The Evolution of Vertebrate Blood Clotting

The original draft of Finding Darwin's God contained a longer, more detailed account of the way in which the vertebrate blood clotting system might have evolved. This documents contains part of that original draft (with references) and a number of diagrams. My editor insisted that the blood clotting section of my draft was already too long and too technical, and that the manuscript would benefit from paring down the details. For the general reader, I agree that this was a wise editorial decision.

However, a number of readers have asked me to place the more detailed description on the Web, and that's what this document represents. As you will see, my description was originally planned to follow a portion of the text that explains the evolution of clotting in an invertebrate - the lobster. If you have a copy of my book, the lobster clotting system is described on pages 158-161.

Ken Miller
Brown University



Imitating Rube

Does the human clotting system lend itself to the same kind of analysis? At its core, the actual mechanism of clotting is remarkably simple. A fibrous, soluble protein called fibrinogen ("clot-maker") comprises about 3% of the protein in blood plasma. Fibrinogen has a sticky portion near the center of the molecule, but the sticky region is covered by little amino acid chains with negative charges. Because like charges repel, these chains keep fibrinogen molecules apart.

When a clot forms, a protease (protein-cutting) enzyme clips off the charged chains. This exposes the sticky parts of the molecule, and suddenly, fibrinogens (which are now called fibrins) start to stick together, beginning the formation of a clot.

The protease that cuts off the charged chains is called thrombin. So, just like the lobster clotting system, the heart of the reaction involves just two molecules: fibrinogen and thrombin. But, unlike the lobster, there's a lot more to this machine. It turns out that thrombin itself exists in an inactive form called prothrombin. So it, just like fibrinogen, has to be activated before it can start the clotting process. What activates prothrombin? Here's where life gets really interesting. Prothrombin, a protease itself, is activated by another protease called Factor X which clips of part of the inactive protein to produce active, clot-forming thrombin. OK, so what activates Factor X? Believe it or not, there are still more proteases, two of them, actually, called Factor VII and Factor IX, that can switch on Factor X. What switches them on? Here's where a good teacher goes the the blackboard, and so will I:

Now, "beauty" is a word that most people wouldn't think to put in the same sentence with "biochemistry," but the biochemistry of this pathway is beautiful indeed. Start at the bottom, with the conversion of soluble fibrinogen into clot-forming fibrin. As you look up, you can trace this process to one of two different external stimuli, both of which make good sense. At the upper left, the pathway can be started with damage to a cell surface, something that happens whenever blood is exposed to the air or a foreign object at the surface of a wound. At the upper right, tissue factor, a soluble protein found in most tissues but not in the bloodstream, activates the pathway. This is where clotting starts from an internal hemorrage, a broken vessel within the tissues of the body. So, both of these ultimate stimuli lead to the same set of clot-forming proteins (Factor X, thrombin, and fibrin), but neither does it directly. Instead, each activates a "cascade" of intermediate factors, nearly all of them proteases, which eventually activate clot formation.

It sure does look pretty, but why a cascade? Why couldn't we have a simpler pathway, like the lobster, where something like tissue factor activated clotting directly? Well, we could, but a complex pathway, even if it drives biochemistry students to distraction, has advantages of its own. For one thing, the multiple steps of the cascade amplify the signal from that first stimulus. If a single active molecule of Factor XII could activate, say, 20 or 30 molecules of Factor XI, then each level of the cascade would multiply the effects of a starting signal. Put 5 or 6 steps in the cascade, and you've amplified a biochemical signal more than a million times. Clotting with fewer steps would still work, but it would take longer to produce a substantial clot, and would be much less responsive to smaller injuries.

