table of contents

A Scenario for the Evolution of Hemostasis
by Kevin O'Brien

This is a revised version of an article that originally appeared in the talk.origins newsgroup. It was composed in response to the following challenge:

Let's start with the clotting system. Tell us all how the IC core, consisting of fibrinogen, prothrombin, Stuart factor, and accelerin, could have evolved.

It is granted that prothrombin and the Stuart factor may have evolved from a common ancestor A by duplication and divergence, if A was autocatalytic.

References specific to talk.origins have been removed and slight editorial changes have been made. - Richard Harter, editor.

First of all, some ground rules.

Remember Behe's own definition of IC? If it evolved then it isn't IC. Remember your own statements concerning IC and function? The core may be IC now, but if its components served other non-IC functions before they became co-opted into the blood clotting system, they could still have evolved "piecemeal".


Let's establish some terminology. Except for fibrinogen/fibrin and prothrombin/thrombin, I will use the official names for each component, the factor names. After I use an official name for the first time, I will abbreviate it for simplicity. For example Factor X, also called the Stuart factor, will be given the abbreviation, fX. Similarly, Factor V, also called accelerin, will be given the abbreviation, fV, and so on.

With two exceptions, all of these factors are enzymes, specifically proteases. That means they attack other proteins, trying to cleave the peptide bond. Some proteases chew proteins to pieces, while others - like the serine proteases that make up the hemostatic pathways - are more discrete, only make one or a few cuts. Like all enzymes, proteases possess an active site where the reaction occurs.

Proteases come in two forms, an active and an inactive form. It is a standard convention to refer to inactive enzymes as zymogens and to use "pro-" as a prefix. In this article inactive forms will be given the prefix "pro-" and active forms the suffix "a". Thus pro-fX signifies pro-Factor X, i.e., the Factor X zymogen. Likewise the active form of Factor X would be fXa.

Although Factor V and VIII are not enzymes, they also come in an active and inactive forms. While I will not refer to their inactive forms as zymogens, the standard convention still requires the use of the "pro-" prefix and the "a" suffix.

For all the coagulation proteins, activation is achieved by proteolytic cleavage. That is, some part of the zymogen protein structure is cut out and removed by another protease, or a single cut is made that allows the zymogen to refold itself into a new shape. Either way, the active site of the zymogen is hidden away and must be exposed before the enzyme can become active.

It should be pointed out, however, that with a few exceptions protease zymogens still have some low level of activity, so they are never completely inactivated.

Modification of the IC core

Finally I would add a new component to your "IC core": activated protein C (APC). This protein is another serine protease and it attacks Factor V; in essence it's another regulatory protein in the central blood clotting system that prevents too much fV being activated, so that too much thrombin isn't made.

The Ancestral Factor A

It is now generally accepted as a fact by molecular biologists that all the serine proteases involved in the blood clotting system evolved from a common ancestor by gene duplication and subsequent differentiation. They have even reconstructed the DNA sequence of the hypothetical common ancestor, using the DNA sequences of Factors VII, IX and X, APC and prothrombin. In turn the reconstructed ancestral sequence has been used as a basis for attempting to determine when the factors diverged [Krawczak et al. 1996]. (This sequence may be wrong, however, because it is contradicted by later analysis, which I will cite later.) Though I will not discuss this study in more detail, I will have more to say about this subject later. One thing I can say now, however, is that there is not just one common ancestor for prothrombin and Stuart Factor, but several, at least six of which come directly between prothrombin and Stuart Factor.

Overview of the hemostatic (clotting) process

The hemostatic (clotting) process has are four phases.

In phase 1 biomolecules released by the damaged tissue cause the blood vessel to constrict so as to diminish blood flow to the wound site.

In phase 2 other molecules cause platelets, which are covered with sticky cell adhesion molecules, to bind to the wound to form a temporary plug called a white thrombus. This also serves as a foundation for laying down collagen to form a new basement membrane. However, the plug is delicate and many platelets break under the strain, releasing yet more biomolecules directly into the wound site itself.

In phase 3 the actual clotting process is initiated to build the fibrin network that will trap red blood cells, creating a red thrombus.

Finally, phase 4 involves dissolving the clot after the wound has healed.

In our discussion we will concentrate on events in phase 3. This phase starts in two separate ways, known as the extrinsic and intrinsic pathways.

The extrinsic pathway

In the extrinsic pathway (which most biochemists consider nowadays to be the more important of the two) Factor VII zymogen is converted to active Factor VII by proteolytic cleavage. fVII is a serine protease that is capable of autoactivation. pro-fVII has unusually high activity for a zymogen, while fVIIa has unusually low activity for a protease [Dickinson et al. 1998]. What this means is that approximately 1% of the circulating fVII is active [Kalafatis et al. 1997], because both the zymogen and the active enzyme can activate pro-fVII, but since both have (relative to other active serine proteases) fairly low activity, only a tiny proportion of pro-fVII becomes activated. This aids in regulation, because fVIIa activates the Factor IX and Factor X zymogens, which in turn convert prothrombin to thrombin (directly or indirectly), so the less fVIIa there is around, the less likely a clotting event will be started by accident.

