The evolution of Hedgehog

PLoS has recently published a highly speculative but very interesting paper on how a particular signaling pathway, the Hedgehog pathway, might have evolved. It's at a fairly early stage in hypothesis testing, which is one of the things that makes it interesting — usually all you see published is the product of a great deal of data collection and experiment and testing, which means the scientific literature gives a somewhat skewed view of the process of science, letting the outsider mainly see work that has been hammered and polished, while hiding the rougher drafts that would better allow us to see how the story started and was built. It's informative in particular for those who follow the creationist “literature”, which often crudely apes the products of actual working science, but lacks the sound methodological underpinnings. In particular, creationism completely misses the process of poking at the real world to develop ideas, since they begin with their conclusion.

So take this description as a work in progress — we're seeing the dynamic of building up a good working model. As usual, it starts on a sound foundation of confirmed, known evidence, makes a reasonably hypothesis on the basis of the facts, and then proposes a series of research avenues with predicted results that would confirm the idea.

Hedgehog is an extremely important family of signaling molecules in multicellular animals. As you might expect, if you're multicellular, your cells need mechanisms to pass messages to one another in development — cells in a tissue all need to work together as a team, and they need ways to coordinate. They need a way to tell each other, "Hey, gang, we're all part of the spleen! Let's build supporting structures for blood cells!" In development, they're also needed for patterning; some Hedgehog molecules, for instance, are used to define the location of the midline of an animal, so that organs will be built bilaterally. Lose hedgehog, the animal loses a sense of that important dividing line, and next thing you know, it's building one eye instead of two, and putting it right where it shouldn't.

What the Hedgehog molecule itself is is a message, sent to other cells, to trigger changes in gene expression in the targeted cells. What it takes is a source cell with the machinery to secrete the message, a molecule (Hedgehog) with unique characters that will make it a specific key, a receptor or sensor on the target cell, and an effector that can be modified by the sensor, and which can then modify gene expression (usually by a cascade of intermediate molecules) to change the cell's pattern of development. Here's a diagram of this process for Hedgehog, or if that's not enough, go read my previous summary of how Hedgehog works:

(Click for larger image)
Schematic overview of the Hedgehog signaling pathway. Signal-secreting cells (left) release the morphogen protein Hh after
modifying it through the addition of two lipid molecules. A C-terminal cholesterol moiety is added via the activity of an intein domain within Hh itself,
whereas the protein Ski/Rasp attaches an N-terminal palmitic acid. Lipid-modified Hh is released from the producing cell with the aid of the Disp
protein. Signal-receiving cells (right) bind Hh via the transmembrane protein Ptc, perhaps with the assistance of the iHog/Boi family of proteins. Hh
binding to Ptc leads to the de-repression of the GPCR-related protein Smo. Smo subsequently initiates intracellular signal transduction events, which
involve proteins such as Cos2, Fu, and Su(fu), that lead to changes in target gene expression. The inhibition of Smo by Ptc is of particular interest
here; it occurs nonstoichiometrically, in a manner that appears to rely on a catalytic activity in Ptc.

Here's the cast list, just to help you keep everything straight.

• On the secreting side, there are a set of molecules involved in exporting Hedgehog, and also in modifying the protein, with one of the key players being the molecule Dispatched. I'm not going to say anything more about this part of the pathway, but notice the similarity between Dispatched and Patched: they're part of the same family.

• Hedgehog is the message. It's a small protein with a lipid, cholesterol, attached to it. It's secreted into the extracellular space, where it can diffuse to nearby cells.

• Patched is the receptor. When it binds Hedgehog, it no longer interacts with the next molecule in the pathway, Smoothened. When it is not binding Hedgehog, it inhibits Smoothened, preventing it from doing its job.

• Smoothened is the effector. When Patched is not interfering with it, it switches on a chain of proteins that lead to the activation of specific genes in the nucleus.

It's a kind of double-negative regulation: Hedgehog inhibits Patched, preventing it from inhibiting Smoothened, which turns on genes. It's also a complicated, Rube Goldbergish molecular machine that, superficially, looks like it had to have been very difficult to evolve, of the kind creationists love to harp on. Nothing just comes out of nowhere, though, and it turns out that each piece here has interesting evolutionary antecedents.

Start with Patched. This protein is ancient; related members of this family of genes can be found in prokaryotic bacteria, which are clearly not multicellular, and clearly not animals. What is it doing in bacteria? It's a transporter, a little pump that the cell uses to secrete excess byproducts of its metabolism, in particular a class of lipids called hopanoids, which are steroid-like molecules. There should be a little light-bulb above your head right now: remember that Dispatched is involved in the secretion of Hedgehog, which is covalently modified with the addition of a steroid-like molecule.

