1/ Our paper is out today in @cp-cellsystems.bsky.social. We used protein language models and deep learning to identify candidate protease allergens in the human microbiome.
www.cell.com/cell-systems...
Posts by Carrie Sokol
Why does this matter?
If microbes in or on us can produce functional allergens, this could help explain:
• Why certain microbiota associate with allergic disease
• How environmental exposure and microbial ecology intersect
• Why allergen exposure isn’t just “outside in” (8/x)
Huge thanks to the incredible collaborative team who made this possible.
This is just the beginning.
The allergenic universe may be far larger - and far more biologically integrated - than we imagined. 🧬🤖 (7/x)
#Allergy #Immunology #Microbiome #MachineLearning #AI #Neuroimmune
More broadly, this work reframes allergens not as a static list…
…but as a predictable functional category.
And once something is predictable, it becomes discoverable.
And once it’s discoverable, it becomes targetable. (6/x)
That means we now have:
✔️ A discovery pipeline for identifying protease allergens
✔️ Experimental validation of microbiota-derived allergens
✔️ Evidence that our allergen universe is likely much broader than we thought (5/x)
The model predicted novel candidate allergens from gut and oral microbes.
So we tested two of them experimentally and found that these previously undescribed hits from our candidate list (one serine and one cysteine protease) showed bona fide allergenic activity. (4/x)
We asked: “What proteins out there look like allergens?”
Including proteins produced by the human microbiota.
(3/x)
We focused on serine protease allergens — a major class of environmental triggers.
Instead of cataloging known allergens, we trained ML models to identify the features that distinguish allergenic vs non-allergenic proteases.
Not just sequence similarity — functional logic. (2/x)
What if we’ve dramatically underestimated the allergenic universe?
Our new collaborative paper asks a simple but radical question:
Can we teach AI what makes a protease an allergen — and then use it to discover new ones? (1/x)
🔗 www.sciencedirect.com/science/arti...
Thanks for the highlight @immunoah.bsky.social ! I had an amazing time sharing our new science!
👨🔬🎉Kicking off my bluesky presence by posting our lab's achievement! Our lab's first study is now out in Cell ! We show that IFN-λ induces ZBP1/CASP-8/GSDMC-dependent pyroptosis in gut epithelial cells impairing tissue repair post-colitis.#IBD #Immunology 🧵https://authors.elsevier.com/a/1k2OAL7PXqSL5
In Rio for ACMI 2024! All barrier (or mucosal, but we can fight over it) all the time!
But many questions remain! What controls the function of these cells? Is this pathway truly conserved in humans? Does this IL-3 pathway explain why some dupilumab non-responders do so well with JAK inhibitors? What questions do you have? (15/x)
So, we think that GD3 cells and their IL-3 act as a neuroimmune rheostat involved in allergen function. Ramp it up – develop allergen hyperresponsiveness. Tone it down – develop allergen ignorance. (14/x)
All of this was upstream of allergen-induced DC migration and Th2 differentiation. If you blocked this pathway at any step, you lost the initiation of the adaptive immune response to allergens. (13/x)
Il3ra signaling in neurons led to JAK2 phosphorylation, which drove the immediate itch response, and STAT5 phosphorylation, which drove the downstream transcription of Tac1 (gene for Substance P and a couple other neuropeptides). (12/x)
IL-3 acted upon Il3ra-expressing sensory neurons that were enriched in the PEP1 subset (these neurons aren’t stereotypically associated with itch, but they are required for protease-allergen induced itch and they release the neuropeptide Substance P). (11/x)
GD3 cells – but not other gd cell types – produced IL-3 under homeostatic conditions and after T cell receptor activation. This GD3-derived IL-3 was necessary for the itch response to allergens. (10/x)
www.nature.com/articles/s41...
*Now, DETCs aren’t present in humans because hominoids have a mutation in the gene that is required for their development. But, GD3s don’t require that gene (Skint1), which means that they might represent the epidermal gd T cell subset seen in humans. (9/x)
There are several gamma/delta cell types in the skin – the usual suspects in rodents are the dermal g/d cells and DETCs*, but there is another very poorly recognized epidermal g/d T cell subset that we simplistically referred to as GD3s (3rd skin g/d cell). (8/x)
To start his hunt for this putative cell, Cameron Flayer examined the immediate itch response to various protease allergens (plant, animal, fungal, and venom) and unexpectedly for us found that gamma/delta (g/d) T cells were required for allergen-induced itch. (7/x)
We hypothesized that an innate immune cell sets the allergen-responsiveness threshold of sensory neurons to allergens. This idea was inspired by work by Brian Kim’s group showing that IL-4 and IL-13 primed sensory neuronal responses to histamine. (6/x)
www.sciencedirect.com/science/arti...
This question about the inter-individual heterogeneity of allergic responses initiated a major focus in the lab. What determined the set-point or activation threshold of sensory neurons to allergens? (5/x)
This finding was very cool, but it presented a problem. If we all have sensory neurons (we all get itchy every once in a while, don’t we), and if we are all surrounded by the same amount of allergens, why aren’t we all allergic? (4/x)
We found that this substance P activated allergic-skewing dendritic cells to start up the entire allergic immune response. (3/x)
Four years ago, we found that the immune system doesn't directly detects allergens. Instead, it is the sensory nervous system that directly detects protease allergens leading to an itch response and the release of a neuropeptide called substance P. (2/x)
www.sciencedirect.com/science/arti...
Alert! New science from the lab is hot off the press!
www.nature.com/articles/s41...
Before I describe the findings, let’s set the stage.
The lab has always been interested with how allergens are detected. They don’t infect us, so why do we detect them and who does the detecting? (1/x)