ABSTRACT
Birth is a critical period for the developing brain, a time when surging hormone levels help prepare the fetal brain for the tremendous physiological changes it must accomplish upon entry into the ‘extrauterine world’. A number of obstetrical conditions warrant manipulations of these hormones at the time of birth, but we know little of their possible consequences on the developing brain. One of the most notable birth signaling hormones is oxytocin, which is administered to roughly 50% of laboring women in the United States prior to / during delivery. Previously, we found evidence for behavioral, epigenetic, and neuroendocrine consequences in adult prairie vole offspring following maternal oxytocin treatment immediately prior to birth. Here, we examined the neurodevelopmental consequences in adult prairie vole offspring following maternal oxytocin treatment immediately. Control prairie voles and those exposed to 0.25 mg/kg oxytocin were scanned as adults using anatomical and functional MRI, with neuroanatomy and brain function analyzed as voxel-based morphometry and resting state functional connectivity, respectively. Overall, anatomical differences brought on by oxytocin treatment, while widespread, were generally small, while differences in functional connectivity, particularly among oxytocin-exposed males, were larger. Analyses of functional connectivity based in graph theory revealed that oxytocin-exposed males in particular showed markedly increased connectivity throughout the brain and across several parameters, including closeness and degree. These results are interpreted in the context of the organizational effects of oxytocin exposure in early life and these findings add to a growing literature on how the perinatal brain is sensitive to hormonal manipulations at birth.
BACKGROUND
Oxytocin (OXT) is a potent and pleiotropic hormone that surges at birth to help facilitate the tremendous changes that both mammalian mothers and their offspring must accomplish upon delivery (1–3). As of 2019, 29.4% of laboring women in the U.S. received OXT to induce labor (4), and according to available survey data, this figure raises to ∼50% of birthing women in America when considering OXT used to either induce and/or augment labor (5). This obstetric practice is of interest to neuroscience because there is evidence OXT can cross the placenta (6) and a growing literature suggests the neonatal brain is particularly sensitive to OXT around the time of birth, when OXT receptor (Oxtr) expression begins to accelerate (7) and OXT neurons in the brain undergo intense remodeling (8). Indeed, the long-term, developmental effects of OXT manipulations in early life are well-documented (9,10), which suggests the perinatal period may be a sensitive period with regard to the impact of OXT.
Some initial studies suggested higher rates of autism spectrum and attention deficit / hyperactivity disorder amongst children born to women whose labors were induced with OXT; however, meta-analysis of these findings suggest that any such conclusions remain premature (11). While there have been conflicting reports as to whether OXT administered to induce / augment labor is associated with increased rates of autism spectrum disorder or autistic-like behavior in offspring, considerations of dose add an important degree of nuance to this topic (12,13). Thus, regardless of whether the consequences of obstetrically administered OXT raise to the level of a neurodevelopmental disorder, the question of whether OXT affects offspring neurodevelopment is of great public health relevance given its widespread use.
Previously, we investigated the impact of maternally administered OXT on offspring neurodevelopment and behavior using the socially monogamous prairie vole (14). We found that fetal physiology was indeed sensitive to maternally administered OXT and that, in the fetal brain, such OXT dose-dependently increased methylation of the Oxtr promoter. In adulthood, OXT-exposed offspring of both sexes were found to demonstrate a broadly gregarious phenotype such that they exhibited more spontaneous alloparental care toward unrelated pups and spent more time in close social contact with opposite-sex adults. Male voles exposed to OXT also showed increased density of OXT receptor in the central amygdala, insular cortex, and parietal cortex, while showing decreased vasopressin receptor density in the ventral pallidum.
Because OXT is a pleiotropic hormone, we opted for a broad survey of the brain in the present study, carrying out whole-brain resting functional connectivity to further characterize the scope of the neurodevelopmental consequences of OXT exposure at birth. Here, we used magnetic resonance imaging (MRI) to scan the brains of adult male and female prairie vole offspring originally born to pregnant females treated with 0.25 mg/kg OXT on the expected day of delivery. We examined both anatomical measures (voxel based morphometry (VBM) and diffusion-weighted imaging (DWI)) as well as functional measures (resting state functional connectivity (rs-fMRI) and several graph theory measures detailed below).