At the same time, there needs to be enough fVIIa around to quickly initiate clotting when necessary. When damage occurs, tissue cells that are still intact find themselves exposed to blood flow. Many of these cells possess a membrane protein called tissue factor. Tissue factor has a number of different functions [Carmeliet and Collen 1998], but in this case it serves as a binding site for both pro-fVII and fVIIa [McVey 1994]. Binding the zymogen or active protease to these receptors changes their shape in such a way that their activity is greatly enhanced [Higashi and Iwanaga 1998]. This allows them to autoactivate large amounts of zymogen, mostly through interactions between bond proteases instead of with plasma zymogens [Neuenschwander et al. 1993]. It also enhances their recognition of and activity for activating Factor IX and Factor X zymogens. However, they do not remain bound for long and when they dissociate they loose their enhanced activity. This acts as a further level of regulatory control.

The intrinsic pathway

The second pathway (now considered to be less important) is the intrinsic pathway. It begins with the proteolytic activation of Factor XII zymogen. Another protease also capable of autoactivation, pro-fXII must bind to a cell surface to be activated, either platelets or vascular endothelial cells [Tans and Rosing 1987]. (Some researchers believe they have identified the membrane protein the factor binds to, but their claims have yet to be finally verified [Joseph et al. 1996].) In binding, the zymogen undergoes a conformational change making it easier for another surface-bound fXIIa or some other plasma protease to activate it. Surface-bound fXIIa then in turn activates the Factor XI zymogen, which must also bind to the cell surface. pro-fXI needs the cofactor kininogen to bind to a cell surface; kininogen itself binds to pro-fXI. At the same time, surface-bound fXIIa also converts the zymogen prekallikrein to active kallik rein, a protease that can in turn convert surface-bound pro-fXII to surface-bound fXIIa, which in turn converts more prekallikrein, and so on. This process is called the reciprocal activation mechanism and it seems to be more important for activating pro-fXII than fXIIa alone is. Prekallikrein also needs to be bound to the cell surface to be activated and also needs kininogen to bind to that surface.

As an interesting side-light, prothrombin competes with kininogen to bind to pro-fXI [Baglia and Walsh 1998]. If it succeeds, the complex still needs to bind to a platelet cell surface, but it is then activated by thrombin instead of surface-bound fXIIa. In this way, once thrombin is activated, the entire reciprocal activation mechanism can be heavily supplemented.

When activated, fXIa uncouples itself from the cell surface and kininogen, and become a plasma enzyme. Another serine protease, it activates Factor IX zymogen, which is also a serine protease. Also known as the Christmas factor, fIXa in turn activates the Factor X zymogen. At first it can do this alone, but later when sufficient amounts of thrombin have been created, thrombin begins activating Factor VIII. This protein is not an enzyme, but in its inactive form its binding site is covered, so it must be proteolytically activated to uncover that site. fVIIIa in turn binds to sites on platelets and vascular endothelial cells, where it binds up fIXa [Brinkman et al. 1997]. This concentrates fIXa where it is needed the most, where it can activate large amounts of pro-fX. It is also believed to enhance the activity of fIXa [Kalafatis et al. 1997].

Once activated, fXa begins converting prothrombin to thrombin. It's at this point that things get interesting. The main function of thrombin is to cleave small peptides off the ends of soluble fibrinogen molecules to create insoluble fibrin. These strands begin to adhere to the platelets, forming the fibrin plug, which in turn traps red blood cells. Another enzyme called fibrinase then creates cross-linkages between the strands, forming a hard clot. However, thrombin also activates pro-fVIII, which accelerates the activity of fIXa; thrombin also activates pro-fXI, thereby supplementing that which is produced by the reciprocal activation mechanism. Thrombin also activates Factor V, a protein like fVIII that has to be proteolytically activated, but is not itself an enzyme. fVa then binds to receptors on platelets and vascular endothelial cells, where it binds up fXa, concentrating it in a limited area and enhancing its activity [Rosing and Tans 1997]. fVa can in fact increase the conversion of prothrombin to thrombin more than 10,000 fold.

But it doesn't stop there. Thrombin can also destroy fVa and it binds to the endothelial cell receptor thrombomodulin, where it initiates the protein C pathway, which in turn leads to the formation of activated protein C (APC), itself a serine protease. APC inactivates fVa and to a lesser extent fVIIIa, thus destroying the fVa/fXa and fVIIIa/fIXa complexes. fVIIIa is also unstable and tends to break down easily, releasing fIXa in the process. These actions depress the activation of thrombin, either directly or through the down-regulation of fXa activation. Despite all these different affects, however, they all work together in a negative feedback loop, where effects tend to smooth themselves out and reach an equilibrium where they are neither too fast nor too slow. The result is a system that can respond rapidly to injury, creating large amounts of needed material very quickly, yet then settles down to a constant level of production that can in turn be shut off equally quickly. Though not perfect, it is nonetheless an excellent system.