Michael Behe is in awe of the the intricate complexity of this system, and so am I. And he is also correct in pointing out that if we take away part of this system, we're in trouble. Hemophiliacs, for example, are unable to synthesize the active form of Factor VIII. This means that they are unable to complete the final step of one of the pathways, and that's why hemophilia is sometimes known as the bleeder's disease. Defects or deficiencies in any of the other factors are equally serious. No doubt about it - clotting is an essential function and it's not something to be messed with. But does this also mean that it could not have evolved? Not at all. The key to understanding the evolution of this intricate system, as Russell Doolittle has pointed out, is the fact that the clotting factors share an exquisite and revealing similarity.

Building the Machine

To paraphrase Darwin, the notion that evolution could have produced a system as intricate as the blood clotting cascade seems, we might freely confess, "absurd in the highest possible degree." This is especially true if you believe, as Behe seems to, that clotting is not possible until the entire cascade of factors is assembled.

But we already know that evolution doesn't start from scratch, and it doesn't need fully-assembled systems to work. Remember the lobster system as an example. Blood clotting evolved there from two pre-existing proteins, normally found in separate compartments of the body, that had a fortuituous interaction when damage to a blood vessel brought them together. Once that interaction was established, natural selection did the rest.

Could something like this have happened here?

Remember, we're not starting from nothing. We're starting about 600 million years ago in a small pre-vertebrate. with a low-volume low-pressure circulatory system. Just like any small inverterbate with a circulatory system, our ancestral organism would have had a full compliment of sticky white cells to help plug leaks. In addition, that ancestral system would have had something else. Most of the time, hemorrage starts with cell injury, meaning that cells are broken in the vicinity of a wound and their contents are dumped out. That means, among other things, that all of a cell's internal signalling molecules are suddenly spilled out into the damaged vascular system. Included among the contents are a whole slew of internal signalling molecules, including prominent ones like cyclic adenosine monophosphate (abbreviated: cAMP), all dumped into the tissue surrounding a wound.

Why would a sudden gusher of cAMP in a wound be significant? Well, it turns out that vertebrates use cAMP as a signalling molecule to control the contractions of smooth muscle cells, the very sort of muscle cells that surround blood vessels. Therefore, the release of internal cAMP from broken cells would automatically cause smooth muscles around a broken vessel to contract, limiting blood flow and making it more likely that the blood's own sticky white cells would be able to plug the leak. That means that we already have some ability to limit damage and plug leaks in a primitive, low-pressure system. Not a bad place to begin.

Our next step is to consider the nature of blood itself. For reasons relating to osmotic pressure, the tendency of water to move across cell membranes, blood plasma is a viscous, protein-laden solution. And it's also important to note that the extracellular environment of ordinary tissue is very different from blood. These spaces are laden with protein signals, insoluble matrix molecules, and extracellular proteases that cut and trim these molecules to their final shapes and sizes. In fact, such proteases constitute one of the major forms of extracellular signalling. So the tissues of our ancestral vertebrate would be laden with protein-cutting enzymes for reasons completely unrelated to clotting.

Keeping all of this in mind, what would happen when a blood vessel broke in such an organism?

Well, protein-rich plasma flows into an unfamiliar environment, and sticky white cells quickly "glom" up against the fibers of the extracellular matrix. Tissue proteases, quite accidentally, are now exposed to a new range of proteins, and they cut many of them to pieces. The solubility of these new fragments vary. Some are more soluble than the plasma proteins from which they were trimmed, but many are much less soluble. The result is that clumps of newly-insoluble protein fragments begin to assumulate at the tissue-plasma interface, helping to seal the break and forming a very primitive clot. (Could one object that this is too primitive and too nonspecific to work? That it wouldn't be sufficient to seal breaks? Well, it turns out that you can't make this objection for the very simple reason that this is pretty much the clotting mechanism used today by a large number of invertebrates. Works for them, and therefore there is no reason why it wouldn't have worked for the ancestors of today vertebrates, either!)

Now we get down to business. A mutation duplicates an existing gene for a serine protease, a digestive enzyme produced in the pancreas. Gene duplications happen all the time, and they are generally of such little importance that they are known as "neutral" mutations, having no effect on an organism's fittness. However, the original gene had a control region that switched it on only in the pancreas. During the duplication, the control region of the duplicate is damaged so that the new gene is switched on in both the pancreas and the liver. As a result, the inactive form of the enzyme, a zymogen, is relesased into the bloodstream.

