Serpin Polymerization Is Prevented by a Hydrogen Bond Network That Is Centered on His-334 and Stabilized by Glycerol*

Polymerization of serpins commonly results from mutations in the shutter region underlying the bifurcation of strands 3 and 5 of the A-sheet, with entry beyond this point being barred by a H-bond network centered on His-334. Exposure of this histidine in antithrombin, which has a partially opened sheet, allows polymerization and peptide insertion to occur at pH 6 or less when His-334 will be predictably protonated with disruption of the H-bond network. Similarly, thermal stability of antithrombin is pH-dependent with a single unfolding transition at pH 6, but there is no such transition when His-334 is buried by a fully closed A-sheet in heparin-complexed antithrombin or in (cid:1) 1 -antitrypsin. Replace- ment of His-334 in (cid:1) 1 -antitrypsin by a serine or alanine at pH 7.4 results in the same polymerization and loop-peptide acceptance observed with antithrombin at low pH. The critical role of His-334 and the re-formation of its H-bond network by the conserved P8 threonine, on the full insertion of strand 4, are relevant for the design of therapeutic blocking agents. This is highlighted here by the crystallographic demonstration that glycerol, which at high concentrations blocks polymerization, can replace the P8 threonine and re-form the disrupted H-bond network with His-334.

The serpin family of serine protease inhibitors (1) provides a clear example of the way in which dysfunction and disease can result from conformational instability (2,3). The inhibitory function of the serpins is dependent on a triggered opening of the five-stranded A ␤-sheet of the molecule to allow the insertion of the cleaved reactive center loop as an additional strand in the center of the sheet (4 -6). As a consequence, the serpins are particularly vulnerable to mutations affecting the critical region of the molecule underlying the point of entry of the loop, between strands 3 and 5 of the sheet (7). Mutations in this shutter region (Fig. 1a) allow the aberrant opening of the A-sheet, with the risk of the insertion, into its lower half, of the reactive loop of another molecule to give intermolecular linkage and polymerization of the serpin. Even minor changes in the shutter region of ␣ 1 -antitrypsin (8) and antichymotrypsin (9) result in their polymerization and intracellular aggregation with consequent lung and liver disease and similarly with C1-inhibitor (10) and antithrombin mutations (11) resulting in angioedema and thrombosis. But the best example of the critical function of this region comes from recent investigations of a novel form of familial neurodegenerative disease due to the aggregation within neurons of a brain-specific serpin, neuroserpin (12). The polymerization and aggregation of neuroserpin results from mutations in its shutter region. Two of these mutations affect residues previously identified in other serpin diseases, Ser-53 and -56 (template numbering (13) is used throughout this paper). The dysfunction accompanying mutations of serines 53 and 56 is readily explicable by the alteration in packing of residues underlying the A-sheet at the point where a sliding movement opens the gap between strands 3 and 5A (2,7). But a third and novel mutation in neuroserpin (14,15) has also focused attention on histidine 334 at the point of bifurcation of strands 3 and 5 ( Fig. 1, b and c). This conserved histidine is centrally placed in the shutter region and is less directly involved in the critical packing interactions of sheet opening than either serine 53 or 56, but surprisingly the mutation of His-334 results in a much more severe neurodegenerative disease. An explanation for this is likely to be the key H-bond network centered on His-334, which bridges strands 3 and 5 of the A-sheet and notably links His-334 directly to the base of Ser-53 and indirectly to Ser-56 ( Fig. 1, b and c).