Complexity of the hemostatic system

I described the hemostatic system in this much detail because I wanted to impress upon everyone just how complex this system is. It isn't just limited to the four molecules mentioned in the challenge; the whole system is so interdependent that deficiency in most areas leads to devastating diseases. For example, the inability to make fVIII leads to hemophilia type A [Martin 1983, pg 596], whereas the inability of APC to deactivate fVa leads to thrombophilia [Rosing and Tans 1997]. By Behe's own definition, this whole system -- not just the core -- could be IC.

Evolution of Hemostasis - Initial Assumptions

So how could such a system develop piecemeal slowly over time? In essence by taking advantage of three basic mechanisms: gene duplication, adaptation of an old component to a new function, and natural selection of mutagenetically altered genes. Let me try to explain in more detail as I go along.

Every evolutionary scenario has to make certain assumptions, otherwise every scenario has to go back to abiogenesis for its starting point. I will try to keep the number of assumptions to a minimum (and I will indicate future ones with asterisks), but one I must start with is that chordates evolved from invertebrates.

Hemostatis in Invertebrates

If this is true, then we would expect the common ancestor of all chordates to have, with some modification, the basic invertebrate body plan. This includes a central body cavity filled with a fluid called hemolymph. Hemolymph is very similar to blood, but it has significant differences as well. For example, the oxygen carrier in hemolymph uses copper instead of iron. Since the organs are suspended in this fluid, it bathes and saturates them, insuring that their tissues get their fair share of oxygen and nutrients. There is no need for a closed circulatory system, because the invertebrate's body movements circulate the fluid. Aiding in the circulation is a single artery running the length of the body cavity, usually on the dorsal side. It is open at both ends and has openings along its length. At the end closest to the head it thickens into a primitive one-chamber heart. The heart pumps, drawing in fluid and forcing it out again, helping to keep it in circulation.

Because of this, invertebrates are not as susceptible to thrombosis as vertebrates are with their closed circulatory systems. Clots that try to form in the artery are flushed out before they have a chance to anchor themselves. Even so, the artery can still function even if it is blocked. Besides, since body movement is the main method for circulating the hemolymph, a non-functional artery would barely impair an invertebrate. Clots that form in the body cavity itself would be destroyed by macrophages. So invertebrates are less concerned with thrombosis and so their hemostatic system should have fewer regulatory pathways. In point of fact, invertebrate hemostasis can be complex [Ghidalia et al. 1989], but the first primitive invertebrates could have gotten by with a much simpler system [Zimmermann 1983]. All they would need is a protein to form insoluble fibers to fill the damaged area, enzymes to initiate fiber formation and to cross-link the fibers, and scavenger cells to clean up after the clot is formed and remove the clot after the damage has healed.

It is generally believed that sponges (phylum Porifera) are an excellent model for studying the organization and evolution of metazoans [Muller 1998]. Of all the systems they do share in common with vertebrates and invertebrates alike, what is interesting is that they do not have a hemostatic system. Which makes sense, since they have neither blood nor hemolymph. This would indicate that hemostasis was invented in invertebrates and passed on to vertebrates. However, they do have scavenger cells, called amoebocytes, which are also present, in one form or another, in all known invertebrates. They also proteases, including serine proteases. As such, it is reasonable to **assume** that these enzymes and cells were present in ancestral metazoans, and that when invertebrates split off from those metazoans they retained these cells and proteases. They were probably used initially to defend the organism from attack by bacteria: the amoebocytes would eat them and the proteases would digest them. In fact, fibrinogen was probably developed first to take advantage of the fact that lipopolysaccharides found on the membranes of gram-negative bacteria, can actually activate certain serine protease zymogens in arthropods. These active proteases could then convert fibrinogen into fibrin, which would have engulfed the bacteria, isolating and immobilizing them so that they may be more easily digested [Iwanaga and Kawabata 1998]. Only by accident would it subsequently be discovered that this same system could also plug damage to the body cavity and thus prevent the loss of hemolymph until the damage could be repaired. As such, this is an example of the second of the three mechanisms involved in the evolution of hemostasis: adaptation of old components to new functions. If fibrinogen was developed to form clots around bacteria, this same function would serve equally well to plug damaged tissue.

