Thread by @dsquintana

Very happy to announce that our new paper on oxytocin pathway gene networks in the human brain has just been published in @NatureComms nature.com/articles/s4146…

Here’s the story behind this study, how we did it, and what we found...

[THREAD]

Oxytocin-like signaling is an ancient system that goes as far back down the tree as earthworms & sea squirts ncbi.nlm.nih.gov/pubmed/26392129 Oxytocin production occurs similarly across mammals but the *location* of oxytocin receptors in the brain can vary, even in closely related species

Take the prairie and meadow vole: closely related species but different patterns of oxytocin receptor expression in the brain, which is thought to reflect the social organization of these species ncbi.nlm.nih.gov/pubmed/17118932

Experimentally increasing oxytocin receptor expression modulates social behavior in prairie voles (Keebaugh & @lyoun03), providing evidence for a link between expression patterns and behavior ncbi.nlm.nih.gov/pmc/articles/P…

But what about expression patterns in humans? Published studies had only investigated expression in a few brain regions ncbi.nlm.nih.gov/pubmed/1657300 but ideally we want whole brain expression—what if you miss a high expression region?

Without access to post-mortem brains, I couldn’t do this myself. But then one day in our lab meeting, @jarekrokicki presented the @AllenInstitute human brain atlas and I saw an opportunity.

Here’s the @SFFNORMENT multimodal imaging lab, including some of the study co-authors

Here’s how @jarekrokicki, co-first author of our study, analysed our data from the raw Allen gene expression dataset

Many genes are involved in oxytocin signaling, but we focused on 3: oxytocin receptor (OXTR), CD38 (oxytocin secretion) & OXT (structural gene for oxytocin). We then extracted data for 54 brain areas (AAL atlas), finding ↑ expression in subcoritical, temporal & olfactory areas

We also took a closer look at expression patterns in the hypothalamus and thalamus. Check out the high expression of OXT in the supraoptic (SON) and paraventricular (Pa) nuclei [panels c & f] — precisely what you would expect based on rodent studies

We also looked at the association between oxytocin pathway genes and dopaminergic, muscarinic acetylcholine, vasopressin, and opioid gene sets [a].

The OXTR/CD38 gene cluster (central black square in matrix) was highly expressed in the brain compared to the rest of the body [d]

We then applied weighted gene co-expression analysis (WGCNA) on all 94 genes in the full oxytocin signaling pathway, identifying a co-expression module of 28 genes with OXTR & CD38. This module was highly enriched in waist-hip ratio GWAS & feeding behavior and cognition genesets

We also looked at the correlation between oxytocin pathway gene expression in the brain against ALL 20,737 protein coding genes. Several of the top 10 correlated genes are involved in metabolic regulation (NTSR2, GLUD1, GLUD2, Spexin)

While the Allen dataset is spatially rich (samples were taken from from 363+ brain regions), these data are derived from only six donors. So we cross-checked these patterns against the GTEx database (less spatial precision but 88+ donors). Things looked fairly similar

We also calculated measures of differential stability for gene expression, which calculates how similar expression patterns are between the 6 donors. OXTR and CD38 were among the top decile of 20,737 genes suggesting strong reproducibility from donor-to-donor

There’s only one female donor in the Allen sample. But the expression patterns in this donor were still strongly related to the other five. There was also not much evidence of sex differences of oxytocin pathway gene expression in the GTEx dataset across 12 brain regions

Next, we created voxel-by-voxel maps of gene expression based on the Allen data and then correlated these maps with meta-analytically derived brain activation patterns of cognitive states from NeuroSynth neurosynth.org

Neurosynth is an amazing resource based on a corpus of 14000+ fMRI studies. Not only can you meta-analytically generate cognitive state maps for specific states (e.g., fear) but you can also “decode” cognitive states, given a brain map—reverse inference nature.com/articles/nmeth…

We correlated our expression maps with cognitive state maps and identified the 15 strongest relationships with OXTR, CD38, & OXT maps. Identified cognitive states can be broadly categorized as anticipatory, appetitive, and aversive

The strongest correlations with OXTR/CD38 are interesting, but what if thousands of other genes had *even stronger* correlations? We looked at this and found oxytocin pathway genes were among at least the top 2.5% of genes (out of 20,737) for most of the 15 cognitive states

OXTR and CD38 was also more highly expressed in regions associated with social vs. non-social cognitive states [c, d]

So while oxytocin is certainly involved in social cognition, it seems to be associated with several other anticipatory, appetitive, and aversive cognitive states in addition.

The strong correlation with the “taste” cognitive state, enrichment in waist-hip GWAS + feeding behavior genesets, & high correlation with “metabolic” genes also suggests that oxytocin plays a role in energy regulation. Early work on intranasal oxytocin & food intake is promising

In 2016, Thomas Insel, who was the Director of the NIMH, wrote a commentary on guiding the translation of oxytocin neuroscience to the clinic. The whole article is great, but one section really stuck with me since I first read the paper

We now have a map of oxytocin receptors in the human brain, which genes are co-expressed with oxytocin genes, and associated psychological states.

HUGE thanks to all co-authors, especially @jarekrokicki & @larswestlye

Another thing that was really interesting in the context of oxytocin trials was that we found evidence for high oxytocin receptor expression in the olfactory region, which is where intranasally administered oxytocin is thought to enter the brain nature.com/articles/mp201…