You know you're in America when...

... you order a chicken caesar salad and an iced coffee, and have trouble finishing both of them.

Especially when you're me.

(They were both nice)

Musing on apoptosis, stem cells, dual rail signalling, such like and so forth

On my walk into Mountain View this evening, I picked up a copy of the US edition of New Scientist. On page 14, 3rd July 2004 issue, the article Embryonic stem cells 'should be dead' jogged a memory of an excellent conversation early this year, when doseybat and I visited livredor in Dundee.

We were chatting about an old idea of mine about modelling dependencies between genes (I'll spare you the details of that, because it's actually not directly relevant). Anyway, livredor was describing some of the characteristics of P53, which somehow seems to know when a cell is trying to behave like more than one cell line at once, triggering apoptosis. The interesting thing was that there seemed to be no direct connection between P53 and what it was sensing, because it appeared somehow to 'know' that something was wrong, despite many, many possible triggers. I started asking naive questions, working from a computer science/information theory point of view, just to try to understand what was going on. What I didn't know at the time was that a large proportion of signals within cells operate in pairs; i.e. one protein for 'on', and a different protein for 'off'. Apparently, in a normal cell, you never typically see both molecules present at the same time, just one or the other. And, these signals form networks that are at least to a point (to my engineer's eyes, at least) somewhat analogous to signals in networks of digital logic gates. Interestingly, this exact same 'dual rail', one signal for on, another separate signal for off, principle is actually commonly used in the design of asynchronous digital circuits. I mentioned that I'd recently been to a workshop presentation, where dual rail signalling had been modified to use 'both on' to represent an error state, that (so it turns out) naturally propagates around the circuit if you happen to use the default (obvious) designs for and, or and not elements.

It occurred to me that if you built a digital circuit from dual rail logic with several, distinct operating modes, each selected by a separate dual-rail input (i.e. several such inputs, with exactly one input active at any time), then heavily optimised the logic to remove gates that would only be necessary to arbitrate cases where multiple modes might be simultaneously active, you might have something analogous to what goes on inside a cell after it has been specialised. It is the nature of genetic algorithms to optimise anything you throw at them, so in a case where all normally functioning cells only ever have one mode operational at any time, it would be very likely indeed that evolution would knock out unnecessary 'logic'. OK, so what would happen if, say, two mode select inputs are simultaneously active? Well, in the case of dual rail digital logic, without the extra hardware necessary to handle these cases, you'd get a (probably large) number of cases where both rails are turned on simultaneously. I wondered if cells might do this, and if P53 might pick up on that somehow. Anyway, livredor pointed out that there is a membrane through which these signalling molecules pass, with 'on' signals passing in the opposite directions to 'off' signals. Hmm, I thought. Then she pointed out that if you block the channels in that membrane (she mentioned a recently discovered drug that did just that), you turn on P53, and hence the demise of the cell, rather extremely. Smoking gun, thought I.

Anyway, I didn't really think much more about this until I saw the article in New Sci today. Apparently, in stem cells (which are basically cells where pretty much all modes are turned on at once, until they specialise), contrary to expectations, all the usual things that turn on programmed cell death are present big time, but the cells don't die. This is really fascinating, because if my theory is correct, if all the modes are simulataneously active in a cell, as with a stem cell, you will inevitably get exactly the behaviour that was reported.

I suppose I'm mostly posting this as a prod to livredor -- I really am fascinated to know whether this hunch is right. It feels right from an engineering point of view, but, of course, someone will have to prove that it is true by inventive labwork, which is (realistically) well outside my sphere.