SLC or OR gene superfamilies: exploration for MorphMap paper (ORFs, need to check CRISPR)

[AnneCarpenter]

Anna Greka - Anne emailed (who works on ion channels) to see if she knows someone to talk with - Dec 15 https://mail.google.com/mail/u/0/#sent/KtbxLzfphpSXKvqjLksXdBkZFbLTKPBNnB

Eitan Hoch - We could contact (working on a single SLC gene): https://mail.google.com/mail/u/0/popout?ver=1uc31lzgyq4bq&q=slc%20genes&qs=true&qid=7AD9A531-CA56-4FEA-BCA3-F3E12F0B2C4C&aqid=0EFF949C-A964-4282-B723-90BA8849E4E6&search=query&cmembership=1&silk=E6C0CD25-2274-41A5-AD3C-A6E239256E92&th=%23thread-f%3A1562673576686256012&qt=city.1.gene.1.genes.1.genetic.1.genetics.1.genome.1.lake.1.salt.1.slc.1&cvid=1

Found one paper: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3082935/

Could email the JUMP cell painting email list

[AnneCarpenter]

Anna Greka suggested talking to Jen Pluznick at Hopkins, emailed her today. Also emailed the JUMP Cell Painting email list.

[AnneCarpenter]

This really is a dream team we assembled in very quick time! I'm copying below some emailed insights from our OR experts below that indicate it IS worth pursuing this to see where it goes.

That being the case, I've taken a stab to assign tasks mainly to Niranj and Evotec, and they can report on results in github and in this email thread. Once those groundwork tasks are done, we will feel comfortable investing the SLC/OR experts' time to dig into the biological relationship.

At that time, I will ask this group for volunteers (from among yourselves and/or subordinates) who want to shift from consulting to really contributing... in other words, as a group, volunteers will conspire to take full ownership of digging in and producing a few paragraphs of a story and figures for the paper, using Niranj and Evotec as resources to answer queries but otherwise driving what analyses (and possibly wet lab experiments) make sense.

I hope you all have a lovely and real break from your work over the holidays. All the best!!

Anne

On Wed, Dec 20, 2023 at 4:22 AM Bolek Zapiec <[email protected]> wrote:
Hi Anne,

Thank you for bringing this fascinating discovery to our attention. I can certainly echo Jen in saying that this is indeed of potential interest and even if the underlying reason for the phenotypic convergence of the OR and SLCs (along with some of the other ORFs) turns out to be due to the overexpression of membrane proteins resulting in ER-stress, it’s still worth looking into on two levels I can think of.

From the Olfaction point of view, heterologous expression of ORs has been a challenge in de-orphaning the vast majority of mammalian ORs, and has limited their utility in commercial applications in the Smell & Taste industry (e.g. Matt Roger’s group in Firmenich). Hiro Matsunami has been leading the charge for a while now in understanding the ways ORs can be made to work in cell lines, but generalizing the challenge to membrane proteins and developing a standardized assay (like Cell painting) for profiling this phenotype may offer a key to unlocking this box and e.g. with the help of combination screening understanding what components (chaperones, ORF UTRs, etc) allow ORs to be expressed & trafficked without backing up in the ER & Golgi.

From the Cell Painting perspective, if it turns out that this is a convergent phenotype, then it’s again a beautiful case example for how Cell Painting is able to detect convergent pathways and the inner-life of a cell. In this case, if it turns out that folding/trafficking of challenging membrane proteins is resulting in this convergent phenotype, it provides a training set for a robust classifier of the protein misfolding & ER-/ Golgi-stress response and can be readily tested by comparing the Cell Painting fingerprint with that of another protein known to trigger this response. I personally would be interested in having such a classifier for profiling the intracellular activity of our compounds.

It might not be the kind of surprising discovery that might get the cover of CNS, like genetic co-regulation of ORs and SLCs might have, but I wouldn’t dismiss it as an experimental artifact alone as this new perspective might shed new light on this fundamental process that is of value to both science & industry.

Since the data/findings were shared in confidence I certainly won’t share this further at this point, but I’d glad to participate in any further meetings/email discussions about identifying what’s the underlying cause behind these data.

Best,

Bolek

Dr. Bolek Zapiec

Associate Director

Bioassay Automation & Compound Logistics
Discovery Pharmacology
Discovery & Development Technologies

signature_4053395742

Merck Healthcare KGaA | Frankfurter Straße 250 | 64293 Darmstadt | Germany

From: Jennifer Pluznick <[email protected]>
Date: Tuesday, 19. December 2023 at 22:34
To: Anne Carpenter <[email protected]>, Bolek Zapiec <[email protected]>
Subject: RE: SLC/OR families

Hi Anne,
I can’t speak to the SLCs, of course, but some thoughts on the ORs (just copying you and my OR colleague Bolek so as not to spam everyone):

…it could be interesting (at least to me!) if some ORs do this and others don’t, especially if you can uncover how those two groups differ. If these are two novel groups of ORs, that could be quite interesting.

However, there may be a hint in the literature: as was said, when ORs are exogenously expressed without ‘help’ (without special N-terminal tags or co-expression of chaperones), they get stuck in the ER, and are thought to induce an unfolded protein response. Hiro Matsunami (Duke) has shown that a subset of ORs are better behaved (https://elifesciences.org/articles/21895); he refers to these as “overrepresented ORs” (oORs) because they are well-expressed even when key chaperones are knocked out. I wonder if the reason you got this signal from some ORs and not others is that the oORs in your study did not cause as much ER stress?

