinct from female-directed USVs (Figure 3C, right). To quantify the difference between female-directed and optogenetically elicited USVs for each mouse, we estimated the Maximum Mean Discrepancy (MMD) (Gretton et al., 2012) between distributions of latent syllable representations as in Goffinet et al., 2019. In addition, a baseline level of variability in syllable repertoire was established for each mouse by estimating the MMD between the first and second halves of female-directed USVs emitted in a recording session (Figure 3D). A paired comparison revealed significantly larger differences between female-directed and optogenetically elicited USVs than expected by variability within the female-directed recording sessions alone (Figure 3D, two-sided, continuity-corrected Wilcoxon signed-rank test, W = 5, p<0.01), or than expected by across-day variability in female-directed recording sessions from control animals (gray points, Figure 3D, female-directed USVs were recorded on 2 different days from N = 10 control mice, p=0.003 for difference between female-directed vs. female-directed in control mice and opto vs. female-directed in experimental mice, Mann Whitney U test). In conclusion, many USVs elicited by optogenetic activation of POA neurons resemble female-directed USVs, although a subset differs in their acoustic features from USVs found in the animals’ female-directed repertoires.

Figure 3. Acoustic characterization of ultrasonic vocalizations (USVs) elicited by optogenetic activation of preoptic area (POA) neurons.

Figure 3.

(A) The variational autoencoder (VAE) takes spectrograms as input (left), maps the spectrograms to low-dimensional latent representations using an encoder (middle), and approximately reconstructs the input spectrogram using a decoder (right). (B) (Left) Dimensionality reduction techniques such as PCA or UMAP can be used to visualize the resulting latent vectors. (Right) Interpolations in latent space correspond to smooth USV syllable changes in spectrogram space exhibiting realistic dimensions of variation. (C) UMAP projections of latent syllable representations of female-directed USVs (red) and optogenetically elicited USVs (blue) from two example mice. (D) Maximum Mean Discrepancy (MMD) was calculated between distributions of latent syllable representations to generate three comparisons: female-directed USVs from the first half of the recording session vs. female-directed USVs from the second half, all female-directed USVs vs. opto-USVs (N = 16 experimental mice), and female-directed USVs from two different recordings sessions in N = 10 control mice (gray points). Note that larger MMD values indicate that distributions are more dissimilar (E) UMAP projections of latent descriptions of female-directed (red) and optogenetically elicited USVs (blue) for all mice (N = 15). (F) UMAP projections from panel E, color-coded by total energy (left) and frequency bandwidth (right). Example spectrograms of opto-USVS and female-directed USVs are plotted below, and the location of each example USV in UMAP space is indicated by the colored dots on the grayscale UMAP projection on the bottom left. See also Figure 3—source data 1.

Figure 3—source data 1. Source data for Figure 3D.

We next sought to understand in more detail exactly how these acoustically unusual optogenetically elicited USVs differed from natural USVs. When the latent representations of these two types of USVs were plotted together for all mice in our dataset, it became clear that optogenetically-elicited USVs and female-directed USVs are largely acoustically overlapping except in one region of the UMAP representation (upper middle portion of Figure 3E, dominated by blue points). Despite this outlying region of acoustically distinct optogenetically elicited USVs, we conservatively estimate that only 20% of condition information (optogenetically elicited versus female-directed) can be predicted by latent syllable descriptions, consistent with largely overlapping distributions of natural and optogenetically elicited USVs (0.20 bits, fivefold class-balanced logistic regression). We then re-plotted UMAP representations of the USVs, with each USV syllable color-coded according to syllable energy (i.e. amplitude, Figure 3F, left) or frequency bandwidth (Figure 3F, right). This analysis revealed that the acoustically unusual optogenetically elicited USVs tended to be louder and had greater frequency bandwidths than female-directed USVs. Visual inspection of spectrograms of optogenetically-elicited USVs also confirmed that those that did not overlap acoustically with natural USVs tended to be louder and have greater frequency bandwidths (Figure 3F, bottom, opto 1 and opto 2), while optogenetically elicited USVs that overlapped with natural USVs did not possess these unusual acoustic features (Figure 3F, bottom, opto 3). To determine whether the differences between optogenetically elicited and natural USVs were consistent across mice, we summarized each recording session by the mean latent representation of its syllables, and then summarized the shift from natural to optogenetically elicited syllable repertoires by the corresponding vector between summary points. A shuffle test revealed significantly larger alignment between these vectors than expected by chance (mean cosine similarity = 0.50, p<1e-5), indicating that optogenetically-elicited USVs differed from female-directed USVs in a manner that was consistent across mice. In summary, optogenetic activation of the POA elicits USVs whose acoustic features are largely overlapping with those of female-directed USVs produced by the same animal, despite the artificiality inherent to optogenetic stimulation.

Activating PAG-projecting AmgC/M neurons transiently suppresses USV production

We then explored how PAG-projecting amygdala neurons contribute to vocalization. We began with a viral strategy designed to express ChR2 in PAG-projecting AmgC/M and CeA neurons, by injecting a Cre-dependent AAV driving ChR2 expression targeted to the amygdala and then injecting AAV-retro-Cre into the PAG (Figure 4). Surprisingly, given the strong transsynaptic labeling of both the CeA and AmgC/M achieved with modified rabies tracing from the PAG vocal gating circuit, we found that this viral strategy failed to label neurons in the CeA and instead only labeled AmgC/M neurons, whose cell bodies reside medial to the CeA and dorsal to the medial amygdala (Figure 4A, Figure 4—figure supplement 1). To ensure that this labeling pattern was due to restricted tropism of the AAV-retro-Cre virus and not to inaccurate targeting of the CeA, we repeated the injections of the AAV-retro-Cre virus in the PAG of a Cre-dependent tdTomato reporter mouse. Again, we observed cell body labeling in the AmgC/M but not in the CeA (Figure 4—figure supplement 2), suggesting that in contrast to the modified rabies virus used in the transsynaptic tracing from the PAG vocal gating circuit, the AAV-retro-Cre virus can infect AmgC/M but not CeA neurons.

Figure 4. Activating AmgC/M-PAG neurons transiently suppresses ultrasonic vocalization (USV) production.

(A) (Left) Viral strategy used to express ChR2 in AmgC/M-PAG neurons. (Right) Confocal image of representative AmgC/M-PAG cell body labeling achieved with this viral strategy. (B) (Left) Spectrogram showing a representative trial in which optogenetic activation of AmgC/M-PAG neurons suppresses USV production during the laser stimulation period. (Right) Group data quantified for N = 8 mice. Gray shading above and below the mean represents S.E.M. (C) Confocal image and quantification of in situ hybridization performed on AmgC/M-PAG neurons (GFP, shown in white), showing overlap with VGlut2 (red) and VGAT (green). DAPI in blue, N = 2 mice. (D) Left: viral strategy used to express ChR2 in the periaqueductal gray (PAG) axon terminals of AmgC/M-PAG neurons. Right: Quantification of the number of USVs produced in the 1 s period prior to optogenetic stimulation (pre), the 1 s period of optogenetic stimulation (laser), and the 1 s period following optogenetic stimulation (post). Data for each mouse were normalized by dividing the pre, laser, and post measurements by the total number of USVs produced during the pre-laser period. Group averages are shown for mice in which AmgC/M-PAG neurons were optogenetically activated (N = 12, dark blue), mice in which the PAG axon terminals of AmgC/M-PAG neurons were optogenetically activated (N = 3, light blue), control mice in which the blue laser was shined over the mouse’s head but not connected to the optogenetic ferrule (N = 9, gray), control mice in which GFP was expressed in AmgC/M-PAG neurons (N = 4, green), and control mice in which the laser was triggered but not turned on (N = 6, black). Error bars represent S.D. Please note that the decay in USV rates over time in the control groups reflects the natural statistics of USV production (increasing probability that a bout will end as time progresses). See also Figure 4—figure supplements 12, Figure 2—figure supplements 12, and Figure 4—source data 1.

Figure 4—source data 1. Source data for Figure 4C and D.

Figure 4.

Figure 4—figure supplement 1. Extent of cell body labeling of AmgC/M-PAG neurons.

Figure 4—figure supplement 1.

AmgC/M-PAG neurons were labeled by injecting AAV-retro-Cre into the caudolateral periaqueductal gray (PAG) and AAV-FLEX-GFP into the AmgC/M. (A–C) Three planes of coronal section are shown that cover the rostral to caudal extent of AmgC/M-PAG cell body labeling.
Figure 4—figure supplement 2. Comparison of hypothalamus and amygdala cell body labeling achieved after transsynaptic tracing from the periaqueductal gray (PAG) vocal gating circuit versus.

Figure 4—figure supplement 2.

AAV-retro-Cre injection into the caudolateral PAG. Representative coronal sections are shown for POA and amygdala cell body labeling observed after monosynaptic rabies-based tracing from GABAergic PAG neurons (right panels, GFP-labeled cells) or AAV-retro-Cre injection into the caudolateral PAG of an Ai14 reporter mouse (middle panels, tdTomato-labeled cells, see Materials and methods). A-C show different planes of section as illustrated in the atlas images in the left panels. Note that although both AmgC/M and CeA neurons are labeled after transsynaptic tracing from PAG-GABA neurons (B3 and C3), CeA neuronal labeling is absent after injection of AAV-retro-Cre (C2). Scale bars, 250 µm. All brain atlas images were obtained from the Allen Mouse Brain Reference Atlas, (https://mouse.brain-map.org/experiment/thumbnails/100048576?Image_type=atlas).

To test whether PAG-projecting AmgC/M neurons influence USV production, we first tested the effects of optogenetically activating these neurons in isolated mice. Optogenetic activation of AmgC/M-PAG neurons failed to elicit USV production and also did not drive any other obvious behavioral effects (Videos 2, 3, 4). However, when AmgC/M-PAG neurons were optogenetically activated in male mice that were actively courting females and vocalizing, USV production was immediately and reversibly suppressed (Figure 4B, N = 8 mice). This suppressive effect was restricted to the period when AmgC/M-PAG neurons were being optogenetically stimulated, and USV production rebounded following the end of the optogenetic stimulation period (Figure 4B). After using in situ hybridization to confirm that most AmgC/M-PAG neurons are GABAergic (~92% AmgC/M-PAG neurons express VGAT, Figure 4C), we used a similar intersectional viral strategy to express ChR2 selectively in GABAergic AmgC/M-PAG neurons (Figure 4D, AAV-retro-FLEX-ChR2 injected into the PAG of a VGAT-Cre mouse). With this strategy, we found that optogenetic activation of GABAergic AmgC/M-PAG neurons robustly suppressed the male’s USV production during courtship encounters with a female (Figure 4D, N = 4 male mice). Finally, we tested the effects on vocal behavior of optogenetically activating the axon terminals of GABAergic AmgC/M-PAG neurons within the PAG (Figure 4D). Such bilateral terminal activation was also sufficient to suppress USV production (in N = 3 of 3 males; Figure 4D).

Video 2. Optogenetic activation of AmgC/M-PAG neurons causes no obvious behavioral effects in the absence of a social partner.

Download video file (2.2MB, mp4)

An example male mouse with ChR2 expression in AmgC/M-PAG neurons is shown alone in a chamber with no social partner. Optogenetic activation of AmgC/M-PAG neurons with pulses of blue light does not elicit ultrasonic vocalization (USV) production or any other obvious behavioral response. Video is shown at the top, a spectrogram (bottom) showing the audio recording is synchronized to the video, and pitch-shifted audio (80 kHz to 5 kHz transformation) is included to place any USVs that may have occurred within the human hearing range.

Video 3. Optogenetic activation of AmgC/M-PAG neurons causes no obvious behavioral effects in the absence of a social partner.

Download video file (2MB, mp4)

An example male mouse with ChR2 expression in AmgC/M-PAG neurons is shown alone in a chamber with no social partner. Optogenetic activation of AmgC/M-PAG neurons with pulses of blue light does not elicit ultrasonic vocalization (USV) production or any other obvious behavioral response. Video is shown at the top, a spectrogram (bottom) showing the audio recording is synchronized to the video, and pitch-shifted audio (80 kHz to 5 kHz transformation) is included to place any USVs that may have occurred within the human hearing range.

Video 4. Optogenetic activation of AmgC/M-PAG neurons causes no obvious behavioral effects in the absence of a social partner.

Download video file (2MB, mp4)

An example male mouse with ChR2 expression in AmgC/M-PAG neurons is shown alone in a chamber with no social partner. Optogenetic activation of AmgC/M-PAG neurons with pulses of blue light does not elicit ultrasonic vocalization (USV) production or any other obvious behavioral response. Video is shown at the top, a spectrogram (bottom) showing the audio recording is synchronized to the video, and pitch-shifted audio (80 kHz to 5 kHz transformation) is included to place any USVs that may have occurred within the human hearing range.

One possibility is that activating AmgC/M-PAG neurons suppresses USV production by putting the mouse into a fearful or aversive state, rather than through a direct suppressive effect of AmgC/M-PAG neurons on the PAG vocal gating circuit. To test this idea, we carefully examined the non-vocal behaviors of male mice during optogenetic activation of AmgC/M-PAG neurons. Mice exhibited neither freezing nor fleeing during optogenetic stimulation of AmgC/M-PAG neurons and, more notably, they usually continued to follow and sniff the female during the laser stimulation periods (Video 5; distance between male and female did not increase during optogenetic stimulation, Figure 2—figure supplement 2, panel B). We also confirmed that the change in USV production rates driven by the optogenetic activation of AmgC/M-PAG neurons was different from the change in spontaneous USV rates over time in mice that did not receive laser stimulation (Figure 4D, black trace), the change in USV rates over time in GFP control mice (Figure 4D, green trace), and the change in USV rates over time in AmgC/M-PAG-ChR2-expressing mice that were connected to a dummy ferrule that only shined blue light over their head (Figure 4D, gray trace; p<0.01 for differences between ChR2 groups vs. control groups during laser time, p>0.05 for differences between groups in post-laser period; two-way ANOVA with repeated measures on one factor, p<0.01 for interaction between group and time, followed by post-hoc pairwise Tukey’s HSD tests). Finally, we performed real-time place preference tests in which AmgC/M-PAG neurons were optogenetically activated when mice were in one of two sides of a test chamber (AmgC/M-PAG neurons were labeled with either the AAV-retro-Cre or the AAV-retro-ChR2 viral strategies). This experiment revealed that activation of AmgC/M-PAG neurons does not drive a negative place preference (Figure 2—figure supplement 2, panel A). In summary, activating AmgC/M-PAG neurons transiently and selectively suppresses USVs produced by male mice during courtship, an effect that cannot be accounted for by the mouse being put into a fearful or aversive state.

Video 5. Optogenetic activation of AmgC/M-PAG neurons transiently suppresses ultrasonic vocalization (USV) production.

Download video file (1.1MB, mp4)

A male mouse which has ChR2 expressed in AmgC/M-PAG neurons is shown interacting with and producing USVs directed at a female social partner. Optogenetic activation of these neurons with pulses of blue light transiently suppresses USV production without suppressing non-vocal courtship behavior. Video is shown at the top, a spectrogram (bottom) showing the optogenetically elicited USVs is synchronized to the video, and pitch-shifted audio (80 kHz to 5 kHz transformation) is included to place the USVs within the human hearing range.

Axonal projections of POAPAG and AmgC/M-PAG neurons

To further characterize the anatomy of POAPAG and AmgC/M-PAG neurons, we used intersectional methods to label these neurons with GFP and tdTomato respectively and traced their axonal projections throughout the brain (Figure 2—figure supplement 3, AAV-retro-Cre injected into caudolateral PAG, AAV-FLEX-GFP into POA, AAV-FLEX-tdTomato into AmgC/M). We observed dense projections from both POAPAG and AmgC/M-PAG neurons to a variety of dopaminergic cell groups, including the VTA, SNc and retrorubral/A8 region. We also note that AmgC/M-PAG neurons provide input to the lateral preoptic area (Figure 2—figure supplement 3, top left), while POAPAG neurons provide input to the same region in which AmgC/M-PAG cell bodies reside (Figure 2—figure supplement 3, middle left). As expected, we also observed dense and overlapping terminal fields from both of the cell groups within the caudolateral PAG (Figure 2—figure supplement 3, bottom right).

Synaptic interactions between POAPAG and AmgC/M-PAG neurons and the PAG vocal gating circuit

The functional and anatomical experiments described above establish that two different populations of inhibitory forebrain neurons provide input to the PAG vocal gating circuit, one of which (the POA) promotes USV production in the absence of any social cues, while the other (the AmgC/M) suppresses spontaneous USVs produced by male mice during courtship. To understand how two different GABAergic and presumably inhibitory inputs to the PAG can exert opposing effects on vocal behavior, we performed ChR2-assisted circuit mapping experiments in brain slices to characterize the properties of POA and AmgC/M synapses onto PAG-USV neurons and nearby GABAergic PAG neurons.

