Crystal structure of cleaved human alpha 1-antichymotrypsin at 2.7 A resolution and its comparison with other serpins

The crystal structure of proteolytically modified human alpha 1-antichymotrypsin (ACT), a member of the serpin superfamily, has been solved by Paterson search techniques and refined to an R-factor of 18.0% at 2.7 A resolution with mean deviations from standard bond lengths and angles of 0.013 A and 3.1 degrees, respectively. The final model consists of 374 amino acid residues, 126 solvent molecules and five sugar residues. Asn70 could be identified unambiguously as a glycosylation site and Asn104 is probably also glycosylated. The structure of cleaved ACT is compared with cleaved alpha 1-antitrypsin (alpha 1 PI) and with plakalbumin, which are prototypical models for cleaved and intact serpins, respectively. Cleaved ACT is very similar to cleaved alpha 1 PI; in particular, it has strand s4A, which is liberated by proteolysis, inserted as the middle strand in beta-sheet A. ACT and alpha 1 PI differ locally only at sites of insertions, except at the segment s3C-turn-s4C, which is displaced by several angstr?m units. This region of ACT is involved in DNA binding.     

DNA strongly impairs the inhibition of cathepsin G by alpha(1)-antichymotrypsin and alpha(1)-proteinase inhibitor

This paper explores the possibility that neutrophil-derived DNA interferes with the inhibition of neutrophil cathepsin G (cat G) and proteinase 3 by the lung antiproteinases alpha(1)-proteinase inhibitor (alpha(1)PI), alpha(1)-antichymotrypsin (ACT), and mucus proteinase inhibitor (MPI). A 30-base pair DNA fragment ((30bp)DNA), used as a model of DNA, tightly binds cat G (K(d), 8.5 nM) but does not react with proteinase 3, alpha(1)PI, ACT, and MPI at physiological ionic strength. The polynucleotide is a partial noncompetitive inhibitor of cat G whose K(i) is close to K(d). ACT and alpha(1)PI are slow binding inhibitors of the cat G-(30bp)DNA complex whose second-order rate constants of inhibition are 2300 M(-1) s(-1) and 21 M(-1) s(-1), respectively, which represents a 195-fold and a 3190-fold rate deceleration. DNA thus renders cat G virtually resistant to inhibition by these irreversible serpins. On the other hand, (30bp)DNA has little or no effect on the reversible inhibition of cat G by MPI or chymostatin or on the irreversible inhibition of cat G by carbobenzoxy-Gly-Leu-Phe-chloromethylketone. The polynucleotide neither inhibits proteinase 3 nor affects its rate of inhibition by alpha(1)PI. These findings suggest that cat G may cause lung tissue destruction despite the presence of antiproteinases.     

Proteolytic inactivation of alpha 1-proteinase inhibitor and alpha 1-antichymotrypsin by oxidatively activated human neutrophil metalloproteinases.

Human neutrophils use the H2O2-myeloperoxidase-chloride system to generate chlorinated oxidants capable of activating metalloproteinase zymogens that hydrolyze not only native and denatured collagens, but also the serine proteinase inhibitor (serpin) alpha 1-proteinase inhibitor (alpha 1 PI). To identify the metalloenzyme that hydrolyzes and inactivates alpha 1 PI, neutrophil releasates were chromatographed over gelatin-Sepharose and divided into fractions containing either progelatinase or procollagenase. The gelatinase-containing fraction cleaved alpha 1 PI in a manner inhibitable by native type V, but not type I, collagen. Conversely, while the collagenase-containing fraction also cleaved alpha 1 PI, this activity was inhibited by type I, but not type V, collagen. Because type I and V collagens are competitive substrates for collagenase and gelatinase, respectively, each of the metalloproteinase zymogens were purified to apparent homogeneity and examined for alpha 1 PI-hydrolytic activities. Both purified gelatinase and collagenase inactivated alpha 1PI by hydrolyzing the serpin within its active-site loop at the Phe352-Leu353 and Pro357-Met358 bonds, albeit with distinct kinetic properties. Furthermore, purified collagenase, but not gelatinase, cleaved a second serpin, alpha 1-antichymotrypsin, by hydrolyzing the Ala362-Leu363 bond within its active-site loop. These data demonstrate that human neutrophils use chlorinated oxidants to activate collagenolytic metalloproteinases whose substrate specificities can be extended to members of the serpin superfamily.     

