Out of randomness, order

I really shouldn’t need help to spot a paper from the Elowitz lab, but Marc Kirschner had to remind me to write on this one (Sprinzak et al. 2010 Cis-interactions between Notch and Delta generate mutually exclusive signalling states. Nature 465 86-90 PMID: 20418862).  Obviously I haven’t been paying enough attention to cell-cell communication.  This paper proposes a very interesting model of how the Notch signaling system works.

Notch, like so many pathways involved in patterning, was discovered in Drosophila.  It has the fascinating ability to make patterns of different cell types out of a homogeneous sheet of cells.  Somehow, one favored cell “decides” to become a new cell type — a neuron, perhaps — and all the cells around it are then “told”, through the Notch pathway, that this cell type is taken and they have to become something else.  This is not patterning on the scale of the Hox genes — you be a thorax, and I’ll be a leg — but cell-by-cell fine structure.

How Notch does this has been a mystery for years.  A lot of work has gone into understanding what signals actually result when Notch is activated.  This is complicated because all the cells you’re studying carry both Notch and the ligand (the canonical Notch ligand is called Delta) and they interact whether they’re both in the same cell or in two neighboring cells.  When they’re both in the same cell, Delta inactivates Notch; when Delta and Notch are in different cells next to each other, the result is Notch activation.

It’s not too hard to understand why Delta would behave differently depending on the geometry of the situation.  Here’s a simple-minded picture of what’s going on.  When Notch (blue) is side-by side with Delta (red), it’s inactivated.  When they’re across from each other, the interaction is physically different and the result is activation.  But both of these events are typically happening in the same cell at the same time, and so the question is, how does the cell integrate the two sets of signals it’s getting.

Sprinzak et al. set out to separate the cis (same cell) and trans (different cell) activities of Notch-Delta so that they could be studied independently.  First, they made a cell line that always expresses Notch, but expresses Delta only when it is induced by doxycycline.  (The Notch they used was manipulated so that it signals to a YFP reporter, but not to normal Notch targets — this is to avoid any downstream complications).  Then they made a fusion protein out of the back end of an antibody (IgG) and the front end of Delta, and adsorbed it onto a glass plate.  This gives them the tools to ask: if the only signaling going on is trans signaling, what happens?

It’s kind of boring, actually.  More trans-Delta, more activation.  Eh.  But it gets much more interesting when they ask what happens in the presence of cis-Delta.  This is a more complicated experiment, because having cis-Delta in the cells means that you have to worry about trans-activation from cell to cell.  To minimize this problem they use a low cell density so that most cells aren’t touching another cell.  This means that most of the trans-activation is still coming from the IgG-Delta on the plate, and this allows them to compare the trans-only activation (above) with trans-activation in the presence of cis-Delta.  The way they modulate the amount of cis-Delta is to express a lot of it, then watch it decay; it’s labeled, so they can quantify how much is in each cell.

Now, instead of a graded response (more trans-Delta gives more signaling) we see a sharp threshold response — nothing happens until the level of cis-Delta gets low enough, and then suddenly you get a dramatic, sharp transition from no signaling to lots of signaling.  The density of IgG-Delta on the plate doesn’t affect the cis-Delta threshold, nor does it change how sharp the threshold is, though it does affect the height of the “on” response.

These are surprising data, but happily Sprinzak et al. can explain the main features of the observations with a simple model.  We need to make one novel assumption, which is this: Notch and Delta in the same cell inactivate each other.  (We know from earlier work that cis-Delta inhibits Notch, but not that the inhibition is mutual; finding out if this is true, and if so what the mechanism is, will presumably be the focus of some intensive future work in many labs.)

As we saw in a previous post (in which I discussed a paper that came out about a month after the Sprinzak et al. paper, by the way) if you have two proteins that bind each other tightly, each of which fluctuates slightly in concentration, you will end up with an all-or-none situation: if you have slightly more Notch than Delta, you will use up all the Delta and be left with only Notch (or vice versa).  This is analogous to what sometimes happens in Scrabble, when you use up all your vowels and are left with only consonants, or (again) vice versa.  So a cell is forced to be in one of two states: a “sending” state (high Delta, low Notch) or a “receiving” state (high Notch, low Delta).

This model neatly explains what was seen in the experiments.  If you have no cis-Delta to start with, you get a graded response to trans-Delta.  If you have a lot of cis-Delta, you have no free Notch and you can’t receive signal (but you can send signals to the cells around you through the free cis-Delta on your surface).  If cis-Delta gets low enough, you suddenly can’t send any more, but now you can receive; this is the threshold at which you start to see Notch signaling.  The threshold is the point at which cis-Delta and Notch concentrations match.

Just as in the Balaban paper I discussed earlier, this system looks as if it might exploit the noise in a system of two proteins that interact tightly to provide a phenotypic switch.  The switching between cell states only requires mutual inhibition — there is no need for transcriptional or other cell-internal feedbacks.  The added wrinkle here is that the cells communicate with each other, and Notch signaling affects the expression of Delta.  So the phenotype of Cell A will influence the phenotype of neighboring Cell B.  But in this model each cell intrinsically has only two choices — it can be in sending mode, or in receiving mode.

Not only does this model explain the results in Sprinzak et al., it also explains older, puzzling data on Notch/Delta mutants.  A fly with only one functioning copy of the Notch gene will give you a messed-up pattern of wing veins, for instance.  A fly that has only one copy of the Delta gene will give you a different pattern that is equally messed up.  But if you cross the two flies together and get a double mutant, the pattern is magically back to normal.  In the model Sprinzak et al. propose, this is easy to explain: when the level of cis-Delta goes down to 1/2 of normal, the amount of Notch required to overcome it also goes down to 1/2 of normal.  So if both proteins are reduced in concentration to the same degree, the point at which cis-Delta and Notch balance stays the same.  Neat.

Sprinzak D, Lakhanpal A, Lebon L, Santat LA, Fontes ME, Anderson GA, Garcia-Ojalvo J, & Elowitz MB (2010). Cis-interactions between Notch and Delta generate mutually exclusive signalling states. Nature, 465 (7294), 86-90 PMID: 20418862