Javier Corral, University of Murcia, Centro Regional de Hemodonación, Ronda de Garay S/N, Murcia 30003, Spain. Tel.: +34 96 8341990; fax: +34 96 8261914. E-mail: email@example.com
Summary. Background: The metastable native conformation of serpins is required for their protease inhibition mechanism, but also renders them vulnerable to missense mutations that promote protein misfolding with pathological consequences.
Objective: To characterize the first antithrombin deficiency caused by a large in-frame insertion.
Patients/Methods: Functional, biochemical and molecular analysis of the proband and relatives was performed. Recombinant antithrombin was expressed in HEK-EBNA cells. Plasma and recombinant antithrombins were purified and sequenced by Edman degradation. The stability was evaluated by calorimetry. Reactive centre loop (RCL) exposure was determined by thrombin cleavage. Mutant antithrombin was crystallized as a dimer with latent plasma antithrombin.
Results: The patient, with a spontaneous pulmonary embolism, belongs to a family with significant thrombotic history. We identified a complex heterozygous in-frame insertion of 24 bp in SERPINC1, affecting strand 3 of β-sheet A, a region highly conserved in serpins. Surprisingly, the insertion resulted in a type II antithrombin deficiency with heparin binding defect. The mutant antithrombin, with a molecular weight of 59 kDa, had a proteolytic cleavage at W49 but maintained the N-terminal disulphide bonds, and was conformationally sensitive. The variant was non-inhibitory. Analysis of the crystal structure of the hyperstable recombinant protein showed that the inserted sequence annealed into β-sheet A as the fourth strand, and maintained a native RCL.
Conclusions: This is the first case of a large in frame-insertion that allows correct folding, glycosylation, and secretion of a serpin, resulting in a conformationally sensitive non-inhibitory variant, which acquires a hyperstable conformation with a native RCL.
In a closed and highly pressured circulatory system, a rapid and potent response of the hemostatic system is vital to avoid the otherwise fatal consequences of bleeding caused by vascular damage. Thus, a complex network of serial proteolytic reactions rapidly generates thrombin and a fibrin clot . However, such a procoagulant potential must be correctly controlled, to prevent inappropriate, excessive or mislocalized clotting of blood, which may cause thrombotic disorders. Endogenous anticoagulants are required for this control. Among them, antithrombin plays a key role, as it inhibits several procoagulant serine proteases such as FIXa, FXa and thrombin [2,3]. Accordingly, antithrombin deficiency significantly increases the risk of thrombosis, and complete deficiency causes embryonic lethality .
Antithrombin is a member of the serpin superfamily, which comprises a wide variety of serine protease inhibitors. Members of the serpin superfamily modulate different proteolytic cascades in a wide array of physiological systems by a common inhibitory mechanism that necessitates a remarkable conformational change and requires that serpins fold into a metastable native conformation . Unfortunately, this metastable conformation also renders serpins vulnerable to different environmental and genetic factors [6,7]. Thus, missense mutations may promote misfolding, leading to inactive, hyperstable conformations (latent or polymers) with potential pathological consequences .
Mutations in the SERPINC1 gene may result in two types of antithrombin deficiencies . Type I deficiencies are featured by the absence of mutant protein in plasma. These mutations can be caused by unstable mRNA, premature stop codons, and/or abnormal folding of the protein that is not secreted. Type II deficiencies are characterized by the secretion of a variant protein with impaired or null inhibitory activity. These mutations usually affect functional domains, such as the heparin binding site or the reactive centre loop (RCL), or have a pleiotropic effect.
As expected for such a conformationally sensitive protein, insertions or deletions affecting SERPINC1, including those maintaining the open reading frame, have been identified in patients with type I deficiency . There is only one type II deficiency caused by three nucleotides in-frame deletion: antithrombin London, which lacks Arg393 (the P1 residue of the RCL). This mutant is secreted but obviously has no inhibitory activity, and is associated with a high risk of thrombosis .
Here we describe the first type II antithrombin deficiency caused by a large in-frame insertion in the coding region of the SERPINC1 gene. Biochemical characterization and a 2.8 Å resolution X-ray crystal structure reveal unique features that account for its reduced affinity for heparin and its substrate behaviour towards thrombin. The implications of this mutant in the thrombotic disease and the polymerization process of serpins are discussed.