This causes no problem for the organism - most pancreatic proteases are inactive until a small piece near their active sites can be cut away by another protease. However, when damage to a blood vessel allows plasma to seep into tissue, suddenly our previously inactive plasma serine protease is activated by tissue proteases, increasing the overall protein-cutting activity at the site of the hemorrage. Blood clotting is enhanced, so our duplicate gene (with the mistargeted protein) is now favored by natural selection.

That plasma protease gene is now subject to the same witches' brew of copying errors, rearrangements, and genetic reshuffling that affect the genes for every other cellular protein. Over time, bits and pieces of other genes are accidentally spliced into the plasma protease sequence. Because the selective value of the plasma protease is pretty low (it doesn't help clotting all that much), most of these changes make very little difference. But one day, through a well-understood process called "exon shuffling," a DNA sequence known as an "EGF domain" is spliced into one end of the protease gene. EGF stands for epidermal growth factor, a small protein used by cells throughout the body to signal other cells. EGF is so common that just about every tissue cell has "receptors" for it. These receptors are cell surface proteins shaped in such a way that they bind EGF tightly.

The fortuitious combination of a EGF sequence with the plasma protease changes everything.

In a flash, the tissue surrouding a broken blood vessel is now teeming with receptors that bind to the new EGF sequence on our serum protease. As a result, high concentrations of the circulating protease bind directly to the surfaces of cells near a wound. The proteases are activated in the same way, but now their proteolytic activities are highly localized. The production of a clot of insoluble protein fragments is now faster and more specific than ever. Organisms with the new EGF-protease can clot their blood much more quickly than before, and therefore are favored by natural selection. To emphasize its role in the clotting process, that cell surface protein with the EGF receptor is called Tissue Factor.

What happens next? Well, remember the case of the lobster in which a duplicate of a circulating protein (vitellogenin) became specialized to produce a clot-forming protein (lobster fibrinogen)? Once we have a situation in which every hemorrage activates a protease bound to tissue receptors, a gene duplicate of one of the major plasma proteins would then be under strong selective pressure to increase its ability to interact with the bound protease. Fibrinogen, the soluble protein that now is now the primary target of proteolysis in the clotting cascade, clearly arose in this way. Natural selection would favor each and every mutation or rearrangement that increased the sensitivity of fibrinogen to the plasma protease, dramatically enhancing the ability of the new protease to form specific clots of insoluble protein.

There is no doubt that these three steps, each one supported by classic Darwinian mechanisms, would have been sufficient to fashion a rudimentary clotting system. This would leave us with system in which circulating plasma contains both an inactive serine protease and its fibrinogen target. The protease would activated by contact with tissue factor, and the active protease, in turn, would cleave sensitive sites in fibrinogen to form a clot. This system wouldn't be nearly as quick, as responsive, or as sensitive as the current system of vertebrate clotting, but it would work a little better than the system that preceeded it, and that's all that evolution requires.

Adding Complexity

Could evolution take this rudimentary system and produce a multilayered cascade of factors? Just watch. Most serine proteases, including trypsin and thrombin, are auto-catalytic. That means that some extent they can activate themselves, in many cases by cleaving a few amino acids to switch on their active sites. So, we could diagram the actual functions of our ancestral plasma protease (which we'll call protease A) like this:

As we have seen, the inactive form of the protease (A) is changed into the active form (A*) when two things happen: it is bound to tissue factor (TF) and it is activated by tissue proteases, including our protease itself (that's the autocatalytic part). This means - and this is important - that our protease is actually involved in cutting two things: Fibrinogen, and also itself, converting A's inactive precursor protein into A*.