METHODS
Subjects
Prairie vole offspring (Microtus ochrogaster) were generated as previously described (14). All procedures were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee of Northeastern University. On the expected day of delivery, pregnant females were either injected intraperitoneally with OXT (0.25 mg/kg, ‘OXT’) or left undisturbed (‘Control’). Offspring were only included if they were delivered within 24 hours of OXT treatment. Offspring were raised by their birth parents as we previously observed no effect of maternal-OXT treatment on offspring outcomes (14). At 20 days of age, OXT and Control offspring were weaned into same-sex sibling pairs and were left to mature. Upon reaching adulthood (postnatal days 60-70), OXT and Control offspring underwent three neuroimaging scans: aT1-weighted anatomical scan for voxel based morphometry scan (VBM), an awake, resting state functional scan (rs-fMRI), and an anesthetized diffusion-weighted imaging scan (DWI), as detailed below. Subject offspring consisted of 17 Control females, 19 Control males, 17 OXT females, and 16 OXT males. From these, 7 Control females, 12 Control males, 13 OXT females and 13 OXT males were ultimately included in the rs-fMRI analyses after removing subjects due to motion artefact or technical difficulties.
Neuroimaging
All neuroimaging measures were collected using a Bruker BioSpec 7.0T/20-cm Ultra Shield Refrigerated horizontal magnet (Bruker, Billerica, MA). A 20-G/cm magnetic field gradient insert (inner diameter 12 cm) was used to scan anesthetized subjects using a quadrature transmit/receive volume coil (inner diameter 38 mm). Imaging sessions began with an anatomical scan with the following parameters: 20 slices; slice thickness, 0.70 mm; field of view, 2.5 cm; data matrix, 256 × 3 × 256; repetition time, 2.5 seconds; echo time (TE), 12.0 ms; effective TE, 48 ms; number of excitations, 2; and total acquisition time, 80 seconds.
Voxel Based Morphometry (VBM)
The following procedures were adapted for use in the vole from those described previously for rats (15). For each subject, the atlas (image size 256 × 256 × 63) (H × W × D) was warped from the standard space into the subject image space (image size 256 × 256 × 40) using the nearest-neighbor interpolation method. In the volumetric analysis, each brain region was therefore segmented, and the volume values were extracted for all 111 regions of interest (ROIs), calculated by multiplying unit volume of voxel (in mm3) by the number of voxels using an in-house MATLAB script. To account for different brain sizes, all ROI volumes were normalized by dividing each subject’s ROI volume by their total brain volume.
Diffusion-weighted Imaging (DWI)
The following procedures were identical to those described previously (16,17). Diffusion-weighted imaging (DWI) was acquired with a spin-echo echo-planar imaging (EPI) pulse sequence with the following parameters: repetition time/TE, 500/20 ms; 8 EPI segments; and 10 noncollinear gradient directions with a single b-value shell at 1000 seconds/mm2 and 1 image with a b-value of 0 seconds/mm2 (referred to as b0). Geometrical parameters were as follows: 48 coronal slices, each 0.313 mm thick (brain volume) and with in-plane resolution of 0.313 × 3 × 0.313 mm2 (matrix size, 96 × 3 × 96; field of view, 30 mm2). The imaging protocol was repeated 2 times for signal averaging. DWI acquisition took 35 to 70 minutes. DWI included diffusion-weighted three-dimensional EPI image analysis producing fractional anisotropy (FA) maps and apparent diffusion coefficient. DWI analysis was implemented with MATLAB (version 2017b) (The MathWorks, Inc., Natick, MA) and MedINRIA version 1.9.0 (http://www-sop.inria.fr/asclepios/software/MedINRIA/index.php) software.
Each brain volume was registered with the three-dimensional MRI Vole Brain Atlas template (Ekam Solutions LLC, Boston, MA) allowing voxel- and region-based statistics (18). In-house MIVA software was used for image transformations and statistical analyses. For each vole, the b0 image was coregistered with the b0 template (using a 6-parameter rigid-body transformation). The coregistration parameters were then applied on the DWI indexed maps for each index of anisotropy. Normalization was performed on the maps providing the most detailed and accurate visualization of brain structures. Normalization parameters were then applied to all indexed maps and then smoothed with a 0.3-mm Gaussian kernel. To ensure that preprocessing did not significantly affect anisotropy values, the nearest neighbor option was used following registration and normalization.
Resting State Functional MRI (rs-fMRI)
We used the same equipment and scanning protocols as in our recent work; for complete details see (18–20). Data were analyzed as 111 nodes corresponding to brain regions specified in a vole-specific atlas (18). Pearson’s correlation coefficients were computed per subject across all node pairs (6105), assessing temporal correlations between brain regions. Then, r-values’ (−1 to 1) normality were improved using Fisher’s Z-transform. For each group, 111×111 symmetric connectivity matrices were constructed, each entry representing the strength of edge. An |Z|=2.3 threshold was used to avoid spurious or weak node connections (21).