The ultimate origin of fibrinogen remains obscure. However, we know that, compared to cytochrome C or collagen, it is a phylogenetically young protein [Zimmermann 1983]. We also know that it shares structural similarities with other clotting proteins, such as casein (milk protein) which is the protein responsible for milk clotting [Jolles et al. 1978]. Therefore it is likely that it was produced either by mutagenesis of an existing gene or by the creation of a new gene by gene duplication or exon shuffling. Gene duplication is when the whole gene sequence -- exons, introns, pseudogenes, regulatory sites, etc. -- is carbon copied and transferred to another location on the same chromosome or a different chromosome. I'll have more to say about this later. Proteins created in this process show clear phylogenetic relationships between each other, but genes created by exon shuffling do not. This process involves the recombination of exons from different genes mediated by the introns [Cooper 1997, pg 140]. Since each exon encodes a single functionally distinct protein domain, this process results in chimeric genes that possess different functions from different proteins. These disparate functions can either combine into a single new function or coordinate so that the protein can perform several different functions simultaneously. There are examples of these genes in nature (see for example Collet and Candy 1998), including some hemostasis serine protease factors [Citarella et al. 1996], and we are just beginning to use this process to create new protein-based therapeutics in the drug industry [Arnold 1998]. Since the structural similarities between fibrinogen and other clotting proteins do not seem to be phylogenetic, it is more reasonable to **assume** that fibrinogen formed by exon shuffling between one or more of these clotting protein genes and one or more structural protein genes, than to invoke gene duplication.

We also know that, despite its phylogenetic youth, fibrinogen is over 1.5 billion years old [Ghidalia 1989]. Structurally it was probably more similar to invertebrate fibrinogen (called coagulogen [Zimmermann 1983; Iwanaga and Kawabata 1998]) than to vertebrate fibrinogen; we know that coagulogen and fibrinogen split off from each other a little more than 1.5 billion years ago [Ghidalia 1989], probably at the same time the common chordate ancestor split off from the invertebrates. Later, about one billion years ago, fibrinogen (in chordates) diversified into three different chain types, this time by gene duplication [Doolittle 1976; Crabtree et al. 1985]. After this point, fibrinogen's role as part of innate immunity waned as the chordate immune system developed, but it's role as a repair protein became more important as a closed circulatory system developed. Finally, the three chains diversified into the modern alpha, beta and gamma chains, mostly by the loss of introns [Crabtree et al. 1985].

The ultimate origin of the serine proteases is also obscure; whether they all derive from a single common ancestor or (as I personally believe) they stem from several structurally similar enzymes produced by exon shuffling, or whether some other explanation will turn out to be true, is as yet uncertain. (Iwanaga and Kawabata [1998], however, present evidence that they might -- with some exceptions like subtilisin -- all derive from a single common ancestor.) However, as has already been mentioned, it is an established fact that vertebrate serine proteases in general are descended from a single common ancestor [Iwanaga and Kawabata 1998; see for a description and phylogenetic tree, or cladogram]. The only question that remains is how did the specific hemostatic protease factors evolve? Like everything else they evolved by a combination of chance and natural selection; chance that the mutagenic events that caused the diversification of the duplicated gene would create a needed function, and then the selection of that function to be retained. As long as the different stages of the modern vertebrate hemostatic system could have been managed by a fewer number of more general proteases (and that seems to be the case), then any duplication and diversification which created factors that were more specific for certain steps would actually be improving the process rather than impairing it, whereas those that had less specificity or even no specificity would do the process no harm.

Ancestral Biochemicals

For example, Iwanaga and Kawabata [1998] show that APC and thrombin shared a common ancestor (which I will call Common Ancestor 9a or ca9a), which makes sense because both can deactivate fV, so it is highly likely that fV (or its common ancestor) was around when ca9a was around. They also show that ca9a shared a common ancestor with Factor C, the initiator of arthropod hemolymph coagulation (ca8a). Meanwhile, fX and fIX appear to share a common ancestor (ca10a), which also makes sense in that fVII can activate both. However, since there is no evidence that fIX can activate thrombin directly, this would indicate that this gene copy lost that ability while retaining its "autoactivation" capability, only changed to recognize fX rather than itself. ca8a then seems to share a common ancestor with fVII (ca9c), which again makes sense since fVII can activate both, being as it previous activated their common ancestor; this might also be a true case of retained autoactivation that recognizes all zymogens of the same immediate family because of that family's shared ancestry. Meanwhile, ca9c itself shares a common ancestor (ca8b) with a group of compliment serine proteases (indicating that some factors of the immune system are related to most factors of the hemostatic system), and ca8b in turn shares a common ancestor with ca8a (ca7a). Since this serine protease serves as the common ancestor for fX (the Stuart Factor) and thrombin (among others), this enzyme would correspond to Factor A, the hypothetical common ancestor for fX and thrombin. Finally, ca7a in turn shares a common ancestor (ca6a) with a family of complement inactivaters.

Meanwhile, fXI and kallikrein share a common ancestor (ca7c), which makes sense since they are both activated by fXII. ca7c in turn shares a common ancestor (ca6b) with the family of acrisomal proteases. And ca6b shares a common ancestor (ca5c) with ca6a itself.

In yet another extended family, fXII shares a common ancestor (ca7e) with the whole family of plasminogen activators. ca7e then shares a common ancestor (ca6c) with the trypsin super-family. ca6c in turn shares a common ancestor (ca5d) with the tryptase family. Finally, ca5d shares a common ancestor (ca4c) with ca5c.

ca4c shares a common ancestor (ca3c) with a family of trypsin-like enzymes; ca3c then shares a common ancestor (ca2b) with another family of trypsin-like enzymes.

