Guy S. Salvesen, Program in Apoptosis and Cell Death Research, The Burnham Institute, La Jolla, California, USA.

Guy Salvesen was invited to act as rapporteur for this Horizon Symposium. In this article, he brings together some of the threads of the Meeting to give a personal 'snapshot' of the state of play in this exciting interdisciplinary field.

Proteins are the toughest biological polymers. Peptide bonds are stable for hours in boiling concentrated acid, yet they last no more than microseconds in the presence of an effective protease. Until about five years ago (in the days when spam was a kind of food), one of the most frequent e-mails I would receive from strangers had a version of the question "I think I have a protease somewhere in my preparation - how do I get rid of it?" Now that question has been replaced by "I think I have a protease somewhere in my preparation - should I work on it?" To me, this more than anything else signifies the widespread interest in the field of proteolysis, in all its facets. No longer is proteolysis considered a menace by all but a dedicated few. Even the most peripherally interested biologist knows now that proteases are far from simple degradation machines, and that most of the workings of biology contain essential proteolytic steps and pathways (Box 1).

The study of proteolysis is ancient, going back at least to the nineteenth century with the description of pepsin by Schwann in 1836, and trypsin by Corvisart in 1856. Proteases and their inhibitors are frequently used by structural biologists, protein engineers and protein folders as experimental models. Proteases have a prominent role in the manufacture of cheese and washing powder. A breed of the most successful therapeutic drugs - from the point of view of patients, physicians and pharmaceutical companies - targets a human protease.

So, given this wealth of information and experimental, industrial and therapeutic success, why hold a Horizon Symposium on proteolysis? Well, proteolysis researchers have to admit that there are large gaps in knowledge and great setbacks in drug design (see 'The trouble with inhibitors'). Of course, most fields of study have gaps in knowledge, but in the protease field we know where the problems are - we just can't get to them. But there have also been gains in new protease families, catalytic types and biological mechanisms during the past five years. And this Symposium aimed to set the ancient field of proteolysis in the context of new discoveries, new protease types and potential new therapeutic strategies in the light of past experience, and naturally also in the context of the post-genomic landscape on which we all now toil.

How to be a protease
Proteases have learned to cope with the legendary stability of peptide bonds through a number of apparently unconnected mechanisms. There are the serine proteases, the cysteine proteases, and the N-terminal nucleophile hydrolases that use a catalytic threonine residue. Collectively, these catalytic classes share the property of having an enzyme side-chain that participates in catalysis ('covalent catalysis'; Fig. 1). By contrast, the other known protease classes, the aspartic proteases and metalloproteinases, carry out peptide-bond hydrolysis without covalent participation - primarily by generating a highly reactive water molecule.

		Figure 1 | Protease catalytic types. The five protease catalytic types partition essentially into two catalytic mechanisms: covalent and non-covalent (general acid-base) catalysis. The fundamental similarity is stabilization of the tetrahedral intermediate in catalysis, which is shown here. The dichotomy is whether a protease participated covalently during catalysis (Ser, Cys or Thr proteases), or polarizes a water molecule to accomplish catalysis (Asp or metalloproteinases). Nucleophiles that participate in covalent catalysis (Nuc) are serine, cysteine or threonine, and the base is usually, but not always, a histidine. In general acid-base catalysis, the acids and bases are the side-chains of aspartic residues, or glutamic residues and zinc in the case of metalloproteinases. In this rendition, the shading designates the enzyme surface; atoms originating from the protease or protease-bound water are shown in red; and atoms originating from the substrate are shown in black. The fundamental distinction is central to the evolution of natural inhibitors, as well as chemistries available for the design of small-molecule inactivators.

All proteases share a quest to accelerate the formation of the transition state in peptide-bond hydrolysis. This state, sometimes called the tetrahedral intermediate, is a prerequisite for peptide-bond scission in all protease mechanistic classes. In fact, tetrahedral-intermediate formation might be the most important job a protease does. It is also probably the most difficult to simulate because protein engineers and chemists have yet to invent or evolve an independent structure or fold that can efficiently catalyse peptide-bond cleavage. Clearly there are examples in which catalytic antibodies, for example, have evolved to efficiently hydrolyze activated esters, but these are very different from the resonance-stabilized peptide bond. I think this is a surprising gap in our knowledge, and remains fertile ground for future work.