Materials and methods
Functional and genetic analysis
Activity, antigen levels and heparin affinity of plasma antithrombin from the patient, and PCR amplification and sequencing of SERPINC1 gene, were performed essentially as previously reported . Thrombophilic tests including protein C, protein S, antiphospholipid antibodies, FV Leiden and prothrombin G20210A were also performed.
Polyacrylamide gel electrophoresis in denaturing and non-denaturing conditions (both in the presence and absence of 6 m urea) was performed essentially as indicated elsewhere . After separation, proteins were transblotted onto a polyvinylidene difluoride membrane. Antithrombin was immunostained with rabbit anti-human antithrombin polyclonal antibody (Sigma-Aldrich, Madrid, Spain), followed by donkey anti-rabbit IgG–horseradish peroxidase conjugate (GE Healthcare, Barcelona, Spain), with detection via an ECL kit (Amersham Biosciences, Piscataway, NJ, USA).
Separation of antithrombin variant from plasma according to the pI was performed using the 3100 OFFGEL Fractionator and the OFFGEL Kit pH 4–7 (Agilent Technologies, Cernusco, Italy) following the manufacturer’s protocol. Briefly, 20 μL of plasma from the patient and a control were loaded on 24 cm immobilized pH gradient strips (IPG 4-7). This system achieves pI-based fractionation by electro-elution. The resulting 24 soluble fractions were electrophoresed using SDS-PAGE and antithrombin was identified by western blot, as indicated above.
Plasma from the patient and a plasma pool of 100 healthy blood donors (20 μL) were treated with 0.6 U N-glycosidase F (Roche Diagnostics GmbH, Mannheim, Germany) at 37 °C for 15 h, after a previous denaturing step (5 min at 95 °C in 150 mm sodium phosphate buffer, pH 7.4). Samples were resolved by SDS-PAGE, as described above.
Plasma protein purification
Variant antithrombin from the patient was purified by heparin affinity chromatography at pH 7.4 on HiTrap Heparin columns (GE Healthcare), as previously described , using an ÄKTA Purifier (GE Healthcare) in 100 mm Tris-HCl and 10 mm citric acid, in a gradient from 0.15 to 2 m NaCl. Fractions with variant antithrombin were applied to a HiTrap Q column (1 mL) (GE Healthcare). Proteins were eluted with 100 mm sodium phosphate buffer pH 6.0, in a gradient from 0 to 1 m NaCl. The fractions containing variant antithrombin were pooled, desalted over 5 mL HiTrap desalting columns (GE Healthcare) and stored at −70 °C. Protein purity was checked by 8% SDS-PAGE, as described above.
Wild-type latent antithrombin was obtained by purifying α-antithrombin from 2 L of plasma from healthy subjects by heparin affinity on a HiTrap Heparin column (GE Healthcare). Fractions containing pure antithrombin were pooled and incubated in 50 mm NaH2PO4, pH 6.0, 40% glycerol at 50 °C for 72 h. Purity and stability were verified by 8% SDS-PAGE and non-denaturing PAGE in the presence of 6 m urea as described above.
Edman sequencing of the mutant purified from the patient’s plasma and of the purified recombinant variant were performed under reducing conditions with Applied Biosystems Procise 494 equipment (Foster City, CA, USA).
A solution of 3,5-dimethoxy-4-hydroxycinnamic acid (10 g/L) in acetonitrile (ACN)/water/trifluoroacetic acid (TFA) (50:50:0.1 by vol.) was chosen for protein analysis. Experiments were carried out on a Voyager-DE™ STR Biospectrometry workstation (Applied Biosystems, Madrid, Spain), equipped with an N2 laser (337 nm). Recorded data were processed with data explorer™ Software (Applied Biosystems).
Antithrombin samples were also digested in 100 mm NH4HCO3 (pH 7.8) containing trypsin (ratio enzyme:substrate, 1:50) at 37 °C for 16 h. Peptide mixtures from in situ digestion of proteins were desalted in a GELoader tip packed with 0.5 μL POROS-10 R2 (PerSeptive Biosystem) slurry.