The potential role of His-334 as a guardian of sheet opening is indicated in the crystallographic structures of two serpins, antithrombin and heparin cofactor II (16,17). Both of these serpins are exceptional in having, in their native states, a partial opening of the A-sheet with an initial insertion of the reactive loop (to a level of P14, see Fig. 1a). In both, the hydrogen bonding between strands 3 and 5 does not commence until His-334 (at the level of insertion of P8 in cleaved serpins), which appears as the first barrier to further opening of the sheet (Fig. 1b). The consequences of the loss of this barrier and of opening the A-sheet are potentially 2-fold. It can allow the complete insertion of the uncleaved reactive loop of the molecule to give the inactive latent form or it can allow the incorporation of the reactive center loop of another molecule resulting in serpin polymerization. Transition to the latent state takes place as a physiological mechanism in the plasminogen activator inhibitor-1 (18), which exceptionally among the serpins has a glutamine rather than a histidine at 334 (19). Transition to the latent form is also a significant pathological mechanism in the conformationally unstable mutants of antithrombin in which a premature conversion to the latent conformation can result in a catastrophic decrease in antithrombin activity with massive thrombosis (20). More commonly though, with most serpins such as ␣ 1 -antitrypsin, the opening of the A-sheet predominantly results in the formation of loop-sheet polymers with the insertion of the P7-3 portion of the loop of one molecule into the lower half of the A-sheet of the next molecule (21). Here we show how studies of both types of * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The transition in antithrombin and ␣ 1 -antitrypsin confirm the critical function of His-334 as a barrier to premature opening of the A-sheet.

MATERIALS AND METHODS
Materials-Restriction enzymes and T4 DNA ligase were purchased from New England Biolabs, and oligonucleotides were synthesized by MWG-Biotech. The expression vector pQE31 and Escherichia coli strain SG13009 (pREP4) were from Qiagen. Isopropyl-␤-D-thiogalactopyranoside was from Melford Laboratories Ltd. (Suffolk, England). Kanamycin sulfate was from Roche Molecular Biochemicals. Heparin-Sepharose, HiTrap Q-Sepharose, and HiTrap chelating columns were from Amersham Biosciences. Trypsin and ampicillin were from Sigma. The substrate S-2222 was from Chromogenix. Peptides encoding P14-3 (acetyl-SEAAASTAVVIA) and P14-9 (acetyl-SEAAAS) of antithrombin reactive loop were synthesized by the Department of Biochemistry, University of Cambridge, Cambridge, UK, and an analogue peptide (FLEAIG) of P7-2 of antitrypsin reactive loop, where proline at P2 position was replaced by a glycine, was synthesized by MWG-Biotech. The peptide formyl-MLF was purchased from Sigma. High affinity heparin pentasaccharide (H5*), which has an extra sulfate group and higher affinity for antithrombin (22), was a gift from Dr. Maurice Petitou (Sanofi Research, Toulouse, France).
Purification and Preparation of Native and Latent Antithrombin-Native antithrombin was purified from frozen plasma by precipitation of plasma with dextran sulfate and calcium chloride (23,24) after which the supernatant was diluted with an equal volume of equilibration buffer (50 mM Tris-HCl, 10 mM sodium citrate, 5 mM sodium EDTA, 0.4 M NaCl, pH 7.4). This mixture was applied to a heparin-Sepharose affinity chromatography column previously equilibrated with the same buffer, and the column was thoroughly washed with equilibration buffer. The antithrombin was eluted using a gradient from 0.4 to 2 M NaCl in the equilibration buffer. Antithrombin peaks were further purified by anion exchange chromatography on a HiTrap Q-Sepharose column. Antithrombin was concentrated in 10 mM Tris-HCl, 50 mM NaCl, 1 mM EDTA, pH 7.4, and snap-frozen.
Latent antithrombin was prepared using the glycerol method (25). Briefly, native antithrombin was incubated with 40% glycerol in 50 mM phosphate buffer (pH 6) for 24 -36 h at 50°C, which gave nearly 100% latent formation. After incubation, the samples were diluted 4-fold with 10 mM Tris-HCl, pH 7.4, and 1 mM EDTA loaded onto a HiTrap-Q-Sepharose column and washed with the same buffer. Latent antithrombin was eluted with a 0 -0.5 M NaCl gradient in 10 mM Tris-HCl, pH 7.4, and 1 mM EDTA. The concentration of antithrombin was determined using an extinction coefficient of 6.5 (26).