Cascades or pathways - explosive or progressive?

The concept of a proteolytic cascade originated in the pioneering work on blood coagulation by Davie, Ratnoff and MacFarland around 40 years ago. Since that time, plenty of information has been gathered about the components that support coagulation in vitro and in vivo. Significantly, these two approaches do not always match, and some putative coagulation proteases have dropped from modern depictions of blood coagulation. For example, a physiological role for Hageman factor (FXII) and prekallikrein in the coagulation pathway is unlikely because individuals lacking this activity have no bleeding complications, even though both proteases accelerate blood coagulation in vitro. Nonetheless, blood coagulation stands as the model proteolytic cascade in which information is passed through a pathway by sequential activations of protease zymogens, with a minimal pathway requiring the proteases Factor VII, Factor X and prothrombin (see 'Signalling scissors: new perspectives on proteases'). Now, other proteolytic cascades are recognized, and the discussion at the Symposium illustrated our current knowledge of cascades using the examples of coagulation (serine proteases), apoptosis (cysteine proteases) and matrix remodelling (serine proteases and metalloproteinases).

It could be that most people go into a discussion on cascades expecting enlightenment as to how a conserved mechanism can be represented by a variety of unrelated functional processes. Hence, protease cascades, much like other cascade events, might even be thought to exist for a common reason and to share a common mechanism. Surprisingly, the discussions on known proteolytic cascades came to a different conclusion. It might not be appropriate to think that all cascades have a common function or even a common progression. It might not even be appropriate to think of them as cascades, but sometimes in some cases just as linear pathways or even connecting networks (see 'Complex cascades and the search for substrates').

The initial notion of a cascade emphasized the rapid amplification of proteolytic activity. It was reassuring to hear that this concept remains valid, as beautifully exemplified by the speedy formation of thrombin visualized by in vivo substrate imaging. A clot takes longer to form, but the terminal clotting protease is activated within seconds of tissue injury in vivo. Another important feature of a cascade is that it provides multiple regulation points, presumably for fine tuning. Most proteolytic cascades have inhibitors that target the activated proteases and can act either as 'buffers' or 'thresholds' to hold in check a pathway that has become inappropriately activated. The problem is that this is very difficult to verify in vivo, so the multiple-regulation-point concept awaits some experimental tests, and remains pretty much an item of faith.

Considerable biochemical and genetic evidence now undermines the existence of a cascade, or even a proteolytic pathway, in extracellular matrix remodelling. Perhaps the concept of a redundant network of interactions is more appropriate in this case, although the audience considered that many of the metalloproteinases once thought to degrade matrix might be limited processing enzymes whose substrates are cytokines and growth factors. It is even unclear whether the plasminogen activator/plasmin/metalloproteinase pathway exists anywhere but in vitro. The 20-year-old concept that the matrix must be broken down to allow cell migration (in normal or metastatic instances) is under close scrutiny and revision to the possibility that the proteolytic degradation of the extracellular matrix might not be required for migration at all. All of which brings us to one of the discussion highlights of the meeting - the quest to identify the true substrates of proteases.

The great substrate hunt

Decades ago, the way to determine the specificity of a new protease was to throw it onto the oxidized insulin B chain and sequence the products. In this way, it is possible to determine what can be called intrinsic subsite occupancy - the inherent preference of a protease to bind specific side-chains in its specificity pockets. This simple procedure produced much of the early data on specificity, and one could argue that we haven't progressed much further than this. Certainly, the current rapid-throughput techniques that use positional scanning of synthetic peptide substrate libraries, substrate phage display and their ilk have accelerated discovery. But most attendees of the Symposium thought that these methods have yet to deliver reliable information on natural protease substrates. Like other enyzme families, such as kinases, the problem remains - can one predict natural substrates from intrinsic subsite occupancy? The resounding answer of "No!" is tempered by only a few cases in which it has proven successful (caspases being the best example), but the relative tolerance of most proteases for multiple sequences precludes this predictive method. Consequently, the only way to determine the natural substrates is to go looking for them by finding which proteins are bound to proteases in vivo by genetic methods, or by directly determining the products of proteolysis ex vivo.