Meanwhile, two of the other arthropod coagulation factors -- Factor B and preclotting enzyme -- share a common ancestor (ca5a), which in turn shares a common ancestor (ca4a) with the last known arthropod coagulation protein, Factor G. ca4a then shares a common ancestor (ca3a) with the snake-locus serine protease of Drosophila melanogaster, which in turn shares a common ancestor (ca2a) with the Easter serine protease also of Drosophila melanogaster. Finally, ca2a shares a common ancestor (ca1) with ca2b.

The Evolutionary Scenario

This may sound somewhat confusing at first, but it should become clearer once I describe my scenario. Also, I chose this form of designation on purpose. The letters after the numbers are of no importance, but the numbers indicate the number of generations between the first common ancestor and the final protease. If ca1 represents the first generation, then thrombin would be a tenth generation protease -- nine common ancestors (including ca1) plus thrombin itself. fIX and fX are eleventh generation proteases, whereas fXII is only an eighth generation protease. In and of itself, this does not establish relative ages, but it does give some idea of the order in which these enzymes were developed. More important is the spacing between the various common ancestors on the phylogenetic tree; the greater the distance the longer the amount of time between divergent events. As such, while I cannot assign exact dates to each common ancestor (Iwanaga and Kawabata [1998] does not provide that kind of data) I can say in a relative sense when each protease appeared.

It could also be pointed out that there are two ways of interpreting cladograms. The conservative, or "purist", interpretation is that cladograms represent periods of stasis punctuated by rapid divergent events. As such, the horizontal lines would represent the lifespan of the protease in question, while the vertical lines would represent rapid divergence events into two new and different proteases. One problem with this interpretation is that it ignores mutagenic events that occur in the horizontal lines that could produce significant changes over hundreds of millions of years, thus creating a different protease from what you started with. Another problem is that, when information is scarce, some proteases can seem to have appeared much earlier than they probably did, simply because we do not know about any divergent events between the protease in question and its last known common ancestor. Another problem is that this interpretation assumes that evolution occurs almost exclusively cladogenetically, with almost no anagenetic evolution in the horizontal lines. Of course, the purists claim that cladograms should not be used to determine when and how proteases diverged, but only to establish their phylogenetic relationships. Still, it is because of these and other problems, and the recognition that properly detailed cladograms can in fact tell us when proteases diverge, that leads me to adopt the second, more liberal, interpretation. This claims that the horizontal lines represent a lineage slowly evolving from one nearly identical protease to the next by the slow accumulation of mutations, while the vertical lines represent points where the protease before it splits by gene duplication into two identical copies, that then evolved along separate lines. As such, the horizontal lines immediately after the split represent two different populations of the same protease species, which then gradually diverge because of different mutagenic events. For this essay that is how I will interpret the cladogram provided by Iwanaga and Kawabata [1998]. Later when I detail my scenario, I will use the word "split" to indicate when a common ancestor undergoes gene duplication and the word "diverging" to indicate mutagenesis along the horizontal lines.

All that leaves then is where did fV and fVIII come from? It is known that fV and fVIII have homologous sequences, so they are believed to be descended from a common ancestor [Church et al. 1984]; this is further supported by the fact that APC is able to deactivate both fV and fVIII because of shared sequences in what are called the terminal A domains [Walker and Fay 1990]. However, the factors also share sequence homology with ceruloplasmin [Church et al. 1984], a blood protein that binds up copper. It is not a transport protein, because once it binds copper it doesn't let it go; its purpose is to sequester copper to prevent copper toxicity. This would have been a very valuable protein in invertebrate hemolymph. In any event, APC also binds to ceruloplasmin, which shares 60% and 40% homology in the APC binding sequences of fVIII and fV, respectively [Walker and Fay 1990]. This suggests that the common ancestor of these three proteins gave rise to ceruloplasmin first, then fVIII and finally fV. Finally, structural studies of the A domains of the three proteins have all but proven the phylogenetic relationship of these proteins. They demonstrate that the factors create a triangular arrangement with their three A domains just as in ceruloplasmin, the factors bind copper in their A domains just as in ceruloplasmin, and the cleavage sites for APC on the factors are solvent exposed as in ceruloplasmin, along with other similarities [Villoutreix and Dahlback 1998].

So with the background information out of the way, let me present my scenario without further ado. It begins at some point during metazoan evolution, when multicellular organisms had not yet developed sophisticated body plans, but were further along than simply being a collection of cells. At this point the variety of serine proteases was low. One of them was ca1. At a certain point in time it split. One lineage diverged to ca2b first, which then split; the other to ca2a very soon after, which also split. One lineage of ca2b diverged into ca3c, which then split. This evolution almost certainly went hand-in-hand with continued metazoan evolution. One of the ca3c lineages diverging to ca4b, which was the common ancestor for one family of trypsin-like enzymes. Meanwhile, one of the ca2a lineages diverged to ca3a; the other diverged into the Easter serine protease of Drosophila melanogaster. The second of the ca3c lineages then diversified into ca4c, which then split. One of these lineages very quickly diverged into ca5d, which then split. Meanwhile, the second ca2b lineage diversified into ca3b, which was the common ancestor of another family of trypsin-like enzymes. The second lineage of ca4c diversified into ca5c, which then split, while at nearly the same time one lineage from ca5d diverged into ca6c, which also split.