Analysis of the plasma elimination kinetics and conformational stabilities of native, proteinase-complexed, and reactive site cleaved serpins: comparison of alpha 1-proteinase inhibitor, alpha 1-antichymotrypsin, antithrombin III, alpha 2-antiplasmin, angiotensinogen, and ovalbumin.

Proteinase inhibitors of the serpin superfamily may exist in one of three distinct conformations: the native form, a fully active protein with the reactive site loop intact; the proteolytically modified form in which inhibitory capacity is abolished; and the proteinase-complexed form, a stable equimolar complex between the inhibitor and a target proteinase. Here, the specificity and kinetics of the plasma elimination of different serpin conformations are compared. Proteinase-complexed serpins were rapidly cleared from the circulation. However, the native and modified forms were not cleared rapidly, indicating that the receptor-mediated pathways which recognize the complexes fail to recognize the native and modified forms. This result suggests that significant structural differences exist between modified and proteinase-complexed serpins. The structural differences were probed by using transverse urea gradient gel electrophoresis, a technique that allows comparisons of the conformational stabilities of proteins. With the exception of the noninhibitory serpins ovalbumin and angiotensinogen, the modified and proteinase-complexed serpins were both stabilized thermodynamically compared to the native forms. In addition, the proteinase component of the serpin-proteinase complex was usually thermodynamically stabilized. These data are used to compare the conformations of serpin-proteinase complexes with those of native and modified serpins; they are discussed in terms of a model whereby serpins inhibit proteinases in a manner similar to that described for other types of protein inhibitors of serine proteinases.     

Purification and characterization of a novel cysteine proteinase (periodontain) from Porphyromonas gingivalis. Evidence for a role in the inactivation of human alpha1-proteinase inhibitor

Periodontal disease is characterized by inflammation of the periodontium manifested by recruitment of neutrophils, which can degranulate, releasing powerful proteinases responsible for destruction of connective tissues, and eventual loss of tooth attachment. Although the presence of host proteinase inhibitors (serpins) should minimize tissue damage by endogenous proteinases, this is not seen clinically, and it has been speculated that proteolytic inactivation of serpins may contribute to progression of the disease. A major pathogen associated with periodontal disease is the Gram-negative anaerobe Porphyromonas gingivalis, and in this report, we describe a novel proteinase that has been isolated from culture supernatants of this organism that is capable of inactivating the human serpin, alpha1-proteinase inhibitor, the primary endogenous regulator of human neutrophil elastase. This new enzyme, referred to as periodontain, belongs to the cysteine proteinase family based on inhibition studies and exists as a 75-kDa heterodimer. Furthermore, periodontain shares significant homology to streptopain, a proteinase from Streptococcus pyogenes, and prtT, a putative proteinase from P. gingivalis. Clearly, the presence of this enzyme, which rapidly inactivates alpha1-proteinase inhibitor, could result in elevated levels of human neutrophil elastase clinically detected in periodontal disease and should be considered as a potential virulence factor for P. gingivalis.     

Physical mapping of four serpin genes: alpha 1-antitrypsin, alpha 1-antichymotrypsin, corticosteroid-binding globulin, and protein C inhibitor, within a 280-kb region on chromosome I4q32.1.

Alpha 1-antitrypsin (alpha 1AT; protease inhibitor [PI] locus), alpha 1-antichymotrypsin (alpha 1ACT; AACT locus), corticosteroid-binding globulin (CBG; CBG locus), and protein C inhibitor (PCI; PCI locus) are members of the serine protease inhibitor (serpin) superfamily. A noncoding PI-like (PIL) gene has been located 12 kb 3' of the PI gene. The PI, PIL, and AACT loci have been localized to 14q32.1, the CBG locus has been localized to 14q31-14q32.1, and PCI has been mapped to chromosome 14. Genetic linkage analysis suggests tight linkage between PI and AACT. We have used pulsed-field gel electrophoresis to generate a physical map linking these five serpin genes. The order of the genetic loci is AACT/PCI-PI-PIL-CBG, with a maximum distance of about 220 kb between the AACT/PCI and PI genes. These genes form a PI cluster at 14q32.1, similar to that of the homologous genes on murine chromosome l2. The close proximity of these genes has implications for disease-association studies.     

