Uncovering the brain-wide pattern of synaptic input to vasopressin-expressing neurons in the paraventricular nucleus of the hypothalamus

Abstract

Arginine vasopressin (AVP) is a neuropeptide critical for the mammalian stress response and social behavior. AVP produced in the hypothalamus regulates water osmolality and vasoconstriction in the body and in the brain it regulates social behavior, aggression, and anxiety. However, the circuit mechanisms that link AVP to social behavior, homeostatic function, and disease are not well understood. This study investigates the circuit configurations of AVP expressing neurons in the rodent hypothalamus and characterizes synaptic input from the entire brain. We targeted the paraventricular nucleus (PVN) using retrograde viral tracing techniques to identify direct afferent synaptic connections made onto AVP-expressing neurons. AVP neurons in the PVN display region-specific anatomical configurations that reflect their unique contributions to homeostatic function, motor behaviors, feeding, and affiliative behavior. The afferent connections identified were similar in both sexes and subsequent molecular investigation of these inputs shows that those local hypothalamic inputs are overwhelmingly non-peptidergic cells indicating a potential interneuron nexus between hormone cell activation and broader cortical connection. This proposed work reveals new insights into the organization of social behavior circuits in the brain, and how neuropeptides act centrally to modulate social behaviors.

INTRODUCTION

AVP is a small neuropeptide hormone well known for its roles in regulating plasma osmolality^1,2, arterial blood pressure, circadian rhythms, and social behaviors^3–8. AVP is primarily produced by neurons in the paraventricular nucleus of the hypothalamus (PVN), suprachiasmatic nucleus of the hypothalamus (SCN), and supraoptic nucleus of the hypothalamus (SON) where magnocellular neurosecretory neurons project to the pituitary^9–13. AVP-expressing neurons in the PVN regulate social behaviors including social approach⁸, social play^7,14 aggression¹⁵, social cognition¹⁶, anxiety-like behaviors¹⁴, and monogamy¹⁷ through central projections in the brain. The goal of the work described here is to reveal the sources of synaptic input to AVP-expressing neurons that enable social behaviors and, therefore, we focus on the AVP-dependent circuits of the PVN.

Here, we use conditional genetics to target AVP-expressing neurons in the PVN (${\text{AVP}}_{\text{PVN}}^{+}$) along with retrograde rabies-based monosynaptic tracing^20–22 to identify the synaptic connections made onto ${\text{AVP}}_{\text{PVN}}^{+}$. By imaging the intact brain, we quantify the brain-wide patterns of synaptic input to ${\text{AVP}}_{\text{PVN}}^{+}$, and then express these inputs in a standardized and reproducible reference frame. Afferent connections are primarily local and similar in both sexes, with the majority found in the adjacent hypothalamic subnuclei and ventrally in the thalamic regions and pallidum. Identification of these inputs shows that hypothalamic inputs to the PVN likely target non-AVP producing cells preferentially indicating a possible interneuron “gateway” between hormone cell activation and broader cortical connection.

While the coarse connectivity of hypothalamic circuits has been established¹⁸, the PVN contains a diverse and interwoven collection of cell types that are believed to provide unique contributions to circuit function and behavior. The work here establishes the circuits in which ${\text{AVP}}_{\text{PVN}}^{+}$ neurons are directly embedded and represents a key step towards understanding AVP-dependent brain functions. We have identified the direct synaptic inputs that constrain from where ${\text{AVP}}_{\text{PVN}}^{+}$ neurons receive information. This work provides a foundation for future studies exploring the electrophysiological function of ${\text{AVP}}_{\text{PVN}}^{+}$ neurons, as well as studies identifying the multisynaptic neural circuits that translate sensory and interoceptive input through ${\text{AVP}}_{\text{PVN}}^{+}$ neurons to, ultimately, influence social behaviors.

RESULTS

We verified our genetic strategy first by crossing the AVP-ires-cre line with a tdTomato reporter line¹⁹ which allowed us to visualize accurate AVP expression in neurons throughout the brain (Figure 1b,c). Next, we injected a conditional AAV-FLEX-TVA-mCherry virus, without the AAV-FLEX-RG virus required for transsynaptic circuit mapping, to determine the specificity of ‘starter neurons’ to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons. This approach allowed us to verify that the viral strategy was faithful^20–22 with 92% of all GFP-expressing ‘starter neurons’ co-expressing tdTomato (Figure 1e–g). Because AAV-FLEX-TVA-mCherry weakly expresses mCherry and could potentially be confused with tdTomato expression, we further confirmed the specificity of starter neurons through anti-AVP immunostaining (Supplementary Figure 1). Consistent with our previous results (Figure 1g) we found 7% of GFP-expressing starter neurons did not express tdTomato nor AVP, while 93% of starter neurons co-expressed both tdTomato and AVP. All identified starter neurons that expressed tdTomato were co-labeled by anti-AVP immunostaining and all starter neurons labeled by anti-AVP immunostaining also expressed tdTomato.