Network Analyses
Graph theory network analysis was generated using Gephi, an open-source network and visualization software (22). For all groups, the absolute values of their respective symmetric connectivity matrices were imported as undirected networks and a threshold of |Z|=2.3 was applied to each node’s edges to avoid spurious or weak node connections (23).
Betweenness Centrality
Betweenness centrality analyzes occurrences where a node lies in the path connecting other nodes (24). Let be the number of pathways from i to j going through k. Using these measures of connection, the betweeness of vertex k is:
Degree Centrality
Degree centrality indicates the number of associations of a specific node (25). Non-weighted, binary degree is defined as: where n is the number of rows in the matrix in the adjacency matrix A and the elements of the matrix are given by Aij, the number of edges between nodes i and j.
Closeness Centrality
Closeness centrality measures the average distance from a given starting node to all other nodes in the network (26). Closeness is defined as: where d(y,x) is the distance between vertices x and y and N is the number of nodes in the graph.
Statistics
Normality tests of control females, control males, OXT females and OXT males were performed to examine if parametric or non-parametric assumptions were required for future analysis. Shapiro-Wilk’s tests were performed to examine normality assumption for degree, closeness and betweeness centrality values. Regional p-values that were greater than 0.05 were assumed to be normal. A corresponding list of nodes that classified a region is detailed in table S1. After assumptions of normality were validated, one-way ANOVA tests were used to compare differences in degree, closeness and betweenness centralities between groups. When necessary, a nonparametric Kruskal-Wallis test was performed if there was evidence against normality assumption. Statistical differences between groups were determined using a Mann-Whitney U test (a = 5%). The following formula was used to account for false discovery from multiple comparisons:
P(i) is the p value based on the t test analysis. Each of 111 regions of interest (ROIs) (i) within the brain containing V ROIs. For graph theory measures, statistical analyses were calculated using GraphPad Prism version 9.0.0 for MacOS (GraphPad Software, San Diego, California USA, www.graphpad.com). For post-hoc analyses of graph theory parameters, Holm-Šídák test (for parametric) and Dunn’s (for nonparametric) were used after correction for multiple comparisons.
RESULTS
We observed a number of differences in VBM measures; however most were small to very small in effect size (Figure 1 and Tables 1-3). Within the Control group, there was only a single sex difference in regional volume; within the OXT group, however, OXT females had larger volumes in 8 of 17 cortical regions and smaller volumes in 11 brainstem / cerebellar regions compared to OXT males (Table 1). Similarly, Control males had larger volumes in 9 of 17 cortical regions and smaller volumes in 9 brainstem / cerebellar regions compared to OXT males (Table 2). Comparing within females, OXT treatment at birth resulted in smaller volumes in 4 of 17 cortical regions (Table 3). Control animals generally had larger amygdalar volumes than OXT animals (4 of 6 subregions in males; 2 of 6 in females). When morphometry data from all 111 brain regions were loaded into a PCA, the overall explanatory value of dimensions 1 and 2 was modest (29.4% and 15.9% respectively) and there were impacts of both sex and treatment on dimension 1, with male sex (F(1,69) = 4.98, p = 0.029) and OXT treatment (F(1,69) = 16.06, p < 0.001) leading to greater values (Figure 2).
We observed a broad, albeit subtle pattern of effects in OXT-exposed males in DWI measures. In terms of FA, Control animals showed small but widespread sex differences, with females having greater FA than males across brain regions, a pattern not present in OXT animals due to increased FA among OXT-exposed males (Figure 3A, Tables 2 and 3). Indeed, there were no differences in either FA or ADC between OXT males and OXT females, whereas there were 62 and 48 such regions amongst Control animals. When FA data from all 111 brain regions were loaded into a PCA, there were no effects of either sex or treatment detected (Figure 3A). In terms of ADC, Control males had greater ADC values than OXT males across 84 brain regions (Table 5). While we observed widespread ADC differences between Control males and females, we observed no sex differences in ADC in the OXT-exposed condition. The PCA for ADC revealed a main effect of OXT exposure (F(1,65) = 5.80, p = 0.019) and a trend toward an interaction between sex and treatment (p = 0.087). Post-hoc analysis revealed that Control males having greater dimension 1 values than both OXT males (p = 0.026) and OXT females (p = 0.042). Thus, across the brain, OXT males’ ADC values more closely resembled Control females and OXT females than they did Control males.