Preparation of Recombinant Antitrypsin Variants-Human ␣ 1 -antitrypsin cDNA was amplified by polymerase chain reaction and inserted into the expression vector pQE31 as previously described (27). Mutagenesis was carried out by a two-step polymerase chain reaction. The recombinant ␣ 1 -antitrypsin was expressed with an MRSHHHHHH tag at the N terminus and purified from the soluble fraction of E. coli lysate. Briefly, expression plasmids were transformed into SG13009 (pREP4) cells and grown in 2 liters of 2ϫ tryptone-yeast extract medium at 37°C until A 600nm ϭ 0.8 -1.0; then isopropyl-␤-D-thiogalactopyranoside was added to a concentration of 1 mM, and the culture was transferred to 30°C for a further 3 h. The cells were collected by centrifugation, resuspended in buffer A (25 mM phosphate buffer, pH 8.0, 0.5 M NaCl, and 1 mM ␤-mercaptoethanol), and disrupted by sonication. The supernatant of the cell lysate was loaded onto a HiTrap nickel-chelating column (5 ml), and after washing to baseline with buffer A, the bound protein eluted as a shouldered peak with an imidazole gradient (0 -0.2 M). The fractions containing antitrypsin were collected, dialyzed against buffer B (10 mM Tris-Cl, pH 8.0, 1 mM EDTA, and 1 mM ␤-mercaptoethanol), and loaded onto a HiTrap Q-Sepharose column (5 ml). The column was then washed with a NaCl gradient (0 -0.5 M) in buffer B. ␣ 1 -Antitrypsin was eluted as the major peak (second peak) around 0.2 M NaCl. The fractions were pooled, concentrated, and snap-frozen in liquid nitrogen. The protein concentrations were determined using an extinction coefficient of 5.3 (28). All ␣ 1 -antitrypsin variants were confirmed to be in pure monomeric form by SDS and native PAGE.
Complex Formation between Antitrypsin Variants and Trypsin-␣ 1 -Antitrypsin was mixed with trypsin (1:2 molar ratio) at room temperature (22 Ϯ 1°C) for 15 min. Samples were then mixed with reduced SDS-loading buffer and heated at 95°C for 5 min in a PCR block. SDS-gel electrophoresis (29) was performed in a 12% gel, and the protein was visualized with Coomassie Blue.
Stoichiometries and Rates of Inhibition-Stoichiometries of inhibi-tion were determined by incubating increasing concentrations of ␣ 1antitrypsin with a fixed concentration of trypsin (1 M). Reactions were incubated at room temperature (22 Ϯ 1°C) for 30 min. The residual amidolytic activity was determined by the addition of 0.1 mM S-2222 substrate in PBS with 0.1% PEG 8000. Linear regression analysis of the decrease in protease activity with an increasing concentration of ␣ 1antitrypsin yielded the estimates for the stoichiometry of inhibition as the intercept on the abscissa. The rates of inhibition of trypsin by recombinant ␣ 1 -antitrypsin variants were determined at room temperature by a discontinuous assay procedure as previously described (27).
Briefly, under pseudo first-order conditions, 10 l of 0.5 M ␣ 1 -antitrypsin variants were mixed with 10 l of 20 nM trypsin in PBS with 0.1% PEG 8000. The residual protease activity was determined at timed intervals by diluting the reaction mixture into 1 ml of the assay buffer containing 0.1 mM S-2222 substrate. The observed rate constant, k obs , for the reaction was obtained from the slope of a semilog plot of the residual protease activity against time, and the second-order rate constant, k app , was calculated by dividing k obs by the initial ␣ 1 -antitrypsin concentration.
Thermal Stability-Circular dichroism (CD) experiments were performed using a JASCO J-810 spectropolarimeter. Samples of ␣ 1 -antitrypsin or antithrombin were prepared in 0.1 M sodium acetate for pH 4 -6, 0.1 M sodium phosphate for pH 6 -8, 0.1 M Tris for pH 8 -9, and 0.1 M glycine for pH 9 -10. All buffers were filtered, and samples were centrifuged before the experiment. Thermal unfolding experiments were performed by monitoring the CD signal at 222 nm between 25 and 95°C using a heating rate of 2°C/min at a concentration of 0.25 mg/ml for antitrypsin variants and 0.5 mg/ml for antithrombin. Melting points (T m ) were calculated using an expression for a two-state transition as described previously (30). All the results are the average of three experiments. The pH-dependent antithrombin thermal melting points were treated as a single transition between pH 5 and 8, and the pK a of this transition was fitted by GraFit (version 3.0, Erithacus Software Ltd.).