These events occurred rather quickly, probably within 200 million years. Though a couple of minor families have been created as well, it is at this point or just before that the common ancestors of the three main groups of serine proteases have been created. ca3a will eventually form the bulk of the invertebrate coagulation system, while ca5c would generate the hemostatic and compliment mega-family and ca6c would generate the trypsin mega-family. This probably represents a stage in metazoan development when different body plans, along with specific tissue types, are being "experimented" with. Amoebocytes may or may not have been created by this time, but fibrinogen was still some ways off. ca3a, 5c and 6c, along with ca3b and ca4b, were probably carrying out specific, if broadly-based, functions in certain tissues. Whether by accident or design, ca5c may have ended up in the tissue that would eventually form the circulatory system.

Another 100 million years would see even more diversification. One lineage of ca6c diverged into ca7f, which was the common ancestor of the trypsin and elastase super-families, while one lineage of ca5c diverged into ca6a, then split. The second lineage of ca5c then diverged into ca6b, which split, while one lineage of ca3a diverged into ca4a, which also split. The second lineage of ca3a diverged into the snake-locus serine protease of Drosophila melanogaster. One lineage of ca6b diverged into ca7d, which was the common ancestor of the acrosomal family of proteases. Finally, one lineage of ca6a diverged into ca7a (Factor A), which split.

At this point or just before it the stage was set for the diversification of invertebrates from the rest of the metazoans. Though perhaps not fully developed, their basic body plan was established. One of the very few other metazoan body plans that would also survive from this period would be that of the sponges. Amoebocytes definitely would exist by now, since they are found in both Porifera and invertebrates. The proto-invertebrates would also have ca4a, 7a and 6b, at least, to serve their metabolic needs. However, fibrinogen probably was not yet around.

Next apparently the second lineage of ca6c diverged into ca7e, which split, while very quickly afterward one lineage of ca7a diverged into ca8a as the second diverged into ca8b, both of which split. At this point or shortly before chordates probably split off from the invertebrates. In invertebrates, one lineage of ca4a would diverge very shortly after this into ca5a, which was the common ancestor of proclotting enzyme and Factor B, while the other lineage would diverge into Factor G, all several hundred millions years from now. However they seem to appear within a few million years of each other (in the order given above), suggesting that at that time evolution may have been actively selecting for a more complex hemolymph coagulation system. If this is true, then it is interesting that Factor C, which would be diverge from one lineage of ca8a, appears somewhat later than these other three. This may have something to do with its combined coagulation/compliment fixation activities; Iwanaga and Kawabata [1998] do suggest that Factor C is a recent attempt in arthropods to form a protein that acts as both. Perhaps another hundred million years of evolution will see the separation of coagulation and compliment fixation in arthropods, if this trend continues. Nonetheless, as this point in my scenario invertebrates would be limited to ca4a to initiate coagulation, which would also mean that fibrinogen in the form of coagulogen was now present as a hemolymph protein. ca8a may have also been involved, either as a cofactor or as an alternative route to coagulation initiation. The rest of the common ancestors either went on to form other serine proteases that would not be involved in hemolymph coagulation or became extinct.

At this point we will leave the invertebrates and concentrate on the chordates. They would have initially retained virtually everything of their invertebrate ancestors (eg, body plan, metabolic systems, innate immunity system) because of how successfully everything worked together (the only difference being the permanent presence of a dorsal notochord even in the adult stage). As such, the proto-vertebrates would have also inherited amoebocytes for defense and general clean-up, fibrinogen for defense and damage repair, and serine proteases for general and specific metabolic activities, especially ca4a (which may have eventually gone extinct), ca8a and 8b, and ca7e. As their evolution continued in a different direction as that of the invertebrates, the first new protease to appear would be the divergence of ca9a (the common ancestor of thrombin and protein C) from the second lineage of ca8a. At this point, however, it may have replaced ca4a as the initiator of hemolymph coagulation for two reasons. The first is that it might have been more efficient; the second is that it might have been self-regulating, since it would probably have the ability to both activate its own zymogen (no need for lipopolysaccharides anymore) and to inactivate itself by destroying its own active form (most proteases can do both).