In the analyses of functional connectivity (Figures 4-8), OXT males stood out as having a widespread pattern of greater connectivity. The vast majority of connections across all groups arose from positive correlations. Male sex and OXT treatment both increased the proportion of significant connections for both positive and negative connections (chi-square p < 0.001 for both effects). Whereas Control females were found to have significant functional connectivity in 5% of all possible connections, OXT females had significant connectivity in 7.2% of connections. Whereas Control males had connectivity in 6.3% of connections, OXT males had significant connectivity in 12.5% of connections (Figure 4A). Similarly, male sex and OXT treatment at birth both increased the strength of connectivity among region-region pairs whose activity was significantly correlated, but only for positively correlated pairs (p < 0.001 for both effects, Figure 4B). Thus, OXT led to more regions significantly functionally connected and stronger correlations in such regions among males and to a lesser extent among females (Figures 4,5). In examining the pattern of results within correlation matrices, we observed a concentration of stronger connectivity among regions of the same cluster (Figure 4D and E).
This led us to examine the strength of intra- vs. inter-cluster connectivity as a function of sex and birth treatment, which revealed that male sex and OXT treatment increased intra-cluster connectivity (e.g. central amygdala and medial amygdala) to a greater degree than inter-cluster connectivity (e.g. central amygdala and dentate gyrus, Figure 4C). As shown in Figure 5, when comparing the strength of connectivity at the level of regional clusters (e.g. all subregions of the amygdala), we observed a treatment effect in the medulla and olfactory system, however there were no significant post-hoc differences. There was also two treatment by sex interactions such that OXT males had weaker connectivity than OXT females across the thalamus (p < 0.002) and stronger connectivity than OXT females across the cortex (p < 0.001).
We next examined three indices from analyses based in graph theory: betweenness, closeness, and degree (see above for definitions of each). In terms of betweenness, male sex and OXT treatment both increased betweenness in the basal ganglia (Figure 6, p < 0.01 for both effects). In terms of closeness, we observed similar albeit more widespread effects, with male sex and OXT treatment and both increasing closeness across several regional clusters (Figure 7). Lastly, similar effects were found in terms of degree, with OXT males having greater degree across a wide swath of regional clusters (Figure 8).
DISCUSSION
Here we describe how, in prairie voles, exposure to exogenous OXT at birth can impact neurodevelopment in ways that impact neural anatomy and functioning into adulthood. Overall, anatomical differences, while widespread, were generally small, whereas differences in functional connectivity, particularly among OXT-exposed males, were larger. Anatomically, OXT at birth led to a slight reduction in amygdalar volume and OXT males in particular had slightly smaller cortices and slightly larger brainstem / cerebellums (Tables 1-3). OXT at birth led to males resembling females in terms of FA and ADC (Figure 3). However, functionally, OXT males showed marked differences from all other groups. OXT at birth led males to display robustly increased functional connectivity throughout the brain (Figures 4-8). This was particularly the case across the cortex in males in terms of the strength, closeness, and degree of connectivity (Figures 5, 7, and 8). Both the number (Figure 4A) and strength (Figure 4B) of positively correlated connections were greater in males and OXT-exposed animals. This was true both within cluster and, to a lesser extent, between clusters as well (Figure 4C). When these effects were examined regionally, few regions stood out (Figure 5); thus, we view these effects as reflecting broad, brain-wide differences. In terms of graph theoretical analyses of functional connectivity: the basal ganglia stood out for both sex and OXT increasing betweenness, whereas such differences were more widespread for closeness and degree. What does this notably broad increase in functional connectivity mean for the OXT males? We have at present only a few hints.
Interestingly, the observed effects of OXT at birth extended beyond brain regions with dense expression of the OXT receptor. The robust and widespread changes in OXT males’ neural physiology (i.e. functional connectivity) were greater than the changes observed in neuroanatomy (i.e. VBM and DWI). If OXT males are continuously experiencing high levels of communication between brain regions, we would expect that to eventually produce changes in anatomical connectedness. Either our anatomical measures of connectivity were insufficiently sensitive to detect these changes, or the relatively small changes in anatomy we did detect are sufficient to produce functional changes that are comparatively more robust. The functional connectivity scans were undertaken with subjects lightly anesthetized, so it is unlikely that OXT males were responding differentially to the conditions of scanning. Furthermore, we have observed no evidence of stress reactivity being affected by perinatal OXT in our previous studies. OXT acts as a pleiotropic hormone around the time of birth to coordinate the transition from fetal to neonatal life (1), so by affecting neurodevelopmental trajectory, a single OXT exposure could produce such widespread differences across the brain.