Supplementary table 2 from the paper above lists oORs (and the counterparts, uORs). In theory, you could use these lists to see if this tracks with which ORs gave you the signal vs not…..however: the Matsunami study was done in mice and the gene names are completely different….in some/many cases the orthologs aren’t known with certainty between mice and humans, as there are 1000 murine ORs and 400 human ORs, and they are all similar to one other. Fwiw, the 3 well-conserved ORs that I was looking for and didn’t see on your list (OR6B1, OR51E1, OR51E2 in humans, Olfr78, Olfr558 and Olfr449 in mice) are known to be oORs.

Best

Jen

From: Anne Carpenter <[email protected]>
Sent: Tuesday, December 19, 2023 4:06 PM
To: Alexander Ehrmann <[email protected]>; Andrey Zinovyev <[email protected]>; Arne Monsees <[email protected]>; Bolek Zapiec <[email protected]>; Niranj Chandrasekaran <[email protected]>; Fauman, Eric B <[email protected]>; Helena Batoulis <[email protected]>; Holger Hennig <[email protected]>; Jennifer Pluznick <[email protected]>; Rahul Valiya Veettil <[email protected]>; Tim James <[email protected]>; William Charles <[email protected]>
Subject: SLC/OR families

Thanks again for your precious time helping us figure this out! Below are the many notes/questions and ideas for analysis or experiments for next steps. We will prioritize and keep this group posted.

If I understand correctly and the ER stress hypothesis is true, then this is a not-so-exciting result in the sense that it's already known that overexpressing membrane proteins yields a (to some degree artifactual) impact on cells? Or might it still be the case that the fact that a particular subset do this and others don't tell us something interesting?

Anne

[AnneCarpenter]

For ER stress, worth noting this paper: Lebedeva et al Cell Death and Disease (2021)12:770 ; https://doi.org/10.1038/s41419-021-04045-4

They made isogenic human cellular models (in hTERT RPE1 cells) of Congenital disorders of glycosylation (CDG) and deglycosylation (CDDG), which have ER stress as an underlying biological mechanism. The genes involved are PMM2 and DPAGT1 and NGLY1. Among 1049 drugs screened, they found some that reverse the morphological impact using Cell Painting (minus SYTO14 and WGA - they included only mitochondria, the actin cytoskeleton, endoplasmic reticulum, and nuclei). 58 compounds in the primary screen. They chose those that worked across all cell lines for deeper followup.

On the cell lines themselves before the drug screen, they did some validation studies demonstrating ER stress: see section starting "In order to establish the ER stress profiles of the cellular models" involving staining for particular markers.

"All mutant cell lines exhibited a flat, extended morphology (Fig. 3C) reminiscent of cellular senescence that was not seen in the isogenic parental line" and slower proliferation.

"A majority of the active compounds in our screen, #s 3, 4, 6, 9, 15, 25, are reported to affect microtubules"... "through modulation of JAK/STAT and growth factor signaling among others"

(this task is waiting for Evotec/Niranj for next steps)

[AnneCarpenter]

I want to be sure the steps above are still on Evotec's radar (@auranic ?)

[auranic]

@AnneCarpenter I believe we successfully completed all these steps. I apologize if the results were overlooked in the email flow. Let me summarize them here and compile the information into a PowerPoint presentation. This way, if necessary, we can include some slides in the Supplementary materials. Please allow me a couple of days for this.

[AnneCarpenter]

Great, thank you!

[niranjchandrasekaran]

Notebook

Many SLC-OR connections are still present in the new ORF data. There aren't any OR genes in CRISPR. The top left cluster is the one with both SLC and OR connections. All the other clusters are SLC or OR exclusive.

ORF

CRISPR

[AnneCarpenter]

We will very likely include this story in the paper, so we want to be sure we define the genes included in the plots based on the latest data (vs selecting the genes from a prior version). Have we done this already? If so then I think the next step is to follow the notes from my comment above: #6 (comment)

[niranjchandrasekaran]

Notebook

I checked the images on Alán's tool: broad.io/orf and I didn't find any artifacts. Here are all he SLC and OR genes in the dataset: https://lite.datasette.io/?metadata=https://raw.githubusercontent.com/broadinstitute/monorepo/main/libs/jump_rr/metadata/orf_matches.json&install=datasette-json-html&parquet=https://zenodo.org/api/records/12766577/files/orf.parquet/content#/data/content?_filter_column_1=Gene%2FCompound&_filter_op_1=startswith&_filter_value_1=SLC&_filter_column=Match&_filter_op=startswith&_filter_value=OR&_sort=rowid

To check with features contribute to the similarity between SLC and OR genes, I ran the same analysis I did for MYT1-RNF41 (#3 (comment)) to generate correlation between feature groups within each compartment and imaging channel. Since we are not comparing just a pair of features but one cluster to another, I report the median similarity across all pairs.

Here are the results:

These results suggest that there are strong similarities between the two clusters, not only in the ER channel. Maybe that can rule out the ER stress hypothesis? @AnneCarpenter

[AnneCarpenter]

I don't think we expected the ER channel to dominate the features just because we have stressed the ER, if that is what you are asking. I think the other queries and followups are still worth pursuing.

[niranjchandrasekaran]

Notebook

It looks like a bunch of OR genes are all on the same plate. But they are at least not from the same row or column. We may want to mention this in the SLC-OR section of the paper. But many other SLC and OR genes are not on the same plate and are still strongly correlated. So I am not overly concerned about this.

[AnneCarpenter]

See our draft paragraph here in the paper: https://docs.google.com/document/d/160QYCeJXMJMPgvYcirkkKaY99r6dZq6FMqYGE61CfFc/edit#bookmark=id.l2u8zkksn0ba

[dselinger-plex]

Adding a test comment here.

[niranjchandrasekaran]

Included in the morphmap paper