Next, a lineage of ca8b appears to have diverged into ca9c, which split. ca9c probably possessed two important activities: the first was the ability to enhance both its activation and activity while in contact with a cell surface (ca8b probably had this same activity, since the compliment proteins it also gave rise to have this activity as well); the other was the ability to activate ca9a (if ca7a was capable of autoactivation, this would have been passed on to its descendents, ca8a and 8b, which would in turn pass them on to their descendents, ca9a and 9c; the descendents of ca9a would eventually loose this capability as they became parts of regulated pathways, but the descendents of ca9c would have retained the capability as part of their essential function). ca9a's ability to autoactivate would have been hampered by the need for the free-swimming enzymes to find each other, so the formation of active ca9a would have been slow to build up. The presence of a protease that would bind to the area of damage itself, thus concentrating itself there, and which would have had its own activity enhanced by binding to the cell surface, would have allowed for a large amount of activated protease to be present right where it would be needed the most, hence enhancing the rate at which ca9a would be activated. Also the contemporary descendents of ca6b (the common ancestor of fXI and kallikrein) would be around as well. Since they would be closely related to ca7a they might still retain the ability to autoactivate its descendents, especially ca9c. ca6b probably had the ability to enhance its activity and its activation by binding to cell surfaces (its descendents have that ability), so like ca9c it could have concentrated in the area of damage and helped to activate even more ca9c. Compared to modern rates, however, the overall rate would still would have been slow, because the pro-ca9a would still have to find the activated ca9c bound to the surrounding cells. This may explain why the second phase of hemostasis is to block the site of damage with cells, so as to give the activation of the proteases enough time to build up enough overall activity to begin to clot the damage. Also, in the case of both ca9c and ca6b activity would drop off once the factor disassociated itself from the cell surface; if this happened quickly or easily, then this would also represent the first attempt to impose some sort of multi-level regulation on the whole system.

It would then appear that a lineage of ca7e diverged into ca8c. ca8c was the common ancestor of the plasminogen activator family, which has no known function in hemostasis, but the other lineage of ca7e would diverge into fXII. It seems that fXII appeared about 100 million years before even thrombin, making it one of the earliest hemostatic factors to appear. Its lineage probably possessed the ability to autoactivate in contact with cell surfaces; whether or not it also possessed the ability to activate any of the contemporary descendents of ca6b (the common ancestor of fXI and kallikrein) is uncertain. If it did, it is doubtful that it retained this ability as a form of "autoactivation" since it is a fourth generation protein from the last common ancestor shared with ca6b. This may simply be a case of fortunate chance, in which a dormant functional domain serving as an intron is mutagenetically reactivated, or maybe exon shuffling donated the necessary functional domain from another protease. We may never know. However, once it had that ability natural selection would have retained it as another means of more quickly activating large amounts of ca9a to form the hard clot faster.

(The second lineage of ca5d would have diverged into ca6d next, but it leads to the tryptase family and so will not be considered in this discussion.)

Next apparently the second lineage of ca8b diverged into ca9b, but this led to the C1 compliment protein family. While this shows that at least some of the compliment serine proteases are closely related to the hemostasis serine proteases, they will have no further part to play in this discussion. Soon afterwards, however, one lineage of ca9c diverged into ca10a, while the other diverged into fVII. Its linear ancestors almost certainly possessed or developed all the functions that it would itself possess, including the ability to activate ca10a, which is the common ancestor of fIX and fX. Later the second lineage of ca6b would diverge into ca7c, which was the common ancestor of fXI and kallikrein. (The second lineage of ca6a diverged into ca 7b, which was the common ancestor of the compliment inactivater family.)

From this point on there were apparently no more gene duplication events, though Iwanaga and Kawabata [1998] do not provide enough information to be certain. In any event, while the modern proteases had yet to appear, by this time their lineages were established. It was probably during this time that first fVIII and then fV diverged from ceruloplasmin. The lineages had pretty much all the activities and functions of their modern descendents, which meant that modern hemostasis was more or less in place. However, it still suffered from one major drawback: since fIX and fX were free-swimming, their concentrations were too diffuse for enough to be activated or to activate enough thrombin to respond quickly to damage (though they were much faster now than they had been in the past). The development of membrane-bound binding proteins that would immobilize fIX and fX and enhance their activity would solve that problem, making hemostasis more efficient.


In conclusion, it should now be clear (even if the details seem confusing) that even a system as complex as hemostasis could have evolved piecemeal by starting with a system already in place to perform a different function (innate immunity) that is not IC (there were many other mechanisms in place that were just as effective), co-opting it to perform a new function, then improving it by slowly adding new components that enhance both the efficiency and the regulation of the system, often by modifying the original components themselves, while at each stage of evolution a fully functional hemostatic system would already be in place. And as long as the new components were created by gene duplication, followed by subsequent mutagenesis controlled by natural selection, the system would not be in danger of being destroyed and could take advantage of the splitting or improving of functions that could result. The possibility that hemostasis might be IC now is no bar from its having evolved to its current state.


Arnold FH. "When blind is better: Protein design by evolution." Nature Biotechnology 1998; 16(7):617-8.

Baglia FA, Walsh PN. "Prothrombin is a cofactor for the binding of factor XI to the platelet surface and for platelet-mediated factor XI activation by thrombin." Biochemistry 1998; 37(8):2271-81.

Brinkman HJ, Koster P, Mertens K, van Mourik JA. "Dissimilar interaction of factor VIII with endothelial cells and lipid vesicles during factor X activation." Biochemical Journal 1997; 323 ( Pt 3):735-40.

