Oxytocin gene networks in the human brain: A gene expression and large-scale fMRI meta-analysis study

Oxytocin is a neuropeptide involved in animal and human reproductive and social behavior, with implications for a range of psychiatric disorders. However, the therapeutic potential of oxytocin in mental health care suggested by animal research has not been successfully translated into clinical practice. This may be partly due to a poor understanding of the expression and distribution of the oxytocin signaling pathway in the human brain, and its complex interactions with other biological systems. Among the genes involved in the oxytocin signaling pathway, three genes have been frequently implicated in human social behavior: OXT (structural gene for oxytocin), OXTR (oxytocin receptor), and CD38 (central oxytocin secretion). We characterized the distribution of OXT, OXTR, and CD38 mRNA across the brain, identified putative gene pathway interactions by comparing gene expression patterns across 20737 genes, and assessed associations between gene expression patterns and mental states via large-scale fMRI metaanalysis. In line with the animal literature, expression of the three selected oxytocin pathway genes was increased in central, temporal, and olfactory regions. Across the brain, there was high co-expression with several dopaminergic and muscarinic acetylcholine genes, reflecting an anatomical basis for critical gene pathway interactions. Finally, fMRI meta-analysis revealed that the oxytocin pathway gene maps correspond with motivation and emotion processing.


Introduction
Oxytocin is an evolutionarily conserved neuropeptide implicated in an array of social and reproductive behaviors, and its role in complex behavioral traits and in the pathophysiology of mental health conditions has attracted considerable attention (1). Oxytocin is mostly synthesized in neurons in the supraoptic nucleus and the paraventricular nucleus, and released both systemically and centrally. Human research has shown beneficial effects of intranasal oxytocin on performance on tests assessing social cognition (2), gaze to the eye region (3), and the retrieval of social cues (4). Moreover, single nucleotide polymorphisms in oxytocin pathway genes have been associated with social behavior and psychiatric disorders (5). Emerging evidence also points to the oxytocin system's role in energy metabolism (6), with relevance for metabolic functions (7).
While central oxytocin receptor (OXTR) mRNA localization in the rodent brain is well-described (8), its anatomical distribution across the human brain is poorly understood, as investigations have tended to sample very few brain regions (9, 10). Partly attributed to this limited understanding of human central OXTR gene expression (11), the translational promise of oxytocin provided by preliminary clinical trials (e.g., 12, 13) has yet to be fulfilled. The distribution of OXTR mRNA across the brain provides a proxy for the distribution of central oxytocin binding (14, 15), allowing for a detailed mapping of the anatomical geography of the oxytocin system in the human brain. Seminal animal work using histochemistry and immunohistochemistry revealed high concentrations of OXTR mRNA in the hypothalamus, amygdala, olfactory bulb, ventral pallidum, and the dorsal vagal nucleus (16, 17). Further, experimentally increasing (18) or decreasing (19) OXTR expression in the prairie vole nucleus accumbens modulated partner preference behavior, suggesting an intriguing correspondence between the spatial distribution of OXTR mRNA, its functional neuroanatomy, and behavioral relevance.
In addition to the anatomical distribution of OXTR mRNA, characterizing its interactions with other key elements of the oxytocin signaling pathway and biological systems beyond this pathway is critical for determining oxytocin's behavioral and functional relevance. Along with OXTR, CD38 and oxytocin-neurophysin I (OXT) genes in the oxytocin signaling pathway have been implicated in human social behavior (5). Specifically, CD38 is involved in central oxytocin secretion (20), and OXT encodes the oxytocin prepropeptide containing the nonapeptide oxytocin and the carrier protein neurophysin-I (21). OXT mRNA is highly expressed in human paraventricular nucleus of the hypothalamus, the lateral hypothalamic area, and the supraoptic nucleus, and there is evidence of co-expression with OXTR mRNA and the μ and κ types opioid receptor mRNA (22), providing a putative avenue for interactions between the oxytocin and opioid pathways .
The interactions with the oxytocin system extend beyond the opioid pathway, including the dopaminergic (23, 24) and muscarinic acetylcholine (9, 25) circuits, with possible implications for social behavior and psychiatric disorders. For instance, the dopamine D2-receptor subtype (DRD2), has been implicated in various putative intermediate phenotypes in psychiatric illness, including motivational processing (26) and pair bonding in animal models (27, 28). Moreover, the muscarinic acetylcholine M4 receptor (CHRM4) has been associated with schizophrenia (29) and implicated in cognitive flexibility (30), social impairment (31), and dopamine release (32). Finally, oxytocin can also bind to AVPR1A receptors (33), which has also been linked to social functioning (34). However, the mRNA co-expression of these systems, and potentially others, with the oxytocin system is not well characterized. While brain regions with high oxytocin pathway gene expression in humans have been identified (9, 10, 22), inference of specific mental states from single brain regions is elusive. For instance, commonly observed increases in medial and lateral frontal region activity during emotion and pain processing seem to be better explained by more general sustained attention processes (35). Leveraging data from more than 11,000 fMRI studies, NeuroSynth (35) allows for reverse inference of mental states based on a given brain gene expression map with high specificity. Establishing the specific mental state correlates of oxytocin pathway genes will provide a deeper understanding of the central human oxytocin system and its relevance for brain functions and mental health.
By leveraging the Allen Human Brain Atlas, which offers a uniquely comprehensive gene expression survey from six neurotypical adult human brains, we first characterize the anatomical distribution of mRNA expression of OXT, OXTR and CD38, which are key constituents in the oxytocin pathway.
Second, we explore putative gene interactions by identifying mRNA maps with overlapping anatomical distributions with our target genes across all protein coding genes in the database, additionally focusing on selected dopamine, muscarinic acetylcholine, and vasopressin gene sets. Third, we decode the mental state relevance of the selected oxytocin genes using quantitative CRHM5) (9, 25), and a vasopressin set (AVPR1A, AVPR1B) (34,37 For out-of-sample validation, central oxytocin pathway median expression profiles for 10 distinct brain regions were extracted from the Genotype-Tissue Expression (GTEx) project database (39). Although this dataset offers less spatial precision, the data is derived from a larger dataset of donors (mean sample size for mRNA expression across brain regions = 131.7, range = 88 to 173). For this comparison, median gene expression for these 10 distinct GTEx regions were calculated using Allen dataset (Fig. S2).
Spearman's rank correlation coefficients (r s ) were calculated to assess the relationship between oxytocin pathway expression profiles from the two databases. Sex differences in central oxytocin pathway gene expression were also examined in the GTEx sample. Data and analysis script availability. mRNA expression data is available from Allen Human Brain Atlas (http://human.brain-map.org) and GTEx (http://gtexportal.org). The Matlab script for producing the brain region-specific data, the resulting dataset, and the R script used for statistical analysis is available at https://osf.io/jp6zs/. imaging maps with motivation-related topics (e.g., "reward", "anticipation", "incentive"), emotion-processing, and aversive-related topics (Fig. 5a, Table   S3). Figure 5b shows the full distribution of correlation coefficients for each mental state term across all 20737 gene maps, with labelled oxytocin pathway genes and their absolute rank for each term (Table S3). Notably, the OXTR map was ranked among the top 100 out of 20737 genes for several functional imaging maps (i.e., the top 0.5%).

