While we're on the what's-going-on-inside-cancer-cells topic, there's another new Nature paper that makes interesting reading on the subject. This one also confirms some earlier work, but does a pretty thorough job of it, and sheds some new light as well on that breast cancer mutation work that I blogged about the other day.

This group has taken around 100 cells at a time from tumor samples, and applied single-nucleus sequencing techniques to them. After correcting for a number of factors (see the paper, of course, if you're really into this stuff), they can use this technique to get a good read on the genomic copy number of the cells, which is of particular interest for unstable things like tumor samples. (Comparison of single-cell data versus the averages from samples of millions of cells were also made; the correlations are quite good).

The cells were taken from a breast cancer sample ("triple negative" ductal carcinoma) and from an associated metastatic liver tumor from the same patient. (Given that situation, I'm guessing that these were post-mortem). Each sample was dissected into physical zones, and the cells were then flow-sorted and subjected to sequencing. This revealed a number of interesting patterns.

For one thing, the original tumor sample showed different cell populations across its diameter. There were standard diploid cells in the entire sample, but one population type (less than diploid) was found only one end of the sample, fading out gradually towards the middle, with two other varieties (near-tetraploid) showing up at the other end. On closer inspection, almost all of those garden-variety diploids were normal cells, and most of those were white blood cells, immunocytes that had infiltrated the tumor.

Of the cancer cells themselves, half of them sorted out into three distinct clonal populations, and the metastatic tumor was found to be purely derived one of these. Looking these over, it appears that these three groups emerged very early in the process, and the various mutations (and there were many) still traced back to these early branch points. One of them (arising much later in the process) was clearly more like to break loose and resettle than the others, a pattern that has been seen in other studies. Trying the same thing with another set of tumors from a different patient, they found in this case that the tumor had emerged from a clonal expansion of a single aneuploid cell line, and the metastatic tumor was from one of the later resulting mutants (and had hardly evolved since).

But what about the rest of the tumor cells in these samples? Those turned out to be the pseudodiploid population that they'd seen in the initial sorting, and these were all over the place genetically. No family-tree relationship could be drawn between them (in contrast to aneuploids), indicating that they hadn't been doing the clonal-exapansion thing. They seem to be the result of some ongoing genetic instability in the tumor population, generating a steady stream of one-of-a-kind messed-up cells with missing chunks of chromosomes. You have to wonder if a lot of the 1700 mutations that the full-genome sequencing work picked up were from these cells, and that the other clonally-similar lines showed less variation.

If that's true, it would probably be good news - perhaps all these mutations aren't as evenly spread across the worrisome cells types as you might fear, and especially among the ones that go metastatic. It's a pity that we can't yet do whole-genome sequencing on single nuclei - combining the population breakdown this copy-number technique gives with the hardcore sequence data would really tell the story, and tell us which cells we have to try to kill off first. And finding the root causes of all the genomic instability would be a big advance, too - I wrote about this back in 2002, and it's still relevant and still not well worked out.