Stephen Quake: Single Cell Genomics

Stephen Quake was here a few days back and gave a nice talk summarizing about several decades of research. The first part of his talk was devoted of what he probably most well known for: single cell genomics. By developing “small plumbing systems”, AKA “microfluidics”, Stephen was able to develop a platform where lots of interesting reactions could be run parellel. His early successes included using his system to sequence his own genome and to sequence the genomes of uncultured microorganisms. He also covered recent work in which individual cells were separated and sequenced. I forget exactly how it was done (RNA->cdDNA + sequencing, or simply PCR off raw DNA), but the end result is a set of SNPs for each cell. Cells and SNPs can then be clusters and from this clustering, groups of similarly mutated cells can be recovered. These can be constructed in to a phylogenetic tree of clones and this tree can then be used fo infer the clonal evolution of the cells in the population with one goal being to reconstruct the types and relationships of clonal members constituting a tumor mass. Its a nice idea and seems to work, however, the second set of technologies he showed were simpler and even more powerful.

The second big idea is the sequencing of free-floating DNA in the bloodstream as a diagnostic tool. Stephen introduced the idea of getting prenatal testing where it is desirable to genotype the baby for possible aneuploid issues (like trisomy 21). To do so cells need to be harvested from the amniotic fluid which involves a long needle and some risk. Might there be an easier way? Indeed. Stephen proceeded to show us that circulating DNA in the bloodstream could be used for this, and many other purposes. Remarkably, as a by product of celular lifecycle, lysed cells leak their nucleic acid content into the blood stream. This may sound counterintuitive since it is the job of the macrophage to scoop up and process debris, but there are sufficient numbers of dying cells at all times to leave about 1000 genome equivalents of DNA per ML of circulating blood. It turns out, somewhat amazingly that a small number of baby-derived cells also make it into the mothers blood stream and some of these may lyse as well with the consequence that a small fraction of the circulating blood belongs to the baby. Now if you simplet isolate DNA and map it back to chromosomes you can use a simply binning procedure to look for aneuploidy. In the case of a control, there should be equal amounts of chromosome 21 relative to the other chromosomes (normalized by chrome size, of course). In the case of a mother of a baby with trisomy 21, the total chromosome 21 content will go up relative to the other chromosomes in proportion to the amount of baby DNA in the bloodstream. As you can imagine this is an extra chromosome relative to 23*2 - 46. So already you are talking about a difference in baby DNA representing a change of 47 chromosome to 46 in the healthy baby. Multiply this by the fact that the total DNA content is only some tiny fraction (lets say 1/1000 for argument’s sake) of circulating DNA and you see how the counts work. You need to be able to detect a 1 part in 1/(47,000) difference. If the fraction of baby DNA is smaller, the numbers get correspondingly smaller. But remarkably you can EASILY surpass this using modern sequencing where a single run can have tens to hundreds of millions of reads and therefore the test can run to an arbitrary precisions by sequencing deeper. He showed that this technique could be used to detect trisomy 21 in a study, and that this is now the basis for a commercially/clinically available prenatal test. its MUCH faster and easier than karyotyping although I imagine the karyotypes will remain the gold standard for a while.

The DNA-counting idea was then extended to RNA so show how tissue-specific RNA expression could be used to determine the presence of lysed cells of a certain tissue. What can this be used for? Well, for one if can be used to track organ transplant failure on the idea that a rejected organ’s cells will release a lot more DNA/RNA into the bloodstream and allow an observer to track the rise. A second application he showed was the ability to track fetal RNA! What? That sounds crazy to me. He showed data suggesting it was possible to pick up on important trimester-dependent gene expression like chorionic gonatropinc simple from the mother’s blood. This seemed to me to be at the limits of detectability - after all, how many cells in the fetus could be lysing and releasing RNA (or perhaps there is active RNA transport - seems unlikely) - but that was the data he showed. And he doubled down when asked about gene expression later in pregnancy - for example could certain expression signatures indicating neurodevelopmental disorders be picked up. “Easily,” he said.

While the single-cell stuff is cool, the non-invasive testing stuff is off the chain. The power of modern sequencing never ceases to astound and these techniques push forensic ability to the limits through the use of technology.