Given that optogenetic activation of GABAergic AmgC/M-PAG neurons suppresses USV production, we predicted that GABAergic AmgC/M neurons directly inhibit PAG-USV neurons. To test this idea, we performed whole-cell voltage clamp recordings from PAG-USV neurons while optogenetically activating AmgC/M-PAG axons within the PAG. Briefly, AAV-FLEX-ChR2 was injected into the AmgC/M of a VGAT-Cre;Fos-dsTVA crossed mouse in order to express ChR2 in GABAergic AmgC/M-PAG axon terminals within the PAG. After four weeks, we used the CANE method (Rodriguez et al., 2017; Sakurai et al., 2016; Tschida et al., 2019) to infect PAG-USV neurons with a pseudotyped CANE-rabies virus driving the expression of mCherry (CANE-RV-mCherry, Figure 5A–B, see Materials and methods). We visually targeted our recordings to mCherry-expressing PAG-USV neurons and optogenetically activated AmgC/M terminals in the presence of TTX and 4AP in order to isolate monosynaptic pathways (Figure 5C–D). Activating AmgC/M-PAG terminals evoked inhibitory postsynaptic currents (IPSCs) in a majority (16/29) of the mCherry-labeled PAG-USV neurons from which we recorded (mean current = 180.3 pA at 0 mV in TTX/4AP). These evoked IPSCs were completely abolished by application of the GABAA receptor antagonist gabazine (Figure 5E–F). No optogenetically elicited EPSCs were detected when recording at −70 mV, the chloride reversal potential. These findings support the idea that AmgC/M-PAG activity suppresses ongoing USV production by directly inhibiting PAG-USV neurons.

Figure 5. AmgC/M neurons provide direct inhibition onto PAG-USV neurons.

Figure 5.

(A) Viral strategy (left) and schematic (right) for whole-cell patch clamp recordings from fluorescently identified CANE-tagged PAG-USV neurons while optogenetically activating AmgC/M-PAG axons. (B) Example image of overlap of neurobiotin and mCherry-labeled PAG-USV cells with ChR2-expressing AmgC/M-PAG axon terminals in the PAG. (C) Example of light-evoked IPSCs at different voltages from one PAG-USV cell recorded in TTX/4AP while stimulating AmgC/M-PAG axons (left). Inhibitory postsynaptic currents (IPSCs) were abolished by bath gabazine application (right). (D) The peak magnitude of light-evoked currents at different membrane voltages for the same cell as (C) shows that the current reverses around the reversal potential of chloride and is abolished by gabazine. Currents were identified as IPSCs in this manner based on their reversal behavior and, for a subset of cells, by disappearance in gabazine. (E,F) Light-evoked IPSCs recorded in TTX/4AP (observed in n = 16 of 29 CANE-tagged cells from nine mice) were abolished by application of gabazine (n = 10 cells also recorded in gabazine, N = 10 cells, p<0.001, paired t-test). IPSC amplitude refers to the peak of the light-evoked current at 0 mV holding potential. Error bars represent S.E.M. See also Figure 5—source data 1.

Figure 5—source data 1. Source data for Figure 5D and F.

Given that activating GABAergic POAPAG neurons elicits vocalization (Figure 2), and that the majority of PAG-USV neurons are glutamatergic (Tschida et al., 2019), we hypothesized that POAPAG axons act via local GABAergic interneurons in the PAG to disinhibit PAG-USV neurons. To test this hypothesis, we first performed whole-cell patch clamp recordings from GABAergic PAG neurons while optogenetically activating POAPAG axons within the PAG. GABAergic PAG neurons were labeled by injecting AAV-FLEX-mCherry into the PAG of a VGAT-Cre mouse, while AAV-FLEXa-ChR2 was injected into the POA to express ChR2 in POAPAG axon terminals within the PAG (Figure 6A–B). After waiting 4 weeks to achieve functional expression of ChR2 in POAPAG axon terminals, we cut brain slices from these mice and recorded optogenetically evoked currents from fluorescently identified VGAT+ PAG neurons (see Materials and methods). Optical stimulation of POAPAG axons with blue-light-evoked IPSCs in the majority (26/36) of voltage clamped GABAergic PAG neurons from which we recorded (mean current = 328.8 pA at 0 mV) (Figure 6C). These evoked IPSCs persisted upon application of TTX/4AP and were blocked by gabazine, indicating that POAPAG axons make inhibitory synapses directly onto GABAergic PAG neurons (Figure 6D).

Figure 6. POA neurons provide direct inhibition onto VGAT+ PAG neurons, which provide direct inhibition onto PAG-USV neurons.

(A) Viral strategy (left) and schematic (right) for whole-cell patch clamp recordings from fluorescently identified VGAT+ PAG cells while optogenetically activating POAPAG axons. (B) Example image of mCherry-labeled VGAT+ neurons with ChR2-labeled POAPAG axon terminals in the PAG. (C, D) Light-evoked inhibitory postsynaptic currents (IPSCs; observed in n = 26 of 36 VGAT+ neurons recorded from 11 mice) persisted in TTX/4AP and were abolished by bath application of gabazine (n = 10 cells recorded at baseline, n = 22 cells recorded in TTX/4AP, and n = 13 cells also recorded in gabazine including the following pairs: 6 cells recorded in both baseline and TTX/4AP, 3 cells recorded in both baseline and gabazine, and 10 cells recorded in both TTX/4AP and gabazine, p=0.03, one-way ANOVA comparing baseline vs. TTX+4-AP vs. SR-95531, followed by a post-hoc t-test revealing a significant difference between TTX+4-AP vs. SR-95531, p<0.018). IPSC amplitude refers to the peak of the light-evoked current at 0 mV holding potential. Error bars represent S.E.M. (E) Viral strategy (left) and schematic (right) for whole-cell recordings from fluorescently identified CANE-tagged PAG-USV neurons while optogenetically activating local VGAT+ PAG neurons. (F) Example image of mCherry-labeled CANE-tagged PAG-USV neurons and ChR2-labeled VGAT+ PAG neurons. (G,H) Light-evoked IPSCs recorded in TTX/4AP (observed in n = 13 of 16 CANE-tagged cells from four mice) were abolished by gabazine application (N = 10 cells also recorded in gabazine, p<0.001, paired t-test). IPSC amplitude refers to the peak of the light-evoked current at 0 mV holding potential. Error bars represent S.E.M. See also Figure 6—figure supplement 1 and Figure 6—source data 1.

Figure 6—source data 1. Source data for Figure 6D and H.

Figure 6.

Figure 6—figure supplement 1. POA neurons provide direct inhibition onto few PAG-USV neurons.

Figure 6—figure supplement 1.

(A) Viral strategy (left) and schematic (right) for whole-cell patch clamp recordings from fluorescently identified PAG-USV cells while optogenetically activating POAPAG axons. (B) Light-evoked inhibitory postsynaptic currents (IPSCs) were observed in 1 of 23 mCherry-labeled PAG-USV neurons recorded from seven mice. See also Figure 6—figure supplement 1—source data 1.
Figure 6—figure supplement 1—source data 1. Source data for panel B of Figure 6—figure supplement 1.

To test whether these GABAergic PAG neurons synapse onto PAG-USV neurons, as predicted of a disinhibitory circuit mechanism, we injected AAV-FLEX-ChR2 into the PAG of a VGAT-Cre;TVA crossed mouse in order to express ChR2 in local VGAT+ neurons (Figure 6E). After 2 weeks, we used CANE to selectively infect PAG-USV neurons with CANE-RV-mCherry (Figure 6). Several days later, we visually targeted mCherry-expressing PAG-USV neurons for whole-cell recordings while optogenetically activating local GABAergic PAG neurons in the presence of TTX and 4AP (Figure 6F). Optogenetically activating local VGAT+ neurons evoked IPSCs in almost all (13/16) of the PAG-USV neurons from which we recorded (mean current = 579.9 pA at 0 mV in TTX/4AP) and these currents were completely abolished by application of gabazine (Figure 6G–H). This experiment confirms the presence of a functional connection between local inhibitory neurons and the PAG-USV neurons that gate USV production.

We also performed whole-cell recordings from mCherry-labeled PAG-USV neurons while optogenetically activating POAPAG axons within the PAG (AAV-FLEX-ChR2 injected into the POA of a VGAT-Cre;Fos-dsTVA crossed mouse, CANE method used to infect PAG-USV neurons with CANE-RV-mCherry as described above; Figure 6—figure supplement 1, panel A, see Materials and methods). After first confirming that we could optogenetically evoke IPSCs in mCherry-negative cells in each slice, we visually targeted our recordings to mCherry-expressing PAG-USV neurons. Optogenetic activation of POAPAG terminals evoked IPSCs in only 1 of 23 PAG-USV neurons (Figure 6—figure supplement 1, panel B). Although there are caveats to interpreting a low probability of synaptic connection in brain slices, POAPAG neurons appear to provide fewer or weaker synaptic inputs to PAG-USV neurons than to nearby GABAergic PAG neurons, supporting the idea that POAPAG neurons primarily act through PAG interneurons to disinhibit PAG-USV neurons and promote USV production.

Discussion

Here, we used a combination of monosynaptic rabies tracing, optogenetic manipulations of neural activity in freely behaving animals, and optogenetics-assisted circuit mapping in brain slices to elucidate the functional relevance and synaptic organization of descending inputs to the PAG vocal gating circuit. We identified two populations of forebrain inhibitory neurons, one located in the preoptic hypothalamus and the other in a central-medial boundary zone within the amygdala, that drive opposing effects on vocal behavior. Optogenetic activation of POAPAG neurons drives robust and long-lasting bouts of vocalization in the absence of any social cues normally required to elicit vocalizations, and the acoustic features of optogenetically elicited USVs shared many features with spontaneously produced social USVs. In contrast, optogenetic activation of a VGAT+ population of AmgC/M-PAG neurons transiently suppressed USV production in male mice during active courtship without disrupting other non-vocal courtship behaviors. Further, activation of AmgC/M-PAG neurons did not elicit fearful or aversive behavior, indicating that the effect on vocal behavior was not driven or accompanied by a generalized change in behavioral state. Finally, we paired optogenetic activation of descending POA or AmgC/M inputs to the PAG with whole-cell recordings from PAG-USV or GABAergic PAG neurons to investigate how these POA and AmgC/M inputs drive opposing effects on vocal behavior. These slice experiments support a model in which AmgC/M-PAG neurons directly inhibit PAG-USV neurons to suppress vocalization, while POAPAG neurons directly inhibit GABAergic PAG interneurons, which in turn inhibit PAG-USV neurons, resulting in a net disinhibition of PAG-USV neurons that promotes vocalization (Figure 7). To our knowledge, this is the first study to reveal the synaptic and circuit logic by which forebrain afferents to the PAG influence the decision to vocalize, a key behavior for communication and survival.

Figure 7. Model of bidirectional descending control of the periaqueductal gray (PAG) vocal gating circuit.

Figure 7.

Inhibitory neurons within the POA provide direct input to inhibitory neurons within the PAG, which in turn provide direct input to PAG-USV neurons. In this manner, activation of POAPAG neurons disinhibits PAG-USV neurons, which provide excitatory input to downstream vocal premotor neurons and drive USV production. Conversely, inhibitory neurons within the AmgC/M provide direct inhibitory input to PAG-USV neurons. Hence activation of AmgC/M-PAG neurons reduces PAG-USV activity and transiently suppresses USV production.

We observed that, when optogenetically activated, POAPAG neurons act directly through the PAG to elicit USV production in both male and female mice (Figure 2), confirming and extending a recent report that activation of GABAergic POA neurons elicits USVs in both sexes (Gao et al., 2019). These findings contrast with the behavioral observation that female mice in general produce fewer USVs than males. For example, female mice produce only about 1/5 of the total USVs recorded during male-female courtship interactions (Neunuebel et al., 2015), and we observed that female mice vocalize at lower rates than males when encountering novel female social partners (unpublished observations). Taken together, these findings suggest that different levels of POAPAG activity in males and females might contribute to sex differences in vocal behavior but, when strongly activated by optogenetic methods, POAPAG neurons in males and females are similarly potent in their ability to activate the downstream PAG vocal gating circuit and elicit USVs. Although it remains unknown which factors might drive differential activation of male and female POAPAG neurons, it is possible that sex differences in the density (Campi et al., 2013; Gorski et al., 1978; Orikasa and Sakuma, 2010; Panzica et al., 1996), synaptic organization (Raisman and Field, 1971), and gene expression patterns (Moffitt et al., 2018; Xu et al., 2012), including those of sex hormone receptors (Cao and Patisaul, 2011) of POA neurons, might all contribute to this sexually dimorphic behavior (for a review, see Lenz et al., 2012). More broadly, our findings add to a growing body of literature indicating that male and female brains contain latent circuits for sex-typical behaviors that can be unmasked by artificial neural activation but that are gated in a sex-specific manner during natural behavior (Clyne and Miesenböck, 2008; Gao et al., 2019; Rezával et al., 2016; Wei et al., 2018).

We also found that similar to activation of POAPAG neurons, optogenetic activation of Esr1+ POA neurons was sufficient to elicit USV production. Although a previous study reported that activation of Esr1+ POA neurons promotes mounting (Wei et al., 2018), we failed to observe mounting when we optogenetically activated either POAPAG neurons or Esr1+ POA neurons in male and female mice. Although the reasons for this discrepancy remain uncertain, one possibility is that our use of lower intensity optical stimulation can account for this difference (3–5 mW, 10–20 Hz vs. 10 mW, 40 Hz in Wei et al., 2018), and that the level of Esr1+ POA neuronal activation required to elicit USV production is lower than the threshold to elicit mounting. An interesting possibility is that different projection-defined subsets of Esr1+ POA neurons contribute to distinct aspects of courtship behavior, similar to what has been described for the contribution of projection-defined subsets of galanin-expressing POA neurons to distinct aspects of parental behavior (Kohl et al., 2018; Wu et al., 2014). Notably, though, a recent study found that ablation of VGAT+ POA neurons did not affect the numbers of social USVs produced by male and female mice, although the acoustic features of male courtship USVs were altered following ablation of these neurons (Gao et al., 2019). In contrast, ablation or silencing of POA neurons greatly reduces non-vocal consummatory courtship behaviors including mounting and ejaculation (Bean et al., 1981; Floody, 1989; Wei et al., 2018). These findings are consistent with the idea that POAPAG neurons promote the production of USVs during later stages of courtship (Gao et al., 2019), which differ acoustically from USVs produced in earlier phases of courtship (Hanson and Hurley, 2012; Keesom et al., 2017; Matsumoto and Okanoya, 2016; White et al., 1998). These findings also suggest that other neuronal populations that lie upstream of the PAG vocal gating circuit, and that are potentially interconnected with the POA, serve to promote USV production during the early phases of courtship and in other behavioral contexts.

Here we employ a newly described VAE-based unsupervised modeling method to compare the acoustic features of optogenetically elicited USVs to each animal’s repertoire of female-directed USVs. Although synchronous optogenetic activation of POAPAG neurons is likely quite different from the natural activity patterns of these neurons, the majority of optogenetically elicited vocalizations fall within the distribution of naturally produced USVs. This finding provides further experimental support for a model in which the PAG-USV neurons that are disinhibited by input from the POA gate USV production but do not directly pattern the acoustic content of vocalizations. The VAE also allowed us to identify and interrogate the acoustic features of optogenetically elicited vocalizations that fell outside the natural acoustic distribution. We found that these unusual optogenetically elicited USVs were louder and greater in frequency bandwidth, which we speculate may arise because synchronous optogenetic activation of POA neurons activates the PAG vocal gating circuit more strongly than occurs during natural behavior. However, whether altering the intensity or duration of POA stimulation systematically influences the acoustic features of optogenetically-elicited USVs remains to be tested. Another possibility is that because the mice in our experiments were singly tested in the absence of social partners and thus typically were not moving at high speeds during optogenetic stimulation, the USVs elicited by optogenetic stimulation of POAPAG neurons may be more similar acoustically to spontaneous USVs emitted in response to stationary social cues, such as female urine, rather than in response to a mobile female partner. Interestingly, a previous study found that the USVs produced by males in response to female urine were louder and had greater frequency bandwidth than those produced to female social partners (Chabout et al., 2015), reminiscent of the difference between female-directed and optogenetically evoked USVs in our dataset.

The current study also identifies a novel population of GABAergic AmgC/M-PAG neurons that lie at a boundary zone between the CeA and the medial amygdala and that project to PAG-USV neurons (i.e. AmgC/M-PAG neurons). Although this population of cells remains to be characterized comprehensively at a molecular and physiological level, our data show that transiently activating these neurons transiently suppresses USV production without driving fearful or aversive responses. Additionally, optogenetically activating AmgC/M-PAG neurons suppresses vocalization without interrupting non-vocal courtship behaviors more generally, providing additional support for the idea that PAG-USV cells are specialized neurons that gate USV production but that do not control non-vocal aspects of courtship.

We found that AmgC/M-PAG neurons make inhibitory synapses on PAG-USV neurons, which in turn gate vocalizations by exciting downstream vocal-respiratory pattern generating circuits (Tschida et al., 2019). Thus, the AmgC/M to PAG pathway provides a monosynaptic substrate through which vocalizations can be rapidly and effectively suppressed. We anticipate that such descending inhibitory inputs onto PAG-USV neurons act rapidly to suppress vocalization in behavioral contexts (in the presence of predators, conspecific competitors, etc.) in which vocalizing is risky or otherwise adverse, although this idea remains to be tested. We also note that while optogenetic activation of AmgC/M-PAG neurons transiently suppressed vocalization without obvious effects on non-vocal social behaviors and movement, it is possible that AmgC/M projections to other PAG cell types modulate diverse behaviors in addition to vocalization. Although POAPAG neurons are also GABAergic, we found that optogenetically activating these neurons promotes rather than suppresses USV production, likely through a disynaptic disinhibition of PAG-USV neurons mediated by local PAG interneurons. Consistent with the idea that disinhibition within the PAG is important for vocal production, work in primates has shown that pharmacological blockade of GABA receptors lowers the threshold for vocalization and elicits spontaneous vocalizations as well (Forcelli et al., 2017; Jürgens, 1994; Lu and Jürgens, 1993). Indeed, disinhibition of glutamatergic projection neurons has emerged as a prominent circuit motif within the PAG for releasing a variety of behaviors, including freezing (Tovote et al., 2016), pup grooming (Kohl et al., 2018), and antinociception (Morgan and Clayton, 2005). Our results support a model in which PAG-USV neuronal activity is tightly regulated by descending inputs as well as inputs from local GABAergic PAG neurons, which in turn integrate a variety of behaviorally relevant forebrain inputs to appropriately gate PAG-USV activity and hence USV production. More generally, such disinhibitory circuit motifs in the PAG may provide a failsafe mechanism that carefully regulates the behavioral contexts in which crucial but potentially costly behaviors, including vocalization, are produced.

