Prairie vole atlas

As we aimed to quantify the oxytocin-stained neuronal elements within identified brain nuclei in the prairie vole model, we started with composing an annotated brain atlas of the prairie vole based on serial coronal sections of a male animal brain, stained for visualization of cytoarchitecture and myelinated fiber tracks (See Methods). The current pdf version of the atlas, The Histochemical Brain Atlas of The Prairie Vole (Microtus ochrogaster), released with this manuscript (https://doi.org/10.18130/V3/LSAONY) contains 33 annotated brain atlas templates, representing 200 µm thick coronal plates (Fig. 1). Each plate includes superimposed images of Nissl and Myelin sections that were used for demarcating and annotating brain regions on templates. The plates also contain oxytocin immuno-stained sections that we used for fiber and cell mapping analysis, as described below. Annotated templates are provided as vector images, which can be copied by future users on their own material (Ramos and Erisir 2024).

Brain regions containing OXT+ cell bodies and dendrites

We used The Histochemical Atlas of Prairie Vole Brain as a reference template to map the locations of oxytocin-producing cell bodies in the prairie vole brain. In DAB-labeled sections, oxytocin immunostain (OXT+) fills the somata in their entirety, as well as the dendrites emanating from the soma (Fig. 2A-E). Most cells have a multi-polar structure, with dendrites extending in different directions. Semi-quantitative scoring and computation of a Staining Density Index (see Methods) for each region allowed assessments of the prevalence of OXT+ cells.

Fig. 2

OXT+ cell bodies in prairie vole brain primarily reside in the hypothalamus. A An OXT-stained coronal section through the hypothalamus reveals darkly stained cell bodies in the PVH, SON, and LH. The superimposed inset marks the region on a corresponding atlas section in red and applies to panels A to E. Scale bar = 100 μm. B Upon higher magnification, the OXT+ PVH cell bodies are multipolar. Those that are close to the ventricle extend dendrites into the ventricular lining (black arrows). Scale bar in B = 500 μm and applies to panels B to E. C) In the SON, OXT+ somata are fusiform and multipolar, and project dendrites toward the ventral surface of the brain. D Ventricular lining surrounding lateral ventricles occasionally displayed solitary OXT+ somata, which extend processes toward the ventricle. These are referred to as ectopic cells in the text as they cannot be identified as belonging to any brain nuclei. The example in panel D is close to BNST, the only extra-hypothalamic region that consistently contains OXT+ cells. E The presence of OXT labeling in AVPV cells is unique to the prairie vole. F Light-sheet microscope image of the hypothalamus at 5 × objective in the horizontal plane, thick white arrows illustrate cell body groups within the PVH, LH, SON, and Tu, thin white arrows in the lower left corner indicate anterior (a), posterior (p), and lateral (l) directionality of the image, maximum intensity projection across 1100 μm, scale bar = 500 μm

The hypothalamic regions had the highest density of OXT+ cells (Fig. 2A & F). The OXT+ cells are densest in the PVH and SON (Fig. 2B, C), followed by the tuberal nucleus (Tu) and periventricular hypothalamic preoptic part (PVpo). OXT+ cells are sparser in other regions, including the lateral hypothalamus (LH) and medial preoptic area (MPOA). An extrahypothalamic region, the BNST, consistently contains OXT+ cell bodies that closely line the ventricle and occasionally appear ectopic, and as such, do not belong to one particular brain region (Fig. 2D). Dendrites of cells in PVH, SON, and BNST often are located close to and sometimes within the ventricular lining (Fig. 2B–D). As described before (Kelly et al. 2018; Kenkel et al. 2021), the distinction between magnocellular and parvocellular cells that are typically seen in other rodent models, such as mice and rats, is not as apparent in the prairie vole brain. Although the rostral part of the PVH primarily encompasses the magnocellular cell group and the caudal part encompasses the parvocellular group of the prairie vole, no differences were seen within the cell morphology of the PVH. For the full list of cell body locations see Supplemental Table 1. Finally, light-sheet microscopy corroborated that the densest cell body localizations reside in the hypothalamus (Fig. 2F).

To reveal potential sexual dimorphisms, cell body SDI scores from 3 male and 3 female vole brains were compared using Welch’s t-tests. No statistical differences were found (Supplemental Fig. 2a), suggesting that sexual dimorphisms seen in behaviors modulated by OXT cannot be explained by the localization or the density of oxytocinergic cells.

