Biological Units: The Strength of A Single Cell

Several recent papers have proposed methods of seeing huge numbers of individual RNA molecules within a cell. I suspect that half the appeal of these methods is the beautiful images they generate. But they’re significant largely because they allow us to pick apart biology at an increasingly tiny scale — in this case, see how an individual cell functions, rather than how the average cell functions.

Laws of Averages

It’s easier to analyse a bigger sample. Biochemical procedures to find and identify proteins and RNAs are sensitive and powerful, but if you’re looking for “everything”, you can always look more, and deeper, if you have more molecules. More cells means more molecules means more chances to find something cool.

But the thing is, if your experiment requires a million cells in order to get enough sample, then you can’t necessarily say that any specific cell in your sample had the particular pattern you observe. You know the average pattern, but you don’t know how variable it is. In particular, seeing a million copies of a transcript in a million cells probably doesn’t mean that each cell has one copy of that transcript — more likely, some cells have a thousand, and a whole lot have zero.

We’re getting a mean without a variance. To get variance, we need to look at single cells.

Beautiful Pictures and Fuzzy Analysis

There are a few different ways of measuring RNA in an individual cell. One platform, called fluidigm, lets you purify RNA from a single cell in a tiny tube, and then do PCR — basically downsizing all the biochemical methods I talked about before.

But more often, people use pictures to see individual cells. They flag a certain molecule with a certain color, and look for how bright that color is, in how many cells, to get a sense for how variable expression is of one gene in a sample.

There are two problems with this. The first is that it’s easier to count how many times a certain sequence came up in your sample than it is to measure exactly how many fluorescent molecules have clustered in a certain spot in a cell. A strong signal probably means high expression, but a weak signal could be low expression, or bad binding, or a bad sample, or any combination of the above.

Sadly, there’s only one way around that, and that’s to optimize your staining. Get a dye that you know works. Do enough samples and enough controls that you are confident in your results.

The bigger problem is that with this method, it’s pretty easy to see one RNA in a cell. Or two, or three. But with sequencing you can, quite literally, look at everything at the same time. You can’t do that with imaging.

Paint by Number

That’s exactly what these techniques try to get around. There are several ways they do this, but the commonality between all of them is that they use superimposed images to develop a bar-code for each RNA molecule in a cell.

So, whether by pooling tags or sequencing, each transcript corresponds to a specific series of colors coming from a spot in the cell. You can read off the colors in order and identify which transcript is in that spot.

A pooled tag approach takes multiple pictures of the same cell with different dyes — a certain RNA transcript will be red and then green and then blue, another will read green, green, blue. And so on. You can use established dyes that are very efficient and let you see things at a very high resolution. You can be sure that you catch the molecules you’re looking for, but you need to know what you’re looking for.

A sequencing approach actually runs a sequencing reaction within a fixed cell. In this case, each base is color-coded, so red-red-green would quite literally correspond to TTA, for instance. You aren’t using good probes, or really any probes at all, so you can’t be sure that something isn’t in there and just in bad shape for sequencing. But you could find things you didn’t even know to look for — new mutated versions, perhaps, which would be of particular interest in cancer.

Back to Averages

In the end, with any of these approaches, the idea would be to analyse thousands — if not millions — of cells. Because we are still looking for averages. Knowing that something happens in at least one cell isn’t so useful — maybe the cell is just strange. Knowing that something happens in 90% of cells, or even 70 or 30% of cells, however, is huge.

And with these methods, by using images, we can get a sense of the variation as well as the average. That’s the strength of looking at a single cell.