By exploiting selective genetic access to PAG-USV neurons as a point of entry into central circuits for social and courtship vocalizations, we have begun to map the brain-wide architecture and synaptic organization of circuitry for a complex, natural behavior. In addition to the inputs from the POA and AmgC/M that were the focus of this study, our transsynaptic tracing identified a number of forebrain regions whose projections converge onto the PAG vocal gating circuit, consistent with the idea that the PAG integrates a wide variety of social, environmental, and interoceptive information to gate vocalization in a context-appropriate manner. Given that context-dependent vocal gating is a hallmark of human vocalizations, including speech (Stivers et al., 2009), it will be of great interest in future studies to more fully describe the neuronal populations whose inputs to the PAG shape vocal behavior. We note that vocal behavior is not simply binary: in addition to deciding whether or not to vocalize, an animal must produce vocalizations that are appropriate for a given situation. The elucidation of circuit and synaptic mechanisms through which forebrain inputs to the PAG vocal gating circuit influence USV production represents an important first step toward understanding how forebrain-to-midbrain circuits regulate the production of vocalizations across different behavioral contexts to enable effective communication.

Materials and methods

Key resources table.

Reagent type
(species) or
resource
Designation Source or
reference
Identifiers Additional information
Strain, strain background (Mus musculus, C57BL/6J) C57 Jackson Labs RRID:IMSR_JAX:000664
Strain, strain background (Mus musculus, B6N.129S6(Cg)-Esr1tm1.1(cre)And/J) Esr1-Cre Jackson Labs RRID:IMSR_JAX:017911
 Strain, strain background (Mus musculus, B6J.129S6(FVB)-Slc32a1tm2(cre)Lowl/
MwarJ)
VGAT-Cre Jackson Labs RRID:IMSR_JAX:016962
Strain, strain background (Mus musculus, B6;129S6-Gt(ROSA)26Sortim14(Cag-tdTomato)Hze/J) Ai14 Jackson Labs RRID:IMSR_JAX:007908
Strain, strain background (Mus musculus, B6;129-Fostm1.1Fawa/J) Fos-dsTVA Jackson Labs RRID:IMSR_JAX:027831
Recombinant DNA reagent AAV2/1-hSyn-Flex-Chr2-eYFP Addgene (K. Deisseroth) RRID:Addgene_26973
Recombinant DNA reagent AAV-pgk-retro-Cre Addgene (P. Aebischer) RRID:Addgene_24593
Recombinant DNA reagent AAV2/1-pCAG-flex-GFP Addgene (H. Zeng) RRID:Addgene_51502
Recombinant DNA reagent AAV2/1-pCAG-flex-Tdtomato Addgene (H. Zeng) RRID:Addgene_51503
Recombinant DNA reagent AAV-flex-oG Duke Viral Vector Core
Recombinant DNA reagent EnvA-G-RV-GFP Rodriguez et al., 2017 (DOI: 10.1038/s41593-017-0012-1), Sakurai et al., 2016 (DOI: 10.1016/j.neuron.2016.10.015)
Recombinant DNA reagent CANE-RV-mCherry Rodriguez et al., 2017 (DOI: 10.1038/s41593-017-0012-1), Sakurai et al., 2016 (DOI: 10.1016/j.neuron.2016.10.015)
Recombinant DNA reagent AAV-flex-TVA-mCherry Rodriguez et al., 2017 (DOI: 10.1038/s41593-017-0012-1), Sakurai et al., 2016 (DOI: 10.1016/j.neuron.2016.10.015)
Commercial assay or kit HCR v3.0 Molecular Instruments
Chemical compound, drug Gabazine Tocris Cat# 1262 (10 µM)
Chemical compound, drug TTX Tocris Cat# 1069 (2 µM)
Chemical compound, drug 4AP Sigma-Aldrich Cat# 275875 (100 µM)
Software, algorithm MATLAB Mathworks RRID:SCR_001622
Software, algorithm ImageJ NIH RRID:SCR_003070
Software, algorithm ZEN Zeiss RRID:SCR_013672
Software, algorithm Spike7 CED RRID:SCR_000903
Software, algorithm pClamp Molecular Devices RRID:SCR_011323
Software, algorithm IGOR Pro WaveMetrics RRID:SCR_000325
Other NeuroTrace 435/455 Invitrogen/Thermo Fischer Scientific Cat# N21479 (1:500)

Contact for reagent and resource sharing

Further information and requests for resources and reagents should be directed to the co-corresponding authors, Katherine Tschida (kat227@cornell.edu) or Richard Mooney (mooney@neuro.duke.edu).

Experimental models and subject details

Animal statement

All experiments were conducted according to a protocol approved by the Duke University Institutional Animal Care and Use Committee (protocol # A227-17-09).

Animals

For optogenetic activation and axonal tracing experiments, the following mouse lines from Jackson labs were used: C57 (C57BL/6J, Jackson Labs, 000664), Esr1-Cre (B6N.129S6(Cg)-Esr1tm1.1(cre)And/J, Jackson Labs, 017911), VGAT-Cre (B6J.129S6(FVB)-Slc32a1tm2(cre)Lowl/MwarJ, Jackson Labs, 016962), Ai14 (B6;129S6-Gt(ROSA)26Sortim14(Cag-tdTomato)Hze/J, Jackson Labs, 007908). Fos-dsTVA mice (B6,129-Fostm1.1Fawa/J, Jackson Labs, 027831) were used for activity-dependent labeling of PAG-USV neurons employed in the transsynaptic tracing experiments and in whole-cell recording experiments. In a subset of whole-cell recording experiments, VGAT-Cre homozygous mice were crossed to Fos-dsTVA homozygous mice. Note that male Esr-1-Cre mice were often smaller and less healthy than their female littermates. While later weaning allowed them to grow to normal size, these animals still had lower survival rates after surgeries than any other animals used in this study, particularly when bilaterally implanting ferrules in the PAG for optogenetic stimulation of axon terminals.

Method details

Viruses

The following viruses and injection volumes were used: AAV2/1-hSyn-FLEX-ChR2-eYFP (Addgene), AAV-pgk-retro-Cre (Addgene), AAV-hsyn-retro-FLEX-ChR2 (Addgene), AAV-FLEX-GFP (Addgene), AAV-FLEX-tdTomato (Addgene), AAV-FLEX-oG (Duke Viral Vector Core). EnvA-ΔG-RV-GFP, CANE-RV-mCherry, and AAV-FLEX-TVA-mCherry were produced in house as previously described (Rodriguez et al., 2017; Sakurai et al., 2016; Tschida et al., 2019). The final injection coordinates were as follows: POA, AP = 0.14 mm, ML = 0.3 mm, DV = 5.5 mm; AmgC/M, AP = −1.5 mm, ML = 2.3 mm, DV = 4.6 mm; PAG, AP = −4.7 mm, ML = 0.7 mm, DV = 1.75 mm. Viruses were pressure-injected with a Nanoject II (Drummond) at a rate of 4.6 nL every 15 s.

Transsynaptic tracing from PAG-USV and GABAergic PAG neurons

To selectively infect PAG-USV neurons with viruses, ds-Fos-TVA males were given social experience with a female (30–60 min) that resulted in high levels of USV production (500–5000 USVs total). Males were then anesthetized (1.5–2% isoflurane), and the caudolateral PAG was targeted for viral injection. For transsynaptic tracing from PAG-USV neurons, the PAG was injected with a 4:1:1 mixture of CANE-LV-Cre, AAV-FLEX-TVA-mCherry, and AAV-FLEX-oG (total volume of 300 nL). After a wait time of 10–14 days, the PAG was then injected with EnvA-ΔG-RV-GFP (100 nL, diluted 1:5), and animals were sacrificed after waiting an additional 4–7 days.

To transsynaptically label inputs to GABAergic PAG neurons, the caudolateral PAG of VGAT-Cre mice was injected with a 1:1 mixture of AAV-FLEX-TVA-mCherry, and AAV-FLEX-oG (total volume of 100 nL). After a wait time of 10–14 days, the PAG was then injected with EnvA-ΔG-RV-GFP (100 nL, diluted 1:5), and animals were sacrificed after waiting an additional 4–7 days.

We note that because our goal was to identify long-range inputs onto PAG-USV and GABAergic PAG neurons, we used survival times that prioritized visualization of afferent cell bodies in distant locations rather than the integrity of the starter cell populations (which die off over time). Hence, we do not include quantification of starter cell populations within the PAG, as these cannot be meaningfully related to the numbers of cells that provide monosynaptic input to PAG-USV and GABAergic PAG neurons.

In vivo optogenetic stimulation

Custom-made or commercially available (RWD) optogenetic ferrules were implanted in the same surgeries as viral injection just above target brain locations and were fixed to the skull using Metabond (Parkell). Neurons or their axon terminals were optogenetically activated with illumination from a 473 nm laser (3–15 mW) at 10–20 Hz (50 ms pulses, 2–10 s total) or with phasic laser pulses (1–2 s duration). Laser stimuli were driven by computer-controlled voltage pulses (Spike 7, CED). For stimulation of POA cell bodies or axon terminals, the laser was triggered manually at regular intervals while the animal was alone in the chamber. For stimulation of AmgC/M neurons or terminals, the laser was triggered manually each time the mouse began vocalizing for several seconds toward a female social partner.

Post-hoc visualization of viral labeling

Mice were deeply anesthetized with isoflurane and then transcardially perfused with ice-cold 4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4 (4% PFA). Dissected brain samples were post-fixed overnight in 4% PFA at 4°C, cryoprotected in a 30% sucrose solution in PBS at 4°C for 48 hr, frozen in Tissue-Tek O.C.T. Compound (Sakura), and stored at –80°C until sectioning. To visualize viral labeling post-hoc, brains were cut into 80 µm coronal sections, rinsed 3x in PBS, and processed for 24 hr at four degrees with NeuroTrace (1:500 Invitrogen) in PBS containing 0.3% Triton-X. Tissue sections rinsed again 3 × 10 mins. in PBS, mounted on slides, and coverslipped with Fluoromount-G (Southern Biotech). After drying, slides were imaged with a 10x objective on a Zeiss 700 laser scanning confocal microscope.

Floating section two-color in situ hybridization

In situ hybridization was performed using hybridization chain reaction (HCR v3.0, Molecular Instruments). Dissected brain samples were post-fixed overnight in 4% PFA at 4°C, cryoprotected in a 30% sucrose solution in RNAse-free PBS (i.e. DEPC-PBS) at 4°C for 48 hr, frozen in Tissue-Tek O.C.T. Compound (Sakura), and stored at –80°C until sectioning. 80 µm thick coronal floating sections were collected into a sterile 24-well plate in DEPC-PBS, fixed again briefly for 5 min in 4% PFA, then placed in 70% EtOH in DEPC-PBS overnight. Sections were rinsed in DEPC-PBS, incubated for 45 min in 5% SDS in DEPC-PBS, rinsed and incubated in 2x SSCT, pre-incubated in HCR hybridization buffer at 37°C, and then placed in HCR hybridization buffer containing RNA probes overnight at 37°C. The next day, sections were rinsed 4 × 15 min at 37°C in HCR probe wash buffer, rinsed with 2X SSCT, pre-incubated with HCR amplification buffer, then incubated in HCR amplification buffer containing HCR amplifiers at room temperature for ~48 hr. On the final day, sections were rinsed in 2x SSCT, counterstained with DAPI (Thermo Fisher, 1:5000), rinsed again with 2x SSCT, then mounted on slides and coverslipped with Fluoromount-G (Southern Biotech). After drying, slides were imaged with a 10x or 20x objective on a Zeiss 700 laser scanning confocal microscope.

Cells were scored from two to seven sections of tissue per brain region from each animal, and the absence or presence of staining within cells was quantified manually by comparing labeling within cells to background labeling in nearby regions known to have lower levels of expression of a given RNA transcript than the region of interest. Neighboring control regions with lower levels of transcript expression were present in the same coronal sections as the regions of interest and were determined by consulting the Allen Brain Atlas ISH Data (https://mouse.brain-map.org/search/index; experiment 72081554 for VGAT expression, experiment 79591677 for Esr1 expression, and experiment 73818754 for VGlut2 expression). These control regions were as follows for the following target regions and transcripts: (1a) POA VGAT: control region, fornix; (1b) AmgC/M VGAT: control region, thalamus; (1 c) CeA VGAT: control region, thalamus; (2a) POA Esr1: control region, fornix; (3a) POA VGlut2: control region, caudate putamen; (3b) AmgC/M VGlut2: caudate puteman; (3 c) CeA VGlut2: control region, caudate putamen.

USV recording and analysis

To elicit USVs, single-housed males or females were presented with a freely moving female, either in a novel test chamber or in the home cage. USVs were recorded with an ultrasonic microphone (Avisoft, CMPA/CM16), amplified (Presonus TubePreV2), and digitized at 250 kHz (Spike 7, CED). USVs were detected using codes modified from the Holy lab (http://holylab.wustl.edu/) using the following parameters (mean frequency >45 kHz; spectral purity >0.3; spectral discontinuity <0.85; min. USV duration = 5 ms; minimum inter-syllable interval = 30 ms). To elicit USVs for tagging of PAG-USV neurons using CANE (for transsynaptic tracing and slice experiments), FosTVA males were given social experience with a female (30–60 min session), either in their home cage fitted with an acoustically permeable lid or in a test chamber that had no lid and allowed easy microphone access. Sixty minutes from the start of the session, FosTVA males were anesthetized and taken for injection of the PAG with viruses (see above), such that injections began approximately 2 hr from the start of USV production.

Real-time place preference tests

Mice were lightly anesthetized to connect the 473 nm laser to the optogenetic ferrule, then mice were placed in the center of a custom-made two-sided test chamber, illuminated with infrared light only. The side of the chamber in which each mouse received optogenetic stimulation was chosen randomly for each place preference test. When the mouse was in the selected side, it received continuous 10 Hz optogenetic stimulation using the minimum laser power that had either elicited or inhibited USV production for that same mouse. Place preference was scored over a 20-min test period as the proportion of the total time that the mouse spent in the stimulated side of the chamber.

Quantification of optogenetically elicited body movements

The mouse’s position was measured using custom Matlab codes that detected and tracked the centroid of the mouse’s body position across video frames (Logitech webcam, 30 frames per second), and speed of movement was calculated as the change in position across pairs of frames. To align movement with optogenetic activation of POA or AmgC/M neurons, we first estimated the temporal offset between the webcam video and USV audio by calculating the time of the peak cross-covariance between the high-pass filtered webcam audio and the low-pass filtered USV audio. This offset was then used to align the mouse’s movement to the onset of each optogenetic laser stimulus. To measure the effects of optogenetic stimulation on the distance between an interacting male and female mouse, the position of each mouse was tracked manually in every 6th frame, and the distance between mice was scored as the distance from the center of the male’s head to the base of the female’s tail.

Comparison of acoustic features of optogenetically elicited USVs to female-directed USVs

A total of 52,821 USV syllables were segmented automatically with MUPET 2.0 using default parameter settings (Van Segbroeck et al., 2017). Of these syllables, 23,805 came from recordings of 15 mice recorded under both natural and optogenetic conditions (56% natural USVs). The remaining 29,016 syllables came from recordings of a control group of 10 mice used to establish across-day syllable repertoire variability. False positives (noise) from the experimental group were manually removed by visual inspection of spectrograms, with 79% of the original syllables retained. Syllables were analyzed using Autoencoded Vocal Analysis v0.2 (Goffinet et al., 2019), a Python package for generating low-dimensional latent descriptions of animal vocalizations using a VAE (Kingma and Welling, 2013). Briefly, the VAE jointly trains two probabilistic maps: an encoder and a decoder. Spectrograms are encoded into low-dimensional ‘latent’ representations which can be subsequently decoded to approximately reconstruct the original spectrograms. Both encoding and decoding distributions are parameterized by convolutional neural networks. We trained a VAE on spectrograms of single USV syllables from both experimental and control groups using the following parameters: min_freq = 30e3, max_freq = 110e3, nperseg = 1024, noverlap = 512, spec_min_val = −5.0, spec_max_val = −1.5, mel=False, time_stretch=True, within_syll_normalize=False. Each input spectrogram was 128-by-128 pixels (16,000 dimensions) and the VAE converged on a parsimonious representation of only five dimensions. To visualize these five-dimensional spaces, the latent representations of syllable spectrograms are projected into two dimensions using the UMAP algorithm (McInnes et al., 2018). To quantify differences in syllable repertoires, we estimate the Maximum Mean Discrepancy (Gretton et al., 2012) between distributions of latent syllable representations as in Goffinet et al., 2019. First, a baseline level of variability in syllable repertoire was established for each mouse by estimating MMD between the first and second halves of female-directed syllables emitted in a recording session. Then MMD between each mouse's natural and optogenetically elicited repertoires was estimated. A paired comparison test revealed significantly larger differences between optogenetic and natural repertoires than expected by variability within the natural condition recording sessions alone (two-sided, continuity-corrected Wilcoxon signed-rank test, W = 9, p<5e-3). We then estimated MMD between female-directed syllable repertoires recorded on different days, using the set of 10 control mice.