To address potential species differences, we compared the localization of cells in each brain region of the prairie vole to previous studies in the mouse, hamster, and rat (Swaab et al. 1975; Laurent et al. 1989; Whitman and Albers 1998; Otero-García et al. 2016; Son et al. 2022). While most OXT+ cell localizations are similar, we noted two differences. First, similar to hamsters and unlike other rodents, the prairie vole has OXT+ cells within the median preoptic nucleus (MEPO) (Whitman and Albers 1998). Second, the prairie vole has a group of OXT+ cells located in the anteroventral periventricular nucleus (AVPV), which has not been noted in any other species (Fig. 2E). Interestingly, AVPV is among the regions in which processes project towards the ventricle and occasionally through the ependymal cell layer in the prairie vole (Fig. 2E).

Brain regions containing OXT-labeled axons

Next, we used DAB-labeling of OXT with brightfield microscopy to examine all major nuclei containing OXT axonal projections. On coronal sections, two distinct projection paths emerging from the PVH can be observed: one consists of a dense bundle of axons that courses dorso-laterally, and another that contains sparse and thin axonal fibers that project more ventro-medially (Fig. 2A). The dorso-lateral axons extend horizontally above the anterior hypothalamic nucleus (AHN) and make a sharp ventral turn at the lateral hypothalamus (Fig. 2A, Fig. 3A). PVH axons seem to merge with another stream of axons emerging from the SON and these course along the ventral surface of the brain along the retrochiasmatic nucleus (RCH) and above the optic tract (opt) towards the median eminence (ME) (Fig. 3B). The axons of this dorso-lateral stream are thick and display swellings throughout their course (Fig. 3A).

Fig. 3

Locations of OXT+ fibers within the prairie vole LH (A), RCH (B), AHN (C), and MPOA (D). The superimposed inset marks the region on a corresponding atlas section in red. Scale bar = 50 μm and applies to panels A to D. E) Illustration of fiber density data (Supplemental Table 3) over a sagittal section adapted from Allen Mouse Brain Atlas (mouse.brainmap.org). The density of fibers (Supplemental Table 3) is represented in tertiles (dense, sparse, and very sparse, which was determined within the cerebral cortex (orange), hindbrain (yellow), hypothalamus (green), striatum/pallidum (blue), midbrain (teal), and thalamus (pink). See Supplemental Table 3 for full list of brain regions, their SDIs, and their classifications into the subregions (cerebral cortex, hindbrain, etc.),

The ventro-medial axons emerging from the PVH follow a less defined course, winding throughout the AHN (Fig. 3C). These axons do not appear to be joining the OXT axon stream projecting towards the median eminence and are very fine displaying many varicosities (Fig. 3C). Further, there are axons in other regions of the hypothalamus, anterior and posterior to the appearance of the PVH and SON, which show thick fibers with large swellings such as the MPOA (Fig. 3D).

The OXT+ axons are dense in many non-hypothalamic subcortical brain regions, including the periaqueductal gray (PAG), ventral tegmental area (VTA), pons, nucleus reuniens of the thalamus (RE), BNST, and AcbSh and AcbC (Fig. 3E. Note: classifications into “dense”, “sparse”, and “very sparse” were made based on tertiles that can be derived from the pooled data in Supplemental Table 3. Dense regions: SDIs = 0.46–1; sparse regions: SDIs = 0.07–0.42; very sparse regions: SDIs = 0–0.05). Various sub-cortical regions appeared to have dense fiber staining, while labeled axons were extremely rare in cortical regions. Interestingly, OXT+ axons were observed touching all major ventricles and were prominent in many areas that border the ventricles, such as the lateral septum (LS), BNST, paraventricular nucleus of the thalamus (PVT), and PAG (Fig. 3E; for the full list of axonal locations and the SDIs of individual subjects see Supplemental Table 2; for all axonal locations and the pooled SDIs see Supplemental Table 3).

Sexual dimorphisms for OXT+ fibers were examined across 105 brain regions in 3 male and 3 female prairie vole brains. Regions that only had one score in either males or females were excluded because a mean value could not be calculated from one observation. While our sample size was not sufficient for statistical power in population comparisons, comparison of SDIs within each brain region revealed two regions that may be sexually dimorphic: the dorsal endopiriform cortex (EPd), and the prelimbic cortex (PrL) (Supplemental Fig. 2b). Unlike in male brains, female brains did not display any fibers in Epd (0.17 ± 0.026 vs. 0 ± 0; mean ± SD; Welch’s t-test, t(2) = -11.7, p = 0.007)) Males also had higher OXT fiber density in the PrL than females (0.25 ± 0.07 vs. 0.04 ± 0.06; Welch’s t-test, t(3.95) = −3.88, p = 0.02; Supplemental Fig. 2b). These results suggest that perhaps males utilize more OXT in the EPd and PrL than females, and this may play a role in downstream behavioral outcomes.