Figure 1: Experimental design and strategy verification.

a) Schematic of functional circuits for ${\text{AVP}}_{\text{PVN}}^{+}$ neurons in the hypothalamus. ${\text{AVP}}_{\text{PVN}}^{+}$ neurons impact biological processes including sexual behavior (motor control), water balance homeostasis (osmolality), homeostatic stress, and circadian clocks. b) Transgenic constructs for the knock-in AVP-ires-cre line and the cre-dependent tdTomato reporter line. c) Rendering of AVP expression in the brain generated by crossing the AVP-ires-cre and R26-LSL-tdTomato transgenic lines. d) Viral constructs used for circuit mapping include flexed AAV-FLEX-TVA-mCherry, flexed AAV-FLEX-RG, and SADΔG-EGFP (TVA-dependent rabies with GFP). e) Viral Injection protocol for validating the specificity of rabies labeling restricted to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons by excluding expression of the glycoprotein. The timing of viral injections is indicated below. f) Endogenous tdTomato fluorescence in ${\text{AVP}}_{\text{PVN}}^{+}$ neurons. Overlap showing tdTomato expression (red, i), GFP expression (green; ii), and merged signal (ii). Arrowheads indicate cell bodies of ‘starter neurons’ that express both tdTomato and GFP and therefore indicate specific infection g) Fraction of rabies-infected neurons that also express tdTomato (yellow).

Spatial distribution of synaptic inputs to AVP neurons

To identify the afferent inputs to ${\text{AVP}}_{\text{PVN}}^{+}$ we injected a conditional AAV-FLEX-RG, in addition to the injection protocol described above, to facilitate packaging and transsynaptic retrograde transmission of the SADΔG-EGFP virus from starter neurons²² (Figure 2a, b). Whole brains were then imaged intact, digitally reconstructed as a 3D volume, GFP-expressing neurons were identified in a semiautomated fashion, and each data set was aligned to the Allen Mouse Brain Common Coordinate Framework²³ (AMBF; Figure 2c; see methods). The vast majority of synaptic inputs to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons were restricted to subcortical regions of the brain and, in particular, inputs were densest in hypothalamic areas adjacent to the PVN (Figure 2c). The number of neurons labeled retrogradely by GFP scaled linearly with the size of the initial infected population (R² = 0.77, p < 0.0001; Figure 2d).

Figure 2: Brain-wide identification of synaptic inputs to ${\text{AVP}}_{\text{PVN}}^{+}$.

a) Viral injection protocol of retrograde tracing experiments. In addition to AAV-FLEX-TVA-mCherry, AAV-FLEX-RG was injected to allow transsynaptic retrograde transmission from starter neurons. b) Sagittal representation of anatomical inputs to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons located in the PVN (red) and outside the PVN (cyan). Each point represents a single identified neuron mapped to the Allen Mouse Brain Common Coordinate Framework (AMBF). c) Linear regression of the number of GFP-expressing cells in the PVN compared to GFP-expressing cells located outside the PVN for each brain (R^{^}2 = 0.77, p < 0.0001)

Each viral injection was unilateral, and the identified synaptic inputs were primarily located ipsilateral to the injection area (77.42 +− 6.10 SEM, p = 0.07; Supplemental Figure 2a). The distribution of inputs was centered on the injection site with a roughly normal distribution in the three cardinal directions (Supplemental Figure 2b). Spreading outward from the point of injections, the number of cells decline significantly after approximately 1.5 mm, with the largest spread being dorsal from the PVN (Supplemental Figure 2b). An example of the scope of individual inputs to ${\text{AVP}}_{\text{PVN}}^{+}$ is shown in Figure 3. While the majority of inputs were located in the hypothalamus (Figure 3a), more distant inputs were identified in regions including the bed nucleus of the stria terminalis (BNST), medial amygdala (MeA), thalamus, (TH), orbital cortex (ORB), and periaqueductal grey (PAG) (Figure 3b–g).

Figure 3: Synaptic input to ${\text{AVP}}_{\text{PVN}}^{+}$. neurons are dense in subcortical regions.

a) Sagittal representation of total inputs to ${\text{AVP}}_{\text{PVN}}^{+}$. color coded based on gross region: cyan = hypothalamus, indigo = thalamus, purple = cerebral nuclei, red = hippocampal formation, teal = midbrain, green = hindbrain, yellow = isocortex, orange = olfactory areas. b-g) Visualization of cells in six regions with synaptic input. bed nucleus of stria terminalis (BNST; b) paraventricular nucleus (PVN; c), medial amygdala (MeA; d), thalamus (TH; e), hippocampus (HIPP; f), periaqueductal grey (PAG; g).