What about Smoothened? It is also a member of a distinguished family, sharing similarities to g-protein coupled receptors (GPCRs). These molecules turn up all over the place; I've mentioned their role in transducing signals in vision and taste. Basically, these are molecules in the cell membrane that can detect a stimulus and trigger a cascade of protein activation events that lead to changes in cellular metabolism.

That's it, and the hypothesis should be clear. Evolution mixed and matched pieces from a couple of different pathways to build this elaborate signaling pathway in a series of small, easy steps, each one conferring a sensible advantage to the organism. Here's a hypothetical illustration of its evolution.

(Click for larger image)
A parsimonious scenario for the evolution of the Ptc/Smo system. We hypothesize that during the transition to multicellularity, a
pre-existing lipid homeostasis system took on a new function in signaling. Initially, an ancient lipid transporter diversified; one of its descendents
came under the transcriptional control of a GPCR that sensed the same lipid (i.e., forming a negative homeostatic feedback loop). Then, the fortuitous
addition of a protein moiety to the lipid in question brought the system under the control of gene expression; a neighboring cell could now secrete
the lipid at will (by coupling it to the protein moiety). Because the combined lipid-protein molecule would block the transporter, this meant that the
sending cell was capable of changing the perceived homeostatic state of the receiving cell, which would have established a graded (quantitative)
mode of cell-cell communication.

It's straightforward. Prokaryotes had a protein Patched for pumping out a lipid X; they also had a GPCR Smoothened that could sense lipid X, and which had a role in regulating genes involved in X metabolism.

One of the genes involved in X metabolism that Smoothened would regulate is Patched. This is the coupling step, and makes sense selectively. If the cell is swimming in lots of X, turn on the protein that will help it get rid of excess X.

The next step is the addition of feedback. If lots of Patched molecules are produced, it would be efficient to halt additional production, so the evolution of an evolutionary linkage between Patched and Smoothened is favored. This would prevent runaway production of proteins in this particular pathway beyond what is needed.

The final step occurs when this pathway is coopted in the evolution of multicellularity. Lipid X is the primitive signal to ramp up a cellular pathway to a particular level and then stop; cells can use this signal to trick neighbors by dosing them with X, causing them to switch on a particular bank of genes. Even more cleverly, by modifying X with a covalently bound protein (which prevents the cell from processing X), they can make an even stronger signal. This part requires a little twist in your perception; we have a protein bias and tend to think of the protein as the important part, which was modified by adding a lipid…this suggests that primitively, the lipid was the signal, and the protein was an add-on.

I like it. It's a very pretty model, and it makes a lot of sense. It's also built on an established pattern of molecular homology and function, so it's not simply conjured out of thin air — there are good reasons to think it is a reasonable explanation.

Science doesn't just stop there, with mere plausibility. The next step is to test the idea, and the authors make a whole series of predictions of what ought to be seen if their idea is valid.

1. There ought to be deeper similarities between the bacterial transporters and Patched. The bacterial proteins are known to function as trimers (three identical proteins have to group up to do their job), and also to use a proton gradient to power their pumping action. Does Patched have similar requirements?

2. Can these bacterial transporters be similarly modified functionally by the addition of protein blockers to their substrates?

3. Patched in multicellular animals can still function in the absence of cholesterol on Hedgehog; it has acquired a specific affinity for the protein, not the lipid. However, could there still be a relic binding site for cholesterol in Patched?

4. There are some interesting phylogenetic quirks that should be explored. Nematodes, for instance, have Patched, but no Smoothened. The authors predict that C. elegans Patched should be able to interact with Drosophila Smoothened. (I found this prediction a bit chancy: nematodes could have secondarily lost Smoothened, and it's dangerous to propose that their pathway represents an intermediate step in the evolution of the pathway.)

5. They also make predictions about the precise mechanism of Patched/Smoothened function. Since Patched is argued to derive from a pump that is pushing a molecule away from the cell, there should be a strictly local required interaction between the two proteins — they shouldn't work by Patched shuffling an intermediate small molecule to Smoothened.

6. One proposal for how Patched and Smoothened interact is by way of local lipid dynamics — Patched is changing the lipid environment of its small area of membrane, sucking up some specific constituent, and Smoothened is reacting to that shift.

That's juicy stuff, with lots of interesting details to be teased apart and analyzed. I also note that it's a research program that takes advantage of biochemistry, molecular biology, bacteriology, and comparative biology to answer an interesting question in developmental biology — I'm feeling like I have to go back to school to brush up on same basics to keep up with my field any more.

But also note what's important about doing science. This is an entirely speculative model about the evolution of a specific pathway, but it requires 1) an appreciation of existing information, and 2) a plan for testing and extending our knowledge. It could also be wrong at multiple points, but the process of working out where it fails will help refine the next model.

Hausmann G, von Mering C, Basler K (2009) The Hedgehog Signaling Pathway: Where Did It Come From?. PLoS Biol 7(6): e1000146. doi:10.1371/journal.pbio.1000146