Recombinant expression and purification of antithrombin variant
Recombinant antithrombin was constructed on the β-glycoform S137A antithrombin background in order to reduce glycosylation heterogeneity and to facilitate purification. Site-directed mutagenesis of the pCEP4-S137A antithrombin plasmid was performed using the Stratagene Quik Change Site-Directed Mutagenesis kit (Agilent Technologies)  and the appropriate primers. Human Embryonic Kidney cells expressing the Epstein Barr Nuclear Antigen 1 (HEK-EBNA) were grown in DMEM with GlutaMAX-I medium (Invitrogen, Barcelona, Spain) supplemented with 5% fetal bovine serum (Sigma-Aldrich) to 60% confluence at 37 °C and 5% CO2 in a humidified incubator. Transfection was performed by addition of the plasmid encoding for the antithrombin mutant (200 μg/mL) with a previous incubation for 30 min in serum-free OptiMEM culture medium with Lipofectamine LTX reagent (Invitrogen) according to the manufacturer’s protocol. Twenty-four hours post-transfection the cells were washed with PBS and exchanged into CD-CHO medium (Invitrogen) supplemented with 4 mm l-glutamine and 0.25 mg/mL Geneticin (Invitrogen). Cells were grown for 10 days and culture medium was collected every 42 h for purification.
Protein purification of the recombinant protein was identical to that used for purification of the plasma protein.
Continuous anti-FXa and anti-FIIa activities were also measured with 0.5 μm recombinant wild-type and mutant proteins in the presence and absence of 80 nm pentasaccharide or 5 U unfractionated heparin, using the appropriate chromogenic substrates (S-2765 or S2238, respectively) at a concentration of 200 μm. FXa and FIIa activities were evaluated at a final concentration of 5.4 nm and 4.8 nm, respectively.
The heat capacity (Cp) of samples was recorded over a temperature range of 10–130 °C, by using differential scanning calorimetry (VP-DSC; MicroCal Inc., LLC, Northampton, MA, USA), employing two fixed cells, a reference cell, and a sample cell. Prior to all the measurements, the buffer and protein solutions were degassed. The volume of the calorimetric cells was 0.5 mL and the protein concentration used was 0.2 mg/mL. The measurements were carried out with a microcalorimeter at the scan rate of 60 °C/h using recombinant wild-type and variant antithrombins. Data were analyzed using the origin dsc software that was provided by MicroCal Inc.
Thrombin cleavage of recombinant wild-type and variant antithrombins was carried out by incubating 2 μg of protein with 1 μg of human thrombin (Calbiochem, Merck, Madrid, Spain) at 37 °C for 15 min. Afterwards, reactions were stopped with 0.1 μm AEBSF and loaded in a 10% SDS-PAGE under non-reducing conditions. Coomassie staining was used to detect the different migration of proteins after incubation with thrombin.
Mutant antithrombin was crystallized as a dimer with latent plasma-derived antithrombin. Hanging drops were set up containing 2 μL antithrombin (8.75 mg/mL mutant antithrombin, 8.75 mg/mL latent plasma-derived antithrombin in 20 mm Tris pH 7.4) and 2 μL precipitant (17.0% PEG 4000, 50 mm Na/KPO4 pH 6.7). A single crystal was cryoprotected by passing the crystal sequentially into solutions containing 0–20% glycerol with 19% PEG 4000, 50 nm Na/KPO4 pH 6.7 and flash cooled in liquid nitrogen. A dataset was collected from a single crystal at Diamond Light Source, Station IO4 (Oxford, UK), using an ADSC Q315r detector (wavelength 0.9763 Å), and was processed using the programs iMosflm and Scala. A single solution was found by molecular replacement using the structure of native/latent antithrombin heterodimer (pdb id1E04) as a search model, using Phaser . The structure was refined using Refmac  and rebuilt in Coot . Processing and refinement statistics are given in Table 1, and final coordinates and structure factors are deposited in the Protein Data Bank under accession code 4EB1.