Distribution of synaptic inputs to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons is most dense in thalamus and hypothalamus

We initially segmented the full brain volume into 10 mutually exclusive regions that comprehensively span the mouse brain volume and used these regions to broadly classify the locations of cells bodies that provide synaptic input to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons. More than half of inputs to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons came from within the hypothalamus (56.76 +/− 2.19 SEM, p = 2.36e-06) while roughly one third of the inputs came from thalamus (33.12 +/− 2.35 SEM, p= 0.00068). Smaller but significant populations of inputs were identified in the ventral cerebral nuclei (5.44 +/− 0.64 SEM, p = 0.014) and the hippocampus (2.76 +/− 0.75 SEM; p = 0.15 Figure 4). Input neurons were the densest in the hypothalamus while the thalamus was clearly over-represented when accounting for region size (Figure 4d). Despite being the largest coarse region, the isocortex contained very few identified neurons indicating extremely sparse labeling. Given the small number of neurons identified in olfactory areas, isocortex, midbrain and hindbrain (all less than 1%), they were excluded from subsequent analyses; however, the full datasets including statistical analyses for all regions are provided (Supplementary Materials).

Figure 4: Coarse analysis of synaptic input to ${\text{AVP}}_{\text{PVN}}^{+}$. Neurons.

a) 3D representation of coarse anatomical areas used to segment synaptic inputs. b) Percentage of synaptic inputs coming from each coarse area represented as a fraction of total input from each brain. c) Percentage of cells within each coarse area. d) Density of synaptic input is highly enriched in the hypothalamus and thalamus when controlled for region size. The diagonal line indicates the expected proportion of synaptic inputs based on region size and if inputs were evenly distributed across the brain volume. Color coding of region is constant for all panels.

Region specific distribution of synaptic inputs

Closer investigation of the hypothalamus was conducted using regions defined by the AMBF²³. The bulk of hypothalamic input was restricted to the paraventricular zone (PVZ) with many of those inputs coming from within the PVN itself (9.70 +/− 0.84 SEM, p=0.003; Figure 5). Inputs from the PVN were more than twenty times as dense as any other hypothalamic region when accounting for region size (Supplementary Materials). While a portion of this cell density is due to the initial ‘starter cell’ infection, AVP antibody staining and the sheer number of cells compared to other regions instead shows a high level of local connection. Smaller populations within the PVZ that showed significant inputs to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons include the intermediate part of the PVN (1.40 +/− 0.22 SEM, p= 0.04) and arcuate nucleus (1.17 +/− 0.18 SEM, p=0.04). The next largest sources of inputs to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons come from the medial zone (MZ, 10.03 +/− 0.50 SEM, p= 3.4e-05) and the lateral zone (LZ, 8.76 +/− 0.60 SEM, p= .0005) of the hypothalamus. Within the MZ a significant population of input originated from the descending PVN (PVNd, 2.88 +/− 0.40 SEM, p= 0.03) with smaller populations from the medial preoptic nucleus (MPN, 2.10 +/− 0.25 SEM, p= 0.015) and anterior hypothalamic nucleus (AHN,1.57 +/− 0.16 SEM, p= 0.01) and ventromedial hypothalamus (VHA, 1.51 +/− 0.22 SEM, p= .035). Within the LZ the highest percentage subregions include the lateral hypothalamus (LHA, 3.50 +/− 0.39 SEM, p=.011), zona incerta (ZI, 2.36 +/− 0.21 SEM, p= .003) and tuberal nucleus (TU, 1.97 +/− 1.10 SEM, p= .05). Finally, the periventricular region of the hypothalamus (PVR), the dorsomedial hypothalamus (DMH, 2.51 +/− 1.22 SEM, p= .031) provides significant synaptic input to ${\text{AVP}}_{\text{PVN}}^{+}$ (Figure 5).

Figure 5: Region-specific breakdown of hypothalamic synaptic inputs to ${\text{AVP}}_{\text{PVN}}^{+}$.Neurons.

a) Segmentation of the extended hypothalamus. Left: representation of broad hypothalamic subdivision (inset: proportion of total cells in the hypothalamus). Center Left: Coronal slice visualizing synaptic input location Center Right and Right: proportions of synaptic input to four broad subdivisions of the hypothalamus: Periventricular zone (PVZ); hypothalamic medial zone (MEZ); hypothalamic lateral zone (LZ); periventricular region (PVR). b) Synaptic input from PVZ subregion of hypothalamus. Supraoptic nucleus (SO); periventricular hypothalamic nucleus (Pva), arcuate hypothalamic nucleus (ARH); periventricular hypothalamic nucleus (Pvi), Paraventricular hypothalamic nucleus intermediate part (PVN). c) Synaptic input from MZ subregion of hypothalamus. (Medial mammillary nucleus (MM); supramammillary nucleus (SUM); ventral premammillary nucleus (PMv); posterior hypothalamic nucleus (PH); ventromedial hypothalamic nucleus (VMH); anterior hypothalamic nucleus (AHN); paraventricular hypothalamic nucleus (PVNd); descending division of medial preoptic nucleus (MPN). d) Synaptic input from LZ subregion of hypothalamus. (Inset: RCH =retrochiasmatic area LPO =lateral preoptic area PeF =perifornical nucleus TU =Tuberal nucleus ZI=Zona incerta LHA=Lateral hypothalamic area. e) Synaptic input from PVR subregion of hypothalamus. Ventromedial preoptic nucleus (VMPO); periventricular hypothalamic nucleus (PVp), posterior anterodorsal preoptic nucleus (ADP); ventrolateral preoptic nucleus (PVLPO); subparaventricular zone (SBPV); median preoptic nucleus (MEPO); medial preoptic area (MPO); anteroventral periventricular nucleus (AVPV); periventricular hypothalamic nucleus (Pvpo), preoptic dorsomedial nucleus of the hypothalamus (DMH).