Equilibrium Unfolding of Antithrombin-Equilibrium unfolding was monitored by spectrofluorimetry ( ex ϭ 290 nm, em ϭ 350 nm) in the presence of guanidine chloride (GdmCl). 1 Antithrombin was incubated in a buffer (50 mM phosphate or sodium acetate, 50 mM NaCl, and 1 mM EDTA) containing various concentrations of GdmCl at 25°C for about 16 h before spectral measurements. The final concentration of protein was 20 g/ml. Equilibrium unfolding was fitted to a two-state model, and the midpoints [GdmCl] 1/2 of unfolding curves at different pH were plotted against pH.
Polymerization of Antitrypsin Variants-Antitrypsin variants were incubated at 1 mg/ml in 20 mM Tris-HCl, 50 mM NaCl, pH 7.4, and 1 mM EDTA for 1 h at temperature from 37 to 65°C. Samples were loaded onto an 8% (w/v) native gel. The proteins were visualized by staining with Coomassie Blue.
Peptide Insertion-Antithrombin at 0.5 mg/ml or ␣ 1 -antitrypsin variants at 0.68 mg/ml were incubated at 37°C in the presence of a 100-fold molar excess of the P14-P3 12-mer peptide (acetyl-SEAAASTAVVIA) at different pH values or the 6-mer peptide (FLEAIG) for different time intervals. The results were analyzed by loading the samples onto an 8% (w/v) native gel with 7 M urea.
Crystallization of Antithrombin/Peptides Complexes-Native antithrombin (1 mg/ml) was incubated with 2 mM of P14-8 peptide of antithrombin (acetyl-SEAAAS) and a tri-peptide (formyl-MLF) at 37°C for 24 h. Samples were then washed several times with 10 mM Tris-HCl, pH 7.4, in a concentrator to remove most of the free peptides, with concentration to 14 mg/ml. Crystallization was performed using hanging drop methods as previously described (31) with the following modification. Antithrombin-peptide complexes were mixed with equal amounts of latent antithrombin and equilibrated against 10 -20% PEG 4000 in 50 mM sodium cacodylate buffer, pH 6.8, and 0.2 M NH 4 F, with or without 12% glycerol. Crystals grew to full size in ϳ1 week.
Data Collection and Refinement-Diffraction data to 2.8 Å were collected from a single frozen crystal at Daresbury Synchrotron Radiation Source (station 14-2) and processed using Mosflm (32) and Scala (33). Processing statistics are given in the Table I. Since the crystal of the ternary peptide complex was essentially isomorphous with the structure of the previously solved antithrombin binary-peptide complex (34), the original binary complex coordinates (PDB accession number 1BR8) were used as the starting model. Refinement to 2.8 Å was performed in crystallography NMR software (35) using the maxi-mum likelihood target of Pannu and Read (36). The model was built with O (37). The final refined structure (Table I) (41), and figures for antithrombin ( Fig. 1, a, b, c) were based on the highest resolution native structure PDB 1E04.

RESULTS
Polymers, Latency, and Glycerol-Prolonged heating of antithrombin and ␣ 1 -antitrypsin at temperatures above 50°C results not only in their polymerization (Fig. 2a) but also in a transition of a proportion of molecules to the latent conformation. This transition to the latent form is minor and barely detectable with ␣ 1 -antitrypsin. However, latent antithrombin is readily formed, though it may not be apparent due to the immediate formation of a heterodimer (20) with reversible sheet C linkage of the latent antithrombin to a molecule of active native antithrombin (Fig. 2b). With antithrombin, the addition of glycerol to the incubation buffer suppresses polymer formation but allows the latent transition (20), such that incubation at 50°C of antithrombin in  40% glycerol results in the quantitative preparation of latent antithrombin (25), free of polymers (Fig. 2b).