Whole-cell recordings

Mice that received viral injections 2–4 weeks prior were deeply anesthetized with isoflurane and standard procedures were used to prepare 300-µm-thick coronal slices. The brain was dissected in ice-cold ACSF containing the following (in mM): 119 NaCl, 2.5 KCl, 1.30 MgCl2, 2.5 CaCl2, 26.2 NaHCO3, 1.0 NaHPO4-H2O, and 11.0 dextrose and bubbled with 95% O2/5% CO2. The brain was mounted on an agar block and sliced in ice-cold ACSF with a vibrating-blade microtome (Leica). Slices were incubated for 15 min at 32°C in a bath of NMDG recovery solution containing the following (in mM): 93.0 NMDG, 2.5 KCl, 1.2 NaH2PO4, 30.0 NaHCO3, 20.0 HEPES, 25.0 glucose, 2.0 thiourea, 5.0 Na L-ascorbate, 2.0 Na-pyruvate, 10.0 MgSO4 7H2O, 0.5 CaCl2, and 95.0 HCl. Slices were then moved to a bath of HEPES storage solution containing the following (in mM): 93.0 NaCl, 2.5 KCl, 1.2 NaH2PO4, 30.0 NaHCO3, 20.0 HEPES, 25.0 glucose, 2.0 thiourea, 5.0 Na L-ascorbate, 2.0 Na-pyruvate, 10.0 MgSO4 7H2O, and 0.5 CaCl2, and allowed to gradually reach room temperature over the course of 1 hr, where they remained for the duration. Recordings were performed in ACSF at a temperature of 32°C. For voltage clamp experiments patch electrodes (4–8 MΩ) were filled with cesium internal solution containing the following (in mM): 130 cesium methanesulfonate, 5 QX-314 Br, 10 HEPES, 8 TEA-Cl, 0.2 EGTA, 4 ATP-Mg salt, 0.3 GTP-Na salt, and 10 phosphocreatine. Recordings were made using a Multiclamp 700B amplifier whose output was digitized at 10 kHz (Digidata 1440A). Series resistance was <25 MΩ and was compensated up to 90%. Signals were analyzed using Igor Pro (Wavemetrics). Neurons were targeted using interference contrast and epifluorescence to visualize fluorescent indicators previously expressed via viral injection. ChR2-expressing axon terminals were stimulated by 5–20 ms laser pulses (3–10 mW) from a 473 nm laser delivered via fiber optic inside the recording pipette (Optopatcher, A-M Systems). To confirm the direct nature of optogenetically evoked currents 2 µM TTX (Tocris) and 100 µM 4AP (Sigma-Aldrich) were added to the ACSF and perfused onto slices. To confirm that evoked currents were GABAergic, 10 µM gabazine (Tocris) was applied. Pharmacological agents including were bath applied for 10 min before making recordings.

Code availability

All custom-written Matlab codes used in this study will be made publicly available at the Duke Digital Repository. The latest version of Autoencoded Vocal Analysis, the Python package used to generate, plot, and analyze latent features of mouse USVs, is freely available online: https://github.com/jackgoffinet/autoencoded-vocal-analysis.

Quantification and statistical analyses

Statistics

Parametric, two-sided statistical comparisons were used in all analyses unless otherwise noted (alpha = 0.05). No statistical methods were used to predetermine sample sizes. Error bars represent standard error of the mean unless otherwise noted. Mice were selected at random for inclusion into either experimental or control groups for optogenetic experiments. Mice were only excluded from analysis in cases in which viral injections were not targeted accurately, or in cases with absent or poor viral expression.

Acknowledgements

Thanks to Michael Booze and Bao-Xia Han for additional mouse husbandry, thanks to Jun Takatoh for help with HCR in situ hybridization, and thanks to Shengli Zhao for providing CANE-related viruses. This work is supported by NIH grants DC 013826 (to RM) and MH 117778 (to FW and RM).

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Contributor Information

Katherine Tschida, Email: kat227@cornell.edu.

Richard Mooney, Email: mooney@neuro.duke.edu.

Catherine Emily Carr, University of Maryland, United States.

Catherine Dulac, Harvard University, United States.

Funding Information

This paper was supported by the following grants:

  • National Institutes of Health R01 DC013826 to Richard Mooney.

  • National Institutes of Health R01 MH117778 to Fan Wang, Richard Mooney.

  • National Institutes of Health F31 DC017879 to Valerie Michael.

Additional information

Competing interests

No competing interests declared.

Author contributions

Conceptualization, Data curation, Software, Formal analysis, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing.

Software, Formal analysis, Methodology.

Software, Formal analysis, Methodology.

Conceptualization, Resources, Supervision, Funding acquisition, Project administration, Writing - review and editing.

Conceptualization, Data curation, Software, Formal analysis, Supervision, Investigation, Visualization, Methodology, Writing - original draft, Project administration, Writing - review and editing.

Conceptualization, Resources, Supervision, Funding acquisition, Writing - original draft, Project administration, Writing - review and editing.

Ethics

Animal experimentation: All experiments were conducted according to protocols approved by the Duke University Institutional Animal Care and Use Committee protocol (# A227-17-09).

Additional files

Transparent reporting form

Data availability

Data have been deposited to the Duke Research Data Repository, under the https://doi.org/10.7924/r4cz38d99. We have deposited 4 types of data in the repository: (1) confocal microscope images of in situ hybridization, (2) audio and video files from the mice used in this study, (3) slice electrophysiology data, and (4) custom Matlab codes used for data analysis. All other data analyzed in this study are included in the manuscript and supporting files.

The following dataset was generated:

Michael V, Goffinet J, Pearson J, Wang F, Tschida K, Mooney R. 2020. Data and scripts from: Circuit and synaptic organization of forebrain-to-midbrain pathways that promote and suppress vocalization. Duke Research Data Repository.