The first method is gaining popularity, but a major caveat is that it can only apply to certain proteases that bind their substrates by exosites - substrate-binding sites that lie outside the active site of a protease and are located on specialized substrate-binding modules or domains. The normal mechanism of proteolysis precludes identifying substrates by the type of trapping technologies used for kinases, such as ERK2, and phosphatases, such as PTP1B, because proteases bind their substrates fairly weakly and hydrolyze them very rapidly.

Genetic tactics provide tremendous power in the quest for natural substrates, and systematic approaches of knocking out all proteases in a particular family, followed by analysis of the effect of the knockout on specific protein cleavage are producing strong data, and frequently great surprises. This works when one has a good idea of what the substrates are, but given that worms, flies and mice each have more than 400 proteases in their genomes, this is limited by current technologies. However, most at the Symposium agreed that this would provide strong proof of specific protease/substrate reactions in vivo (see 'Complex cascades and the search for substrates').

The third approach - to determine the products of specific protease cleavage ex vivo - is currently out of reach, but encouraging strides are being made through proteomic technology. This, all agreed, would represent a great advance and is eagerly awaited. Of course, there is the possibility that we were all being a bit too conservative in dismissing chemical diversity approaches so glibly. After all, the intrinsic subsite occupancy of a protease is a vital property leading to drug discovery. If you know the best synthetic substrate you just need to replace the reporter substrate with a chemical 'warhead' targeting the catalytic mechanism and you have an inhibitor - and, more importantly, even maybe a drug lead. Although statistics indicate that the chances of a lead making it to the market as a therapeutic agent are small, even failed compounds generated along the way are useful in that they can become vital reagents for elucidating protease function in experimental systems (for example, batimastat, PPACK and Z-VAD-FMK are in most protease scientists' fridges).

Activation and inhibition

Nature has adopted a common strategy for regulating proteases and proteolytic pathways, irrespective of catalytic class and evolutionary origin (Fig. 2). Most proteases are synthesized and stored as latent forms (zymogens) that await an activation signal. Following activation they have a limited lifespan during which they cleave their substrates before final inhibition and removal. Not all proteases follow this strategy, but enough do that the Symposium set a goal of discussing whether nature's strategy could be adapted to facilitate therapeutic control. We were introduced to what, for me at least, is a new concept in zymogen activation. Rather than a simple on/off switch, it could be that conformational equilibria dominate the zymogen transition, so that proteases can slide back and forth between the active and latent state in response to the surrounding environment. This concept was embraced, and raised the question as to whether the conditions often used for drug screening might not actually be those encountered in vivo. But how researchers can use this information was a little less clear. Combined with this are the many and varied ways by which natural protease inhibitors perform their function: lock-and-key, suicide or allosteric mechanisms that seem completely unrelated to each other. Great biochemical and structural strides have been made in this area, yet very little has permeated into therapeutic design.

		Figure 2 | A proteolytic cycle. The fundamental mechanisms governing activity are conserved in most proteases. Latent protease zymogens await an activation signal. The activator can be an allosteric activator, or another protease. Once active, substrate and inhibitor compete for protease binding, and the outcome is defined by the local concentration of inhibitor. Inhibition is signified by the double-headed arrow, because many cognate protease inhibitors form reversible complexes with their target proteases. A cascade is defined when the substrate of one protease (coagulation Factor Xa, for example) is another protease zymogen (prothrombin, for example).

Perhaps the most important theme regarding zymogen activation and natural protease inhibitors is that many operate by allosteric mechanisms. Yet there are essentially no drugs that operate by allosteric mechanisms, nor apparently are pharmaceutical chemists pursuing them. This led to another avidly discussed issue regarding protease inhibitor design. It is in the nature of large pharmaceutical companies to have a conservative approach to drug design, and to model future drugs on previous successes. The angiotensin-converting enzyme (ACE) inhibitor captopril and HIV protease inhibitors are the paradigms in the protease field and they both target the active site in an intrinsic subsite occupancy mode. So why risk going for allosteric regulators simply because nature points the way? There is one extremely good reason - active site-directed inhibitors are by character competitive with substrate (Fig. 3). Most proteases in vivo probably operate in an environment of substrate saturation, and this will slow down and/or weaken competitive inhibitors by the factor of 1+Σ/ΣK_m, in which ΣK_m is the sum of all values for all the substrates in a cell. By contrast, allosteric inhibitors are not affected by substrate concentration. Perhaps this is why, in addition to the perennial problems of tissue penetration and pharmacokinetic considerations, the inhibitory concentration of a protease inhibitor drug in vivo is always several orders of magnitude greater than in a purified system.

		Figure 3 | Mechanisms of inhibition. This figure illustrates competitive inhibition (Ic) and allosteric inhibition (Ia). Ic competes with substrate for occupation of the enzyme active site. Ia, formally but inappropriately called non-competitive inhibition, binds to the enzyme-substrate (ES) complex or to free E. The major practical difference is that efficiency of inhibition by Ia is unaffected by the concentration of S, whereas Ic is completely dependent on it. P, product.

Proteases: good guys or bad guys?

Another issue that raised its head at the Symposium is whether one could take advantage of the zymogen/active conformation transition that is frequently produced by natural allosteric activators, such as coagulation factors V and VIII and the caspase-activating apoptosome. Although academics investigate these events, it is rare for industrialists to develop them. This could result from the view that proteases are the offenders in disease, although there might not be much formal proof for this. In cancer, for example, many different proteases are upregulated and/or activated and the common approach is to target them for inhibition. We agreed that proteases are frequently prognosticators of poor outcome (prostate-specific antigen is the classic example), but we had to admit to being perplexed about whether they are inherently bad. There are sure to be outside opinions regarding this, and I can take one example from my own line of research in which caspases can be bad guys or good guys depending on context. Too much apoptosis results in inappropriate death of neuronal cells (neurodegeneration) and caspases are the bad guys in this situation. By contrast, it is now clear that many cancers arise because of defects in normal apoptosis control. Turning caspases back on reinstates the normal death mechanisms that have been overcome in a cancer cell's quest for immortality, and in this context caspases are the good guys. But whether the metalloproteinases and cathepsins that are frequently upregulated are by necessity the bad guys remains controversial, and this is clearly an issue requiring a huge investment to resolve on a case-by-case basis.

Academics want to develop allosteric regulators but Industrialists say it is too risky and untested, which became the the first point of disagreement at the Symposium. It is difficult to know what to do about this - each side has traditional goals that often do not overlap. However, governments in Europe and the United States (at least) are trying to get the two sides together by providing funding for academics that require industrial participation. It is part of the 'Big Science' initiatives and promises some interesting and hopefully productive and paradigm-shifting times ahead.

Compartmentalization

If asked a decade ago if proteolysis could occur within a lipid bilayer, most biochemists would have answered with a resounding "No". There is not supposed to be much water in a lipid bilayer, so how could there be hydrolysis? Yet one of the most exciting breakthroughs in proteolysis research has been the identification of three protease families that apparently cleave transmembrane proteins in their transmembrane region (see 'RIPping and folding: regulated intramembrane proteolysis'). In addition, one of the proteases, γ-secretase, has huge potential as a therapeutic target because it is responsible for the formation of Aβ peptide, which is thought to have a key role in the pathogenesis of Alzheimer's disease. The controversy surrounding the identity of the protein presenilin as the proteolytic core of γ-secretase seemed to have been solved. Presenilin is homologous to signal peptide peptidase, the protease responsible for degrading the short transmembrane stubs left in the endoplasmic reticulum membrane following signal peptide removal. Interestingly, there was even a proposal that presenilin/γ-secretase itself might digest residual membrane stubs resulting from the many transmembrane protein shedding events that occur during cell signalling. The other two proteases - S2P, which is involved in cholesterol signalling, and Rhomboid, which is involved in various membrane-protein shedding and signalling events, were not yet considered to be prime drug candidates for human therapy by the Symposium, but they are exceedingly interesting from a biological and mechanistic perspective.

During the presentations on regulated intramembrane proteolysis (RIP) it became evident that the presenters were eager to hear whether biochemists and chemists thought that proteolysis could occur in a membrane. The chemists that were present said there was no reason why not. Perhaps the RIP proteases create their own water channel, or perhaps RIP is so catalytically slow that the odd water molecule finds its way in. What was completely clear was that this new model was in need of some serious biochemistry and enzymology because nobody really knows how efficient RIP is compared with conventional proteolysis. Despite the tremendous gains, efficient and specific (or even just good) assays are needed to help push this field forwards.

As an example of the need for all disciplines to tackle this area, we discussed whether proteolysis really occurs in the plane of the membrane; that is, actually in the lipid bilayer. The answer remains a moving target with data for and against. If RIP actually occurred in a push-cut-pull mechanism, then cleavage would actually be just outside the membrane anyway (see 'Understanding new mechanisms for proteases'). How does this compare with 'sheddases', such as the ADAMS, which are involved in the shedding of membrane-anchored substrates? These can apparently cut within one residue of the membrane insertion site, strangely blurring the distinction between RIP and conventional sheddases.

The lipid bilayer is arguably the most unlikely place to find proteolysis, but paradigm shifts are also occurring with the much-studied proteases that populate lysosomes. Conventionally, these are thought to be responsible for the breakdown of endocytosed proteins and the generation of the final products of proteolysis that load major histocompatibility complex (MHC) class II molecules for antigen presentation. Yet there are increasingly frequent reports of a breakdown in compartmentalization to yield functioning cathepsins in the cytosol and nucleus. Some years ago this hypothesis was heretical, because the cathepsins were known to have acid pH optima. Not only is this probably untrue, but discussants pointed out that the high protein concentration inside cells would greatly stabilize the lysosomal proteases against pH-induced inactivation. Although audience members were prepared to concede that lysosomal membranes are not the inviolate barriers once thought to keep their contents eternally separate from the rest of the cell, they could not agree on how supposedly degradative enzymes could take on a much more limited role in cutting transcription factors or other signalling proteins, as has been proposed recently. This suggests another area of research that needs to be pursued.

The other way to confine proteolytic activity is within self-compartmentalized proteases, as exemplified by the proteasome (see 'Regulation through degradation'). Each organism has a proteasome-like degradation machine. Although often unrelated in mechanism, and their common function seems to be to keep the bulk of cellular proteins away from the destruction that is occurring in the proteolytic chamber. They can even be considered as mini-organelles, went the discussion. As such fundamental housekeeping structures, it is entirely puzzling how a proteasome-directed small molecule could be of help in diminishing cancer. Nevertheless, one such proteasome inhibitor, bortezomib (Velcade; Millennium) has recently been approved by the US FDA for the treatment of multiple myeloma.

Intervention - who to target

The issues of how to intervene therapeutically with proteases cropped up throughout the Symposium, and so it was perfect that the last two sessions dealt with which specific proteases to target and how to target them. As I mentioned earlier, there was not much discussion on the therapeutic activation of proteases, nor on allosteric regulators in general. However, there was considerable discussion about mechanisms that should be chosen for the conventional competitive inhibitor strategies.

Historically, proteases might have had great successes, such as with ACE inhibitors and HIV protease inhibitors, but there have been notable disasters also, as with the first generation of drugs targeting metalloproteinases. Why? Part of the reason is off-target effects, but in hindsight the off-targets are probably many of the new metalloproteinases that have come to light since the programmes that delivered the metalloproteinase inhibitors came up with their lead compounds. An important lesson was learned: all members of a family that could potentially interact with the inhibitor drug must be obtained and tested. Failure to do this risks more off-target mechanisms, and the protease drug discovery platforms cannot risk many more of these setbacks.

Related to the off-target mechanism is whether specific or general inhibitors should be developed. As might be expected, the answer to this question is both yes and no, since it depends on the context and target. A good example of the need for very specific inhibitors is revealed by the trials on matrix metalloproteinase inhibitors. A good example of a nonspecific strategy is the trials that are occurring with cathepsin inhibitors that target Trypanosomiasis. This parasite seems to concentrate the drug, thereby lowering the therapeutic dose below the level that would have side effects on host cathepsins.

Discussants among the industrialists pointed out a very interesting issue relating to target validation. If it is true that multiple proteases are involved in cancer, and if they are indeed the bad guys, then it might be difficult to convince companies to invest in multiple targets, especially if they have good leads from a kinase programme, for example. Again, this is all the more reason for a massive effort on the parts of academics and industry to find out by all possible strategies just how important protease activity is in tumour progression. From this perspective, the clear consensus was that mouse cancer models are fine for early-stage validation of basic mechanisms, but practically useless for drug development because of the clear differences between human and mouse protease specificity and tumour origin. Even human tumours in mice might not be useful, went the discussion, because many tumour-associated proteases arise from the supporting stroma.

Intervention - how to target

Reversible versus irreversible inhibitors - the debate continues. Pharmaceutical companies seem firmly wedded to the dogma that drugs should be reversible. Irreversible inhibitors contain a mechanism-based electrophile, or 'warhead'. The problem is that irreversible inhibitors might covalently modify proteins, rendering them antigenic, or might show less selectivity because of non-specific reactions between the chemical warhead and non-targeted proteins. A reversible inhibitor, no matter how long it is delivered for, will never show the kind of nonspecificity that an irreversible one will over a long therapeutic period. However, it might be quite difficult to engineer inhibitors of cysteine proteases that do not contain a mechanism-based electrophile. The academics argued that it was just a matter of trial and error, and that irreversible inhibitors are much more effective and easier to make, because one can rely on the chemical warhead to do a lot of the work. The industrialists argued that they would not try to optimize an irreversible lead because of putative side effects. This was the only point at which a clear division between academia and industry was apparent, and it was unfortunate that the two sides couldn't seem to meet somewhere. One delegate suggested to just "go for it - who would have ever thought that the proteasome inhibitor bortezomib would turn out to be a good anticancer drug". Good point.

Conclusions

The Horizon Symposia comprise short, expert introductions to specific areas of study followed by extensive and free-ranging discussions. After an initial settling period, during which participants adjusted to the unusual discussion format, a pattern emerged. This was a very diverse group representing academics and industrialists, and there tended to be only a few of the audience who participated in all topics. On the whole, the audience stayed within its speciality and this was probably due to the extreme breadth of the field. With the exception of irreversible versus reversible inhibition, most of the discussion was fairly creative and we often found ourselves in agreement. Surprisingly, to some at least, we agreed that a lipid bilayer was no barrier to proteolysis, and that a low pH optimum is not a barrier to the cytosolic activity of lysosmal proteases. Importantly, it was absolutely obvious where some of the main areas of almost complete ignorance are, and these included natural substrate identification and whether proteases in disease are always bad guys. Both of these questions will require huge resources to answer.

At present, 5-10% of all pharmaceutical targets are proteases. The pharmaceutical industry clearly believes that proteases are important to target, and their way forward is fairly clear. In light of genomic information they can now realistically characterize all members of a family, aiding in drug design and specificity assignments. There is not a lot of interest outside academia in allosteric regulation, but it's probably up to the academics to prove that this is a valid strategy to pursue for drug development. There was plenty of discussion on specific protease mechanisms and the huge strides in mechanism, biology and structure made in the field. This focused mainly on human proteases, and it is clear that much can also be learned from proteases in simpler organisms. The problem is that worms and flies contain almost as many proteases as humans, and so to really cut the complexity one needs to look at yeast. This is fine for basic mechanisms, such as cell cycle and mitochondrial biogenesis, but not good enough for coagulation, cell death, cancer and so on. So the general consensus is that developing approaches for understanding human proteases in their natural context is a priority.