pH Dependence of Transitions-The incubation of antithrombin at 50°C for 16 h over a range of pH values (Fig. 2c) shows the ready formation of polymers up to pH 5.5 but insignificantly so at pH values greater than 6. Above pH 6 there is a much slower conformational transition, with the formation of latent antithrombin (apparent as the heterodimer) and much less polymer formation until pH 10. The effect of pH on the opening of the A-sheet of serpins can also be assessed by the readiness with which they complex with synthetic P14-3 peptides. This is clearly seen in Fig. 2d which shows the formation of the binary complex (seen as a 7 M urea stable component) in antithrombin incubated for 12 h at 37°with the P14-3 peptide. The rapid formation of the complex at pH 5 as compared with pH 6 and 7 indicates that the low pH favors sheet opening and hence peptide annealment.
pH and Thermal Stability-Serpins typically have a well defined thermal transition with antithrombin having a melting point (T m ) of 57.6°C at pH 7.4. Antithrombin differs, however, from the archetypal serpin ␣ 1 -antitrypsin in that its T m undergoes an atypical change with pH (Fig. 3a) with a transition at apparent pK a of 6.0 Ϯ 0.1 (Fig. 3b). A transition at pH 6 was similarly observed with guanidine chloride-induced antithrombin unfolding (data not shown). The inflection point at pH 6 suggested that the thermal stability of antithrombin was dependent on the protonation of an individual histidine residue, with the likely candidate being His-334. This histidine is uniquely exposed in antithrombin, which has a partially opened A-sheet, with His-334 forming the first interlinking H-bonds between s3A and s5A (Fig. 1, b and c). By comparison, in ␣ 1 -antitrypsin, which is more thermally stable with T m 63.4, His-334 is protected by the full closure of strands 3 and 5 in the A-sheet. To confirm that this difference in pH dependence is related to the exposure of His-334, the change in T m was determined with a closed A-sheet conformation of antithrombin. Closure of the A-sheet of antithrombin was induced by addition of the core heparin pentasaccharide, which has been shown to result in full closure of the A-sheet with burial of His-334 to give a conformation of antithrombin superimposable with that of ␣ 1 -antitrypsin (31). As shown in Fig. 3c, addition of the heparin pentasaccharide to antithrombin results in an increased thermal stability with conversion to a pH dependence curve similar to that of ␣ 1 -antitrypsin and without the inflection at pH 6.
Recombinant Replacement of His-334 -To assess the contribution of His-334 to the stability of ␣ 1 -antitrypsin, variants were expressed recombinantly with substitutions of His-334 by serine and alanine and with the variants named here as Wt for the wild type and H334S and H334A, respectively, for the variants. Both ␣ 1 -antitrypsin variants retain their inhibitory activity and form stable complexes with trypsin identical to those of Wt (Fig. 4a), and rates and stoichiometries of trypsin inhibition were unchanged (data not shown). Incubation of H334A for 1 h at a range of temperatures showed substantial polymerization at 45°C, some 5-10°C prior to equivalent polymerization of the recombinant Wt ␣ 1 -antitrypsin (Fig. 4b).
The H334S variant had a smaller increase in polymerization, intermediate between the Wt and H334A forms. The more ready opening of the A-sheet of H334A versus Wt ␣ 1 -antitrypsin is shown by the increased rate of peptide (P7-2) insertion with the H334A variant (Fig. 4c). The inherent change in stability of the three forms is reflected in their thermal stability, with a decrease from a T m of 63.4°C in Wt ␣ 1 -antitrypsin to 59.6°C in H334S and to 58.2°C in H334A. The H334A mutant has a similar pH-dependent T m curve to those of antitrypsin and antithrombin-heparin pentasaccharide complexes (data not shown). This indicates that the transitions between pH 4 and 5.5 (Fig. 3, b and c) do not represent the protonation of His-334, which is consistent with the burial of His-334 both in antitrypsin and heparin-pentasaccharide-complexed antithrombin.