Comparison of brain regions with OXT+ fibers and Oxtr transcripts

To confirm that the animals used in the current study displayed Oxtr expression patterns similar to that were recently described (Inoue et al. 2022), we examined sample brain regions that were reported to express high levels of Oxtr, yet show vast differences in their OXT+ fiber density in our study. In particular, we examined the Oxtr prevalence in the Nucleus Accumbens Shell (AcbSh), Core (AcbC), and Cingulate Cortex (Cg). Although both regions reportedly have high Oxtr levels (Fig. 4A-B), fiber density analysis yields different SDI values between the Acb (pooled SDI Sh: 0.67 and C: 0.53) and Cg (0.05) (SuppTable3). The RNAScope visualization of Oxtr RNA transcripts (Oxtr) in 10 sections from 5 animals revealed that both the Acb (Sh and C) and Cg contain an abundance of Oxtr-expressing cells (Fig. 4A, B). Specifically, 15.4% ± 6.9 of cells in AcbSh, 23.9% ± 13.1 of cells in AcbC, and in the Cg 18.4% ± 5.3 of cells in Cg expressed Oxtr. The puncta density (i.e. counts of Oxtr puncta per cell) was 3.6 ± 1.6 in AcbSh, 3.4 ± 1.8 in AcbC, and 8.4 ± 2.2 in Cg. These results confirm that the Acb and Cg cells both highly expressed Oxtr in our sample brains, comparable to the Inoue et al. 2022Oxtr dataset, where the Cg received a score of 3, and the Acb (Sh and C) received scores of 4 (Fig. 4A, B; Supplemental Table3).

Fig. 4

Correlation of OXT+ fibers and Oxtr across the whole brain. A Confocal images of Oxtr transcript expression (pink) within the AcbSh and AcbC (A), and the Cg (B). Cell nuclei are stained with DAPI (blue). Scale bar = 50 μm. C-G Graphs shown are illustrations of the raw data (brain regions belonging to category are shown as points), regression lines (solid lines) and 95% confidence intervals (shaded areas around regression lines), these are shown for visualization purposes, see H for full statistical analyses. C The density of OXT+ fibers within the thalamus is positively related to density of Oxtr transcript (p = 0.001), but not in hypothalamus (D, p = 0.41), midbrain (E, p = 0.54), striatum/pallidum (F, p = 0.95), or the cerebral cortex (G, p = 0.44). H) Graphical representation of estimated linear trend analysis illustrating that the thalamus (Thal, pink) is the only region where presence of OXT+ fibers predicts Oxtr. The hypothalamus (Hypo, green) and midbrain (Mid, teal) both have negative, non-significant trend values with a range of OXT+ fibers (often high) and little Oxtr. The striatum/pallidum (Stria/Pall, blue) has a trend value of near 0, thus no relationship between OXT+ fibers and Oxtr. The cerebral cortex (CC, orange) has a positive, nonsignificant trend value, indicating typically low OXT+ fibers and varying amounts Oxtr (often high). I The staining density index (see Methods) at all regions in descending order from the highest density of OXT+ fibers (top) to the lowest density of OXT+ fibers (bottom) colored by region category. The relative amount of Oxtr is indicated by the size of the circle marker

In order to analyze the relationship between OXT+ fiber density and Oxtr localization across the prairie vole brain, we compared our OXT+ fiber SDI data with a dataset of previously published quantitative distribution of Oxtr transcripts (Inoue et al. 2022) (Fig. 4C–I). We examined this relationship across 5 parent regions as established by the Allen Mouse Brian Atlas (mouse.brainmap.org): the cerebral cortex, hypothalamus, midbrain, striatum, and pallidum, and the thalamus (for brain regions included in each category, see Suppl. Table 3). This analysis revealed a significant interaction effect between Oxtr and brain region, suggesting that the relationship between OXT+ fibers and Oxtr transcripts differs depending on region (linear regression, F(1,63) = 2.47, p = 0.05). Interestingly, the subregions within each parent category tend to cluster together in relation to their OXT fiber density, and they typically have a similar level of average Oxtr (Fig. 4I). To further probe the differences amongst regions, we utilized linear trend analysis. There was a general directionality of the trends yet, in most areas, the amount of OXT fibers was not correlated with the amount of Oxtr. There was a negative but non-significant relationship in the hypothalamus and midbrain; no relationship in striatum/pallidum, and a non-significant positive relationship in the cerebral cortex (Fig. 4H). Only the thalamus category (which includes RE, PVT and CM) displays a significant relationship between OXT+ fibers and Oxtr transcript (estimate linear trends analysis, Trend = 0.13, p = 0.001; Fig. 4C). This suggests that within the brain structures included in the thalamus category, the amount of Oxtr transcripts can be explained by the amount of OXT+ fibers. That is, the axonal release in the thalamus may constitute a primary source for OXTR protein binding.