Now, let's suppose that a gene duplication occurs in the gene for our protease, producing a new (B) version of the gene:

At first, just like most gene duplications, this is no big deal. Proteins A and B are identical. Each can bind to TF, each can cleave fibrinogen into fibrin, and each can activate itself or its sister serum protease. So nothing has really changed - we've just got two copies of the same gene. But now let's suppose that a mutation in the active site of B changes its behavior, making it a little less likely to cut fibrinogen and a little more likely to activate protease A. In essence, this would change the relationship between these previously duplicate genes to something like this:

Suddenly, the ability of A to bind to TF becomes much less important. If B can saturate all of the available TF-binding sites itself (by virtue of its EGF domain), then the TF-mediated activation of B, combined with B's affinity for A, will result in a rapid activation of A, producing plenty of activated A to convert fibrinogen into clottable fibrin. Sounds good. But why would natural selection favor a mutation like this in B's active site? Simple: it would increase the efficiency of the clotting process by producing a 2-level cascade. Look closely, and you'll see that our 1-step clotting system required a direct interaction with TF to activate each protease. The new 2-step system allows each TF to activate a protease B, each of which in turn can activate scores or hundreds of A's. With so many more active proteases in the neighborhood of the injury, clotting can now occur more quickly, increasing the chances of surviving a hemorrage. Exactly the sort of stuff that natural selection favors.

Step back for a second and think about what we've just seen. A simple gene duplication sets the stage for the selection of active site mutations that would dramatically improve the clotting process. Gene duplications are neutral mutations, the sort that occur all the time and therefore, given enough time, are highly probable. Once the duplication has taken place, any mutation in the active site that shifts the preferences of the active site in the direction I have mentioned will be strongly favored. And that means that a true 2-step system will evolve very quickly.

Two additional points have to be mentioned. The first one is obvious. If gene duplication and subsequent mutation of the duplicate protease can change a 1-step system into a 2-step one, they could certainly change a 2-step system into a 3-step one. This means that increases in biochemical complexity are not only accomodated by evolutionary theory, they are actually predicted by it. The second point is a little more subtle. Early stages in the evolution of a clot forming-system are bound not to work very well. But as the system starts to work better, as it increases in complexity and efficiency, it begins to present a danger to the organism. That danger, simply put, is that clotting might get out of hand. As the clot-forming cascade evolves larger and larger, there is a chance that a small stimulus will start a reaction that might cause all of an organism's blood to clot, or at least enough of it to cause serious problems. Does evolution have an answer to that, too?

Well, it turns out that it does. First, keep in mind that a primitive clotting system, adequate for an animal with low blood pressure and minimal blood flow, doesn't have the clotting capacity to present this kind of a threat. But just as soon as the occasional clot becomes large enough to present health risks, natural selection would favor the evolution of systems to keep clot formation in check. And where would these systems come from? From pre-existing proteins, of course, duplicated and modified. The tissues of the body produce a protein known as a1-antitrypsin which binds to the active site of serine proteases found in tissues and keeps them in check. So, just as soon as clotting systems became strong enough, gene duplication would have presented natural selection with a working protease inhibitor that could then evolve into antithrombin, a similar inhibitor that today blocks the action of the primary fibrinogen-cleaving protease, thrombin.

In similar fashion, plasminogen, the precursor to a powerful clot-dissolving protein now found in plasma, would have been generated from duplicates of existing protease genes, just as soon as it became advantageous to develop clot-dissolving capability.

In short, the key to understanding the evolution of blood clotting is to appreciate that the current system did not evolve all at once. Like all biochemical systems, it evolved from genes and proteins that originally served different purposes. The powerful opportunistic pressures of natural selection progressively recruited one gene duplication after another, gradually fashioning a system in which high efficiences of controlled blood clotting made the modern vertebrate circulatory system possible.


Mining the Biochemical Past

Can we know for sure that this is how blood clotting (or any other biochemical system) evolved? The strict answer, of course, is we cannot. The best we can hope from our vertebrate ancestors are fossils that preserve bits and pieces of their form and structure, and it might seem that their biochemistry would be lost forever. But that's not quite true. Today's organisms are the descendents of that biological (and biochemical) past, and they provide a perfect opportunity to test these ideas.

