Thread by @Datta_Lab

Olfactory sensory neurons (OSNs) use transcription to make flexible predictions about the environment and dynamically change odor responses, now in our new paper led by the amazing @TatsuyaTsuka0 and @davidhbrann! Thread about what we found 👉1/n bit.ly/3pGvY6u

Mammalian OSNs are thought to more or less faithfully tell the brain about what they are detecting via their odor receptors (OR), suggesting that the brain is largely responsible for sorting novel, salient cues from predictable background signals 2/n

In this view OSNs are empty vessels for OR expression, whose response properties solely arise from the single receptor (of ~1000) they express. This is (in part) why folks use heterologous expression or empty neuron-type systems to characterize OR responses to odor 3/n

On a lark we single cell sequenced a bunch of mature mouse OSNs, pretty much expecting the transcriptomes of all OSNs to be more or less the same regardless of what OR was expressed (as development is over). SUPER WRONG!!! 4/n

scSeq of ~770,000 OSNs (across all conditions) revealed that each of the ~1000 OR-defined OSN subtypes harbors a distinct transcriptome that is largely determined by the expressed OR – an *outrageous* amount of transcriptional precision here. 5/n

The main genes that vary across OSN subtypes? A set of ~70 genes that transform odors into spikes. All of your favorite signaling and intrinsic property genes are on this list – K, Ca, Na channels, G protein subunits, cyclic nucleotide gated channels, phosphodiesterases, etc. 6/n

Expression of these genes is systematically organized, and varies in an analog fashion across OSN subtypes – it is like there is a dial being turned across all OSNs that coordinates expression of these ~70 “functional” genes 7/n

Switching environments, optogenetically tweaking OSNs, blocking odors via nares occlusion – all these manipulations systematically moved this rheostat-like transcriptional program in one direction or another, depending upon the chronic activity state of each OSN 8/n

Turns out, OSNs that are more engaged in a given environment (integrated over hours) express higher levels of genes that attenuate odor responses and lower levels of those that boost responses. The opposite is true for less-engaged OSNs. 9/n

So what do these genes actually do to OSN function? We used Act-Seq to measure acute OSN responses in vivo, which shows that expression levels of the ~70 functional genes predict the amplitudes of acute odor responses far better than in vitro defined receptor EC50s! 10/n

Acute in vivo odor responses are weak in environments where a given OSN subtype is highly engaged by ambient odors (as determined by assessing expression of those ~70 genes), but responses to the same odor in the same subtype are amplified when it was less engaged. 11/n

When we switched mice from one environment to another, OSNs were acutely activated, and the degree of activation predicted changes over hours in functional gene expression…which predicted acute odor responses by that OSN in the future. The whole system is a closed loop! 12/n

Finally, all this prolly matters to the brain, as in vivo calcium imaging of OSN axons in the olf. bulb showed that functional responses to monomolecular odors are determined by prior odor exposure history, mirroring the observed transcriptional changes. 13/n

Barlow argued that adaptation is a primitive form of sensory predictive coding – we show that OSNs systematically use gene expression to reflect the past, thereby filtering out the expected to emphasize the new. 14/n

Lots to take away from this, but one real surprise is that essentially the entire OSN array is differentially engaged by air+odor in a typical environment, and so all OSNs are continually sculpting their acute odor responses in response to this barrage of ambient signal 15/n

This is very different from the view that OSNs are generally “off” and only sparsely recruited by environmental odors; instead OSNs are continuously contending with dense ambient cues, and using gene expression to recenter themselves into a reasonable dynamic range 16/n

This also means that peripheral codes for the same odor (and for odor relationships) are constantly changing – more than 50 percent of OSNs change their response amplitudes to a pure odor if we just change the type of cage in which the mouse is housed 16/n

Intriguing here to see proportional and continuous variation in gene expression, chronic activity and acute odor responses, which feels conceptually related to firing rate homeostasis — much of our thinking was shaped by the amazing work from folks in that field 17/n

Findings also may be relevant to large scale brain scSeq datasets from #BICCN, etc — some of the variation in gene expression *within* a given cell type seen there may reflect systematic transcriptional responses to activity, as we see here 18/n

(n.b., this cell-autonomous txn mech. is distinct from fast post-translational or pre-synaptic adaptation mechanisms previously described, possibly reflecting a different function for this slower form of adaptation) 19/n

(hat tips to @betenoire1 and Nicole L’Etoile who saw similar mechs. in worms, and David Coppola, Tim McClintock and @McGannLab, who had seen hints of this in bulk-seq/occlusion data) 20/n

Important note - work would not have been possible without exceptional efforts from @st4nislav, @guitchounts, and amazing collab with @Neuroboz. 21/n

Finally, if you are interested in the intersection of molecular neuroscience and systems-level function, we are looking for a postdoc with molecular biology experience to address related questions in the brain – please reach out, contact info at dattalab.org 22/22