In contrast, in the hypothalamus and midbrain categories, there is no significant relationship between OXT+ fibers and Oxtr (Trend = −0.04, p = 0.41; and Trend = -0.010, p = 0.54). While the fiber density in these areas is typically high, Oxtr density is low, suggesting that OXT+ fibers in these regions are likely axons of passage (Fig. 4D, E). The striatum and pallidum display no significant relationship between OXT+ fibers and Oxtr. These areas typically have high amounts of Oxtr yet moderate to high (> 0.15SDI) levels of OXT+ fibers (Trend = -0.003, p = 0.95, Fig. 4F). Similarly, the cerebral cortex category contains only a few regions with scant amounts of OXT+ fibers, but many regions display high levels of the Oxtr transcript (Trend = 0.03, p = 0.44, Fig. 4G). This mismatch in the amount of OXT+ fibers and Oxtr, particularly in cortex and striatum/pallidum regions with high Oxtr but no or few OXT+ axons, indicates that the receptor localization cannot be explained by the presence of OXT+ fibers alone and that these regions may be getting the bulk of their oxytocin through non-axonal means.

Ultrastructural characteristics of OXT+ axons in regions with differing amounts of Oxtr transcripts

To characterize the fine morphological properties of OXT+ fibers in regions of high or low levels of Oxtr, we examined the LH, RCH, MPOA, and AHN regions immunostained for OXT using transmission electron microscopy. The OXT+ labeling was evident as the appearance of electron-dense DAB chromogen diffusely filling the profiles of neurons, dendrites, and axons (Fig. 5, 6). In regions where diffuse DAB label in profiles was too dense and obscured the organelles within, pre-embedding gold enhanced visualization approach was used, revealing OXT+ profiles with the appearance of irregularly shaped gold deposits (Fig. 6E).

Fig. 5

OXT+ somata in PVH A display clusters of dense cored vesicles (blue arrowheads, DCVs) in the cytoplasm (A,B). Unlabeled DCVs (yellow arrows) were also observed in PVH neuropil (C). Scale bar = .5 μm and applies to (B, C). D Frequency distribution histogram of OXT+ (blue bars, n = 61) and OXT− (yellow bars, n = 60) dense core vesicle areas, blue and yellow line overlays represent density curves for labeled and unlabeled distribution respectively. The dashed line marks the cutoff value for classifying DCVs as oxytocinergic. OXT+ : oxytocin positive cell, DCVs: dense cored vesicles, mit: mitochondria, m: microtubules, d: dendrite

Fig. 6

OXT+ axons of the LH and RCH typically do not contain DCVs, indicating they are likely axons of passage, while axons of the MPOA and AHN often contain DCVs. A An Electron micrograph of OXT+ axons in the LH, and B in the RCH, with no DCVs. C EM image of an OXT+ axon in the MPOA with one DCVs (blue arrowhead), and D an image of an axon in the AHN with two DCV’s (blue arrowheads), and unstained neuropil with DCV’s (yellow arrows), indicating in the MPOA and AHN oxytocin can be directly released to act on OXTR. Pixel resolution = 1134.92 pixels/μm. Scale bar = 1 μm. Mit: mitochondria, v: vesicles, den: dendrite, m: microtubules, syn: synapse

Within OXT+ profiles, darkly stained, large DCVs often appeared in clusters within labeled somata and neuropil (Fig. 5A, B). Because DCVs of various sizes were also observed in unlabeled profiles (Fig. 5C, yellow arrows), we quantified the size of labeled and unlabeled DCVs in the PVH to obtain a size criterion for OXT+ DCVs. The DCVs in labeled PVH cells were uniformly large (0.016μm2 ± 0.005, n = 61; ~ 140 nm in diameter) and these could be distinguished from other DCVs encountered in the OXT− neuropil (0.004μm2 ± 0.002, n = 60; ~ 70 nm in diameter) (Fig. 5B, C, respectively). A cutoff value of 0.008μm2 marks the intersection point of labeled and unlabeled DCV size distributions (Fig. 5D). Only 5% of stained DCVs fall at or below this cutoff value. Thus, vesicles that are smaller than this cutoff are categorized as OXT− in our subsequent analysis.