References

  1. Alger SJ, Riters LV. Lesions to the medial preoptic nucleus differentially affect singing and nest box-directed behaviors within and outside of the breeding season in european starlings (Sturnus vulgaris) Behavioral Neuroscience. 2006;120:1326–1336. doi: 10.1037/0735-7044.120.6.1326. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Balthazart J, Ball GF. Topography in the preoptic region: differential regulation of appetitive and consummatory male sexual behaviors. Frontiers in Neuroendocrinology. 2007;28:161–178. doi: 10.1016/j.yfrne.2007.05.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Bandler R, Shipley MT. Columnar organization in the midbrain periaqueductal gray: modules for emotional expression? Trends in Neurosciences. 1994;17:379–389. doi: 10.1016/0166-2236(94)90047-7. [DOI] [PubMed] [Google Scholar]
  4. Bean NY, Nunez AA, Conner R. Effects of medial preoptic lesions on male mouse ultrasonic vocalizations and copulatory behavior. Brain Research Bulletin. 1981;6:109–112. doi: 10.1016/S0361-9230(81)80033-0. [DOI] [PubMed] [Google Scholar]
  5. Bennett PJG, Maier E, Brecht M. Involvement of rat posterior prelimbic and cingulate area 2 in vocalization control. European Journal of Neuroscience. 2019;50:3164–3180. doi: 10.1111/ejn.14477. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Campi KL, Jameson CE, Trainor BC. Sexual dimorphism in the brain of the monogamous California mouse (Peromyscus californicus) Brain, Behavior and Evolution. 2013;81:236–249. doi: 10.1159/000353260. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Cao J, Patisaul HB. Sexually dimorphic expression of hypothalamic estrogen receptors α and β and kiss1 in neonatal male and female rats. The Journal of Comparative Neurology. 2011;519:2954–2977. doi: 10.1002/cne.22648. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Carrive P. The periaqueductal gray and defensive behavior: functional representation and neuronal organization. Behavioural Brain Research. 1993;58:27–47. doi: 10.1016/0166-4328(93)90088-8. [DOI] [PubMed] [Google Scholar]
  9. Chabout J, Sarkar A, Dunson DB, Jarvis ED. Male mice song syntax depends on social contexts and influences female preferences. Frontiers in Behavioral Neuroscience. 2015;9:76. doi: 10.3389/fnbeh.2015.00076. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Chen P, Hong W. Neural circuit mechanisms of social behavior. Neuron. 2018;98:16–30. doi: 10.1016/j.neuron.2018.02.026. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Clyne JD, Miesenböck G. Sex-specific control and tuning of the pattern generator for courtship song in Drosophila. Cell. 2008;133:354–363. doi: 10.1016/j.cell.2008.01.050. [DOI] [PubMed] [Google Scholar]
  12. Dujardin E, Jürgens U. Call type-specific differences in vocalization-related afferents to the periaqueductal gray of squirrel monkeys (Saimiri sciureus) Behavioural Brain Research. 2006;168:23–36. doi: 10.1016/j.bbr.2005.10.006. [DOI] [PubMed] [Google Scholar]
  13. Duvarci S, Pare D. Amygdala microcircuits controlling learned fear. Neuron. 2014;82:966–980. doi: 10.1016/j.neuron.2014.04.042. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Ehrlich I, Humeau Y, Grenier F, Ciocchi S, Herry C, Lüthi A. Amygdala inhibitory circuits and the control of fear memory. Neuron. 2009;62:757–771. doi: 10.1016/j.neuron.2009.05.026. [DOI] [PubMed] [Google Scholar]
  15. Evans DA, Stempel AV, Vale R, Ruehle S, Lefler Y, Branco T. A synaptic threshold mechanism for computing escape decisions. Nature. 2018;558:590–594. doi: 10.1038/s41586-018-0244-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Fang YY, Yamaguchi T, Song SC, Tritsch NX, Lin D. A hypothalamic midbrain pathway essential for driving maternal behaviors. Neuron. 2018;98:192–207. doi: 10.1016/j.neuron.2018.02.019. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Fecteau S, Belin P, Joanette Y, Armony JL. Amygdala responses to nonlinguistic emotional vocalizations. NeuroImage. 2007;36:480–487. doi: 10.1016/j.neuroimage.2007.02.043. [DOI] [PubMed] [Google Scholar]
  18. Fenzl T, Schuller G. Periaqueductal gray and the region of the paralemniscal area have different functions in the control of vocalization in the neotropical bat, Phyllostomus discolor. European Journal of Neuroscience. 2002;16:1974–1986. doi: 10.1046/j.1460-9568.2002.02261.x. [DOI] [PubMed] [Google Scholar]
  19. Floody OR. Dissociation of hypothalamic effects on ultrasound production and copulation. Physiology & Behavior. 1989;46:299–307. doi: 10.1016/0031-9384(89)90271-0. [DOI] [PubMed] [Google Scholar]
  20. Floody OR, Cooper TT, Albers HE. Injection of oxytocin into the medial preoptic-anterior hypothalamus increases ultrasound production by female hamsters. Peptides. 1998;19:833–839. doi: 10.1016/S0196-9781(98)00029-1. [DOI] [PubMed] [Google Scholar]
  21. Floody OR. Effects on hamster vocalization and aggression of carbachol injections into the MPOA/AH. Physiology & Behavior. 2009;96:294–299. doi: 10.1016/j.physbeh.2008.10.012. [DOI] [PubMed] [Google Scholar]
  22. Forcelli PA, Waguespack HF, Malkova L. Defensive vocalizations and motor asymmetry triggered by disinhibition of the periaqueductal gray in Non-human primates. Frontiers in Neuroscience. 2017;11:163. doi: 10.3389/fnins.2017.00163. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Fu XW, Brudzynski SM. High-frequency ultrasonic vocalization induced by intracerebral glutamate in rats. Pharmacology, Biochemistry, and Behavior. 1994;49:835–841. doi: 10.1016/0091-3057(94)90231-3. [DOI] [PubMed] [Google Scholar]
  24. Gadziola MA, Shanbhag SJ, Wenstrup JJ. Two distinct representations of social vocalizations in the basolateral amygdala. Journal of Neurophysiology. 2016;115:868–886. doi: 10.1152/jn.00953.2015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Gao S-C, Wei Y-C, Wang S-R, Xu X-H. Medial preoptic area modulates courtship ultrasonic vocalization in adult male mice. Neuroscience Bulletin. 2019;35:697–708. doi: 10.1007/s12264-019-00365-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Goffinet J, Mooney R, Pearson J. Inferring low-dimensional latent descriptions of animal vocalizations. bioRxiv. 2019 doi: 10.1101/811661. [DOI]
  27. Gorski RA, Gordon JH, Shryne JE, Southam AM. Evidence for a morphological sex difference within the medial preoptic area of the rat brain. Brain Research. 1978;148:333–346. doi: 10.1016/0006-8993(78)90723-0. [DOI] [PubMed] [Google Scholar]
  28. Gothard KM. Multidimensional processing in the amygdala. Nature Reviews Neuroscience. 2020;21:565–575. doi: 10.1038/s41583-020-0350-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Green DB, Shackleton TM, Grimsley JMS, Zobay O, Palmer AR, Wallace MN. Communication calls produced by electrical stimulation of four structures in the guinea pig brain. PLOS ONE. 2018;13:e0194091. doi: 10.1371/journal.pone.0194091. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Gretton A, Borgwardt KM, Rasch MJ, Scholkopf B, Smola A. A kernel Two-Sample test. Journal of Machine Learning Research. 2012;13:723–773. [Google Scholar]
  31. Hall IC, Ballagh IH, Kelley DB. The Xenopus amygdala mediates socially appropriate vocal communication signals. Journal of Neuroscience. 2013;33:14534–14548. doi: 10.1523/JNEUROSCI.1190-13.2013. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Hanson JL, Hurley LM. Female presence and estrous state influence mouse ultrasonic courtship vocalizations. PLOS ONE. 2012;7:e40782. doi: 10.1371/journal.pone.0040782. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Holstege G. The periaqueductal gray controls brainstem emotional motor systems including respiration. Progress in Brain Research. 2014;209:379–405. doi: 10.1016/B978-0-444-63274-6.00020-5. [DOI] [PubMed] [Google Scholar]
  34. Holy TE, Guo Z. Ultrasonic songs of male mice. PLOS Biology. 2005;3:e386. doi: 10.1371/journal.pbio.0030386. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Janak PH, Tye KM. From circuits to behaviour in the amygdala. Nature. 2015;517:284–292. doi: 10.1038/nature14188. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Jürgens U. Amygdalar vocalization pathways in the squirrel monkey. Brain Research. 1982;241:189–196. doi: 10.1016/0006-8993(82)91055-1. [DOI] [PubMed] [Google Scholar]
  37. Jürgens U. The role of the periaqueductal grey in vocal behaviour. Behavioural Brain Research. 1994;62:107–117. doi: 10.1016/0166-4328(94)90017-5. [DOI] [PubMed] [Google Scholar]
  38. Jürgens U. Neural pathways underlying vocal control. Neuroscience & Biobehavioral Reviews. 2002;26:235–258. doi: 10.1016/S0149-7634(01)00068-9. [DOI] [PubMed] [Google Scholar]
  39. Jürgens U. The neural control of vocalization in mammals: a review. Journal of Voice. 2009;23:1–10. doi: 10.1016/j.jvoice.2007.07.005. [DOI] [PubMed] [Google Scholar]
  40. Jürgens U, Ploog D. Cerebral representation of vocalization in the squirrel monkey. Experimental Brain Research. 1970;10:532–554. doi: 10.1007/BF00234269. [DOI] [PubMed] [Google Scholar]
  41. Jürgens U, Pratt R. Role of the periaqueductal grey in vocal expression of emotion. Brain Research. 1979;167:367–378. doi: 10.1016/0006-8993(79)90830-8. [DOI] [PubMed] [Google Scholar]
  42. Jürgens U, Richter K. Glutamate-induced vocalization in the squirrel monkey. Brain Research. 1986;373:349–358. doi: 10.1016/0006-8993(86)90349-5. [DOI] [PubMed] [Google Scholar]
  43. Keesom SM, Finton CJ, Sell GL, Hurley LM. Early-Life social isolation influences mouse ultrasonic vocalizations during Male-Male social encounters. PLOS ONE. 2017;12:e0169705. doi: 10.1371/journal.pone.0169705. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Kingma DP, Welling M. Auto-Encoding variational bayes. arXiv. 2013 https://arxiv.org/abs/1312.6114
  45. Kohl J, Babayan BM, Rubinstein ND, Autry AE, Marin-Rodriguez B, Kapoor V, Miyamishi K, Zweifel LS, Luo L, Uchida N, Dulac C. Functional circuit architecture underlying parental behaviour. Nature. 2018;556:326–331. doi: 10.1038/s41586-018-0027-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Kyuhou S, Gemba H. Two vocalization-related subregions in the midbrain periaqueductal gray of the guinea pig. Neuroreport. 1998;9:1607–1610. doi: 10.1097/00001756-199805110-00064. [DOI] [PubMed] [Google Scholar]
  47. LeDoux J. The amygdala. Current Biology. 2007;17:R868–R874. doi: 10.1016/j.cub.2007.08.005. [DOI] [PubMed] [Google Scholar]
  48. Lenz KM, Nugent BM, McCarthy MM. Sexual differentiation of the rodent brain: dogma and beyond. Frontiers in Neuroscience. 2012;6:26. doi: 10.3389/fnins.2012.00026. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Lin D, Boyle MP, Dollar P, Lee H, Lein ES, Perona P, Anderson DJ. Functional identification of an aggression locus in the mouse hypothalamus. Nature. 2011;470:221–226. doi: 10.1038/nature09736. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Lu CL, Jürgens U. Effects of chemical stimulation in the periaqueductal gray on vocalization in the squirrel monkey. Brain Research Bulletin. 1993;32:143–151. doi: 10.1016/0361-9230(93)90068-m. [DOI] [PubMed] [Google Scholar]
  51. Ma J, Kanwal JS. Stimulation of the basal and central amygdala in the mustached bat triggers echolocation and agonistic vocalizations within multimodal output. Frontiers in Physiology. 2014;5:55. doi: 10.3389/fphys.2014.00055. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Maggio JC, Whitney G. Ultrasonic vocalizing by adult female mice (Mus musculus) Journal of Comparative Psychology. 1985;99:420–436. doi: 10.1037/0735-7036.99.4.420. [DOI] [PubMed] [Google Scholar]
  53. Manteuffel G, Schön PC, Düpjan S, Tuchscherer A, Bellmann O. Acetylcholine injection into the amygdala elicits vocalization in domestic pigs (Sus scrofa) Behavioural Brain Research. 2007;178:177–180. doi: 10.1016/j.bbr.2006.12.007. [DOI] [PubMed] [Google Scholar]
  54. Matsumoto YK, Okanoya K, Seki Y. Effects of amygdala lesions on male mouse ultrasonic vocalizations and copulatory behaviour. NeuroReport. 2012;23:676–680. doi: 10.1097/WNR.0b013e3283557eea. [DOI] [PubMed] [Google Scholar]
  55. Matsumoto YK, Okanoya K. Phase-Specific vocalizations of male mice at the initial encounter during the courtship sequence. PLOS ONE. 2016;11:e0147102. doi: 10.1371/journal.pone.0147102. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. McInnes L, Healy J, Melville J. UMAP: uniform manifold approximation and projection for dimension reduction. arXiv. 2018 https://arxiv.org/abs/1802.03426
  57. McKinsey G, Ahmed OM, Shah NM. Neural control of sexually dimorphic social behaviors. Current Opinion in Physiology. 2018;6:89–95. doi: 10.1016/j.cophys.2018.08.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Moffitt JR, Bambah-Mukku D, Eichhorn SW, Vaughn E, Shekhar K, Perez JD, Rubinstein ND, Hao J, Regev A, Dulac C, Zhuang X. Molecular, spatial, and functional single-cell profiling of the hypothalamic preoptic region. Science. 2018;362:eaau5324. doi: 10.1126/science.aau5324. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Morgan MM, Clayton CC. Defensive behaviors evoked from the ventrolateral periaqueductal gray of the rat: comparison of opioid and GABA disinhibition. Behavioural Brain Research. 2005;164:61–66. doi: 10.1016/j.bbr.2005.05.009. [DOI] [PubMed] [Google Scholar]
  60. Neunuebel JP, Taylor AL, Arthur BJ, Egnor SER. Female mice ultrasonically interact with males during courtship displays. eLife. 2015;4:e06203. doi: 10.7554/eLife.06203. [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Newman SW. The medial extended amygdala in male reproductive behavior a node in the mammalian social behavior network. Annals of the New York Academy of Sciences. 1999;877:242–257. doi: 10.1111/j.1749-6632.1999.tb09271.x. [DOI] [PubMed] [Google Scholar]
  62. Nyby J. Elicitation of male mouse (Mus musculus) ultrasonic vocalizations: I. urinary cues. Journal of Comparative and Physiological Psychology. 1979;93:957–975. doi: 10.1037/h0077623. [DOI] [PubMed] [Google Scholar]
  63. Orikasa C, Sakuma Y. Estrogen configures sexual dimorphism in the preoptic area of C57BL/6J and ddN strains of mice. The Journal of Comparative Neurology. 2010;518:3618–3629. doi: 10.1002/cne.22419. [DOI] [PubMed] [Google Scholar]
  64. Panzica GC, Viglietti-Panzica C, Balthazart J. The sexually dimorphic medial preoptic nucleus of quail: a key brain area mediating steroid action on male sexual behavior. Frontiers in Neuroendocrinology. 1996;17:51–125. doi: 10.1006/frne.1996.0002. [DOI] [PubMed] [Google Scholar]
  65. Portfors CV, Perkel DJ. The role of ultrasonic vocalizations in mouse communication. Current Opinion in Neurobiology. 2014;28:115–120. doi: 10.1016/j.conb.2014.07.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  66. Raisman G, Field PM. Sexual dimorphism in the preoptic area of the rat. Science. 1971;173:731–733. doi: 10.1126/science.173.3998.731. [DOI] [PubMed] [Google Scholar]
  67. Rezával C, Pattnaik S, Pavlou HJ, Nojima T, Brüggemeier B, D'Souza LAD, Dweck HKM, Goodwin SF. Activation of latent courtship circuitry in the brain of Drosophila Females Induces Male-like Behaviors. Current Biology. 2016;26:2508–2515. doi: 10.1016/j.cub.2016.07.021. [DOI] [PMC free article] [PubMed] [Google Scholar]
  68. Riters LV, Ball GF. Lesions to the medial preoptic area affect singing in the male european starling (Sturnus vulgaris) Hormones and Behavior. 1999;36:276–286. doi: 10.1006/hbeh.1999.1549. [DOI] [PubMed] [Google Scholar]
  69. Rodriguez E, Sakurai K, Xu J, Chen Y, Toda K, Zhao S, Han BX, Ryu D, Yin H, Liedtke W, Wang F. A craniofacial-specific monosynaptic circuit enables heightened affective pain. Nature Neuroscience. 2017;20:1734–1743. doi: 10.1038/s41593-017-0012-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  70. Sainburg T, Theilman B, Thielk M, Gentner TQ. Parallels in the sequential organization of Birdsong and human speech. Nature Communications. 2019;10:3636. doi: 10.1038/s41467-019-11605-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  71. Sakurai K, Zhao S, Takatoh J, Rodriguez E, Lu J, Leavitt AD, Fu M, Han BX, Wang F. Capturing and manipulating activated neuronal ensembles with CANE delineates a hypothalamic Social-Fear circuit. Neuron. 2016;92:739–753. doi: 10.1016/j.neuron.2016.10.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  72. Schönfeld LM, Zech MP, Schäble S, Wöhr M, Kalenscher T. Lesions of the rat basolateral amygdala reduce the behavioral response to ultrasonic vocalizations. Behavioural Brain Research. 2020;378:112274. doi: 10.1016/j.bbr.2019.112274. [DOI] [PubMed] [Google Scholar]
  73. Siebert S, Jürgens U. Vocalization after periaqueductal grey inactivation with the GABA agonist muscimol in the squirrel monkey. Neuroscience Letters. 2003;340:111–114. doi: 10.1016/S0304-3940(03)00071-5. [DOI] [PubMed] [Google Scholar]
  74. Sternson SM. Hypothalamic survival circuits: blueprints for purposive behaviors. Neuron. 2013;77:810–824. doi: 10.1016/j.neuron.2013.02.018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  75. Stivers T, Enfield NJ, Brown P, Englert C, Hayashi M, Heinemann T, Hoymann G, Rossano F, de Ruiter JP, Yoon KE, Levinson SC. Universals and cultural variation in turn-taking in conversation. PNAS. 2009;106:10587–10592. doi: 10.1073/pnas.0903616106. [DOI] [PMC free article] [PubMed] [Google Scholar]
  76. Subramanian, HH, Balnave RJ, Holstege G. Microstimulation in different parts of the periaqueductal gray generates different types of vocalizations in the cat. J Voice S. 2020;1997:022. doi: 10.1016/j.jvoice.2020.01.022. [DOI] [PubMed] [Google Scholar]
  77. Sugiyama Y, Shiba K, Nakazawa K, Suzuki T, Hisa Y. Brainstem vocalization area in guinea pigs. Neuroscience Research. 2010;66:359–365. doi: 10.1016/j.neures.2009.12.006. [DOI] [PubMed] [Google Scholar]
  78. Tovote P, Esposito MS, Botta P, Chaudun F, Fadok JP, Markovic M, Wolff SB, Ramakrishnan C, Fenno L, Deisseroth K, Herry C, Arber S, Lüthi A. Midbrain circuits for defensive behaviour. Nature. 2016;534:206–212. doi: 10.1038/nature17996. [DOI] [PubMed] [Google Scholar]
  79. Tschida K, Michael V, Takatoh J, Han BX, Zhao S, Sakurai K, Mooney R, Wang F. A specialized neural circuit gates social vocalizations in the mouse. Neuron. 2019;103:459–472. doi: 10.1016/j.neuron.2019.05.025. [DOI] [PMC free article] [PubMed] [Google Scholar]
  80. Van Segbroeck M, Knoll AT, Levitt P, Narayanan S. MUPET—Mouse Ultrasonic Profile ExTraction: A Signal Processing Tool for Rapid and Unsupervised Analysis of Ultrasonic Vocalizations. Neuron. 2017;94:465–485. doi: 10.1016/j.neuron.2017.04.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
  81. Vandries LM, Ghorbanpoor S, Cornez G, Shevchouk OT, Ball GF, Cornil CA, Balthazart J. Testosterone or estradiol when implanted in the medial preoptic nucleus trigger short Low-Amplitude songs in female canaries. Eneuro. 2019;6:ENEURO.0502-18.2019. doi: 10.1523/ENEURO.0502-18.2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
  82. Wei YC, Wang SR, Jiao ZL, Zhang W, Lin JK, Li XY, Li SS, Zhang X, Xu XH. Medial preoptic area in mice is capable of mediating sexually dimorphic behaviors regardless of gender. Nature Communications. 2018;9:279. doi: 10.1038/s41467-017-02648-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  83. White NR, Prasad M, Barfield RJ, Nyby JG. 40- and 70-kHz vocalizations of mice (Mus musculus) during copulation. Physiology & Behavior. 1998;63:467–473. doi: 10.1016/S0031-9384(97)00484-8. [DOI] [PubMed] [Google Scholar]
  84. Whitney G, Alpern M, Dizinno G, Horowitz G. Female odors evoke ultrasounds from male mice. Animal Learning & Behavior. 1974;2:13–18. doi: 10.3758/BF03199109. [DOI] [PubMed] [Google Scholar]
  85. Wu Z, Autry AE, Bergan JF, Watabe-Uchida M, Dulac CG. Galanin neurons in the medial preoptic area govern parental behaviour. Nature. 2014;509:325–330. doi: 10.1038/nature13307. [DOI] [PMC free article] [PubMed] [Google Scholar]
  86. Xu X, Coats JK, Yang CF, Wang A, Ahmed OM, Alvarado M, Izumi T, Shah NM. Modular genetic control of sexually dimorphic behaviors. Cell. 2012;148:596–607. doi: 10.1016/j.cell.2011.12.018. [DOI] [PMC free article] [PubMed] [Google Scholar]

Decision letter

Editor: Catherine Emily Carr1

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

Activity dependent circuit tracing tools were used to identify inputs to the periaqueductal gray neurons that gate the production of ultrasonic vocalizations in mice. Optogenetic manipulations showed that two of these inputs have opposing effects on vocalization, while parallel in vitro experiments revealed circuit features underlying these opposing effects.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting your work entitled "Circuit and synaptic organization of forebrain-to-midbrain pathways that promote and suppress vocalization" for consideration by eLife. Your article has been reviewed by a Senior Editor, a Reviewing Editor, and two reviewers. The reviewers have opted to remain anonymous.

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

The reviewers found the idea of investigating inputs to the PAG and their role in vocal production to be interesting, and appreciated the attempts at doing this in a cell-type specific manner. However, there were serious concerns raised about the evidence provided for the paper's claims. In particular, quantification of and verification of the specificity of the results for both the anatomical and the optogenetic experiments were found to be lacking. These are critically necessary for supporting the conclusions regarding the roles of EA and POA in driving vocalizations.

Reviewer #1:

The manuscript by Michael et al. describes how two different regions of the forebrain, the extended amygdala (EA) and the preoptic region (POA) of the hypothalamus, connect and control the production of ultrasonic vocalizations (USVs) via the periaqueductal gray (PAG). USVs are produced during specific interactions with conspecifics and are fundamental for communication. To understand how different internal and external variables influence the production of USVs requires the identification of the pathways carrying information about such variables. This study makes important discoveries into our understanding of the circuits controlling USV production by identifying how two independent pathways communicate to PAG neurons that ultimately control USV production. Interestingly, these two pathways seem to both provide inhibition to the PAG region that controls the production of USVs, but the net effect is the opposite: while the POA promotes the production of vocalizations, the EA turns it off.

In a recent study (Tschida et al., 2019) the same group has identified a circuit in the caudolateral PAG circuit whose activity is necessary and sufficient to produce USVs, as well its output to the brainstem. Using a similar strategy that relies on the genetic labelling of PAG neurons involved in USV production, the authors now investigated how two different input regions may control the production of vocalizations. Through optogenetic manipulations and circuit mapping, the authors provide convincing information about the logic of connectivity between the amygdala/hypothalamus and the PAG, opening up the opportunity to understand how different internal and external variables might control this fascinating behavior. The study is well performed, using state of the art genetic tools to address the problem. However, in my opinion, the authors do not present or quantify a series of results that would make the paper much stronger. Also, despite the fact that the authors use state of the art tools to label PAG neurons that are involved in USV production (and which they described in Tschida et al., 2019), the manipulations performed in the input regions to the PAG are not specific. Together with the lack of some analysis on the reliability of the optogenetic manipulations performed, I wonder if some of the results are indeed caused by the direct manipulation of the EA and POA.

1) The authors start by identifying the monosynaptic inputs onto the PAG-USV neurons (labeled in an activity dependent manner) and onto GABAergic local interneurons (PAG-VGAT) which are in the vicinity of the PAG-USV and which might inhibit them. They found several structures in the forebrain, including the POA and the EA and CeA. I know these tools are routinely used, but it is also known that these tools can be highly variable and dependent on the volume of injection, number of starter cells, etc. There is no quantification of the starter cells and transsynaptic labeled cells for the experiments in Figure 1A and B. Are the animals used males or females? Also, the PAG is a huge structure, as well as the POA and the EA/CeA. I am not familiar with the term caudolateral PAG, but from the coordinates given, this seems very similar to the vlPAG described in Tovote et al., 2016. Is this true?

Also, the authors only show zoom ins from the labeled structures, POA and EA/CeA. Showing the labeled regions in the context of the rest of the brain would make the figure and results easier to understand. Also, from what I gather, the EA is comprised of the BST and CeA. I am confused that the authors do not consider the CeA part of the EA. Providing more information about the exact location of the labeled structures would make the paper much easier to follow.

These issues might seem irrelevant, but the devil is in the details in these experiments, as we know that the PAG is involved in many different behaviors and its architecture is very complex. And the same holds true for the POA and for the EA.

2) On Figure 1C/D the authors show the results of an immunohistological assay to determine the nature of the input to the PAG-USV and PAG-VGAT neurons. There is very little information in the methods about how this was quantified. The authors show results for only 2 mice in each condition. The numbers presented are quantified by analyzing how many slices per animal?

3) The authors use an optogenetic strategy to target the POA neurons that project to the caudolateral PAG (Figure 2). First, this strategy is markedly different from the one used in Figure 1. Primarily, this strategy (Cre dependent viruses and a retrograde AAV carrying CRE) allows to label all neurons in the POA that project to the PAG, not only the PAG-USV and the PAG-VGAT. It would be important to understand how the neurons labeled like this relate to the ones in the previous experiment.

Then they use ChR2 to elicit activity in the POA neurons and record USVs, showing that this manipulation leads to vocalizations in 6 out of 8 males and 3 out of 4 females. First, maybe I am missing something, but I do not see a control (animal implanted with a fiber and light only or animal injected with a control virus and implanted with a fiber and exposed to light). Second, the only data presented is the number of animals per sex that responded. It would be great to see the histology of these animals: was the injection or fiber mis-placed in the animals that it didn't work? Also, what does it mean to have 6 out of 8 males vocalizing? Every time the light was on did you observe the same result? What was the variability across trials? There is no quantification of the results except for some parameters presented in the supplementary data.

In the same Figure 2C/D it is shown the effect of the POA manipulation using the Esr1+ Cre line. Again, no quantification is offered. How do the results of the two POA manipulations compare to each other? The latency to start USVs seems quite different in the two experiments. But again, because only examples are shown, we do not know, what is the variability across experiments.

Are neurons labeled with strategy A and D sending collaterals to other regions? I also find it surprising that the results of stimulating directly the Esr1-POA-PAG terminals in the PAG gave rise to other behavioral results that were not observed when the somas were directly activated at the level of the PAG. How can you explain this?

