Thread by @shalinhnaik

In the absence of a time machine, a single cell cannot be tested more than once in a destructive assay. We introduce 'clonal multi-omics' (SIS-seq + SIS-skew) to bypass this challenge and identify Bcor as a negative regulator of dendritic cell fate cell.com/immunity/fullt…

Starting in 2012, and what I thought would be wrapped up by 2015, it was a long but ultimately rewarding journey. 3 yrs after our preprint bit.ly/3al3yrH, 2 yrs after @KleinLabHMS seminal state-fate, and @arjunrajlab's recent Rewind, finally out in @ImmunityCP

For 15 yrs, the question for me has been the same: what are the patterns of haematopoietic stem and progenitor cell fate heterogeneity, and what controls it. It started when Shortman and I found 'multipotent progenitors' (MPP) were not all multipotent. nature.com/articles/ni1522

Schumacher lab and I developed next-gen cellular barcoding and found this was also true in vivo. The killer experiment for me was Fig 5: siblings of barcoded HSPCs injected into two different mice demonstrated the same lineage bias. Fate was intrinsic! nature.com/articles/natur…

@DawnSLin also found clonal siblings were synchronized in the *timing* of cellular output. They 'know' ahead of time which cells, how many and when to produce them. In fact, I wish I titled this study 'HSPCs are programmed for their clonal kinetics'. cell.com/cell-reports/f…

So here's the rub. Once you test a cell for fate you know what it makes but lose the chance for scRNA-seq. Conversely, destructive scRNA-seq precludes the chance to test that cell for fate. How do you get both? Use the daughters! Enter clonal multi-omics: SIS-seq + SIS-skew

As a side note, wherever there is clonal heterogeneity (stem cells, T cell responses, cancer causing/resistance clones) clonal multi-omics can help identify the genetic correlates. Methods like this

Jaring/Jessica found that single LSKs were heterogeneous for pDC, cDC1 and cDC2 fate (columns) but clonal siblings were conserved (rows). @zalcenstein did RNA-seq on their frozen siblings. Peter, @MelanieBahlo, @folded provided important preliminary analysis.

Then driven by @Luyi_T @mritchieau & help from @shian_su, they found ~500 genes using linear regression that correlated with fate i.e. those genes that predict clonal fate bias towards the DC subsets 'hidden' in the data. Our time machine in action!

But correlation isn't causation, so we did a CRISPR screen to find novel regulators. After blood, sweat and tears, Jaring got it to work, with help from Andrew, Tracy, Adrienne, and @nanopore sequencing (don't ask!) with Quentin. @Luyi_T did the analysis!

Importantly, he 'rediscovered' known genes validating the approach works! He also discovered 30 putative new regulators where 2 gRNAs gave similar results. We chose Bcor to extensively validate (Thanks @drng). BcorKO HSPCs generated more cDC2s and pDCs, but not cDC1s.

@SaraTomei3 joined in 2018 and heroically took on validation. In vitro, BcorKO had higher numbers of cDC2, pDC, but also pre-DCs. With @Luyi_T and @zalcenstein she index-sorted for scRNA-seq and they found DCs made were similar to WT, but there was a unique pre-DC-like cell.

When transferred in vivo, no phenotype. Try again...no phenotype. Again?.. nothing. WTF?! So we thought "our screen was in Flt3L cultures, lets inject Flt3L"... Phenotype!Turns out Bcor is a negative regulator of cDC2 and pDC but only in emergency conditions with high Flt3L.

This was reminiscent of Dan Tenen's CEBP beta role only in emergency granulopoiesis. Any other examples? nature.com/articles/ni1354. @drng (and @bloodgenes) also wonder about its role in aplastic anemia where DC numbers increase.

But wait, what is the cellular mechanism behind this phenotype? That is, what is a clone's fate ordinarily, and how would is this skewed in BcorKO. There can be 4 reasons how clones change their fate: fate gain, fate loss, clonal expansion, clonal contraction, or combinations.

This is another time machine problem, so we developed SIS-skew. In one version we used barcoding (for throughput) and a Cherry-labelling gRNA at 50% transd. for clone-splitting. By assessing barcodes in Ch+ or Ch- DCs, we could examine the above 4 possibilities.

Simple in theory, the challenge of analysis was something else! Over weeks of Zoom meetings during lockdown, Tom Weber finally developed a method SkewID that finds clonal changes in multidimensional space between WT and BcorKO.

Turns out, cDC2 numbers increase mostly through clonal expansion, and pDCs mostly through fate gain (i.e. clones that couldn't make pDCs now can. Amazing!) Side note, SIS-skew could be applied to many questions of clonal perturbation; cytokines, oncogenes, SNVs, reprogramming.

Final thanks to Meredith, Ee Shan and Tim for functional testing, @LevKats and Madison for BcorKO mice (more to come), and @DawnSLin, @drng, Sam Taoudi and others for critique. And to all of my staunch supporters over the especially difficult last 3 years. You know who you are!

...I believe will transform our ability to discover the controllers of clonal destiny.

Of course in the meantime came the tsunami of beautiful single cell papers from @LeeGrimesLab @IdoAmitLab @WJGreenleaf Bertie Gottgens, John Dick, to name a few, incl. the beautiful inferential study by @haas_lab @larsplus