The second largest area providing input to AVP+ neurons is the thalamus (Figure 6) with most inputs located in the from the midline group of the dorsal thalamus (MTN, 13.77 +/− 1.24 SEM, p=.003). This is followed by the medial group of the dorsal thalamus (MED, 10.24 +/− 1.41 SEM, p= .028). The ventral group of the dorsal thalamus (VENT) also made a significant contribution as well (3.22 +/− 0.50 SEM, p= .044) as well. The cerebral nuclei, which includes the striatum (STR) and pallidum (PAL), provides a substantial source of input to the PVN. The majority of these inputs come from the pallidum (4.15 +/− 0.61 SEM, p= .036), specifically the bed nucleus of the stria terminalis (BNST, 2.60 +/− 0.61 SEM; p 0.119) which displayed a sizeable but also variable level of input (Figure 7).

Figure 6: Region-specific breakdown of thalamic synaptic inputs to ${\text{AVP}}_{\text{PVN}}^{+}$. Neurons.

Left: representation of broad thalamic sub-division (inset: proportion of total cells in the thalamus). Center Left: Coronal slice visualizing synaptic input location Center Right and Right: proportions of synaptic input to four broad subdivisions of the thalamus. Intralaminar nuclei of the dorsal thalamus (ILM); geniculate group (GENv); subparafascicular nucleus (SPF); Reticular nucleus of the thalamus (RT); anterior group of the dorsal thalamus (ATN); ventral group of the dorsal thalamus (VENT); medial group of the dorsal thalamus (MED); midline group of the dorsal thalamus (MTN); lateral group of the dorsal thalamus (LAT)

Figure 7: Region-specific breakdown of cerebral nuclei synaptic inputs to ${\text{AVP}}_{\text{PVN}}^{+}$. Neurons.

a) Segmentation of cerebral nuclei. Left: representation of broad cerebral nuclei sub-division (inset: proportion of total cells in the cerebral nuclei). Center and Right: proportions of synaptic input to two broad subdivisions of the cerebral nuclei. Striatum (STR); pallidum (PAL). b) Left: representation of striatum sub-division. Center Left: Coronal slice visualizing synaptic input location. Center Right and Right: proportions of synaptic input to four broad subdivisions of the striatum. Synaptic input from striatal regions: nucleus accumbens (ACB); lateral septal nucleus (LS); medial amygdalar nucleus (MEA); central amygdalar nucleus (CEA); c) Left: representation of pallidum sub-division. Center Left: Coronal slice visualizing synaptic input location. Center Right and Right: proportions of synaptic input to four broad subdivisions of the pallidum. Synaptic input from pallidum regions: globus pallidus internal segment (Gpi); medial septal nucleus (MS); diagonal band nucleus (NDB); bed nuclei of the stria terminalis (BST)

Distribution of ipsilateral-contralateral synaptic contributions

Injections to the PVN were unilateral and the majority of hypothalamic inputs were ipsilateral to the injection site (Supplemental Figure 2c). In contrast, most thalamic inputs were contralateral to the injection site. The fraction of ‘starter neurons’ that were ipsilateral to the injection site was 77.43 +− 6.10 SEM (Supplemental Figure 1a), indicating that contralateral infection due to viral diffusion across the midline is unlikely to be a major source of contralateral inputs. Instead, the contralateral bias of thalamic inputs, as compared to ipsilateral circuitry inside the hypothalamus likely represents a principle of the circuit organization itself.

By comparing our ${\text{AVP}}_{\text{PVN}}^{+}$ neuron-specific approach to existing datasets that were non-specific, it is possible to estimate the enrichment of sources of input to the PVN specific to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons (Supplemental Figure 3). Overall, data showed there was not a significant relationship between regions that provide input to the PVN and those that provide input specifically to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons²³ (linear regression; P=0.98; Wang, et al., 2020). One group could not be determined by the other, and hints that as a distinct population ${\text{AVP}}_{\text{PVN}}^{+}$ neurons may provide a unique circuit contribution than afferent inputs that synapse onto other cellular populations in the PVN.