Functional diversification during evolution of the murine alpha(1)-proteinase inhibitor family: role of the hypervariable reactive center loop

Alpha(1)-proteinase inhibitor (alpha(1)-PI) is a member of the serpin superfamily of serine proteinase inhibitors that are involved in the regulation of a number of proteolytic processes. Alpha(1)-PI, like most serpins, functions by covalent binding to, and inhibition of, target proteinases. The interaction between alpha(1)-PI and its target is directed by the so-called reactive center loop (RCL), an approximately 20 residue domain that extends out from the body of the alpha(1)-PI polypeptide and determines the inhibitor's specificity. Mice express at least seven closely related alpha(1)-PI isoforms, encoded by a family of genes clustered at the Spi1 locus on chromosome 12. The amino acid sequence of the RCL region is hypervariable among alpha(1)-PIs, a phenomenon that has been attributed to high rates of evolution driven by positive Darwinian selection. This suggests that the various isoforms are functionally diverse. To test this notion, we have compared the proteinase specificities of individual alpha(1)-PIs from each of the two mouse species. As predicted from the positive Darwinian selection hypothesis, the various alpha(1)-PIs differ in their ability to form covalent complexes with serine proteinases, such as elastase, trypsin, chymotrypsin, and cathepsin G. In addition, they differ in their binding ability to proteinases in crude snake venoms. Importantly, the RCL region of the alpha(1)-PI polypeptide is the primary determinant of isoform-specific differences in proteinase recognition, indicating that hypervariability within this region drives the functional diversification of alpha(1)-PIs during evolution. The possible physiological benefits of alpha(1)-PI diversity are discussed.     

Reaction of human skin chymotrypsin-like proteinase chymase with plasma proteinase inhibitors

The ability of plasma proteinase inhibitors to inactivate human chymase, a chymotrypsin-like proteinase stored within mast cell secretory granules, was investigated. Incubation with plasma resulted in over 80% inhibition of chymase hydrolytic activity for small substrates, suggesting that inhibitors other than alpha 2-macroglobulin were primarily responsible for chymase inactivation. Depletion of specific inhibitors from plasma by immunoadsorption using antisera against individual inhibitors established that alpha 1-antichymotrypsin (alpha 1-AC) and alpha 1-proteinase inhibitor (alpha 1-PI) were responsible for the inactivation. Characterization of the reaction between chymase and each inhibitor demonstrated in both cases the presence of two concurrent reactions proceeding at fixed relative rates. One reaction, which led to inhibitor inactivation, was about 3.5 and 4.0-fold faster than the other, which led to chymase inactivation. This was demonstrated in linear titrations of proteinase activity which exhibited endpoint stoichiometries of 4.5 (alpha 1-AC) and 5.0 (alpha 1-PI) instead of unity, and SDS gels of reaction products which exhibited a banding pattern indicative of both an SDS-stable proteinase-inhibitor complex and two lower Mr inhibitor degradation products which appear to have formed by hydrolysis within the reactive loop of each inhibitor. At inhibitor concentrations approaching those in plasma where inhibitor to chymase concentration ratios were in far excess of 4.5 and 5.0, the rate of chymase inactivation by both serpin inhibitors appeared to follow pseudo-first order kinetics. The "apparent" second order rate constants of inactivation determined from these data were about 3000-fold lower than the rate constants reported for human neutrophil cathepsin G and elastase with alpha 1-AC and alpha 1-PI, respectively. This suggests that chymase would be inhibited about 650-fold more slowly than these proteinases when released into plasma. These studies demonstrate that although chymase is inactivated by serpin inhibitors of plasma, both inhibitors are better substrates for the proteinase than they are inhibitors. This finding along with the slow rates of inactivation indicates that regulation of human chymase activity may not be a primary function of plasma.     