Discussion
The anatomical distribution of gene expression in the brain is heterogeneous and highly coordinated (41). Whereas dynamic alterations in gene expression are essential in response to environmental demands, and are critically involved in a range of cognitive functions, learning, and diseases (42, 43), the basic organization likely partly reflects the evolutionary conserved modular layout of the brain (44). Indeed, gene-gene co-expression patterns form specific genetic signatures in the brain, representing distinct tissues and biological systems (41), and likely reflect the potential of complex and differential gene-gene interactions with implications for brain disorders and mental health. Here, we leveraged the unique human brain mRNA expression library from Allen Brain Atlas to show that mRNA reflecting specific genes in the oxytocin pathway are highly expressed in central and temporal brain structures, along with the olfactory region. We also show reduced expression of OXTR and CD38 in the cerebellum, consistent with prior animal research (45). Importantly, the observed oxytocin pathway expression patterns from the Allen database, particularly OXTR and CD38, were consistent with oxytocin pathway expression patterns observed in the GTEx database.
Oxytocin pathway genes showed considerable co-expression with DRD2, COMT, DAT1, and CHRM4 genes, providing evidence for putative interactions between dopaminergic and muscarinic acetylcholine systems with the central oxytocin pathways. Exploratory analysis between oxytocin pathway mRNA and 20737 mRNA probes revealed several relationships worth noting in the context of metabolic and feeding regulation, as well as psychiatric disorders. We discovered that OXTR is highly co-expressed with Neurotensin receptor gene (NTSR2), which has been found to regulate ethanol consumption in mice (46), and Glutamate dehydrogenase 1 and 2 (GLUD1, GLUD2), which are involved in energy homeostasis and insulin secretion (47-49). Second, CD38 is highly co-expressed with NTSR2, GLUD1, and SPX (Spexin), with the latter associated with weight regulation (50).
These results are consistent with emerging evidence that the oxytocin system may play a role in the metabolic and feeding dysregulation often observed in severe mental illnesses (7). In regards to psychiatric disorders, there were also strong negative correlations between oxytocin pathway genes and CADPS2, which has been associated with autism (51)  It has been suggested that comprehensive clinical trials need to demonstrate engagement of drug targets, such as OXTR occupancy reflected by regional brain activity changes (11). Without precise targets, it is unclear whether non-significant effects of intranasal oxytocin, beyond insufficient statistical power (74), are due to an inefficacious drug or misidentified drug targets. By identifying accurate oxytocin pathway targets in the human brain, our study provides a tentative oxytocin treatment target map that may help facilitate efforts to better understand oxytocin treatment efficacy (11, 72).
Analysis of data from the Allen Brain Atlas database provides a unique opportunity to map the central oxytocin system and explore co-expression with other systems. Altogether, these results provide a proof-of-principle demonstration of corresponding cognitive and gene expression patterns of a neuropeptide pathway involved in complex human behaviors.    CD38 (b). Each point represents mean expression with standard errors for a given brain region.
Data was collected from the Allen Human Brain Atlas, representing mean expression across six donors, and then voxel-by-voxel volumetric expression maps were created (Fig. S1). Each brain was registered the Montreal Neurological Institute template using Advanced Normalization tools. Statistics for each region were extracted using based on the Automated Anatomical Label atlas. The bolded dashed lines represent mean expression across the all regions with 1 and 2 standard deviations also shown. One sample t-tests were performed to assess which regions expressed mRNA to a significantly greater degree compared to average expression across the brain. Analyses suggest that compared to average brain expression, there is increased expression of OXTR and CD38 in central and temporal brain structures, along with the olfactory region. Lower than average expression is observed in the cerebellum. Central OXT expression patterns are shown in Figure S4. * p < 0.05 (FDR corrected for 54 tests).    The absolute ranking for each oxytocin pathway gene out of 20737 correlations are shown (also see Table S3).