Potential sex differences

The number of brains were split between male (n=6) and female (n=8) groups in order to identify potential sex differences among input populations. The average number of cells per group were compared against each other for all regions containing at least 1% of total number of inputs. Although some trends emerged among areas including the midline group of the dorsal thalamus skewing male and paraventricular nucleus showing higher numbers in females (Figure 8a), no regions displayed a statistically significant sex difference in magnitude. Similarly, a sensitivity index comparing the region-specific discriminability of synaptic input to ${\text{AVP}}_{\text{PVN}}^{+}$ (maximum d’: 1.3) further emphasized the similarity in synaptic inputs to ${\text{AVP}}_{\text{PVN}}^{+}$ of males and females (Figure 8b).

Figure 8: Comparison of synaptic inputs to ${\text{AVP}}_{\text{PVN}}^{+}$. Neurons in male and female mice.

a) The average number of cells identified for each region containing at least 1% of the total synaptic input in male mice (ordinate) and female mice (abscissa). Error bars indicate SEM. No region displayed a statistical difference in magnitude of projection to ${\text{AVP}}_{\text{PVN}}^{+}$. Of male versus female mice (p>0.1 for all region comparisons; ANOVA with repeated measures, followed by unpaired students t-tests for individual regions). b) A sensitivity index was performed to quantify the discriminability of all regions between males and females. The sensitivity index comparing male and female projection patterns fell between 0.5 and 1 for nearly all regions (peak difference at 1.3) indicating strong similarity in the projection patterns in male and female mice.

Molecular Identification of Inputs

${\text{AVP}}_{\text{PVN}}^{+}$ neurons receive extensive input from within PVN and surrounding hypothalamic areas and these regions are molecularly heterogeneous. AVP, oxytocin (OXT) and corticotrophin releasing factor (CRF) are major neuropeptides released by PVN neurons and we investigated the overlap of these markers with rabies⁺ GFP neurons. Immunohistochemical co-labeling demonstrated limited overlap between GFP+rabies cells and anti-AVP (6%; Figure 9a) or anti-OXT (2%; Figure 9b), and no measurable overlap with anti-CRF (0%; Figure 9c). Because rabies infection can alter gene expression in infected neurons²⁴, we investigated whether AVP expression is inhibited by rabies infection in ${\text{AVP}}_{\text{PVN}}^{+}$ starter neurons. Genetically encoded tdTomato in AVP-ires-Cre neurons showed a 100% overlap with AVP identified through immunohistochemistry in rabies⁺ GFP neurons, indicating that, at least, AVP expression was maintained in rabies-infected neurons for the duration of these experiments (Supplementary Figure 4).

Figure 9: Local connectivity with ${\text{AVP}}_{\text{PVN}}^{+}$.

a) Immunofluorescent images of anti-AVP staining (red) Rabies-GFP (cyan) and their merged composite (white). The histogram shows the percentage of cells counted in each group out of all counted neurons. b) Immunofluorescent images of anti-OXT staining (red) Rabies-GFP (cyan) and their merged composite (white). The histogram shows the percentage of cells counted in each group out of all counted neurons. c) Immunofluorescent images of anti-CRF staining (red) Rabies-GFP (cyan) and their merged composite (white). The histogram shows the percentage of cells counted in each group out of all counted neurons.

DISCUSSION

Given the roles of AVP neurons in varied social behaviors and homeostatic responses, we predicted broad synaptic input from diverse brain regions; however, our data revealed the majority of synaptic inputs to ${\text{AVP}}_{\text{PVN}}^{+}$ come from the hypothalamus itself ²⁵. Thus, while the PVN receives synaptic input from wide ranging brain areas, ${\text{AVP}}_{\text{PVN}}^{+}$ neurons primarily integrate information from within the hypothalamus itself. Below we discuss the functions most strongly represented by synaptic inputs to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons including: 1) homeostatic & thermoregulatory functions; 2) feeding; 3) stress; 4) learning and memory; 5) affiliative behaviors; and 6) pain (Figure 10). While the unbiased and brain-wide approach taken here can identify all major sources of direct synaptic input to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons, each of these input populations is connected with a wider circuit²⁰. By identifying the sources of direct input to ${\text{AVP}}_{\text{PVN}}^{+}$, we have constrained the anatomical circuits that support social and neuroendocrine AVP functions and provide a foundation for future anatomical studies regarding the broader multisynaptic circuits in which ${\text{AVP}}_{\text{PVN}}^{+}$ are embedded.

Figure 10: Summary of major afferent inputs to ${\text{AVP}}_{\text{PVN}}^{+}$. Based on known function.

Most input regions fall into the functional aspects of: Learning & Memory (red circle), Stress (green circle), Homeostasis (blue circle), Nociception (yellow circle), Feeding (pink circle) and Affiliative Behavior (purple). The size of the circles generally represents the strength of the afferent connections measured by proportion of total inputs of the regions of which they encompass. Note that numerous regions have been linked to more than one of these general functions. In these cases, regions are included in all relevant categories.