Crystallographic Detail of His-334 Interaction with Glycerol-A crystallographic finding, reported here (PDB accession number 1LK6), provides an unexpected insight with particular relevance to His-334. The finding was incidental to the determination of the structure of a ternary complex of antithrombin as part of a larger 2 series of serpin-peptide complex structures. The ternary complex was formed by incubation of antithrombin with P14-9 and P6-4 synthetic loop peptides in the presence of glycerol. The 2.8 Å structure shows a glycerol molecule sited in the position occupied in six-stranded antithrombin by the side chain of threonine P8 (Fig. 1, d and e). Critically, a hydroxyl of the glycerol forms a hydrogen bond with the ␦ nitrogen of the His-334 imidazole with the bond length of 2.9 Å being identical to that formed by the hydroxyl of the P8 threonine. Moreover in each case the His-334 imidazole is oriented such that its ⑀ nitrogen is maintained at an optimal hydrogen bond distance of 3.1 Å from the main chain amine of Ser-53. But as well as this specific linkage to His-334, a series of hydrogen bonds are also formed with surrounding structures including the main chain oxygen of Phe-333 and the carboxyl group of the P9 residue. Subsequently, we have also shown crystallographically that glycerol insertion, as in Fig. 1, d and e, takes place even upon rapid exposure of formed crystals to glycerol just prior to diffraction.

DISCUSSION
The findings here open a clearer understanding of the mechanisms leading to aberrant conformational changes in the serpins. In particular the pH dependence of these changes strongly supports structural evidence as to the critical role of His-334 in maintaining the metastable inhibitory conformation. In antithrombin, His-334 in strand 5A is clearly seen as a barrier to further insertion of the reactive loop, due to the bridging linkage of its imidazole side chain to Asn-186 in strand 3A and to Ser-53 in the underlying ␤-sheet (Fig. 1, b and  c). The significance of these interactions is highlighted by the identification of mutations in the shutter region causing familial dementias: at His-334, Ser-53, and also at Ser-56, which bonds to Asn-186 (14,15). The mutations predictably result in a laxity in sheet opening at 37°C similar to that induced by incubation of the normal protein at 50°C. The consequence of such incubation (Fig. 2, a-c) is the formation of either intermolecular loop-sheet linkages to give polymers or the monomeric transition to the latent conformation with a full insertion of the reactive loop into the A-sheet. The demonstration here that both polymerization and peptide annealing of antithrombin occurred at pH below 6 (Fig. 2, c and d) together with the structurally known exposure of His-334 in antithrombin, indicates that sheet opening is facilitated by the protonation of His-334. This protonation of the ⑀ nitrogen of the imidazole will break the hydrogen bond that anchors His-334 to Ser-53 in the underlying B-sheet (Fig. 3d). The accompanying acquisition of a positive charge will disfavor the burying of the imidazole with the combined effects giving disruption of the hydrogen bonds that link strands 3 and 5 of the A-sheet.
In keeping with the conclusion that sheet opening is influpentasaccharide complexes were also fitted, but these transitions do not represent the protonation of His-334 as a similar transition was observed with the H334A mutant. enced by the protonation of His-334, the thermal stability of antithrombin was shown to be pH-dependent, with a characteristic transition at pH 6 ( Fig. 3b). In comparison to this, there is no such pH transition in stability in ␣ 1 -antitrypsin in which the histidine is buried in a tightly closed A-sheet. Moreover, when antithrombin is converted to a form with a similarly closed A-sheet, by complexing with heparin, its T m curve loses the inflection at pH 6 and reverts to the typical serpin response as seen with ␣ 1 -antitrypsin (Fig. 3c). Although there are four histidines in antithrombin, only His-334 undergoes a radical change in environment, due to burying, on heparin activation. Evidence that His-334 has the same protective function against aberrant sheet opening in ␣ 1 -antitrypsin is provided by recombinant substitutions at 334 with alanine (H334A) and serine (H334S). These substituted variants of ␣ 1 -antitrypsin have properties similar to that of antithrombin when its pH value is decreased to below 6. The H334A and H334S variants of ␣ 1antitrypsin, as compared with the wild type recombinant, have a large decrease in T m and more readily form polymers and accept synthetic loop-peptides (Fig. 4). The shutter control of sheet opening is a delicately tuned mechanism, and even slight perturbations may have disastrous functional consequences. For example, two relevant natural antithrombin variants with minor shutter mutations and a decrease in T m of just over 1°C result in severe episodic thrombosis (24,42).
