Thread by @SizunJ

Hi #ScienceTwitter fam, delighted to share some new exciting work on #HIV. Where are the cellular reservoirs of HIV? Why are they so hard to eradicate? What are some of the determinants for latency vs transcription in infected cells? Thread alert🔔! biorxiv.org/content/10.110…

A little primer of #HIV and the #AIDS pandemic. Yes, its the OTHER ongoing viral pandemic! Some numbers from @WHO here. >38 million people are living with HIV. Although anti-retroviral therapeutics are great for helping them led a normal-ish life, eradication is still ongoing

Why is eradication so hard? First off, HIV-infected immune cells (such as CD4 T cells and macrophages), infected cells can be in various locations, like lymph nodes, brain or the gut. These are known as tissue reservoirs.

The virus can also stay dormant for 10s of years in the body, without any transcription activity. And the cells they infect are also part of the immune system. That makes it double hard to recognize and get rid of them!

To identify the tissue localization, phenotypes, functions and microenvironments around these infected cells, we needed to develop a method to robustly identify cells harboring viral DNA (the virus integrates into the host DNA upon infection!), viral transcripts and host proteins

We teamed up with Jake Estes and his talented postdoc Chi at @OHSUNews to develop in situ hybridization in tissues without the need for a standard protease digestion, on the MIBI from @IONpathOfficial for high dimensional tissue imaging! This was done w/SIV in a rhesus model

This avoids the standard caveat of combining DNA/RNA and protein detection, since circumventing the protease digestion allows us to preserve the protein epitopes for robust antibody staining and immunohistochemistry

So essentially, we can go from the standard 2-3 color assays to >30 color tissue imaging such as these images performed on adjacent sections of tissues 👇

Using mesmer, this new fancy tool from @NoahGreenwald, @MikeAngeloLab & @davidvanvalen, we can segment cells accurate to perform unsupervised clustering and annotation to obtain cell phenotypes

We identified 14 different cell types, essentially turning a bunch of images into these pretty phenotype maps:

We confirm the infiltration of some immune cell types in SIV-infected lymph nodes, and see a very clear depletion of CD4+ T cells (a hallmark of HIV/SIV infection)

Now that we have the phenotype of each cell, we can compute the "microenvironments" with a scalable computational methodology, Cellular Neighborhoods (CNs) developed by @_Salient and Graham Barlow in their paper with @ChrisSchuerch cell.com/cell/pdf/S0092…

We were able to identify 11 CNs in the tissue context of SIV infection. Its quite amazing to see them recapitulate the tissue architecture and even reflect the biology (eg CD8 T cell infiltration & CD4 T cell depletion in SIV-infected)

We found 2 interesting macrophage related CNs, which actually reflect the different tissue microenvironments these macrophages are in!

What is cool too are the differences in functional marker expression between SIV+(orange) and SIV-(teal), such as DC-SIGN, FoxO1, IL10 and CD169 in these CNs

By looking at the markers within the totality of CNs, we can actually differentiate between Healthy, Acute and Chronic SIV infections! 🤯

Thanks to some nice modeling by @hanimal725, we found various cell-cell interactions that diff between SIV+/-. One of them that was subtle but interesting was this B-Macrophage interaction (purple)

Further recapitulated in the CNs containing B cells or macrophages 👇

What we found was that these B cells and macrophages were making more IL10 in response to infection!

B cell IL10 levels (but not macrophages) correlated with viral RNA levels, showing some viral-sensing going on. Macrophage IL10 levels reflected the amount of M2 (immunosuppressive) phenotype switch. Cool!

This helped us build a model of how immunosuppressive microenvironments are conditioned by virus induced IL10.

Remember, we still have the information for viral integration (viral DNA) and viral transcription (viral RNA). This allows us to try to predict viral reactivation in situ. We found that using a combination of markers within and around an infected cell is best!

What does this mean? The surrounding of an infected cell matters! Cellular identity and function is (in part) determined by their neighbors! Just like you are the average of your best 5 friends

Finally, we used anchor plots (inspired by my good old epigenomics days) to stratify how the surrounding cells of each infected cells diff. Eg latent cells tend to be near blood vessels, while active ones may be triggered by some FDC mediate mechanism: pubmed.ncbi.nlm.nih.gov/18971284/

Now we can build a model for viral latency/activation in tissue reservoirs!

All these are possible thanks to the fantastic team, especially Chi, @xavi_r_c, @MikeAngeloLab, Jake and @GarryPNolan. Super grateful to funding from @gatesfoundation and @NIAIDFunding, @LLSResearch, @US_FDA and many others for believing in us

We are exciting to extend this work to other diseases and imaging platforms, feel free to reach out if you have any questions! Thanks it folks, thank you for coming to my TED talk, and have a great rest of the week!

Also a big thank you @HenriquesLab for your biorxiv latex template!!