The PVN is well known for its secretion of neuropeptides that influence osmolality, blood pressure, and feeding. Consistent with these roles, many regions that synapse onto ${\text{AVP}}_{\text{PVN}}^{+}$ neurons have important roles in homeostasis and thermoregulation. The dorsomedial hypothalamus regulates activity balance and is broadly connected with the suprachiasmatic nucleus which locally produces AVP, is the key regulator of circadian rhythms²⁶, and is involved in food intake control and addiction²⁷. Similarly, the anterior hypothalamus is involved in the regulation of sleep-wake and thermoregulation in conjunction with preoptic areas, which, themselves, provide direct input to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons²⁸. The lateral hypothalamus houses the orexin system which regulates both arousal and feeding to promote energy balance in mammals^29,30, and ^22,24also regulates gastrointestinal function by way of the vagus nerve³¹. Lastly, the tuberal nucleus, also located in the lateral zone of the hypothalamus, regulates feeding by way of somatostatin neurons³². Collectively, these sources of input to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons³³ highlight the critical role of ${\text{AVP}}_{\text{PVN}}^{+}$ neurons in feeding, wakefulness, water balance, and energy homeostasis.

In addition to its role in sleep, ablation of the LHA increases pain related behaviors³⁴. Similarly, the dorsomedial hypothalamus, which our data shows synapses directly on ${\text{AVP}}_{\text{PVN}}^{+}$ neurons, has been linked to modulation of pain with acute stress suppressing pain and chronic stress increasing pain sensitivity³⁵. Studies that inactivate the DMH in conjunction with mild stressors block the sensitivity to mechanical stimulation while activation increases pain sensitivity, confirming that the DMH mediates behavioral hyperalsia³⁶, and direct activation of the zona incerta (ZI) modulates behavioral responses to noxious stimuli ³⁷. Central pain syndrome, a condition resulting from dysfunction in the spinal cord, has been associated with abnormal regulation of the posterior thalamus by way of signally from the ZI³⁸.³⁹. Collectively, these sources of input to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons likely support the established role of AVP in increasing the threshold to painful stimuli⁴⁰.

We identified synaptic inputs, including local PVN input, from regions linked to the stress response ^41–46. In the PVN, both CRH and AVP are released to the median eminence in response to stress, and AVP release synergistically stimulates ACTH creating a stable feedback loop. The nucleus reuniens (RE) contained in the midline nuclear group of the thalamus is a critical node connecting the hippocampus and prefrontal cortex and lesioning the RE before Chronic Mild Stress blocks neural correlates of mild depression, creating a resilience to stress⁴⁷. RE provides extensive input to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons. ⁴³The BNST, which provided the strongest input from the pallidum is, similarly, linked to anxiety-like behaviors in humans⁴⁸ and animals^49–50 which can be sex-specific⁵¹. Thus, we identified significant input likely linked to the regulation and response to stress and^37–42 to changing internal and environmental conditions.

In line with AVP’s causal role in social memory^52–53 and modulatory role on hippocampal-dependent learning^53–55, we identified input directly from regions critical for learning and memory. The midline nuclear group of the thalamus is comprised of distinct nuclei that project to the prefrontal cortex and hippocampus, and also regulate the extinction of conditioned fear memory retrieval through connections with the amygdala⁵⁶. The RE is an integration hub for memory and cognitive function and its dysfunction has been implicated in memory disorders including Alzheimer’s and schizophrenia^56–58. ⁵⁶Input from these regions shows AVP neurons have direct access to an integrative network that regulates attention, memory, and decision-making in the service of cognitive flexibility and action planning⁵⁹.

The role of AVP in social behavior may be best understood in two closely related vole species that differ with respect to monogamy. Prairie voles display monogamous behavior while meadow voles display non-monogamous behavior⁶⁰, and this difference is mediated by AVP signaling and differential expression of the Avpr1a receptor in the ventral forebrain. Indeed, AVP injected centrally promotes affiliative behavior⁶⁰, and viral expression of the Avpr1a receptor in the ventral forebrain conveyed monogamous behavior to a non-monogamous species^60,61. Studies also consistently implicate AVP in affiliative behavior in mice^8,21,62,63. Consistent with this array of AVP-dependent behaviors, three of the strongest inputs we identified are the MPOA, MEA, and the BNST are directly linked to parental, aggressive, and affiliative behaviors^64,65 indicating three sources from which AVP neurons likely integrate sensory and interoceptive cues relevant to affiliative behaviors. While AVP-dependent behaviors including aggression, mating, and parenting often display species specific sex differences^61,65,66, we identified no sex differences in inputs to ${\text{AVP}}_{\text{PVN}}^{+}$; but rather, we found the input to AVP-expressing neurons quite similar in males and females. We interpret this as evidence that known sex differences in AVP dependent behaviors likely arise in other parts of the mediating circuits, some with well-known sex-differences, or from the efferent targets of ${\text{AVP}}_{\text{PVN}}^{+}$^67,68.