How can you explain such long latencies for the vocalizations to start? In some cases, it took more than 5 seconds for the USVs to be produced (minimum latency). Besides the minimum latency, one should have an idea of the spread of the latency across animals and within animals. Why is the number of animals presented in Figure 2—figure supplement 2A different from Figure 2—figure supplement 2B? I am not familiar with this, but do animals vocalize in isolation? What was the schedule of optogenetic stimulation performed? Are you sure that the USVs were caused by the stimulation? With such long latencies and not knowing the spread in the latency within animal after ChR2 activation, I feel a bit uncomfortable making the strong claim that the POA stim is indeed the one causing the USVs. It is really hard to understand exactly the results.

Note: I think the data from Figure 2—figure supplement 2 should come in the main text.

All the POA-PAG experiments are done in a non-specific population, this is, not only to PAG-USV projecting neurons, but to all caudolateral PAG. Also, these neurons might have collaterals to other brain regions. Activation of the terminals might backpropagate and if those cells send collateral to other regions, the effect could be mediated by that. Could you try, in the experiment where you activated the terminals, to block activity of the POA? This way the results obtained by activating the terminals could be only attributed to the projections to the caudolateral PAG (similar to what was performed in Wong et al., 2016, where they applied TTX and lidocaine in the LS to block activity, while stimulating the LS terminals).

4) I am not familiar with the analysis performed to compare the opto induced versus female directed USVs. However, I was wondering if the authors tried to examine how females responded to the opto-USVs. Also, how do these vocalizations compare to the PAG-USV stimulation described in their recent paper? They mention that the latencies are much longer, but nothing mentioned in relationship to the other USV features.

5) The authors then explored the role of the Amygdala-PAG projections. The strategy designed now involves the use of Cre dependent viruses in the amygdala and the injection of retroAAVs-Cre in the PAG. With this strategy the authors mention that only the EA is labeled and not the CeA. Therefore, their manipulations are only of the EA. Again, I am not an expert in amygdala, but in many papers it is referred that the CeA is part of the EA and therefore I am confused. I stress this point because I am having a difficult time understanding the relationship between the areas manipulated in this study and Tovote et al., 2016. In that case, a similar circuit organization is described, where CeA neurons disinhibit vlPAG neurons to produce freezing.

So, I wonder if the results obtained in here can be in some way interpreted in a similar manner to the results obtained in that paper: that what you are inducing is a fear state that would lead to the animal stopping USV production. I know the authors analysed other behaviors and have no indication that the animal is entering a fearful state during the EA-PAG neurons. However, this strategy does not label EA neurons that project specifically to the PAG-USV neurons, it labels all EA-VGAT neurons that project to the caudolateral PAG. In order to make sure the effect is due to the direct inactivation of PAG-USV neurons, could the authors perform the same mono-synaptic tracing done in Figure 1 and then label the EA-PAG projecting neurons with another strategy? For example, are all neurons EA-PAG neurons gabaergic? If yes, they could just inject a non-flexed retroAAV in the PAG and then perform the monosynaptic tracing starting from the PAG-USV. If those two experiments labeled the same population in the EA, then they could argue that the effect of the opto stim of EA-PAG neurons is indeed by inhibiting PAG-USV neurons (with the opto manipulation being done at the terminals, in the PAG, like in Figure 4D).

Reviewer #3:

During many behavioral contexts mice elicit ultrasonic vocalizations (USV) and it has been shown that these utterances are mainly generated by subcortical areas. Specifically, the periaqueductal grey (PAG) is gating vocal production in mammals. Recently, the same lab has published a paper about optogenetically activating/suppressing a subgroup of PAG neurons that had been previously shown to be active during USVs. In the previous article they describe how silencing PAG-USV neurons blocks USV production whereas activating PAG-USV neurons promotes USV production. In this study the authors aim to investigate the anatomical upstream sources that provide input to the PAG-USV neurons and ask whether the upstream inputs are functional i.e. sufficient for exerting vocal production behaviors.

In short, the anatomical tracing of PAG-USV inputs is interesting but not entirely surprising since PAG inputs have been traced before. The cell-type specific tracing adds a new point. The behavioral experiments provide mixed insights into the circuitry. On the one hand they stand in contrast with previous reports and on the other hand they replicate the same findings which have been described in previous reports. This makes the current study less novel and appealing. In addition, data analysis is insufficient and consequently, I have major concerns to promote this study for publication in eLife.

The tracing of the synaptic inputs onto PAG-USV neurons was achieved via activity-dependent labeling. In addition to labeling these neurons, the authors also labeled the upstream inputs onto GABAergic PAG neurons which provide inhibition onto PAG-USV neurons. Anatomical connections of the PAG had been mapped previously and the novelty here lies in the possibility to determine which cell-type (PAG-USV or GABAergic PAG neuron) other areas project to. The authors completely miss out on this and only provide Table 1 which contains insufficient information to appreciate new insights. I am suggesting to supply example images and quantitative data instead of “+” and “-“ and to discuss these results further.

Although neurons were labeled in a range of different brain areas, the authors decided to focus on the hypothalamus and the amygdala. This choice remains elusive is not well motivated. In addition, the authors put up a strawman by arguing that opposing behavioral effects can be hypothesized since the hypothalamus has been shown to be involved in sexual behaviors whereas the amygdala is involved in fear-related behaviors. However, the hypothalamus is also implicated in anxiety behaviors and the amygdala in positive emotional behaviors.

Subsection “Inhibitory neurons in the hypothalamus and amygdala provide input to the PAG vocal gating circuit”: “dense labeling” – This is not clear from the data. In Figure 1 POA labeling seems sparse in both PAG-USV and PAG-VGAT+? Provide quantitative assessment.

The authors claim that optogenetic stimulation of POA-PAG neurons elicits USVs. The provided evidence is not sufficient to underline this claim. Figure 2A and the movie are single occurrences that were aligned with the optogenetic pulse. How can the authors exclude that this did not happen by chance? Also, the onset of the vocalization differs in Figure 2 and the movie. If the vocalizations occur on a regular basis after optogenetic stimulation these data must be shown and quantified. How reliable was this effect? What is the latency? In addition, Gao et al., 2018 already described that optogenetic activation of GABAergic neurons in POA can evoke USVs in mice which make the data not novel.

To narrow down the molecular phenotype of the POA neurons the authors performed in situ hybridization and determined that POA-PAG neurons expressed the Estrogen α receptor. Wei et al., 2018 demonstrate that activation of POA neurons expressing estrogen α receptors results in sexually biased displays. Michael and colleagues hint towards their inability to replicate this result. In subsection “Activating PAG-projecting POA neurons elicits USVs in the absence of social cues” the number of animals is presented but the data itself are not shown and the statistical tests are unclear. One way to address the mismatch between the Wei et al. study and the results shown here is to record from neurons during optogenetic stimulation. How can the authors confirm that the optogenetic stimulation results in a functional activation of the targeted neurons?

The presented timeline of optogenetic stimulation and resulting vocalizations (subsection “Activating PAG-projecting POA neurons elicits USVs in the absence of social cues”) is difficult to understand. Even when counting the different synaptic stages and adding significant conduction velocity delays, the time course is biologically not plausible but simply too long. This observation is not discussed in detail and it remains unclear. How can it be excluded that vocalizations are being produced by chance and that the correlation of stimulation and vocalization occurs by chance?

The authors attempt to quantify the USVs during optogenetic stimulation and in control conditions in an unbiased way and used an unsupervised modeling approach. The data were visualized as UMAPs of latent features (Figure 3C). Based on differences in the visual appearing of these maps and the MMD the authors claim that a subset of USVs are similar and another subset is dissimilar during opto-stimulation versus female presence (subsection “Acoustic characterization of USVs elicited by activation of POA neurons”). This statement is confusing, and it remains unclear what is similar or dissimilar. To address this issue the authors investigated the acoustics in more detail. They argue that opto-USVS tended to be louder and covered a higher frequency bandwidth. This result, displayed in Figure 3F, is difficult to grasp. In Figure 3F it cannot be discriminated whether points are overlaid and therefore, the green data points are not visible in most part of the figure. Another concern is that USVs differ in different social contexts. The authors should perform the same analysis on USVs that were elicited when no female was present and without opto-stimulation to ensure that the observed change is not due to difference in the social context. Temporal organization and usage can be differentiated in multiple putative USV classes also arising from distinct articulatory patterns (Castellucci et al., 2018).

Subsequently, Michael et al., tested the effect of stimulating PAG-projecting amygdala neurons and found that the stimulation of either EA-PAG neurons or even more specifically, GABAergic EA-PAG neurons results in the suppression of vocalizations during ongoing behavior. While this result is intriguing (Figure 4B) the presentation in Figure 4D (right panel) is misleading. Why is the USV count in the pre-condition so much higher than during the post condition? What is the effect of opto-stimulation on neural activity?

The authors add a section about the upstream axonal projections of POA-PAG and EA-PAG neurons. These data do not add to the focus of this study and are not discussed further. What is the point the authors want to make with this?

To test connectivity in detail the authors performed slice recordings and measured synaptic inputs onto different cell types while stimulating others. While they find that some PAG-USV neurons receive inhibitory currents when EA neurons are stimulated, the authors do not perform the symmetrical experiment and stimulate POA-PAG neurons while recording PAG-USV neurons. Instead, VGAT+ neurons are being recorded and it is shown that a subset receives inhibitory current as well. In addition, VGAT+ neurons are being stimulated and PAG-USV were recorded to demonstrate that VGAT+ neuron provide inhibition to these PAG-USV neurons. Unfortunately, the data are less conclusive as described in the text. For example, in subsection “Synaptic interactions between POAPAG and EAPAG neurons and the PAG vocal gating circuit” 'majority (18 out of 27)': Is this statistically significant? What can one compare it to? Is the number of recorded neurons too low?

Results section and Discussion section: “data not shown” – In times of reproducibility of data and open access this statement is unacceptable. If the authors want to add this information to the text and speculate about them, they have to be shown.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your article "Circuit and synaptic organization of forebrain-to-midbrain pathways that promote and suppress vocalization" for consideration by eLife. Your article has been reviewed by Catherine Dulac as the Senior Editor, a Reviewing Editor, and three reviewers. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

We would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). Specifically, we are asking editors to accept without delay manuscripts, like yours, that they judge can stand as eLife papers without additional data, even if they feel that they would make the manuscript stronger. Thus the revisions requested below only address clarity and presentation.

Summary:

Michael et al., use activity dependent circuit tracing tools to identify inputs to PAG-USV neurons, which they previously showed (Tschida et al., 2019) play a crucial role in gating vocal behavior. They use optogenetic manipulations to show that two of these inputs have opposing effects on male courtship vocalizations, then use slice electrophysiology to describe the circuit logic by which these opposing effects come about. In their rebuttal, the authors have strengthened these findings with the addition of several control experiments and analyses. Though forebrain regions such as the hypothalamus and amygdala have previously been implicated in vocal production, presumably through their descending connections to the PAG, these experiments greatly expand the field of knowledge encompassing these circuits and synaptic mechanisms.

Your reviewers concur that your revised manuscript is greatly improved with the more in-depth quantification of the optogenetic manipulations.

Essential revisions:

1) One of the main claims of the paper is that there is an opposing effect of the POA versus Amg gabaergic input: while the POA leads to USVs, the Amg leads to USV suppression. However, if I understand correctly the experiment depicted in Figure 1B, the G and the TVA should only be expressed in gabaergic neurons of the PAG and then the rabies injection should label all neurons that project to those, irrespective of being gabaergic or glutamatergic (because the rabies is not Cre dependent). Then, the subsequent results with immunohistochemistry show that neurons in the POA and Amg are all gabaergic (according to the staining shown in Figure 1C and D). Later on, it is assumed that the POA input is disinhibitory, which makes sense with this result and everything that follows after. However, the fact that PAG GABAergic neurons also receive inhibitory input from the Amg is not pursued at all in the rest of the paper or discussed. Instead, it is assumed that the Amg GABAergic neurons only project to principal PAG neurons (which is depicted in the cartoon of Figure 7 as well). This should be discussed since it assumes that the projections of the Amg are much more complex than just inhibiting USVs. The cartoon should also, therefore, be revised.

2) Figure 2E: In my opinion, this is an important claim of the paper, that the projections from the POA to PAG can induce USVs. I don't think that such a strong claim can be done with a single male (this experiment is necessary to show that the effect is due to the projections of POA neurons in the PAG).

3) The addition to the Materials and methods section is useful: "Cells were scored from 2-7 sections of tissue per brain region from each animal, and the absence or presence of staining within cells was quantified manually by comparing labeling within cells to background labeling in nearby regions known to be negative for a given RNA transcript." But what are those control regions and how was it confirmed that these are "negative" for the given RNA transcript? In the interest of data transparency and reproducibility, this information should be included.

4) The authors state (subsection “Activating PAG-projecting AmgC/M neurons transiently suppresses USV production”) that "Optogenetic activation of AmgC/M-PAG neurons failed to elicit USV production and also did not drive any other overt behavioral effects". This seems like important evidence for the author's claim that the function of AmgC/M-PAG is primarily to suppress vocal behavior, but the data aren't shown. In the interest of data transparency and reproducibility, these data should be included as a supplemental figure.

eLife. 2020 Dec 29;9:e63493. doi: 10.7554/eLife.63493.sa2

Author response


[Editors’ note: the authors resubmitted a revised version of the paper for consideration. What follows is the authors’ response to the first round of review.]

Reviewer #1:

The manuscript by Michael et al. describes how two different regions of the forebrain, the extended amygdala (EA) and the preoptic region (POA) of the hypothalamus, connect and control the production of ultrasonic vocalizations (USVs) via the periaqueductal gray (PAG). USVs are produced during specific interactions with conspecifics and are fundamental for communication. To understand how different internal and external variables influence the production of USVs requires the identification of the pathways carrying information about such variables. This study makes important discoveries into our understanding of the circuits controlling USV production by identifying how two independent pathways communicate to PAG neurons that ultimately control USV production. Interestingly, these two pathways seem to both provide inhibition to the PAG region that controls the production of USVs, but the net effect is the opposite: while the POA promotes the production of vocalizations, the EA turns it off.

In a recent study (Tschida et al., 2019) the same group has identified a circuit in the caudolateral PAG circuit whose activity is necessary and sufficient to produce USVs, as well its output to the brainstem. Using a similar strategy that relies on the genetic labelling of PAG neurons involved in USV production, the authors now investigated how two different input regions may control the production of vocalizations. Through optogenetic manipulations and circuit mapping, the authors provide convincing information about the logic of connectivity between the amygdala/hypothalamus and the PAG, opening up the opportunity to understand how different internal and external variables might control this fascinating behavior. The study is well performed, using state of the art genetic tools to address the problem. However, in my opinion, the authors do not present or quantify a series of results that would make the paper much stronger. Also, despite the fact that the authors use state of the art tools to label PAG neurons that are involved in USV production (and which they described in Tschida et al., 2019), the manipulations performed in the input regions to the PAG are not specific. Together with the lack of some analysis on the reliability of the optogenetic manipulations performed, I wonder if some of the results are indeed caused by the direct manipulation of the EA and POA.

1) The authors start by identifying the monosynaptic inputs onto the PAG-USV neurons (labeled in an activity dependent manner) and onto GABAergic local interneurons (PAG-VGAT) which are in the vicinity of the PAG-USV and which might inhibit them. They found several structures in the forebrain, including the POA and the EA and CeA. I know these tools are routinely used, but it is also known that these tools can be highly variable and dependent on the volume of injection, number of starter cells, etc. There is no quantification of the starter cells and transsynaptic labeled cells for the experiments in Figure 1A and B. Are the animals used males or females?

We have included additional images to supplement the transsynaptic tracing results currently provided in Supplementary file 1 (Figure 1—figure supplement 1, Figure 1—figure supplement 2, Figure 1—figure supplement 3, Figure 1—figure supplement 4). However, because our goal was to identify long-range inputs onto PAG-USV neurons, we used long survival times that prioritized visualization of afferent cell bodies in distant locations rather than the integrity of the starter cell populations (which die off over time). Thus, while we can count the few remaining starter cells, such quantification will not accurately reflect their original numbers. Thus, we would prefer not to provide such quantification, as it cannot be meaningfully related to the absolute numbers of cells that provide monosynaptic input to PAG-USV neurons. We have added language to this effect to the Materials and methods. In addition, we make no claims about the relative densities of these labeled afferents, rather we used functional approaches to explore their relevance to vocalization.

The transsynaptic tracing from GABAergic PAG neurons was carried out from N=2 females and N=4 males, and the two mice used for in situ hybridization (Figure 1C-D) were both males. Due to the difficulty in eliciting USVs from female mice, the transsynaptic tracing from PAG-USV neurons (Figure 1A) was performed exclusively in male mice, and we have clarified these details in the manuscript (subsection “Inhibitory neurons in the hypothalamus and amygdala provide input to the PAG vocal gating circuit”).

Also, the PAG is a huge structure, as well as the POA and the EA/CeA. I am not familiar with the term caudolateral PAG, but from the coordinates given, this seems very similar to the vlPAG described in Tovote et al., 2016. Is this true?

No, the region of the PAG is not the same as that described by Tovote et al. 2016. The caudolateral PAG abuts the vlPAG but is largely dorsal to the vlPAG. We would like to emphasize to the Reviewer that the part of the PAG that contains PAG-USV neurons as characterized in Tschida et al., 2019 and as considered in the current study is a nearby but distinct part of the PAG to the vlPAG studied by Tovote et al., 2016 and others. We have clarified this distinction in subsection “Inhibitory neurons in the hypothalamus and amygdala provide input to the PAG vocal gating circuit” Results section.