Canonical inhibitor-like interactions explain reactivity of alpha1-proteinase inhibitor Pittsburgh and antithrombin with proteinases

The serpin antithrombin is a slow thrombin inhibitor that requires heparin to enhance its reaction rate. In contrast, alpha1-proteinase inhibitor (alpha1PI) Pittsburgh (P1 Met --> Arg natural variant) inhibits thrombin 17 times faster than pentasaccharide heparin-activated antithrombin. We present here x-ray structures of free and S195A trypsin-bound alpha1PI Pittsburgh, which show that the reactive center loop (RCL) possesses a canonical conformation in the free serpin that does not change upon binding to S195A trypsin and that contacts the proteinase only between P2 and P2'. By inference from the structure of heparin cofactor II bound to S195A thrombin, this RCL conformation is also appropriate for binding to thrombin. Reaction rates of trypsin and thrombin with alpha1PI Pittsburgh and antithrombin and their P2 variants show that the low antithrombin-thrombin reaction rate results from the antithrombin RCL sequence at P2 and implies that, in solution, the antithrombin RCL must be in a similar canonical conformation to that found here for alpha1PI Pittsburgh, even in the nonheparin-activated state. This suggests a general, limited, canonical-like interaction between serpins and proteinases in their Michaelis complexes.     

Inactivation of blood plasma alpha 1-proteinase inhibitor as affected by 2 splenic thiol proteinases active in a neutral medium

It has been found that two active in neutral medium thiol proteinases from bovine spleen, cathepsin L and cathepsin H, bring about rapid and irreversible inactivation of alpha 1-proteinase inhibitor (alpha 1PI)--one of the major plasma inhibitors of serine proteinases. The activity of the enzymes studied did not change upon the interaction with alpha 1PI. With stoichiometric proteinase/inhibitor ratio, the inactivation of alpha 1PI under the effect of cathepsin L was instantaneous, while under the effect of cathepsin H it occurred within 30-60 min. The products of alpha 1PI inactivation had an inhibitory effect on the rate of its reaction with cathepsin L. alpha 1PI inactivation under the action of cathepsin L and cathepsin H was accompanied by the decrease in the molecular mass of the inhibitor from 54 kDA to 46 kDa. This was, probably, caused by the hydrolysis of the peptide bond formed by NH2 group of threonine. The 46 kDa fragment did not undergo further degradation. It did not bind to immobilized trypsin but retained antigenic properties. The results obtained show that the limited proteolysis is a mechanism of the inhibitor inactivation. It is suggested that under some conditions thiol proteinases, upon their release from the cell, participate in the control of effective alpha 1PI concentration.     

Interaction of subtilisins with serpins.

Serpins are well-characterized inhibitors of the chymotrypsin family serine proteinases. We have investigated the interaction of two serpins with members of the subtilisin family, proteinases that possess a similar catalytic mechanism to the chymotrypsins, but a totally different scaffold. We demonstrate that alpha 1 proteinase inhibitor inhibits subtilisin Carlsberg and proteinase K, and alpha 1 antichymotrypsin inhibits proteinase K, but not subtilisin Carlsberg. When inhibition occurs, the rate of formation and stability of the complexes are similar to those formed between serpins and chymotrypsin family members. However, inhibition of subtilisins is characterized by large partition ratios where more than four molecules of each serpin are required to inhibit one subtilisin molecule. The partition ratio is caused by the serpins acting as substrates or inhibitors. The ratio decreases as temperature is elevated in the range 0-45 degrees C, indicating that the serpins are more efficient inhibitors at high temperature. These aspects of the subtilisin interaction are all observed during inhibition of chymotrypsin family members by serpins, indicating that serpins accomplish inhibition of these two distinct proteinase families by the same mechanism.     

Matrix metalloproteinase expression and activity following prostaglandin F(2 alpha)-induced luteolysis

Luteal tissue contains matrix metalloproteinases (MMPs) that cleave specific components of the extracellular matrix (ECM) and are inhibited by tissue inhibitors of metalloproteinases (TIMPs). We previously reported a decrease in luteal TIMP-1 within 15 min of prostaglandin F(2 alpha) (PGF(2 alpha))-induced luteolysis. An increase in the MMP:TIMP ratio may promote ECM degradation and apoptosis, as observed in other tissues that undergo involution. The objectives of these experiments were to determine whether 1) PGF(2 alpha) affects expression of mRNA encoding fibrillar collagenases (MMP-1 and -13), gelatinases A and B (MMP-2 and -9), membrane type (mt)-1 MMP (MMP-14), stromelysin (MMP-3), and matrilysin (MMP-7), and 2) PGF(2 alpha) increases MMP activity during PGF(2 alpha)-induced luteolysis in sheep. Corpora lutea (n = 3-10/time point) were collected at 0, 15, and 30 min and 1, 2, 4, 6, 12, 24, and 48 h after PGF(2 alpha) administration. Northern blot analysis confirmed the presence of all MMPs except MMP-9. Expression of mRNA for the above MMPs (except MMP-2) increased significantly (P < 0.05) by 30 min, and all MMPs increased significantly (P < 0.05) by 6 h after PGF(2 alpha) administration. Expression of MMP-14 mRNA increased significantly (P < 0.05) by 15 min post-PGF(2 alpha) and remained elevated through 48 h. MMP activity in luteal homogenates (following proenzyme activation and inactivation of inhibitors) was increased significantly (P < 0.05) by 15 min and remained elevated through 48 h post-PGF(2 alpha). MMP activity was localized (in situ zymography) to the pericellular area of various cell types in the 0-h group and was markedly increased by 30 min post-PGF(2 alpha). MMP mRNA expression and activity were significantly increased following PGF(2 alpha) treatment. Increased MMP activity may promote ECM degradation during luteolysis.     