Despite the intersectional roles of AVP, OT^69–74, CRH^45,75,76 and the vast majority of connections formed from the hypothalamus, most afferent connections to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons do not express AVP, OT, or CRF. Rabies can alter the gene expression of infected cells and, therefore, it is possible that our analysis underestimates the level of overlap between inputs to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons and AVP, OT, or CRF. However, we observed that rabies infected neurons continue to express AVP for the duration of the experimental protocol. Therefore, while we cannot rule out the possibility that OT and CRF are inhibited by rabies infection, at a minimum the majority of inputs to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons come from non- ${\text{AVP}}_{\text{PVN}}^{+}$ neurons²⁴. One intriguing possibility is that inputs to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons provide a gateway linking more diverse and far-ranging areas of the brain to the neuropeptide manufacturing PVN.

AVP and the neurons that produce AVP sit at a critical juncture of stress, homeostasis and behavior. In this study, we reveal the presynaptic input to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons in the PVN. The AVP-specific circuit framework revealed here is a step towards understanding the mechanisms by which AVP influences circuit functions and ultimately guides adaptive behaviors, as well as, to investigate specific sub-circuit functions. Perhaps most surprising result, given the wide array of AVP-dependent behaviors, is that ${\text{AVP}}_{\text{PVN}}^{+}$ neurons primarily from local circuits with non-peptide-expressing cells that reside near the PVN in both thalamic and hypothalamic areas. From these regions, AVP neurons pull together all the information necessary to integrate allostatic load, thermoregulation, metabolic need, and shape social behaviors.

METHODS

Animals:

Adult AVP-ires-Cre⁶² and R26-LSL-tdTomato (Jax strain #:007905) mice (aged 8–14 weeks) were housed in a fully temperature (22°C) and light (12h light/dark cycle) controlled vivarium and given ad libitum access to food and water.All experiments were performed in strict compliance with the National Institute of Health. All animals were handled according to a protocol approved by the UMass Amherst Institutional Animal Care and Use Committee (IACUC; protocols 2440 and 2654).

Viral Injections:

We performed retrograde tracing using a modified rabies virus SADΔG-EGFP (EnvA), which requires two sequential stereotaxic injections to visualize inputs to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons²⁰. Animals were anesthetized with aerosolized isoflurane (2%) and maintained under isoflurane anesthesia for the duration of each surgical procedure. The skull was exposed, and a small craniotomy was opened dorsal to the stereotaxic coordinates. 550 nl AAV-FLEX-TVA-mCherry and AAV-FLEX-RG²⁰ was injected in 10 increments into the left PVN (Bregma −.58, Lateral .23, Depth 4.75). Stock AAV titers ranged from 0.9×10¹³ to 2.3×10¹³ gc/ml and viruses were mixed (1 part AAV-FLEX-TVA-mCherry: 1 part AAV- FLEX-RG) or diluted by 50% in sterile PBS. All AAV viruses were produced by the UNC Vector Core Facility (Chapel Hill, NC, United States). After 14 days 500 nl EnvA virus (Viral Vector Core, Salk Institute, La Jolla, CA) was injected in the PVN at the same coordinates. In all stereotaxic injections, a glass capillary (20-micron diameter tip) coupled to a Narishige hydraulic manipulator⁷⁷ was used. The skin was sutured shut following each viral injection and animals were injected with Meloxicam analgesic (2.5mg/kg) daily for three days post-surgery.

Clearing Process/ Histology:

Ten days after the final injection, animals were terminally anesthetized with isoflurane and perfused with 50ml cold PBS followed by ~20 ml cold PFA (4% in PBS). The brain was extracted and post-fixed in 20 ml hydrogel⁷⁸ (4% acrylamide, 4% PFA, 0.05% bis acrylamide, and 0.25% VA-044 initiator suspended in 0.01 M PBS) at 4 °C overnight.

After 18 – 24 hours incubation in cold hydrogel, the solution was degassed using at least 1 L nitrogen gas. The sample was then incubated at 37°C until polymerization was complete (at least six hours). Excess hydrogel was removed from the brain manually and the tissue sample was incubated to SDS-clearing solution (10 mM sodium dodecyl sulfate in 0.1 M borate buffer, pH 8.5) for ~48 hours at 37 °C prior to MHD-accelerated clearing⁷⁹. After clearing, brains were transferred from PBS to an OptiView imaging solution^78,80 with refractive index 1.45 and incubated at 37°C for 12 – 24 hours before imaging.