Also, the authors only show zoom ins from the labeled structures, POA and EA/CeA. Showing the labeled regions in the context of the rest of the brain would make the figure and results easier to understand.

Lower magnification views of rabies-labeled neurons within the POA and the AmgC/M/CeA are now shown in Figure 1—figure supplement 1.

Also, from what I gather, the EA is comprised of the BST and CeA. I am confused that the authors do not consider the CeA part of the EA. Providing more information about the exact location of the labeled structures would make the paper much easier to follow.

The confusion raised by reviewer 1 here is a result of our previous choice of name for the region of the amygdala that we characterize in this study. The pocket of cells labeled in the amygdala with both transsynaptic tracing (Figure 1B) and our retro-Cre viral strategy (Figure 4A and Figure 4—figure supplement 1) occupies a boundary zone between the medial amygdala and the central amygdala that did not have any name in any brain atlases that we examined. In the revised manuscript, we call this region the “central-medial boundary zone” (AmgC/M) to avoid this confusion.

2) On Figure 1C/D the authors show the results of an immunohistological assay to determine the nature of the input to the PAG-USV and PAG-VGAT neurons. There is very little information in the methods about how this was quantified. The authors show results for only 2 mice in each condition. The numbers presented are quantified by analyzing how many slices per animal?

The data shown in Figure 1C-D are of representative in situ hybridization experiments (i.e., RNA staining), and the absence or presence of staining was quantified manually. We have added this information to the Materials and methods, as well as information regarding the number of sections from which the total number of scored neurons came. Quantification for each in situ hybridization came from N=2 animals per condition, which is standard in the field.

3) The authors use an optogenetic strategy to target the POA neurons that project to the caudolateral PAG (Figure 2). First, this strategy is markedly different from the one used in Figure 1. Primarily, this strategy (Cre dependent viruses and a retrograde AAV carrying CRE) allows to label all neurons in the POA that project to the PAG, not only the PAG-USV and the PAG-VGAT. It would be important to understand how the neurons labeled like this relate to the ones in the previous experiment.

We agree with the reviewer that the AAV-retro-Cre viral strategy is different than the monosynaptic transsynaptic tracing from defined PAG cells types as shown in Figure 1. Unfortunately, we are not aware of a non-rabies-based viral strategy that would allow us to label monosynaptic inputs to defined PAG cell types for long-term labeling and optogenetic manipulations of activity. Because the rabies-based viral strategy kills the cells after the first week following infection, we are not able to apply this strategy to manipulate neural activity in forebrain inputs to defined PAG cell types. Given this limitation, we used the AAV-retro-Cre viral strategy which, as the reviewer notes, will label cells within a given region that project to the caudolateral PAG (with some exceptions, as noted with the absence of labeling within the CeA using the AAV-retro-Cre strategy, as detailed in Figure 4, Figure 4—figure supplement 1, and Figure 4—figure supplement 2).

The AAV-retro-Cre viral strategy will not label an identical set of cells as those labeled by transsynaptic tracing, and given that there is no way to apply both viral strategies within the same animal, it is not possible to directly compare the populations labeled with the two strategies. In general, however, it is reasonable to expect that the retro-Cre viral strategy will label forebrain neurons that provide input to the PAG vocalization circuit, in addition to forebrain neurons that provide input to other nearby cells in the caudolateral PAG.

Despite these practical limitations, we applied in situ hybridization to verify that the forebrain neurons labeled with the AAV-retro-Cre strategy are of the same neurotransmitter-type as those labeled with rabies-based tracing. In addition to observing robust effects on behavior of optogenetic activation of these PAG-projecting neurons, we also provide evidence that GABAergic EA neurons directly inhibit PAG-USV neurons (Figure 5) and that GABAergic POA neurons directly inhibit VGAT+ PAG neurons (Figure 6). Together, we argue that these findings provide strong support for the conclusion that preoptic and amygdala neurons labeled by the AAV-retro-Cre strategy provide functional and behaviorally relevant inputs to the PAG vocalization circuit.

Then they use ChR2 to elicit activity in the POA neurons and record USVs, showing that this manipulation leads to vocalizations in 6 out of 8 males and 3 out of 4 females. First, maybe I am missing something, but I do not see a control (animal implanted with a fiber and light only or animal injected with a control virus and implanted with a fiber and exposed to light).

The reviewer is correct that we did not include GFP controls for the POA optogenetic activation experiments. We now include data from 5 POA GFP control mice, and we also include quantification to illustrate that the USV rates elicited by optogenetic activation of POAPAG neurons are significantly greater than baseline vocalization rates in these same animals.

Second, the only data presented is the number of animals per sex that responded. It would be great to see the histology of these animals: was the injection or fiber mis-placed in the animals that it didn't work? Also, what does it mean to have 6 out of 8 males vocalizing? Every time the light was on did you observe the same result? What was the variability across trials? There is no quantification of the results except for some parameters presented in the supplementary data.

As stated in the Materials and methods, we excluded any animals with mistargeted injections or poor viral expression from further analysis.

In the same Figure 2C/D it is shown the effect of the POA manipulation using the Esr1+ Cre line. Again, no quantification is offered. How do the results of the two POA manipulations compare to each other? The latency to start USVs seems quite different in the two experiments. But again, because only examples are shown, we do not know, what is the variability across experiments.

There is variability in the efficacy of optogenetic activation in eliciting vocalizations across trials for single animals, between animals within a condition, and between conditions as well. We now include detailed quantification of the mean number of USVs elicited per stimulation, proportion of stim trials that elicited USVs, and mean latency from stim onset to first USV onset, in both experimental and control animals (Figure 2F). We want to highlight that some quantification of the latencies to USV onset and durations of USV bouts elicited by optogenetic activation are compared across mice and across conditions in Figure 2—figure supplement 1.

Are neurons labeled with strategy A and D sending collaterals to other regions? I also find it surprising that the results of stimulating directly the Esr1-POA-PAG terminals in the PAG gave rise to other behavioral results that were not observed when the somas were directly activated at the level of the PAG. How can you explain this?

Yes, these neurons send collaterals to other regions, as shown in in Figure 2—figure supplement 3. When we optogenetically activate Esr1+ POA cell bodies, the placement of the ferrule directly over the POA limits the spread of light to POA cell bodies, despite the fact that the AAV-FLEX-ChR2 expression sometimes spreads to adjacent overlying structures, such as the BNST. In the case of the axonal activation experiments, the ferrule placed in the PAG can activate the axon terminals of any PAG-projecting Esr1+ neurons within or nearby to the POA that are infected by the virus. As such, we suspect that the additional behavioral effects observed in the axon terminal activation experiments relate to the small amount of off-target viral labeling in Esr1+ neurons nearby the POA.

How can you explain such long latencies for the vocalizations to start? In some cases, it took more than 5 seconds for the USVs to be produced (minimum latency). Besides the minimum latency, one should have an idea of the spread of the latency across animals and within animals.

For the POA cell body activation experiments, the majority of mice (13/16) exhibited a minimum latency from laser stimulation to USV production of <500 ms (data shown in Figure 2—figure supplement 1B). The POA lies upstream of the PAG, which itself lies upstream of both the premotor and motor neurons important to vocalization. We found previously (Tschida et al., 2019) that average latency from optogenetic activation of PAG-USV neurons to USV onset was approximately 500 ms (minimum latency ~30 ms), and it is thus not surprising that optogenetic activation of the POA elicits USVs at longer latencies than those observed during PAG-USV activation. We also note that mean latencies on the order of seconds from optogenetic activation of the hypothalamus to observed effects of behavior have been reported previously (see Figure 2I and Figure 3H from Wei et al., 2018, showing mean latency of ~5s from POA activation to mounting and pup retrieval, respectively; similarly, see Figure 4I from Lin et al., 2011 showing mean latency of ~4s from VMHvl activation to attack). We have added language to the Results to directly compare the latencies from POA activation to USV production observed in the current study to latencies reported in other work from the onset of hypothalamic activation to subsequent effects on behavior. We have also included additional quantification regarding the variability in the latency from optogenetic stimulation to USV onsets within animals (Figure 2F).

Why is the number of animals presented in Figure 2—figure supplement 2A different from Figure 2—figure supplement 2B?

We used a variety of laser train stimuli to optogenetically elicit vocalizations from each animal, and in Figure 2—figure supplement 1CFigure , we omitted 5 animals from which we didn’t have a sufficient number of trials with the exact stimulation shown (2s-long laser trains). We have clarified the omission of the 5 animals in the figure legend.

I am not familiar with this, but do animals vocalize in isolation? What was the schedule of optogenetic stimulation performed? Are you sure that the USVs were caused by the stimulation? With such long latencies and not knowing the spread in the latency within animal after ChR2 activation, I feel a bit uncomfortable making the strong claim that the POA stim is indeed the one causing the USVs. It is really hard to understand exactly the results.

Mice vocalize rarely or not at all in isolation and we are highly confident that optogenetic stimulation of POA causes vocalizations in isolated mice. Specifically, we established in Tschida et al., 2019, that mice vocalize very little, if at all, in the absence of any social partners or social cues (see Figure 3A from that paper), and this finding is consistent with the broad consensus in the field that mice produce USVs in a variety of social contexts but very few or no USVs when monitored in social isolation. We have provided new quantification throughout Figure 2 (Figure 2B,D,E) to demonstrate that baseline vocal rates outside of periods of optogenetic stimulation are low, as well as control data showing that delivery of blue light to the brain in mice GFP expression in the POA does not elicit vocalization (Figure 2F).

Note: I think the data from Figure 2—figure supplement 2 should come in the main text.

We have left Figure 2—figure supplement1 as is but have included additional quantification within the main Figure 2.

All the POA-PAG experiments are done in a non-specific population, this is, not only to PAG-USV projecting neurons, but to all caudolateral PAG. Also, these neurons might have collaterals to other brain regions. Activation of the terminals might backpropagate and if those cells send collateral to other regions, the effect could be mediated by that. Could you try, in the experiment where you activated the terminals, to block activity of the POA? This way the results obtained by activating the terminals could be only attributed to the projections to the caudolateral PAG (similar to what was performed in Wong et al., 2016, where they applied TTX and lidocaine in the LS to block activity, while stimulating the LS terminals).

We agree that back-propagation is a caveat of any optogenetic terminal experiment. Although we previously considered the POA inactivation experiment suggested by the reviewer, we did not pursue that approach as lidocaine silencing of POA cell bodies is not guaranteed to block the back-propagation of activity from the PAG terminals to other parts of the axonal arborizations of these cells and therefore does not entirely control for this issue.

We now include data that directly address the reviewer’s concern, however. We observed that activating the axon collaterals of Esr1+ POA cells in another region of the brain, the VTA, fails to elicit vocalizations (Figure 2F). This negative result strengthens the case that there is something special about the projection from POA to PAG, rather than recruitment of non-PAG targets of POA-PAG axons.

4) I am not familiar with the analysis performed to compare the opto induced versus female directed USVs. However, I was wondering if the authors tried to examine how females responded to the opto-USVs. Also, how do these vocalizations compare to the PAG-USV stimulation described in their recent paper? They mention that the latencies are much longer, but nothing mentioned in relationship to the other USV features.

We did not characterize the responses of females to optogenetically elicited vocalizations, as this is tangential to the focus of our study. We also have not included a VAE analysis of the data from Tschida et al., 2019, as it would be difficult to draw systemic conclusions from a comparison to those data due to the limited size of the PAG-USV dataset (N=2 mice) in our previous study.

5) The authors then explored the role of the Amygdala-PAG projections. The strategy designed now involves the use of Cre dependent viruses in the amygdala and the injection of retroAAVs-Cre in the PAG. With this strategy the authors mention that only the EA is labeled and not the CeA. Therefore, their manipulations are only of the EA. Again, I am not an expert in amygdala, but in many papers it is referred that the CeA is part of the EA and therefore I am confused. I stress this point because I am having a difficult time understanding the relationship between the areas manipulated in this study and Tovote et al., 2016. In that case, a similar circuit organization is described, where CeA neurons disinhibit vlPAG neurons to produce freezing.

In the new manuscript, we call this region the “central-medial boundary zone” (AmgC/M) to avoid this confusion.

So, I wonder if the results obtained in here can be in some way interpreted in a similar manner to the results obtained in that paper: that what you are inducing is a fear state that would lead to the animal stopping USV production. I know the authors analysed other behaviors and have no indication that the animal is entering a fearful state during the EA-PAG neurons. However, this strategy does not label EA neurons that project specifically to the PAG-USV neurons, it labels all EA-VGAT neurons that project to the caudolateral PAG. In order to make sure the effect is due to the direct inactivation of PAG-USV neurons, could the authors perform the same mono-synaptic tracing done in Figure 1 and then label the EA-PAG projecting neurons with another strategy? For example, are all neurons EA-PAG neurons gabaergic? If yes, they could just inject a non-flexed retroAAV in the PAG and then perform the monosynaptic tracing starting from the PAG-USV. If those two experiments labeled the same population in the EA, then they could argue that the effect of the opto stim of EA-PAG neurons is indeed by inhibiting PAG-USV neurons (with the opto manipulation being done at the terminals, in the PAG, like in Figure 4D).

We disagree with the reviewer’s suggestion that activation of AmgC/M-PAG neurons induces a fearful state in mice. In fact, we directly addressed this possibility extensively (subsection “Activating PAG-projecting AmgC/M neurons transiently suppresses USV production”) and demonstrated that AmgC/M activation elicits neither fleeing nor freezing (Figure 2—figure supplement 2B) and that optogenetic activation of AmgC/M-PAG neurons does not elicit real-time place aversion (Figure 2—figure supplement 2A).

As previously discussed, it is not possible to perform transsynaptic tracing from defined PAG cell types and AAV-retro-Cre labeling from the PAG in the same animal. Unfortunately, we are unaware of a non-rabies-based (and hence non-toxic) viral strategy that affords us the long-term access to forebrain neurons that project to defined PAG cell types that would be necessary to perform the experiment suggested by the reviewer. We believe that the robust effects on behavior observed in the AAV-retro-Cre optogenetic activation experiments provide strong support for the idea that PAG-projecting neurons within the AmgC/M suppress USV production, and our slice data provide additional support to the idea that at least some of the PAG-projecting AmgC/M neurons provide monosynaptic input to PAG-USV neurons, providing a plausible synaptic substrate for our observed behavioral effects.

Reviewer #3:

During many behavioral contexts mice elicit ultrasonic vocalizations (USV) and it has been shown that these utterances are mainly generated by subcortical areas. Specifically, the periaqueductal grey (PAG) is gating vocal production in mammals. Recently, the same lab has published a paper about optogenetically activating/suppressing a subgroup of PAG neurons that had been previously shown to be active during USVs. In the previous article they describe how silencing PAG-USV neurons blocks USV production whereas activating PAG-USV neurons promotes USV production. In this study the authors aim to investigate the anatomical upstream sources that provide input to the PAG-USV neurons and ask whether the upstream inputs are functional i.e. sufficient for exerting vocal production behaviors.

In short, the anatomical tracing of PAG-USV inputs is interesting but not entirely surprising since PAG inputs have been traced before. The cell-type specific tracing adds a new point. The behavioral experiments provide mixed insights into the circuitry. On the one hand they stand in contrast with previous reports and on the other hand they replicate the same findings which have been described in previous reports. This makes the current study less novel and appealing. In addition, data analysis is insufficient and consequently, I have major concerns to promote this study for publication in eLife.

The tracing of the synaptic inputs onto PAG-USV neurons was achieved via activity-dependent labeling. In addition to labeling these neurons, the authors also labeled the upstream inputs onto GABAergic PAG neurons which provide inhibition onto PAG-USV neurons. Anatomical connections of the PAG had been mapped previously and the novelty here lies in the possibility to determine which cell-type (PAG-USV or GABAergic PAG neuron) other areas project to. The authors completely miss out on this and only provide Table 1 which contains insufficient information to appreciate new insights. I am suggesting to supply example images and quantitative data instead of “+” and “-“ and to discuss these results further.

We have included additional images of monosynaptic rabies tracing from the PAG vocal gating circuit in Figure 1—figure supplement 1, Figure 1—figure supplement 2, Figure 1—figure supplement 3, Figure 1—figure supplement 4. We found that the regions labeled by transsynaptic tracing provide input to both PAG-USV neurons and nearby VGAT+ PAG neurons (in both males and females).

Although neurons were labeled in a range of different brain areas, the authors decided to focus on the hypothalamus and the amygdala. This choice remains elusive is not well motivated. In addition, the authors put up a strawman by arguing that opposing behavioral effects can be hypothesized since the hypothalamus has been shown to be involved in sexual behaviors whereas the amygdala is involved in fear-related behaviors. However, the hypothalamus is also implicated in anxiety behaviors and the amygdala in positive emotional behaviors.

We have provided additional context and rationale for focusing on the hypothalamus and the amygdala (subsection “Inhibitory neurons in the hypothalamus and amygdala provide input to the PAG vocal gating circuit”). We agree with the reviewer that both the hypothalamus and the amygdala have been implicated in a variety of complex and distinct emotional states and behaviors, and there is no doubt that each of these brain regions contains sets of neurons that can influence social behavior in complex and often orthogonal ways.

Subsection “Inhibitory neurons in the hypothalamus and amygdala provide input to the PAG vocal gating circuit”: “dense labeling” – This is not clear from the data. In Figure 1 POA labeling seems sparse in both PAG-USV and PAG-VGAT+? Provide quantitative assessment.

