What About BOB-Seq?

An interesting new approach to short read RNA-Seq library preparation bobbed into view this week, using click chemistry to barcode samples before any enzymatic processing.  This bioorthogonal barcode sequencing approach - aka BOB-Seq -  offers great promise in the burgeoning space of perturbing cell lines with drugs or other agents and then profiling the RNA.   Core to the premise of BOB-Seq are two reagents.  The first is a cell-permeable reagent which tags the RNA by acylating the 2' hydroxyl of ribose.  Extraction of the tagged RNA leads into the second reagent, which will undergo a copper-free click reaction with these tags, tacking on a designed oligo.  After quenching any untagged positions with azide, the samples can be pooled prior to purification, poly-A selection, and reverse transcription.  Pooling upstream of those steps saves effort and reagents - and thereby reducing costs and variability.  During first strand cDNA synthesis, the clicked-on oligo serves to barcode and add a PCR primer target via a template switch mechanism.   Amplify the library judiciously with appropriate primers, and BOB-Seq's your uncle - off you go to the sequencer.

Figure 1 from the preprint

Zooming in on BOB-Seq

Click chemistry is an amazing thing - extremely specific condensation reactions conducted under very mild aqueous conditions.  Many require copper ions, which are toxic to cells, though since here the click reaction is occurring on purified RNA that isn't a concern.  The chemistry used here is called strain-promoted azide-alkyne cycloaddition or SPAAC.  

The preprint carries many details on the development of the protocol.  In particular, there was the question of the best way to link the tag to the first strand cDNA.  After all, rather than being properly kabobbed on the phosphate backbone of the template RNA like any respectable oligonucleotide, the bobtags are linked via the 2' hydroxyl on the RNA.  So with this very odd connectivity, there was a question of what would work.  They tried four different approaches: having the bobtag bear an oligo-dT sequence and thereby prime first strand synthesis, ligating the bobtag to the completed first strand, or two different template switch schemes.  The difference in the last two is whether the bobtag oligo is linked to the 2' hydroxyl via the 5' or 3' end. 

After extensive testing of conditions for each method, the 3' template switch scheme (3'DBCO) was found to have the least barcode switching - 99% barcode accuracy based on experiments in which separately bobcoded human and mouse libraries were sequenced together.  So that is the method they pursued for the remainder of the preprint.  Importantly, this approach gives relatively uniform coverage across the transcript body via a random priming method, which might not be expected from the oligo-dT scheme.  

They also explored the effects of different doses of the acylating agent and found that there is a very consistent relationship between the dose and the length of cDNA inserts generated.  This is interpreted as the the acylation adducts triggering reverse transcriptase pausing and thereby encouraging template switching.  

Several existing methods for RNA-Seq attempt to reduce costs by pooling early, typically after first strand cDNA synthesis primed with a barcoded oligo-dT primer.  DRUG-Seq is one such technique; the paper mentions Prime-seq which is similar but differs in some key details (DRUG-Seq doesn't purify RNA, Prime-seq does).  When I was a graduate student, it was a mantra in the Church lab that methods should try to pool as early as possible, since that reduces the complexity of the number of samples to handled. 

Using HEK293 cells and some previously published perturbations, the Sculpta group generated multiple BOB-Seq libraries and sequenced them, finding that their variation in expression scores between replicates was significantly less than public datasets for HEK293 produced by other methods.  A sobering finding is that PCA plots of these datasets strongly clustered by method and this bears out in generally disjoint Venn diagrams of up- and down-regulated genes.  The preprint attributes this to differences in method, but I can't help but wonder if HEK293 cells just show great variability between sites or culture conditions.  It's a lot of expensive work, but ideally a single lab would run the various protocols on the same source material to better understand the drivers of variation.

Commercialization?

I reached out to the senior author of the BOB-Seq preprint, Neal Amin, to inquire about the commercialization plans for BOB-Seq.  Amin is the founder and CEO of Sculpta Bio, a startup which is planning to develop therapeutics which target splicing.  Enroute to that goal, Scupta is touting "ground truth transcriptomics for AI-drug discovery".  These are the impetus for developing BOB-Seq - earlier pooling should enable lower processing costs and more comprehensive splice isoform discovery.  The strong coverage across the whole of the transcripts versus the highly 3'-biased coverage of DRUG-Seq would mean BOB-Seq should be markedly superior at discovering splice junctions.

Amin says they are still working out their plans for BOB-Seq.  Given the company's goals, it could be retained as a platform technology, could be licensed out selectively to partners, could be offered as a service, or it could be out-licensed to an established NGS reagent kit provider for broad access.  

Going with a kit is the angle I hope for, as I can see this being broadly useful to biology.  Sure, it's a little worrisome that competitors might use your technology to go against you should BOB-Seq be placed in the broad marketplace, but I suspect Sculpta will quickly find there are more disorders amenable to splicing modulators than any single company could possible tackle - even with large established partners.

So here's hoping a BOB-Seq kit is available in the near future so more complex transcriptoms can be fully mapped.  Priority species would clearly include bobcats, bobolinks, bobwhites - not to mention bobtail squid and the geranium known as "stinky bob".  

The preprint also notes the method might find value in areas other than bulk RNA-Seq from cell lines.  Reverse transcription tends to be inefficient on samples in a solid matrix, such as for spatial profiling.  For spatial profiling, acylation and bobcoding could occur with spatially arrayed bobtags, followed by extraction so that the reverse transcription could occur in solution.  Similar approaches with bobtags on beads might be used for single cell RNA-Seq, and the preprint also suggests possible advantages of using bobtagging for Formalin Fixed Paraffin Embedded clinical samples.

Anyone interested in this should definitely read the preprint - as I noted before it is rich in detail and there are many thingamabobs mentioned in it that I have skipped over and details I decided to treat as sideshow BOB-Seq.

The BOB-Seq preprint suggests there are still clever library prep methods out there to be discovered.  Click chemistries have not seen widespread use for NGS applications.  Oxford Nanopore uses click chemistry for their rapid adapters.  A quick search revealed a click chemistry used to sequence DNA damage.  There is Click-Seq, a commercialized RNA-Seq method, but in contrast to BOB-Seq it adds the click residues via the first strand synthesis by reverse transcriptase and so can't pool samples as early as BOB-Seq.

Could something similar be applied to DNA sequencing library preparation?  A general fear in this space is that library preparation costs are not plummeting in parallel with DNA sequencing costs.  This is particularly acute if you wish to sequence very small genomes - the cost of library prepping a bacterial genome might well outweigh its share of a NovaSeq flowcell.   BOB-Seq chemistry is targeting the 2' hydroxyl, which is particularly reactive but importantly is the root of why DNA is "deoxy" - it is missing this hydroxyl.  Could the nucleotide bases themselves be targeted in an analogous way?  That's way beyond my very limited chemical knowledge (in my defense, click chemistry was first announced more than two decades after my last undergraduate credits in Chemistry).