Even a general scheme, like the one I've just presented, leads to a number of very specific predictions, each of which can be tested. First, the scheme itself is based on the use of well-known biochemical clues. For example, most of the enzymes involved in clotting are serine proteases, protein-cutting enzymes so-named because of the presence of a highly reactive serine in their active sites, the business ends of the protein. Now, what organ produces lots of serine proteases? The pancreas, of course, which releases serine proteases to help digest food. The pancreas, as it turns out, shares a common embryonic origin with another organ: the liver. And, not surprisingly, all of the clotting proteases are made in the liver. So, to "get" a masked protease into the serum all we'd need is a gene duplication that is turned on in the pancreas' "sister" organ. Simple, reasonable, and supported by the evidence.

Next, if the clotting cascade really evolved the way I have suggested, the the clotting enzymes would have to be near-duplicates of a pancreatic enzyme and of each other. As it turns out, they are. Not only is thrombin homologous to trypsin, a pancreatic serine protease, but the 5 clotting proteases (prothrombin and Factors X, IX, XI, and VII) share extensive homology as well. This is consistent, of course, with the notion that they were formed by gene duplication, just as suggested. But there is more to it than that. We could take one organism, humans for example, and construct a branching "tree" based on the relative degrees of similarity and difference between each of the five clotting proteases. Now, if the gene duplications that produced the clotting cascade occurred long ago in an ancestral vertebrate, we should be able to take any other vertebrate and construct a similar tree in which the relationships between the five clotting proteases match the relationships between the human proteases. This is a powerful test for our little scheme because it requires that sequences still undiscovered should match a particular pattern. And, as anyone knows who has followed the work in Doolittle's lab over the years, it is also a test that evolution passes in one organism after another.

There are many other tests and predictions that can be imposed on the scheme as well, but one of the boldest was made by Doolittle himself more than a decade ago. If the modern fibrinogen gene really was recruited from a duplicated ancestral gene, one that had nothing to do with blood clotting, then we ought to be able to find a fibrinogen-like gene in an animal that does not possess the vertebrate clotting pathway. In other words, we ought to be able to find a non-clotting fibrinogen protein in an invertebrate. That's a mighty bold prediction, because if it could not be found, it would cast Doolittle's whole evolutionary scheme into doubt.

Not to worry. In 1990, Xun Yu and Doolittle won their own bet, finding a fibrinogen-like sequence in the sea cucumber, an echinoderm. The vertebrate fibrinogen gene, just like genes for the other proteins of the clotting sequence, was formed by the duplication and modification of pre-existing genes.

Now, it would not be fair, just because we have presented a realistic evolutionary scheme, supported by gene sequences from modern organisms, to suggest that we now know exactly how the clotting system has evolved. That would be making far too much of our limited ability to reconstruct the details of the past. But nonetheless, there is little doubt that we do know enough to develop a plausible and scientifically valid scenario for how it might have evolved. And that scenario makes specific predictions that can be tested and verified against the evidence.


Doolittle, R. F., (1993) "The evolution of vertebrate blood coagulation: A case of yin and yang," Thrombosis and Haemostasis 70: 24-28.

Doolittle, R. F., and Feng, D. F., (1987) "Reconstructing the evolution of vertebrate blood coagulation from a consideration of the amino acid sequences of clotting proteins," Cold Spring Harbor Symposia on Quantitative Biology 52: 869-874.

Doolittle, R. F., and Riley, M. (1990) "The amino-acid sequence of lobster fibriongen reveals common ancestry with vitellogenin." Biochemical and Biophysical Research Communications. 167: 16-19.

Xu, X., and Doolittle, R. F., (1990) "Presence of a vertebrate fibrinogen-like sequence in an echinoderm." Proceedings of the National Academy of Sciences (USA) 87: 2097-2101.