In EM preparations of regions that contain many OXT+ fibers but few Oxtr transcripts, such as the LH and RCH (Fig. 6A, B), OXT+ axons are uniformly non-myelinated and vary in diameter. To determine if oxytocin is likely released from these axons or if they are primarily axons of passage, we examined the LH and RCH using electron microscopy and quantified the incidences of OXT+ axons that contained OXT DCVs. Only 6% of axonal profiles in the LH (n = 48) and 9% of the axonal profiles in the RCH (n = 44) displayed an OXT DCV. When present, DCVs were sparse, indicating that the labeled fibers are more likely axons of passage en route to median eminence (Fig. 6A, B). In contrast, in the MPOA and AHN, many OXT+ profiles contain at least one large DCV (26% of profiles, n = 54 and 23%, n = 39, respectively) (Fig. 6 C, D), suggesting that oxytocin may be released from axons in MPOA and ANH, providing the neuropeptide ligand for the OXTR that is expressed in these regions.

Regions that contain Oxtr transcripts and no OXT+ fibers are especially puzzling, because the source of the oxytocin that could activate OXTR in these regions is not obvious. A possibility is the delivery of oxytocin to cortical extracellular space via CSF circulation. Here, we provide evidence that the circulating CSF may contain OXT that is directly released from dendrites. Using light and electron microscopy, we observed many instances of PVH cell dendrites extending through the ependymal cell layer and directly contacting the third ventricle (Fig. 7A, B). Of the PVH cell dendritic profiles that are in the ependymal zones, 13% (n = 156) contained DCVs (Fig. 7C, D). Thus, these dendrites are situated to exocytose oxytocin from dense-cored vesicles directly into the CSF. Oxytocin may then readily flow through the subarachnoid space along blood capillaries and reach OXTR in the cortical regions via volume transmission.

Fig. 7

Dendrites of the PVH and SON contain DCVs and are positioned to release OXT into the CSF. A OXT+ cells of the PVH extend their dendrites towards the third ventricle (3V), appearing to cross the ependymal cell layer. Scale bar = 500 μm. B Electron micrograph of an OXT+ dendrite (OXT+ den) with a DCV (blue arrowhead) located at ventricle-side of the ependymal cells (Ep). The dendrite directly contacts the third ventricle (3 V). C-D Dendrites (OXT+ den) within the PVH that contain DCVs (blue arrowhead). Other, smaller DCVs (yellow arrows) are often encountered in unlabeled neuropil in the same region as the OXT+ dendrites and axons. E An immuno-gold labeled (blue asterisks) OXT+ dendrite within the SON, with a high density of large DCVs (blue arrowheads). Pixel resolution = 1134.92 pixels/μm; scale bars on B-E = 1 μm. Tj: tight junction of ependymal cell; mit: mitochondria; cil: cilia; mv: microvilli; at: axon terminal; v: vesicles

Finally, using DAB-labelled and gold-enhanced EM we revealed that 40% (N = 30) of dendrites in the SON contain DCVs (Fig. 7E). These dendrites often appear densely packed with large DCVs (Fig. 7E). Interestingly, we noticed using light microscopy, dendrites of the SON cells extend towards the ventral surface of the brain and appear to project through the ependymal cell layer, indicating the possibility that oxytocin can be released into the CSF from SON cells as well (Fig. 2C).

Whole-Brain imaging of OXT+ staining reveals the extent of ventricular axonal staining

To confirm that ventricular OXT labeling is a prominent feature of the prairie vole brain, which was observed via the brightfield microscopy analysis in regions previously mentioned, such as the BNST, and PAG, we examined two brains that were prepared for whole brain clearing and light-sheet microscopy. Major ventricles have positive OXT staining and OXT+ axons appear to follow along ventricles to terminate in subcortical regions. The brains scanned at 5X magnification revealed a strong fluorescence signal lining 3rd ventricle flanked by the hypothalamus (Fig. 8A), as well as the lateral ventricles throughout their anterio-posterior span (Fig. 8B). However, whether these axons are filled with DCVs was not examined using electron microscopy.

Fig. 8

3D whole-brain light-sheet microscopy verified that OXT+ axons border major ventricles. A The PVH and PVpo both consistently contain cell bodies, and axons projecting from these regions to follow along the third ventricle (3V), with maximum intensity projection across 775 μm. B Representative image of the lateral ventricle (LV), fimbria, and BNST with axonal fibers again bordering a major ventricle, maximum intensity projection across 1554 μm. Both images were derived from 3D whole-brain scan using a light sheet microscope at 5 × objective, are shown in the horizontal plane, and further processed using Imaris software. The superimposed inset marks the region on a corresponding coronal atlas section in red, and thin white arrows in the lower right corner arrows indicate rostral (r), caudal (c), and ventral (v) orientations of the image. Scale bar = 300 μm