Monday, April 10, 2017

10x Launches Mass T-Cell Receptor Decoding

Adaptive immunity is an endlessly fascinating topic which I have not explored very deeply, which is particularly unfortunate given the many parallels to computing.  Combinatorial logic is used to construct a vast array of possible antigen readers, expression logic ensures that only one such reader is expressed in a given cell and hypermutation and evolution are used to optimize these readers to match specific antigens.  All this not only creates weapons to deploy against foreign invaders, but also a memory which effectively records an individual's history of environmental exposures.  Just before I started writing this two tweets highlighted using adaptive immunity profiling to reveal exposure to tuberculosis and cytomegalovirus.  Adaptive immunity is responsible for transplant rejection, with new companies looking to more selectively modulate immunity to enable transplants without shutting the immune system down.  Adaptive immunity also ties into the white hot field of immunotherapy for oncology, exploring whether differences in antigen response underlay variation in immunotherapy success.  To enable profiling adaptive immunity on a mass scale, 10x Genomics has now introduced a single-cell kit for targeted profiling of T-cell receptor variable regions.
T-cells and B-cells constitute the two key arms of adaptive immunity, and each uses specialized receptors to detect antigens, T-cell receptors (TCRs) for T-cells and immunoglobulins for B-cells.  Each of these is assembled by a multi-level combinatorial hierarchy.  First, each is composed of two independent chains, each of which contains a variable region.  These variable regions are in turn constructed by drawing from pools of V(ariable), D(iversity) and J(oining) segments (T-cell alpha chains and immunoglobulin light chains lack the D segments, but are otherwise built along the same lines).  These arrangements are constructed by DNA splicing and are further diversified by localized mutagenesis.  Evolutionary selection on a cellular scale favors the combinations that are functional (the V(D)J splicing can create non-functional configurations, particularly since the process is inherently sloppy).
(image from 10x Genomics blog)

The challenge for assaying the repertoire of T-cell receptors or immunoglobulins in a sample is that one needs to know the alpha/light and beta/heavy pairing for each cell, and each chain has its own messenger RNA.  So by enlisting single cell RNA profiling, it is possible to read these out as discrete pairs.  However, many single cell RNA methods are optimized for transcript counting and report only a small portion of the target RNA.  Furthermore, if the T-cell receptor or immunoglobulin configuration is the only thing valuable to be mined, then all the other messages from the cell are overburden.  

In a phone conversation, 10x described how they have modified their single cell RNA profiling method to capture specific amplicons encompassing the variable regions of T-cell receptors.  Cell encapsulation and reverse transcription in droplets results in the tagging of messages with both a unique molecular identifier and a droplet-specific 5' barcode.  After emulsion breaking, a semi-nested PCR procedure is used to capture full-length V(D)J .  The ultimate amplicon, containing the V(D)J region, unique molecular identifier and droplet barcode, is on the order of 600-700 basepairs.   Amplified material is fragmented to generate a set of fragments suitable for Illumina sequencing.

Each channel on the encapsulation instrument can capture up to 10,000 cells.  According to 10x, for many applications this is much in excess of what is needed.  For example, tumor-infiltrating lymphocyte populations may be highly enriched for only a few clones.  For applications that might want higher numbers of profiles, such as looking at peripheral blood cells, multiple samples can be encapsulated.  Since the T-cell repertoire (a B-cell version is in development) is now represented very efficiently, very little sequencer capacity is needed.  Pricing is around $1200 per sample, which works out to $0.12 per cell at the maximum density.  

It occurred to me that this application might be feasible on long read instruments.  While the cost per basepair is higher, there is an inherent oversampling required by fragmenting the amplicons and so the cost advantage for Illumina is slightly compressed.  Perhaps it would be most useful in an educational setting, with advanced students extracting mouse T-cells and profiling them all in a single lab session.  The folks at 10x said they are currently focused on their core customer base using short read platforms and extending their applications there, but I'd be shocked if someone didn't do this just to demonstrate proof-of-concept. Well, that would be possible if 10x later launches a mouse version of the kit.  The idea of students profiling their own peripheral T-cells would be cool, but seems ethically fraught given the infectious disease information which could potentially be revealed.

10x doesn't yet have any publications with this method; their focus was to get the application up-and-running and get it out to customers.  The T-cell kits will ship this month, so customers will be able to start experimenting imminently.  In today's burgeoning pre-print environment, perhaps some early results will show up this summer.

No comments: