The serpins are an expanding superfamily of structurally similar but functionally diverse proteins. Evolution, mechanism of inhibition, novel functions, and a revised nomenclature.

Published, JBC Papers in Press, July 2, 2001, DOI 10.1074/jbc.R100016200 Gary A. Silverman, Phillip I. Bird, Robin W. Carrell, Frank C. Church, Paul B. Coughlin, Peter G. W. Gettins, James A Irving, David A. Lomas, Cliff J. Luke, Richard W. Moyer, Philip A. Pemberton, Eileen Remold-O’Donnell, Guy S. Salvesen, James Travis, and James C. Whisstock From the Department of Pediatrics, Division of Newborn Medicine, Children’s Hospital, Harvard Medical School, Boston, Massachusetts 02115, Department of Biochemistry and Molecular Biology, Monash University, Melbourne, Victoria 3800, Australia, Departments of Haematology and Medicine, University of Cambridge, Wellcome Trust Centre for Molecular Mechanisms in Disease, Cambridge Institute for Medical Research, Wellcome Trust/MRC Building, Hills Road, Cambridge CB2 2XY, United Kingdom, Division of Hematology-Oncology/Medicine, University of North Carolina, Chapel Hill, North Carolina 27599, Department of Medicine, Box Hill Hospital, Monash University, Melbourne, Victoria 3128, Australia, Department of Biochemistry and Molecular Biology, University of Illinois, Chicago, Illinois 60612, Department of Molecular Genetics and Microbiology, University of Florida College of Medicine, Gainesville, Florida 32610, Arriva Pharmaceuticals, Alameda, California 94501, Center for Blood Research, Harvard Medical School, Boston, Massachusetts 02115, Burnham Institute, San Diego, California 92037, Department of Biochemistry and Molecular Biology, University of Georgia, Athens, Georgia 30602


Serpin Structural Features: Conformations, Inhibitory
Mechanism, and Polymerization Serpin Conformations-Serpins adopt a metastable conformation that is required for their inhibitory activity (5). This conformation consists of a conserved secondary structure comprised of ␤-sheets A, B, and C and at least 7 ␣-helices (most typically have 9, lettered A-I; Fig. 1A). The RSL, which contains the proteinase recognition site, is an exposed, flexible stretch of ϳ17 residues tethered between ␤-sheets A and C. Serpins can undergo major structural rearrangements that involve alternative conformations for the RSL, ␤-sheet A, and the attached strand 1 of ␤-sheet C. Considering only intramolecular structural changes, serpins can convert to the more stable latent form (Fig. 1B). The RSL inserts into the middle of ␤-sheet A to give a fully antiparallel ␤-sheet, and s1C is extracted from ␤-sheet C to provide an exposed "return" from the bottom of the serpin. Serpins in the latent conformation are noninhibitory but can be converted back to the active state by denaturation and refolding. The T m for unfolding of latent PAI1 (SERPINE1) is 17°C higher than that for the native state (reviewed in Ref. 6). The most stable state for inhibitory serpins is the RSL-cleaved form, in which the RSL has fully inserted into ␤-sheet A, as in the latent conformation, but without the need to extract s1C from ␤-sheet C (Fig. 1C). Estimates of the T m for unfolding of such conformations are Ͼ120°C, compared with ϳ60°C for the native state (7).
The most informative serpin structures, from a mechanistic viewpoint, are those of a Michaelis complex between Serpin 1 and trypsin (Fig. 1D) and of a covalent complex between ␣ 1 AT (SER-PINA1) and trypsin (8) (Fig. 1E). This latter structure represents the proteinase after it has been kinetically trapped in the acylenzyme intermediate that forms normally along the peptide bond cleavage pathway. Whereas the bound serpin is almost indistinguishable from that of the RSL-cleaved form (Fig. 1C), the proteinase is grossly distorted (see below).
bonyl of the P1 residue and cleavage of the peptide bond (6). It is likely that only at this stage, with removal of the restraint, does the RSL start to insert into ␤-sheet A and transport the covalently bound proteinase with it. Upon complete loop insertion the proteinase is translocated by over 70 Å, and its active site is distorted (Fig.  1E). The alignment of the active site catalytic triad is altered by as much as 3 Å, and the P1 side chain is removed from the S1 pocket. Also, 40% of the body of the proteinase shows no traceable electron density. Proteinase distortion and hence inactivation results from compression of the proteinase against the base of the serpin as a consequence of the inserted RSL being just the right length. The energy needed to effect the distortion may come from the much greater stability of the cleaved loop-inserted conformation compared with the native-like conformation. The net result of this conformational rearrangement is kinetic trapping of the acyl intermediate due to slowing of the deacylation steps of the normal substrate reaction by 6 -8 orders of magnitude (k 5 in Fig. 2). Because of the small values for k 5 (complex t1 ⁄2 Х hours to weeks), serpin-proteinase complexes in vivo would bind to their receptors and be cleared (complex t1 ⁄2 Х minutes) long before significant complex decay could occur.
The point in transit where the enzyme activity is reduced sufficiently to commit the intermediate to the kinetic trap is not known but in part contributes to the branched nature of the pathway and the ultimate fate of the complex. If, for example, RSL movement is impeded, the enzyme may successfully complete the deacylation step and escape before it is irreversibly trapped. This noninhibitory pathway yields an active proteinase and a cleaved, inactive serpin. The ratio of serpin products (complex versus cleaved) thus reflects a competition between the rate of ester hydrolysis (k 3 in Fig. 2) and that of loop insertion (k 4 in Fig. 2) to the point of proteinase distortion. This ratio is signified also by the stoichiometry of inhibition, which is defined as (k 3 ϩ k 4 )/k 4 , i.e. the number of moles of serpin needed to inhibit 1 mol of proteinase as a kinetically trapped complex.
This mechanism accounts for the requirements for effective inhibition by serpins, which include a critical RSL length, appropriate residues within the loop that are compatible with rapid and favorable burial into ␤-sheet A, and the presence of Ser in the proteinase active site (6). Such a mechanism is adaptable to the inhibition of cysteine proteinases by serpins, with the difference being that the kinetically trapped intermediate is a thiol ester rather than an oxy ester. The detection of CrmA, a serpin that inhibits cysteine proteinases of the caspase family, in the loop-inserted cleaved conformation supports the feasibility of a common inhibitory mechanism (9), whereas the detection of an SDS-stable complex between SCCA1 and cathepsin S (a cysteine proteinase of the papain family) provides evidence for the formation of a stable, covalent thiol ester-type linkage (3). The few convincing reports of reversible inhibition, such as of single-chain uPA by PCI (SERPINA5) (10) or of chymotrypsin by ␣ 2 AP (SER-PINF2) (11) may represent special cases in which unusual stabilization of the initial noncovalent Michaelis-like complex blocks progres-sion to the substrate reaction.
Dimerization and Higher Order Polymer Formation-A negative consequence of the need for a metastable conformation in the active state is that natural mutations, either alone or in combination with environmental factors, can promote inappropriate loop insertion. When this occurs between the RSL of one molecule and the ␤-sheet of another, dimers and higher order oligomers can result. Either through depletion of active serpin or through pathological effects of the polymers themselves, such aggregate formation can lead to disease. The best characterized examples are the emphysema (serpin depletion) and cirrhosis (intracellular inclusions) associated with loop-sheet polymers of the Z or S variants of ␣ 1 AT (12) (see Supplemental Data, Fig. A) and the dementia associated with neuroserpin (SERPINI1) inclusion bodies (see below).

New Serpins and Novel Functions
Understanding the biologic function of serpins remains an ongoing challenge. For example, the biologic functions for many of the human serpins involved in the clotting and fibrinolytic cascades are well documented. However the role of human serpins in some other types of biologic processes awaits further validation (Fig. 3).
Ov-serpins (B Clade)-In 1993 amino acid similarities among chicken ovalbumin (ov), PAI2 (SERPINB2), and MNEI (SERPINB1) led to the identification of a subgroup of the serpin superfamily (13). The N and C termini of the ov-serpins are shorter than the prototypical serpin ␣ 1 AT, and they also lack a classical secretory signal peptide. At present, there are 13 human ov-serpins (see Supplemental Data, Table B). They map to 6p25 and 18q21 and fall into two classes based on a single difference in gene structure (14). Like ovalbumin, many of the 18q21 serpin genes have an exon encoding a polypeptide loop between helices C and D (CD loop) that may contribute to accessory functions.
Unlike ovalbumin itself, most ov-serpins reside intracellularly with a cytoplasmic or nucleocytoplasmic distribution. However, several ov-serpins (PAI2, megsin (SERPINB7), MNEI, maspin (SERPINB5), and the SCCAs (SERPINB3 and -4)) may function extracellularly as they are released from cells under certain conditions. Release may be facilitated by an embedded, noncleaved hydrophobic N-terminal signal sequence and appears to involve both conventional and non-endoplasmic reticulum-Golgi secretory pathways (15). Regardless of how ov-serpins are released from cells, those with RSL cysteine or methionine residues are susceptible to oxidative inactivation and are likely to have a limited half-life in the extracellular milieu.
With the possible exception of maspin, all human ov-serpins are functional, competitive inhibitors of serine or cysteine proteinases. Several members of the group inhibit more than one proteinase, and dual reactive sites (utilization of more than one P1 residue) have been described for PI6 (SERPINB6), PI8 (SERPINB8), PI9, SCCA1, SCCA2, and MNEI (for example see Ref. 16). However, the CD loops of the ov-serpins have the potential to interact with other proteins. For example, the CD loop of PAI2 is required for its cell survival function (17) and is a target for transglutamination (18). Bomapin (SERPINB10; like the chicken ov-serpin, MENT, see below) carries a nuclear localization signal in its CD loop that presumably interacts with a nuclear importin (19).
The physiological functions of ov-serpins are still emerging. PAI2 may play a role in the regulation of extracellular matrix remodeling through the inhibition of uPA, as high PAI2 and low uPA levels correlate with a positive prognosis in breast cancer (20). Also, PAI2 may have a structural role inside some cells (perhaps keratinocytes) as suggested by its ability to spontaneously polymerize and undergo transglutamination (21).
Many ov-serpins reside in proteinase-secreting cells (22). For example, PI9, a potent inhibitor of granzyme B, is also present in cytotoxic lymphocytes. Because PI9 can protect cells against granzyme B-mediated apoptosis, it probably protects cytotoxic lymphocytes from autodestruction due to misdirected granzyme B. A similar cytoprotective role can be envisaged for PI6, PI8, MNEI, PAI2, and the SCCAs. In addition, endogenous or exogenous ov-serpins may protect bystander cells and tissue from proteolytic damage. Studies in rats show that recombinant MNEI delivered to the airways prevents lung injury by neutrophil proteinases and point to its potential in treating inflammatory lung disease (23).
The ability of many ov-serpins to inhibit more than one proteinase and their presence in epithelial cells suggest that they play a role in barrier function or host defense against microbial or viral proteinases. For example, PI9 inhibits Bacillus subtilisin, and PI8 inhibits furin, a subtilisin-related enzyme (24,25). Additional functions of ov-serpins include the regulation of: 1) cell growth or differentiation, as exemplified by the role of megsin in megakaryocyte differentiation (26), 2) tumor cell invasiveness and motility, as shown by the inhibitory role of maspin in breast and prostate tumors (27), and 3) angiogenesis (see below).
MENT-Grigoryev et al. (28) isolated a novel serpin, MENT, from the nuclei of terminally differentiated chicken hematopoietic cells. MENT is an ov-serpin with a CD loop that contains a nuclear localization signal, a lamin-like chromatin binding domain, and an A-T hook DNA binding motif. The molecule has a relatively high pI (9 versus 5-6.5 for that of other serpins) with the majority of positive charges clustering near the CD loop. Thus, MENT appears to utilize the CD loop to bind tightly to nucleosomes with an apparent stoichiometry of 2:1. MENT is the major non-histone chromatin protein in differentiated nuclei and is concentrated in the heterochromatin. MENT induces higher order chromatin compaction when it is expressed ectopically in cells or added to isolated nuclei in vitro. Although MENT contains a viable RSL, target proteinases have yet to be identified.
Neuroserpin-Neuroserpin, which inhibits tPA, uPA, trypsin, and nerve growth factor ␥ in vitro, is secreted from neurons, glia, and neuroendocrine cells (29). Neuroserpin may play a therapeutic role in protecting the brain from ischemic injury. In a rat stroke model, neuroserpin expression was increased in neurons located within the ischemic penumbra, and intracerebral injections of the protein reduced the stroke volume by 64% and the number of apoptotic cells by 50% (30).
In a familial form of early onset dementia and encephalopathy, toxic intraneuronal inclusions contained neuroserpin polymers. Molecular analysis in two pedigrees revealed mutations (S49P and S52R) in the B-helix (31). These mutations are similar to that seen in ␣ 1 AT Siiyama, in which an S53F mutation facilitates premature opening of ␤-sheet A and the formation of loop-sheet polymers. In turn, these polymers precipitate and accumulate in the cytoplasm until normal cellular function is disrupted.
Serpins and Alzheimer's Disease-ACT (SERPINA3) and, to a lesser extent, other serpins are found within the fibrillary amyloid plaques of brains from patients with Alzheimer's disease, one of the most common forms of dementia (reviewed in Ref. 32). Although the pathogenesis of this disorder is complex, the extracellular accumulation of A␤-(1-42)-peptide fibrils may be neurotoxic by binding to low density lipoprotein receptors and interfering with cholesterol metabolism. ACT appears to facilitate fibril formation by serving as a chaperone for the A␤-(1-42)-peptide. The peptide inserts into B-sheets A and C of ACT in which it assumes a ␤-strand conformation. Upon RSL cleavage, A␤-(1-42)-peptide is released into the extracellular milieu, in which the ␤-strand peptide is now more prone to polymerize.

Pigment Epithelium-derived Factor and Other Serpins That May
Interfere with Angiogenesis-PEDF (SERPINF1) is a secreted, noninhibitory serpin that was isolated from retinal pigment epithelial cells but is also detected in liver, lung, heart, spleen, brain, and testis (6). This factor promotes the survival and differentiation of retinal photoreceptors, cerebellar granule neurons, and spinal motor neurons. Dawson et al. (33) show that PEDF inhibits neovascularization of the rat cornea and endothelial cell migration in vitro. In the cell migration assay, PEDF was as potent as other angiogenesis inhibitors such as angiostatin, endostatin, and thrombospondin-1. Moreover, PEDF antagonized the effects of the angiogenesis inducers, VEGF, basic fibroblast growth factor, plateletderived growth factor, and interleukin 8. PEDF expression in the eye increases with rising oxygen tension (just the opposite of VEGF). Thus, VEGF and PEDF appear to counter-regulate blood vessel growth in the eye by enhancing and antagonizing angiogenesis during hypoxic and hyperoxic conditions, respectively. PAI1, maspin, and RSL-cleaved ATIII (SERPINC1) have been shown to interfere with angiogenesis in various assay systems (34 -36). However, it has yet to be determined whether any of these molecules are truly involved in the physiologic or pathologic regulation of blood vessel growth.
Serpins and Host Defense-A loss of function mutation in the Drosophila serpin gene, Spn43Ac, leads to the necrotic phenotype and constitutive expression of the antifungal peptide, drosomycin (37). Normally activation of the antifungal pathway involves proteolytic cleavage of the Toll (Tl) ligand, spaetzle (spz). In turn, Tl activation leads to an increase in both Spn43Ac and drosomycin synthesis via the Rel-Cactus (NF-B-IB-like) pathway. Spn43Ac appears to act in a negative feedback loop by inhibiting proteinases that activate spz. Thus, the constitutive expression of drosomycin and the necrotic phenotype appear to be secondary to unregulated proteolytic activity.
Plant Serpins-The function of these proteins remains obscure. Several studies show that plant serpins are capable of inhibiting serine proteinase targets (38). However, with the exception of a chymotrypsin-like proteinase identified in ragweed pollen, conventional serine proteinase targets are absent in plants. BLAST searches of the A. thaliana genome using the sequence of trypsin as a probe failed to identify any classical chymotrypsin-like homologues. In vivo studies by Yoo et al. (39) show that up-regulation of the Cucurbita maxima Phloem Serpin-1 (CmPS-1) correlates closely with the inability of the piercing sucking aphid Myzus persicae to survive and reproduce on these plants, suggesting a role for plant serpins in host defense.
Targeted Deletions of Mouse Serpins-Several murine orthologues of human serpins have been deleted by targeted mutation in embryonic stem cells. Some of these mutations have failed to reveal an overt phenotype, whereas others show physiologic and structural alterations as well as embryonic lethality (see Supplemental Data, Table C).
Viral Serpins-Serpins are found within a number of genera within the subfamilies of the vertebrate poxviruses and the gammaherpesviruses (see Supplemental Data, Table D). To date, none of the serpins are required for virus growth in cell culture. Within the vertebrate poxvirus genera, each Orthopoxvirus (variola, vaccinia, and rabbitpox) encodes three highly conserved serpins, SPI-1, SPI-2/CrmA, and SPI-3. Each targets different types of proteinases (Table D). The prototypic member of the Leporipoxvirus genus (myxoma virus) also encodes three serpins, SERP1-3. Within the Avipoxvirus genus, the fowlpox virus genome contains 5 serpin genes (Table D). Other vertebrate poxviruses (molluscum contagiosum and ORF viruses) lack serpin genes. All genera of poxviruses encode serpins with putative Asp P1 residues, and the leporipoxviruses, orthopoxviruses, and fowlpox viruses all have a member with a putative Arg P1 residue (Table D). Only the orthopoxviruses have a serpin with a Phe at the putative P1 site. For more information on Orthopoxvirus and Leporipoxvirus serpins see Supplemental Data.

Conclusions
The serpins are a superfamily of genes that are distributed throughout the metazoa and plantae kingdoms. Serpin family mem-bers are identified by a conserved tertiary structure and a unique suicide substrate-like inhibitory mechanism. Serpins reside both intracellularly and extracellularly and are involved in a diverse set of biologic functions that extend beyond the ability of these molecules to irreversibly inhibit target proteinases. The study of serpin function using different biological platforms, such as the nematode (40) and fruit fly, should help identify the role that these molecules play in development, homeostasis, and host defense.