Imaging:

Images were acquired with the Zeiss Z.1 Lightsheet microscope (Carl Zeiss, Jena, Germany). Rabies-labeled eGFP-expressing neurons were excited with a 488 nm laser. A 561 nm and/or 647 nm laser was used to produce an autofluorescence image for subsequent background subtraction and isolation of the GFP signal. Images were collected with a 5-times magnification objective lens with PCO-Edge scMOS cameras (PCO, Kelheim, Germany). The entire brain was imaged in the horizontal orientation from both the dorsal and ventral surfaces. This produced a series of slightly overlapping 3D image stacks for each brain. 3D Image stacks were saved at 1–5 μm resolution and reconstructed to form a 3D image of the entire brain using custom MATLAB scripts (Mathworks, Natick, MA).

Analyses:

Brain Alignment:

3D image stacks created by the lightsheet microscope were reconstructed using a combination of manual input and computer processing using custom MATLAB scripts (Mathworks, Natick, MA). Each reconstructed 3D brain was then aligned to AMBF²³ framework using elastix^81–82.

Cell-counting:

Rabies-labeled cells were identified using a human-trained computer vision algorithm (Ilastik, Heidelberg University). Automatically identified cells (Ilastik, Heidelberg University) were reviewed and confirmed by a human observer. Transformix^81,82 was then used to migrate the identified cells in each individual brain to the position of the accurate brain region in the AMBF This allowed us to identify the number of eGFP-expressing cells in each region of the brain in each animal.

Statistical Analysis:

Statistical analyses were performed using MATLAB. Means are reported with standard error of the mean. Regressions were performed in MATLAB using the “fitlm” and “predict” functions. Unless otherwise noted, all statistical tests were performed using a two-sample t-test with Bonferroni correction when appropriate.

Antibody Characterization:

Rabbit anti-AVP polyclonal antibody (Immunostar, 20069; used a 1:1000), has been previously characterized and produces accurate fluorescent staining at a 1/1,000 – 1/4,000 dilution in rat hypothalamus^83,84. Rabbit anti OXT antibody (Immunostar, 20068, used at 1:1000) demonstrates accurate labeling of rat hypothalamus using indirect immunofluorescent and biotin/avidin-HRP labeling^85,86. Rabbit anti-CRF antibody (Abcam, ab216599, used at 1:1000) demonstrates accurate labeling of CRF-expressing neurons, made against recombinant rabbit synthetic peptide (aa 148–196).

Supplementary Material

Fig S3

Supplemental Figure 3: Comparison of synaptic input to ${\text{AVP}}_{\text{PVN}}^{+}$. Neurons with anterograde connectivity adapted from the Allen Mouse Brain Common Coordinate Framework. The fraction of input to ${\text{AVP}}_{\text{PVN}}^{+}$. Is indicated on the ordinate and nonspecific anterograde input to the PVN is indicated on the abscissa (AMBF). The thalamic nucleus reuniens (RE) and hypothalamic lateral hypothalamus (LHA) and tuberal nucleus (TU) were overrepresented in afferent inputs to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons. The ventral hypothalamus was conversely underrepresented in afferent inputs to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons when compared to the general PVN.

Fig S2

Supplemental Figure 2: Ipsilateral bias and distance of synaptic inputs from injection site. a) Ipsilateral-Contralateral skew of inputs to the PVN (Ipsilateral inputs: 77.426 ± 6.1 SEM; p = 0.035). b) Breakdown of input distance from PVN centroid in three ordinal directions. Top: most synaptic inputs were ipsilateral, posterior, and dorsal with respect to the PVN injection site. Bottom: The overwhelming majority of synaptic inputs to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons were less than 2mm in distance from the PVN injection site. c) Region-specific bias in ipsilateral versus contralateral connectivity. Left: Contrast ratio for ipsilateral versus contralateral input to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons, contrast ratio = (# cells_contra - # cells_ipsi) / (# cells_contra + # cells_ipsi). Right: total number of ipsilateral (cyan) and contralateral (red) neurons identified for each region. Background color indicates coarse brain region: thalamus (purple), hypothalamus (cyan), cerebral nuclei (green). Note the contralateral bias of the thalamic regions and ipsilateral bias of the hypothalamic regions.

Fig S1

Supplementary Figure 1: Validation of circuit mapping specificity to ${\text{AVP}}_{\text{PVN}}^{+}$ neurons. AVP-ires-Cre mice⁶² crossed with R26-LSL-tdTomato were injected with only AAV-flex-TVA-mCherry and SADΔG-EGFP to restrict rabies infection and GFP expression to starter neurons. a) Immunofluorescent images of AVP-ires-Cre mice crossed with R26-LSL-tdTomato. Top panels showing AVP-ires-Cre dependent tdTomato expression (red), anti-AVP immunostaining (blue) and rabies-GFP (cyan) alongside a composite image. Bottom panels show magnification of highlighted box. b) Venn diagram showing the overlap of cellular expression of tdTomato and AVP. All tdTomato-expressing starter neurons were also labeled with anti-AVP immunostaining and 93% of all GFP-expressing starter neurons expressed both AVP and tdTomato.

Data availability statement:

Data available on request from the authors.