Finally, most of the audience came away teeming with ideas. Several of us met at another meeting a few weeks later and the talk was still of what we had discussed at the Horizon Symposium. This was in no small way due to the stimulating format and excellent organization of the meeting. The grappa was pretty good too.

		Box 1 | Proteolytic pathways
		Blood clotting. The classical model for thrombus generation involves a cascade of activation events, involving the sequential activation of a series of zymogens of serine proteases to produce thrombin. Thrombin converts fibrinogen to fibrin, which forms the blood clot. The cascade is initiated through two pathways. In the intrinsic pathway, all zymogens and cofactors are constituents of the blood. By contrast, in the extrinsic pathway, tissue factor, which is exposed on all non-vascular cells, comes into contact with flowing blood to initiate blood coagulation. The concept of these two distinct pathways has sufficed for many years, but recent observations have blurred these distinctions; for example, deficiencies of proteins in the contact phase of the intrinsic pathway do not lead to haemorrhagic disorders, and thrombin can activate factor XI, a zymogen of the intrinsic pathway. Fibrinolysis. To prevent excessive fibrin accumulation, fibrinolysis promotes local dissolution of thrombi and promotes wound healing by re-establishing blood flow. Fibrinolysis is carried out by the plasminogen activation system; serine proteases that covert the inactive zymogen plasminogen to the active serine protease plasmin. There are two plasminogen activators: urokinase-type (uPA) and tissue-type (tPA). Plasmin, in turn, activates localized extracellular proteolytic activity, which catalyses the relatively nonspecific degradation of extracellular proteins. Complement fixation. The complement system is a group of more than 34 serum proteins that is activated by antigen-bound immunoglobulin, or by membrane components on Gram-negative bacteria or fungi. This stimulates inflammation, antigen phagocytosis and, in some cases, direct lysis of cells. Complement fixation is activated through three protease cascade pathways: the classical, the alternative, and the mannan-binding lectin (MBL) pathways. All pathways involve the cleavage of C3, the most important activation step in complement fixation, and have a common final pathway called the terminal pathway, or the membrane attack complex (MAC). Apoptosis. Apoptosis, the most intensively studied form of programmed cell death, is regulated in mammalian cells by the cysteine proteases known as caspases. Like the enzymes of the blood coagulation pathway, caspases are activated by zymogens as a result of an upstream proteolytic cascade (although cleavage-independent activation can occur due to an increase in local concentration). Again, like blood coagulation, two cascades are known: the extrinsic and intrinsic pathways. The extrinsic pathway is initiated by a signal from a 'death receptor', such as from Fas, which results in the activation of caspase-8 or 10. The intrinsic pathway is thought to be triggered by translocation into the mitochondria of a pro-apoptotic Bcl-2 family member; for example, Bax, which allows the release of factors from the mitochondria, such as cytochrome c. Cytochrome c then triggers the formation of a supramolecular structure called an apoptosome, which is the allosteric activator of caspase-9. Activated caspases-8, -9 or -10 then generate the catalytically active forms of caspases-3 and -7, which are the main apoptotic effectors. Gastrulation. During embryogenesis, cell fates along the dorsoventral axis are determined by a gradient of the extracellular morphogen Sp�tzle, which activates the receptor Toll. Sp�tzle carries the signal needed for ventral and lateral development, and is activated by four serine proteases - Nudel, Gastrulation defective, Snake and Easter - sequentially in a proteolytic cascade, similar to blood clotting. Antero-posterior pattern cell fates are determined by Nodal, a protein related to transforming growth factor-β. The maturation of Nodal is thought to be due to the proteolytic activities of two subtilisin-like proprotein convertases, Spc1 (also called Furin) and Spc 4 (also called Pace4). Cell cycle. The selective and programmed degradation of growth regulators, such as many cyclins, cyclin-dependent kinase inhibitors and DNA replication factors, to permit cell-cycle transitions is controlled by proteolysis. Degradation is controlled by the action of the attachment of ubiquitin, which targets the growth regulators or inhibitory proteins for degradation by the 26S proteasome. One such inhibitory protein, securin, is degraded to free the activity of the protease separase, which is required for sister chromatid separation during anaphase.