Carmeliet P, Collen D. "Tissue factor." International Journal of Biochemical and Cell Biology 1998; 30(6):661-7.

Church WR, Jernigan RL, Toole J, Hewick RM, Knopf J, Knutson GJ, Nesheim ME, Mann KG, Fass DN. "Coagulation factors V and VIII and ceruloplasmin constitute a family of structurally related proteins." Proceedings of the National Academy of Sciences USA 1984; 81(22):6934-7.

Citarella F, Ravon DM, Pascucci B, Felici A, Fantoni A, Hack CE. "Structure/function analysis of human factor XII using recombinant deletion mutants. Evidence for an additional region involved in the binding to negatively charged surfaces." European Journal of Biochemistry 1996; 238(1):240-9.

Collet C, Candy J. "How many insulin-like growth factor binding proteins?" Molecular and Cellular Endocrinology 1998; 139(1-2):1-6.

Cooper GM. The Cell: A Molecular Approach 1997; Sunderland, Massachusetts:Sinauer Associates, Inc.

Crabtree GR, Comeau CM, Fowlkes DM, Fornace AJ Jr, Malley JD, Kant JA. "Evolution and structure of the fibrinogen genes. Random insertion of introns or selective loss?" Journal of Molecular Biology 1985;185(1):1-19.

Dickinson CD, Shobe J, Ruf W. "Influence of cofactor binding and active site occupancy on the conformation of the macromolecular substrate exosite of factor VIIa." Journal of Molecular Biology 1998; 277(4):959-71.

Doolittle RF. "The evolution of vertebrate fibrinogen." Federal Proceedings 1976; 35(10):2145-9.

Ghidalia W, Vendrely R, Montmory C, Coirault Y, Samama M, Lucet B, Bellay AM, Vergoz D. "Overall study of the in vitro plasma clotting system in an invertebrate, Liocarcinus puber (Crustacea Decapoda): considerations on the structure of the Crustacea plasma fibrinogen in relation to evolution." Journal of Invertebrate Pathology 1989; 53(2):197-205.

Higashi S, Iwanaga S. "Molecular interaction between factor VII and tissue factor." International Journal of Hematology 1998; 67(3):229-41.

Iwanaga S, Kawabata S. "Evolution and phylogeny of defense molecules associated with innate immunity in horseshoe crab." Frontiers in Bioscience 1998; 3:D973-84 [Available at Jolles P, Loucheux-Lefebvre MH, Henschen A. "Structural relatedness of kappa-casein and fibrinogen gamma-chain." Journal of Molecular Evolution 1978; 11(4):271-7.

Joseph K, Ghebrehiwet B, Peerschke EI, Reid KB, Kaplan AP. "Identification of the zinc-dependent endothelial cell binding protein for high molecular weight kininogen and factor XII: identity with the receptor that binds to the globular "heads" of C1q (gC1q-R)." Proceedings of the National Academy of Science USA 1996; 93(16):8552-7.

Kalafatis M, Egan JO, van't Veer C, Cawthern KM, Mann KG. "The regulation of clotting factors." Critical Reviews in Eukaryotic Gene Expression 1997; 7(3):241-80.

Krawczak M, Wacey A, Cooper DN. "Molecular reconstruction and homology modelling of the catalytic domain of the common ancestor of the haemostatic vitamin-K-dependent serine proteinases." Human Genetics 1996; 98(3):351-70.

Martin DW. "Blood Plasma and Clotting." In: Martin DW, Mayes PA, Rodwell VW Harper's Review of Biochemistry 1983; Los Altos, California: Lange Medical Publications.

McVey JH. "Tissue factor pathway." Baillieres Clinical Haematology 1994; 7(3):469-84.

Muller WE. "Origin of Metazoa: sponges as living fossils." Naturwissenschaften 1998; 85(1):11-25.

Neuenschwander PF, Fiore MM, Morrissey JH. "Factor VII autoactivation proceeds via interaction of distinct protease-cofactor and zymogen-cofactor complexes. Implications of a two-dimensional enzyme kinetic mechanism." Journal of Biological Chemistry 1993; 268(29):21489-92.

Rosing J, Tans G. "Factor V." International Journal of Biochemical and Cell Biology 1997; 29(10):1123-6.

Tans G, Rosing J. "Structural and functional characterization of factor XII." Semin Thromb Hemost 1987; 13(1):1-14.

Villoutreix BO, Dahlback B. "Structural investigation of the A domains of human blood coagulation factor V by molecular modeling." Protein Science 1998; 7(6):1317-25.

Walker FJ, Fay PJ. "Characterization of an interaction between protein C and ceruloplasmin." Journal of Biological Chemistry 1990; 265(4):1834-6.

Zimmermann E. "The evolution of the coagulation system from primitive defense mechanisms [Article in German]." Behring Inst Mitt 1983; (73):1-12.

table of contents

This page was last updated November 18, 1998.
Copyright © 1998 by Kevin O'Brien