Very grateful to my mentor Aparna @bhadurilab.bsky.social , co-authors @elisafazzari @claudianguyen95 @RyanKan20 @yoojuyoun.bsky.social
@amartija.bsky.social Daria Azizad, Brittney Wick and Maximilian Haeussler!

We’re excited to see how our tools can continue to reveal cell fate mechanisms from molecular maps of the human brain. 🧠🗺️✨

And to open this approach to other biological problems, we’ve put our scripts to make your own meta-atlas and modules at github.com/BhaduriLab/d.... (13/13)

Even a subtle KD of these transcription factors produced a modest decrease in module 20 that cascaded into substantial composition differences. (12/13)

With chimeroids, we knocked down FEZF2 & TSZH3 across multiple genetic backgrounds. Consistently, both FEZF2 & TSHZ3 were required to make deep layer neurons – w/ FEZF2 acting at the level of gene expression and TSHZ3 acting at the chromatin level. (11/13)

Interestingly, TSZH3 and the rest of module 20 was less specific to FEZF2 subtype neurons in the developing and adult mouse brain. So we tested our model in human stem cell derived cortical chimeroid models developed by @bolanosanton.bsky.social and @irenefaravelli.bsky.social @ArlottaLab. (10/13)

We validated that at GW16, FEZF2 and TSHZ3 co-express in deep layers of the human cortex, and by GW20, levels of FEZF2 decrease while TSHZ3 expression stays on in these deep layers. (9/13)

However, FEZF2 expression peaks before module 20 activation during development, and half of module 20 genes are candidate targets of FEZF2 (thanks to data from @LodatoS_Lab @ArlottaLab). This included one transcription factor, TSHZ3, linked to autism spectrum disorders. (8/13)

Most notable is module 20, which we determined drives the specification of deep layer neuronal subtypes found in the adult. Module 20 is enriched in human adult FEZF2 subtypes – but it doesn’t actually contain FEZF2. (7/13)

We found modules that may explain how neuronal/glial fates are initiated and refined into cell types found in the adult human cortex, with the spatiotemporal expression patterns predicted for some of these modules validating in primary tissue. (6/13)

We first annotated these modules for their cell type/biological function, then scored how each of these modules act in other datasets – ie the adult human brain, the developing mouse, and the adult mouse. @alleninstitute.bsky.social @danielajdibella.bsky.social (5/13)

We used these atlases to find gene networks that shape cell types.

Step 1: Define networks w/in each indiv’d in our dataset w/ hierarchical clustering.
Step 2: Correlation analysis btwn indiv’d to id which networks co-expressed across the entire meta-atlas – these are our meta-modules. (4/13)

So we merged recently published datasets into two meta-atlases: one for the developing human cortex (7 datasets, 0.6M cells) & one for the adult human cortex (16 datasets, 2.6M cells) – both on the UCSC Cell Browser now:
dev-ctx-meta-atlas.cells.ucsc.edu
adult-ctx-meta-atlas.cells.ucsc.edu
(3/13)

Plenty of work has cataloged the cell types found in both the developing and adult human brain. But it’s hard to generate an atlas with enough scope and depth to clarify how the cell types present in development turn into those found in adulthood (2/13).