alpha1-Proteinase inhibitor forms initial non-covalent and final covalent complexes with elastase analogously to other serpin-proteinase pairs, suggesting a common mechanism of inhibition

Despite several concordant structural studies on the initial non-covalent complex that serpins form with target proteinases, a recent study on the non-covalent complex between the serpin alpha(1)-proteinase inhibitor (alpha(1)PI) and anhydroelastase concluded that translocation of the proteinase precedes cleavage of the reactive center loop and formation of the acyl ester. Because this conclusion is diametrically opposite to those of the other structural studies on serpin-proteinase pairs, we proceeded to examine this specific serpin-proteinase complex by the same successful NMR approach used previously on the alpha(1)PI-Pittsburgh-S195A trypsin pair. Both non-covalent complex with anhydroelastase and covalent complex with active elastase were made with (15)N-alanine-labeled wild-type alpha(1)PI. The heteronuclear single quantum correlation spectroscopy (HSQC) NMR spectrum of the non-covalent complex showed that the entire reactive center loop remained exposed, and the serpin body maintained a conformation indistinguishable from that of native alpha(1)PI, indicating no movement of the proteinase and no insertion of the reactive center loop into beta-sheet A. In contrast, the HSQC NMR spectrum of the covalent complex showed that the reactive center loop had fully inserted into beta-sheet A, indicating that translocation of the proteinase had occurred. These results agree with previous NMR, fluorescence resonance energy transfer, and x-ray crystallographic studies and suggest that a common mechanism is employed in formation of serpin-proteinase complexes. We found that preparations of anhydroelastase that are not appropriately purified contain material that can regenerate active elastase over time. It is likely that the material used by Mellet and Bieth contained such active elastase, resulting in mistaken attribution of the behavior of covalent complex to that of the non-covalent complex.     

Expression and characterization of recombinant Manduca sexta serpin- 1B and site-directed mutants that change its inhibitory selectivity

Hemolymph of Manduca sexta contains a number of serine proteinase inhibitors from the serpin superfamily. During formation of a stable complex between a serpin and a serine proteinase, the enzyme cleaves a specific peptide bond in an exposed loop (the reactive-site region) at the surface of the serpin. The amino acid residue on the amino-terminal side of this scissile bond, the P₁ residue, is important in defining the selectivity of a serpin for inhibiting different types of serine proteinases. M. sexta serpin-1B, with alanine at the position predicted from sequence alignments to be the P₁ residue, was previously named alaserpin. This alanyl residue was changed by site-directed mutagenesis to lysine (A343K) and phenylalanine (A343F). The serpin-1B cDNA and its mutants were inserted into an expression vector, H6pQE-60, and the serpin proteins were expressed in Escherichia coli. Affinity-purified recombinant serpins selectively inhibited mammalian serine proteinases: serpin-1B inhibited elastase; serpin-1B(A343K) inhibited trypsin, plasmin, and thrombin; serpin-1B(A343F) inhibited chymotrypsin as well as trypsin. All three serpins inhibited human cathepsin G. This insect serpin and its site-directed mutants associated with mammalian serine proteinases at rates similar to those reported for mammalian serpins. Serpin-1B and its mutants formed SDS-stable complexes with the enzymes they inhibited. The scissile bond was determined to be between residues 343 and 344 in wild-type serpin-1B and in serpin-1B with mutations at residue 343. These results demonstrate that the P₁ alanine residue defines the primary selectivity of serpin-1B for elastase-like enzymes, and that this selectivity can be altered by mutations at this position.     