We highlight one set of findings in particular because of their relevance to our previous work. As adults, males exposed to OXT via maternal administration at birth had denser OXT receptor distributed along the extent of the agranular insular cortex (14). In the present study, OXT treatment at birth led to the agranular insular cortex registering as a smaller volume in the brains of adult males (Table 3), along with a slight increase in FA (Table 5) and decrease in ADC (Table 8). Moreover, OXT males’ agranular insular cortex had a greater number of functionally connected regions (35 vs. 19 for Control males) and stronger average connectivity among those regions (z-score 3.45 vs. 2.98 for Control males). Changes in the functioning of the agranular insular cortex correspond to the behavioral effects previously observed, such as increased alloparental caregiving (14).
One pattern generally observed in the brains of humans with autism spectrum disorders is diminished long-distance functional connectivity and increased short-distance functional connectivity (27,28). We observed that both male sex and OXT treatment at birth led to increased intra-cluster connectivity (analogous to increased short-distance hyperconnectivity), but also to increased inter-cluster connectivity, though to a lesser extent. Thus, the present results do not entirely resemble the equivalent effects seen in humans with autism spectrum disorders. While some epidemiological studies have suggested a link between autism spectrum disorders and perinatal oxytocin exposure (11), residual confounding by either genetic or environmental vulnerabilities, could explain this apparent association. Indeed, our previous work, we observed a broadly gregarious phenotype in OXT-exposed voles (14); if such results translated to humans, they would support the contention that underlying vulnerabilities bring on both a need for OXT in the mother and susceptibility to autism spectrum disorders in the child. While our previous work found behavioral differences in both males and females exposed to OXT, which is somewhat in contrast to the present study’s findings, we also found males to be much more affected by perinatal OXT in terms of neuroanatomy, which was assessed as the density of OXT and AVP receptors -and that finding matches those of the present study. Numerous previous studies have found sex-, dose-, and region-dependent effects of early life oxytocin manipulation (9,29).
There have been very few studies on the impact of birth interventions and subsequent brain development (1,3). In the realm of neuroanatomy, Deoni and colleagues recently reported that CS results in smaller brain volumes in neonatal mouse pups (30). However, no such findings have been observed in human children or infants (31). In terms of VBM and functional connectivity in the present study, OXT led males toward a more masculinized phenotype. In terms of ADC and FA, however, OXT led males toward a more feminized phenotype. Further work is needed to reveal the meanings of these differences.
This study is not without limitations. Firstly, because the Control group did not receive a vehicle treatment, the effects of injection were not adequately controlled for, which introduced a stress confound of indeterminate magnitude. In a small validation study, adult offspring of saline-treated dams (n = 3 female, 5 male adult offspring) were not found to have meaningful differences in DWI values compared to the un-treated Control animals of the present study (n = 17 female, 19 male adult offspring).
Collectively, these results support the contention that the perinatal brain is sensitive to OXT administered indirectly to the pregnant female. As obstetric care continues to use OXT for labor induction / augmentation in the majority of births in the U.S. (4,5), a more complete understanding of the neurodevelopmental effects of OXT exposure at delivery is imperative. Furthermore, as delayed cord clamping becomes standard practice (32), this will extend both the duration and dose of potential OXT exposure by the neonate, since the vast majority of deliveries now use OXT during the third stage of labor to prevent postpartum hemorrhage (33,34). Because the perinatal period is a sensitive period for brain development in terms of OXT exposure, these practices deserve further investigation.
Supplementary Figure Legends
Table S1. The list of brain regions and their corresponding regional cluster to which they were classified.
Figure S1. The changes in brain volumes between sexes are not consistent, i.e. in some brain areas females are greater (black sign) and others less than males (red sign). The location of these different brain areas affected by OT at birth when comparing females and males can be seen in the probability heat maps in Fig 1. Brain regions that are greater in volume in females than males are organized around the olfactory system (A. glomerular layer; B. anterior olfactory n; B-D. piriform cortices) and prefrontal/limbic cortical areas (B. motor, insular; C. anterior cingulate; D. retrosplenial). Areas that are larger in males vs females are primarily associated with brainstem/cerebellum (E. reticulotegmental n., raphe, pontine reticular n; F. olivary n.; G. vestibular n., gigantocellularis, paraflocculus, F. cuneate n. 7th lobule, medullary reticular n.).
Footnotes
The manuscript text has been edited to include more Discussion as well as a revised Results section to reflect the inclusion of new Figures.