An understanding of the mechanism controlling sheet opening is needed for the design of therapeutic agents to prevent the pathological polymerization of serpins. Although the backbone position of His-334 is at the level of P8 of a fully inserted reactive center loop, the imidazole side chain of the histidine, by its H-bonding to Asn-186, effectively blocks entry to the sheet beyond the level of P12 (Fig. 1, b and c). So entry of the reactive loop or its homologue peptides, to level P10 and beyond, will disrupt the linkage between s3A and s5A and allow full opening of the sheet. Thus insertion of P14-9 or P14-8 peptides will enable the entry of the P7-3 sequence of the reactive loop of another molecule into the lower half of the opened sheet, with the resultant formation of polymers. But normally there will be preferential entry of the molecule's own reactive loop, facilitated by the ability of the side chain of its P8 threonine to re-form the H-bond network with His-334 of s5A and Asn-186 in s3A. The threonine at P8 is conserved in almost all serpins (13) with a notable exception being in human ␣ 1 -antitrypsin where there is a methionine (this explains why the P14-3 homologue peptide of antithrombin more readily inserts into ␣ 1 -antitrypsin than does its own P14-3 peptide (43)). The clear preference for a threonine at P8 is also shown in another particularly relevant structure, that of a shutter mutant (L55P) of ␣ 1 -antichymotrypsin (44). This pathological mutant is seen in a frozen transitional form, with the reactive loop inserted to P12 but with further entry blocked by the insertion into the P8 position of a threonine from the adjacent F helix. Intriguingly this threonine at position 165 at end of the F helix is invariantly conserved in the inhibitory serpins.
A demonstration of the way insertion of the reactive loop to P9 disrupts the His-334-Asn-186 bond between s3 and s5A is seen in our crystallographic structure of a complex of antithrombin with a P14-9 peptide (Fig. 1, d and e). In this structure, stabilized by a further insertion of a P6-4 tripeptide, a gap has opened between His-334 and the side chain of Asn-186 that is normally bridged by hydrogen bonding to the side chain of the P8 threonine. However, the unexpected finding in the structure is the presence in the P8 position of a glycerol molecule, oriented similarly to that of the side chain of the threonine normally at P8. The glycerol, as with the threonine, is hydrogen-bonded to His-334 with the precise bond distances that ensure the anchoring of the imidazole ring to the main chain of Ser-53. Thus the glycerol has in effect re-formed the bonding network that stabilizes His-334. Although this new network does not link to Asn-186 in s3A, the presence of the glycerol will impede the insertion of external peptides in the P8-P7 position. This, together with the potential for glycerol to also insert into other vacant strand positions, provides an additional explanation for the efficacy of glycerol in the protection of antithrombin against polymerization (Fig. 2b). This protective effect of high concentrations of glycerol had been assumed to be due to its decreasing the rate of diffusion and hence of the collisions between individual molecules required for polymerization. But whereas polymerization will result from random intermolecular collisions, the monomeric transformation to the latent form is due to the ordered entry of the molecule's own reactive loop into the sheet. This entry of the optimally oriented side chains of the reactive loop, including the P8 threonine, should readily displace any in situ glycerol molecules. Thus the latent transition is less affected by the presence of glycerol (Fig. 2b). Nor is it as dependent as polymerization on the opening of the A-sheet as the latent transition continues to occur at higher pH values when the His-334 network is intact and the sheet is predictably closed (Fig. 2c). The likely limiting factor in the latent transition is not so much the opening of the A-sheet as the release of the intact loop by thermal dissociation at its distal hinge, s1C (25,45). This is consistent with previous observation that replacement of glutamine with histidine at 334 in plasminogen activator inhibitor-1 does not decrease latent transition (46). The prevention of polymerization by peptides that insert into the opened A-sheet has been demonstrated in vitro and shown to be selectively achievable (34,(47)(48)(49)(50). The therapeutic challenge is to convert these peptides into pharmacologically effective in vivo agents. The identification here of a focal point for sheet opening gives encouragement as to the feasibility of designing smaller and more effective peptide blockers of polymerization to prevent the pathological polymerization of serpins. Specifically, the demonstration that glycerol can readily act as a surrogate for the critical side chain of the P8 threonine, opens the prospect of achieving an ultimate aim in the field, the development of non-peptide blocking agents.