Role of the connectivity of secondary structure segments in the folding of alpha(1)-antitrypsin

The native form of serpins (serine protease inhibitors) is metastable, which is critical to their biological functions. Spontaneous conversion from the native form of serpins into a more stable conformation, called the "latent" form, is restricted. To examine whether the connectivity of strand 1 of beta-sheet C to the hydrophobic core is critical to the serpin's preferential folding to the metastable native conformation, we designed a circularly-permuted mutant of alpha(1)-antitrypsin, the prototype serpin, in which strand 1C is disconnected from the hydrophobic core. Conformation of the circular permutant was similar to that of the latent form, as revealed by equilibrium unfolding, limited proteolysis, and spectroscopic properties. Our results support the notion that rapid folding of the hydrophobic core with concomitant incorporation of strand 1C into beta-sheet C traps the serpin molecule into its native metastable conformation.     

alpha(1)-Proteinase inhibitor mutants with specificity for plasma kallikrein and C1s but not C1

Coagulation and complement proteinases are activated in sepsis, and one approach to therapy is to develop proteinase inhibitors that will specifically inhibit these proteinases without inhibiting activated protein C, a proteinase that is beneficial to survival. In this study, we made mutants of the serpin alpha(1)-PI, designed to mimic the specificity of C1-inhibitor. The P3-P2-P1 residues of alpha1-PI were changed from IPM to LGR and PFR, sequences preferred by C1s and kallikrein, respectively. Inhibition of C1s, kallikrein, factor XIIa, and activated protein C was assessed by SDS-PAGE, and by determination of the k(app) and SI. alpha(1)-PI-LGR inhibited C1s with a rate of 7790 M(-1)s(-1), but only minimal inhibition of C1 in a hemolytic assay was observed. Kallikrein, factor XIIa, and activated protein C were inhibited with rates of 382,180 M(-1)s(-1), 10,400 M(-1)s(-1), and 3500 M(-1)s(-1), respectively. alpha(1)-PI-PFR was a poor inhibitor of C1s, factor XIIa, and activated protein C, but had enhanced reactivity with kallikrein. Changing the P4' residue of alpha(1)-PI-LGR Pro to Glu reduced the activity with C1s, consistent with the idea that C1s requires hydrophobic residues in this region of the serpin for optimal interaction. The data provide insight into the requirements for kallikrein and C1s inhibition necessary for designing inhibitors with appropriate properties for further investigation as therapeutic agents.     

Mutant recombinant serpins as highly specific inhibitors of human kallikrein 14

The reactive center loop (RCL) of serpins plays an essential role in the inhibition mechanism acting as a substrate for their target proteases. Changes within the RCL sequence modulate the specificity and reactivity of the serpin molecule. Recently, we reported the construction of alpha1-antichymotrypsin (ACT) variants with high specificity towards human kallikrein 2 (hK2) [Cloutier SM, Kündig C, Felber LM, Fattah OM, Chagas JR, Gygi CM, Jichlinski P, Leisinger HJ & Deperthes D (2004) Eur J Biochem271, 607-613] by changing amino acids surrounding the scissile bond of the RCL and obtained specific inhibitors towards hK2. Based on this approach, we developed highly specific recombinant inhibitors of human kallikrein 14 (hK14), a protease correlated with increased aggressiveness of prostate and breast cancers. In addition to the RCL permutation with hK14 phage display-selected substrates E8 (LQRAI) and G9 (TVDYA) [Felber LM, Borgo?o CA, Cloutier SM, Kündig C, Kishi T, Chagas JR, Jichlinski P, Gygi CM, Leisinger HJ, Diamandis EP & Deperthes D (2005) Biol Chem386, 291-298], we studied the importance of the scaffold, serpins alpha1-antitrypsin (AAT) or ACT, to confer inhibitory specificity. All four resulting serpin variants ACT(E8), ACT(G9), AAT(E8) and AAT(G9) showed hK14 inhibitory activity and were able to form covalent complex with hK14. ACT inhibitors formed more stable complexes with hK14 than AAT variants. Whereas E8-based inhibitors demonstrated a rather relaxed specificity reacting with various proteases with trypsin-like activity including several human kallikreins, the two serpins variants containing the G9 sequence showed a very high selectivity for hK14. Such specific inhibitors might prove useful to elucidate the biological role of hK14 and/or its implication in cancer.