We have included additional images (Figure 1—figure supplement 1, Figure 1—figure supplement 2, Figure 1—figure supplement 3, Figure 1—figure supplement 4Figures) that will allow readers to compare the density of labeling within the POA as compared to other forebrain inputs to the PAG that were labeled via transsynaptic tracing. We focused on the preoptic area in part because we found pronounced effects on vocal behavior in our initial experiments with this region, not because the preoptic area contained the highest density of transsynaptically-labeled neurons of all regions providing input to the PAG vocalization circuit. We also note that the efficacy of rabies-based transsynaptic tracing declines substantially as the distance between the upstream region and target region increases, and hence, the true density of a given distal input might be underestimated by considering transsynaptic tracing alone.

The authors claim that optogenetic stimulation of POA-PAG neurons elicits USVs. The provided evidence is not sufficient to underline this claim. Figure 2A and the movie are single occurrences that were aligned with the optogenetic pulse. How can the authors exclude that this did not happen by chance? Also, the onset of the vocalization differs in Figure 2 and the movie. If the vocalizations occur on a regular basis after optogenetic stimulation these data must be shown and quantified. How reliable was this effect? What is the latency? In addition, Gao et al., 2018 already described that optogenetic activation of GABAergic neurons in POA can evoke USVs in mice which make the data not novel.

We have provided additional quantification of the data we collected from these mice including the reliability of the optogenetic activation in eliciting USVs, variability in the latency of these effects within and across mice, as well as data from a variety of controls (Figure 2F). We established in Tschida et al., 2019 (Figure 3A) that mice vocalize very little if at all in the absence of a social partner, and we now show clearly in Figure that optogenetic activation of the POA elicits levels of USV production significantly greater than baseline.

We are aware that Gao et al., 2019 have shown that optogenetic activation of GABAergic POA neurons elicits USV production, and we are pleased that our results are in line with this previous observation. We have extended this observation by demonstrating that activation of Esr1+/GABAergic POA neurons that project to the PAG is sufficient to elicit USVs. Given that GABAergic POA neurons comprise distinct subsets of neurons with heterogeneous projection targets within the brain, we feel that this additional detail is both novel and important, and we emphasize how our work relates to previous findings in subsection “Activating PAG-projecting POA neurons elicits USVs in the absence of social cues”.

To narrow down the molecular phenotype of the POA neurons the authors performed in situ hybridization and determined that POA-PAG neurons expressed the Estrogen α receptor. Wei et al., 2018 demonstrate that activation of POA neurons expressing estrogen α receptors results in sexually biased displays. Michael and colleagues hint towards their inability to replicate this result. In subsection “Activating PAG-projecting POA neurons elicits USVs in the absence of social cues” the number of animals is presented but the data itself are not shown and the statistical tests are unclear. One way to address the mismatch between the Wei et al. study and the results shown here is to record from neurons during optogenetic stimulation. How can the authors confirm that the optogenetic stimulation results in a functional activation of the targeted neurons?

Using our optogenetic stimulation parameters, we were able to elicit robust USV production but never observed mounting of a conspecific following optogenetic activation of POAPAG neurons. Given the robust effect on vocal behavior of optogenetic activation of POA neurons (of which we have provided additional quantification and controls in Figure 2 to convince the reviewers of these effects), we are certain that the optogenetic stimulation parameters applied in our study were sufficient to reliably activate POA neurons. As stated in the Discussion, we believe that our use of lower light intensity and stimulation frequencies could account for the discrepancy between our findings and those of Wei et al.

The presented timeline of optogenetic stimulation and resulting vocalizations (subsection “Activating PAG-projecting POA neurons elicits USVs in the absence of social cues”) is difficult to understand. Even when counting the different synaptic stages and adding significant conduction velocity delays, the time course is biologically not plausible but simply too long. This observation is not discussed in detail and it remains unclear. How can it be excluded that vocalizations are being produced by chance and that the correlation of stimulation and vocalization occurs by chance?

The absence of vocalizations in various control experiments that we now include (Figure 2F) rules out a chance relationship. In the revised manuscript, we have provided quantification showing that the USV production is elicited selectively during and following periods of optogenetic activation, and that baseline rates of vocalization outside of laser stimulation periods were low in these animals (Figure 2B, D, E). We note that mean latencies on the order of seconds from optogenetic activation of the hypothalamus to observed effects of behavior have been reported in other studies as well (see Figure 2I and Figure 3H from Wei et al., 2018, showing mean latency of ~5s from POA activation to mounting and pup retrieval, respectively; similarly, see Figure 4I from Lin et al., 2011 showing mean latency of ~4s from VMHvl activation to attack), adding additional support to the biological plausibility of our observed latencies. Even with direct optogenetic activation of PAG-USV neurons, we observed mean latencies upwards of hundreds of milliseconds (Tschida et al., 2019). We have added mentions of these previous findings in relation to the latencies we observed in the current study (subsection “Activating PAG-projecting POA neurons elicits USVs in the absence of social cues”).

The authors attempt to quantify the USVs during optogenetic stimulation and in control conditions in an unbiased way and used an unsupervised modeling approach. The data were visualized as UMAPs of latent features (Figure 3C). Based on differences in the visual appearing of these maps and the MMD the authors claim that a subset of USVs are similar and another subset is dissimilar during opto-stimulation versus female presence (subsection “Acoustic characterization of USVs elicited by activation of POA neurons”). This statement is confusing, and it remains unclear what is similar or dissimilar. To address this issue the authors investigated the acoustics in more detail. They argue that opto-USVS tended to be louder and covered a higher frequency bandwidth. This result, displayed in Figure 3F, is difficult to grasp. In Figure 3F it cannot be discriminated whether points are overlaid and therefore, the green data points are not visible in most part of the figure.

The data in Figure 3F are not overlaid in a manner that obscures the visibility of green points, and the figure accurately represents the finding that a subset of optogenetically-elicited USVs were louder and had greater frequency bandwidth than female-directed USVs produced by the same mice.

Another concern is that USVs differ in different social contexts. The authors should perform the same analysis on USVs that were elicited when no female was present and without opto-stimulation to ensure that the observed change is not due to difference in the social context. Temporal organization and usage can be differentiated in multiple putative USV classes also arising from distinct articulatory patterns (Castellucci et al., 2018).

Mice produce few or no USVs when tested in the absence of social cues or partners. However, males will often vocalize to the presentation of female urine. The behaviors exhibited by a mouse while investigating urine are relatively similar to those performed by solo-tested mice in our optogenetic activation experiments (no chasing or mounting due to the absence of a social partner, lots of sniffing, and slow, exploratory movements around the test chamber). Interestingly, a previous study found that USVs produced by males in response to female urine were significantly louder and tended to have greater bandwidth than those produced to a live female social partner (Chabout et al., 2015). We thank the reviewer for their insight, and we have added language to the manuscript (Discussion) to highlight the possibility that USVs elicited by optogenetic activation of the POA may be more similar to those produced in response to stationary female cues than to a moving female social partner.

Subsequently, Michael et al., tested the effect of stimulating PAG-projecting amygdala neurons and found that the stimulation of either EA-PAG neurons or even more specifically, GABAergic EA-PAG neurons results in the suppression of vocalizations during ongoing behavior. While this result is intriguing (Figure 4B) the presentation in Figure 4D (right panel) is misleading. Why is the USV count in the pre-condition so much higher than during the post condition? What is the effect of opto-stimulation on neural activity?

Mice typically produce USVs in bouts that last for several seconds. As one progresses further in time through a USV bout, there is an increasing probability that a bout will end. Thus, the decay in USV rates over time (pre vs. laser vs. post) in control animals simply reflects the natural statistics of USV production and is expected. We have clarified this idea in the legend for Figure 4.

The authors add a section about the upstream axonal projections of POA-PAG and EA-PAG neurons. These data do not add to the focus of this study and are not discussed further. What is the point the authors want to make with this?

We included these descriptive data simply for the sake of providing more information regarding the projection patterns of the cells characterized in this study, in case this information is of use to other scientists.

To test connectivity in detail the authors performed slice recordings and measured synaptic inputs onto different cell types while stimulating others. While they find that some PAG-USV neurons receive inhibitory currents when EA neurons are stimulated, the authors do not perform the symmetrical experiment and stimulate POA-PAG neurons while recording PAG-USV neurons. Instead, VGAT+ neurons are being recorded and it is shown that a subset receives inhibitory current as well. In addition, VGAT+ neurons are being stimulated and PAG-USV were recorded to demonstrate that VGAT+ neuron provide inhibition to these PAG-USV neurons. Unfortunately, the data are less conclusive as described in the text. For example, in subsection “Synaptic interactions between POAPAG and EAPAG neurons and the PAG vocal gating circuit” “majority (18 out of 27)”: Is this statistically significant? What can one compare it to? Is the number of recorded neurons too low?

We now include data in which we performed whole-cell recordings in brain slices from PAG-USV neurons while optogenetically activation PAG-projecting POA neurons. We only observed direct synaptic connections in 1/23 cells, and these findings are now reported in Figure 6—figure supplement 1.

Regarding the second point, we have collected additional data from VGAT+ PAG neurons while stimulating POA inputs, although we do not believe that the original number of recorded neurons that we reported is too low. We now report that 26/36 VGAT+ PAG neurons receive direct synaptic input from PAG-projecting POA neurons (subsection “Synaptic interactions between POAPAG and AmgC/M-PAG neurons and the PAG vocal gating circuit”). The result that GABAergic PAG neurons receive direct inhibitory inputs from the POA is quite clear. In our experience using optogenetic circuit mapping techniques in brain slices, finding that two thirds of cells are responsive to optogenetic stimulation is a good yield. We note that such a yield is typical or better than typical for optogenetic slice experiments in PAG. For example, see Figure 2I in the Tovote et al., 2016 Nature paper in which IPSCs were detected in 50% (12/24) of recorded PAG neurons upon optogenetic stimulation of local inhibitory neurons. Extended data Figure 2 from the same paper reports that IPSCs were detected in 3 example PAG neurons in response to terminal stimulation. Obviously, 26/36 responsive cells are a majority and there are many reasons why some cells recorded in brain slices fail to respond to optogenetic stimulation. First, it is unlikely that the POA would provide input to every VGAT+ neuron the PAG, as different PAG inhibitory neurons may be important for different PAG-mediated behaviors. Second, preparing a brain slice severs many connections, while also killing cells and terminals, thus further reducing the number of connections we can detect. Consequently, observing a direct inhibitory connection in 2/3 of recorded cells constitutes good evidence that this connection exists.

Results section and Discussion section: “data not shown” – In times of reproducibility of data and open access this statement is unacceptable. If the authors want to add this information to the text and speculate about them, they have to be shown.

We have removed these mentions of data that are not shown.

[Editors’ note: what follows is the authors’ response to the second round of review.]

Essential revisions:

1) One of the main claims of the paper is that there is an opposing effect of the POA versus Amg gabaergic input: while the POA leads to USVs, the Amg leads to USV suppression. However, if I understand correctly the experiment depicted in Figure 1B, the G and the TVA should only be expressed in gabaergic neurons of the PAG and then the rabies injection should label all neurons that project to those, irrespective of being gabaergic or glutamatergic (because the rabies is not Cre dependent). Then, the subsequent results with immunohistochemistry show that neurons in the POA and Amg are all gabaergic (according to the staining shown in Figure 1C and D). Later on, it is assumed that the POA input is disinhibitory, which makes sense with this result and everything that follows after. However, the fact that PAG GABAergic neurons also receive inhibitory input from the Amg is not pursued at all in the rest of the paper or discussed. Instead, it is assumed that the Amg GABAergic neurons only project to principal PAG neurons (which is depicted in the cartoon of Figure 7 as well). This should be discussed since it assumes that the projections of the Amg are much more complex than just inhibiting USVs. The cartoon should also, therefore, be revised.

The reviewer is correct that our rabies tracing revealed that inhibitory AmgC/M-PAG neurons provide input to PAG-USV neurons as well as to nearby GABAergic neurons. We have revised Figure 7 to include projections of AmgC/M-PAG neurons to both of these PAG cell types. Our optogenetic and slice data support the idea that AmgC/M-PAG neurons suppress USV production via their direct inhibitory inputs to PAG-USV neurons, but it is possible that these cells modulate diverse behaviors via their inputs to other cell types within the PAG, and we have added language to the Discussion to highlight this possibility.

2) Figure 2E: In my opinion, this is an important claim of the paper, that the projections from the POA to PAG can induce USVs. I don't think that such a strong claim can be done with a single male (this experiment is necessary to show that the effect is due to the projections of POA neurons in the PAG).

We have updated the subsection “Animals” to explain that we observed low survival rates for the Esr1-Cre males when we attempted to bilaterally implant the PAG with optogenetic ferrules. In our hands, Esr1-Cre males were smaller and less healthy than their female littermates and had to be weaned later than other animals in order to survive post-weaning. Even with these steps, we still had a low survival rate for Esr1-Cre males after this particularly invasive surgery. At a certain point, due to constraints on the number of animals we had available and the ethical concerns of repeating this experiment when the animals fared so poorly, we made the determination to stop attempting bilateral PAG implants in Esr1-Cre males and rather to focus on rounding out the dataset with female mice.

3) The addition to the Materials and methods section is useful: "Cells were scored from 2-7 sections of tissue per brain region from each animal, and the absence or presence of staining within cells was quantified manually by comparing labeling within cells to background labeling in nearby regions known to be negative for a given RNA transcript." But what are those control regions and how was it confirmed that these are "negative" for the given RNA transcript? In the interest of data transparency and reproducibility, this information should be included.

Neighboring control regions with low levels of transcript expression (relative to regions of interest) were present in the same coronal sections as the regions of interest and were determined by consulting the Allen Brain Atlas ISH Data (https://mouse.brain-map.org/search/index; experiment 72081554 for VGAT expression, experiment 79591677 for Esr1 expression, and experiment 73818754 for VGlut2 expression). These control regions were as follows for the following target regions and transcripts: (1a) POA VGAT: control region, fornix; (1b) AmgC/M VGAT: control region, thalamus; (1c) CeA VGAT: control region, thalamus; (2a) POA Esr1: control region, fornix; (3a) POA VGlut2: control region, caudate putamen; (3b) AmgC/M VGlut2: caudate putamen; (3c) CeA VGlut2: control region, caudate putamen. This information has been added to the Materials and methods.

We’ve also modified the wording describing transcript expression in same-section control regions, from being “negative” for transcript expression to having “low levels” of transcript expression, to acknowledge that an absence of strong labeling via ISH may indicate zero expression of a given transcript.

4) The authors state (subsection “Activating PAG-projecting AmgC/M neurons transiently suppresses USV production”) that "Optogenetic activation of AmgC/M-PAG neurons failed to elicit USV production and also did not drive any other overt behavioral effects". This seems like important evidence for the author's claim that the function of AmgC/M-PAG is primarily to suppress vocal behavior, but the data aren't shown. In the interest of data transparency and reproducibility, these data should be included as a supplemental figure.

To illustrate that optogenetic activation of AmgC/M-PAG neurons did not elicit obvious behavioral changes, we have provided three additional supplementary videos (Video 3, Video 4, Video 5) depicting three different mice in which we optogenetically stimulated AmgC/M-PAG neurons while the mice were alone in a chamber. These videos also include spectrograms and pitch-shifted playback of the audio recordings. It is clear from these videos that the mice do not vocalize upon stimulation and there are no obvious behavioral effects of the stimulation. We would also like to draw the reviewer’s attention to Figure 2—figure supplement 2, in which we demonstrate that optogenetic activation of AmgC/M-PAG neurons does not affect movement speed.

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Data Citations

    1. Michael V, Goffinet J, Pearson J, Wang F, Tschida K, Mooney R. 2020. Data and scripts from: Circuit and synaptic organization of forebrain-to-midbrain pathways that promote and suppress vocalization. Duke Research Data Repository. [DOI] [PMC free article] [PubMed]

    Supplementary Materials

    Figure 1—source data 1. Source data for Figure 1A–C.
    Figure 1—source data 2. Source data for Figure 1D.
    Figure 2—source data 1. Source data for Figure 2C and F.
    Figure 2—figure supplement 1—source data 1. Source data for panels B and C of Figure 2—figure supplement 1.
    Figure 2—figure supplement 2—source data 1. Source data for panel A of Figure 2—figure supplement 2.
    Figure 3—source data 1. Source data for Figure 3D.
    Figure 4—source data 1. Source data for Figure 4C and D.
    Figure 5—source data 1. Source data for Figure 5D and F.
    Figure 6—source data 1. Source data for Figure 6D and H.
    Figure 6—figure supplement 1—source data 1. Source data for panel B of Figure 6—figure supplement 1.
    Transparent reporting form

    Data Availability Statement

    Data have been deposited to the Duke Research Data Repository, under the https://doi.org/10.7924/r4cz38d99. We have deposited 4 types of data in the repository: (1) confocal microscope images of in situ hybridization, (2) audio and video files from the mice used in this study, (3) slice electrophysiology data, and (4) custom Matlab codes used for data analysis. All other data analyzed in this study are included in the manuscript and supporting files.

    The following dataset was generated:

    Michael V, Goffinet J, Pearson J, Wang F, Tschida K, Mooney R. 2020. Data and scripts from: Circuit and synaptic organization of forebrain-to-midbrain pathways that promote and suppress vocalization. Duke Research Data Repository.


    Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd

    RESOURCES