Table 1. Crystals, data processing, refinement and models
a = 69.75 b = 101.75 c = 88.81, α = 90 °β = 105.42 °γ = 90 °
Solvent content (%)
Data processing statistics
Number of protein/water atoms
Average B-factor (Å2)
Reflections in working/free set
r.m.s. deviation of bonds(Å)/angles (°) from ideality
Clinical data and identification of a novel variant antithrombin
The proband, a 51-year-old Caucasian male, suffered a spontaneous pulmonary thromboembolism. Thrombophilic screening only revealed a type II antithrombin deficiency (50% anti-FXa activity and 70% antigen levels). Further analysis by crossed immunoelectrophoresis in the presence of heparin revealed an increased proportion of the form with low heparin affinity, consistent with a type II deficiency with heparin binding defect (Fig. 1A). Further functional analysis revealed an anti-IIa activity of 50% both in the presence and absence of unfractionated heparin.
A familial study confirmed the presence of antithrombin deficiency in other family members who also reported relevant thrombotic history, including pulmonary embolism, stroke, recurrent thrombosis and miscarriages (Fig. 1B).
Sequencing of the coding region of SERPINC1 in the proband only detected an abnormal sequence in exon 4 (Fig. 2A). Analysis of this PCR amplification demonstrated two bands corresponding to the wild-type allele (229 bp) and to the mutated allele (253 bp) (Fig. 2B). Purification and sequencing of the mutated allele showed a complex insertion of 24 bp that involved a deletion of 4 bp, an insertion of 7 bp and a repetition of 21 bp (Fig. 2C). This molecular rearrangement resulted in the in-frame insertion of eight residues, that included the repetition of seven conserved residues of the strand 3A (s3A) (V212-T218), the deletion of residues E209 and L210, and the insertion of three new residues (RTS) (Fig. 2C). This final large insertion was located between strand 3A (s3A) and helix F. Such a molecular reorganization has not been previously described for antithrombin or any other serpin.
Biochemical characterization of plasma variant
SDS-PAGE under reducing conditions and western blot analysis of plasma from the patient detected two bands, but surprisingly, the variant was smaller (54 kDa) than the wild-type antithrombin (58 kDa) (Fig. 3A). Isoelectrofocusing of plasma from the patient by an OFFGEL system confirmed the smaller size of the variant and a significantly higher pI (5.375 for the main wild-type antithrombin and 5.625 for the variant) (Fig. 3B). This variation in the pI, together with the smaller size, might be explained by a reduced glycosylation . However, glycan removal by treatment of plasma with N-glycosidase F caused the same size reduction in wild-type as in mutant antithrombins (Fig. 3C). A potential alternative splicing that might explain the smaller size of the mutant antithrombin despite having an insertion was ruled out by in silico analysis (data not shown). Such an enigma was solved by proteomic analysis and Edman sequencing of the variant purified from plasma. MALDI-TOF mass spectrometric analysis of the variant purified from plasma showed a molecular weight compatible with the nucleotide insertion (59 kDa vs. 58 kDa for wild type) (Fig. S1). Finally, Edman sequencing identified the W49 residue as the N-terminus of the mutant antithrombin purified from plasma when protein was evaluated under reducing conditions.
One additional feature of this variant was its conformational sensitivity. Indeed, a few cycles of freezing/thawing of plasma caused the transition of the mutant native antithrombin to a hyperstable conformation (Fig. 4A). In addition, purification of the variant from plasma by heparin affinity chromatography resulted in protein that eluted at low NaCl concentration and in a hyperstable conformation (Fig. 4B).
Characterization of the recombinant variant
In order to further characterize this mutant, recombinant expression was carried out by transfection of HEK-EBNA cells. This molecule had low heparin affinity according to its elution on heparin affinity chromatography (∼700 mm NaCl), and lacked anticoagulant anti-FXa and anti-IIa inhibitory activity, and no progressive activity was detected (data not shown). Moreover, the purified recombinant mutant protein was hyperstable according to data from calorimetry (Fig. S2), although no cleavage at the N-terminus was produced.
Inhibitory activity of mutant antithrombin
The anti-FXa and anti-FIIa activity in the presence and absence of heparin was measured using plasma from the patient and CD-CHO culture medium from transfected cells with wild-type or mutant plasmid. In plasma samples from the patient the antithrombin activity was 50% and in the cell culture medium of cells transfected with the wild-type or mutant plasmids, we detected negligible activity compared with medium from cells transfected with wild-type antithrombin. This result showed that the variant antithrombin has neither heparin-induced nor progressive inhibitory activity.
Accessibility of the reactive centre loop: thrombin cleavage study
In order to determine if the hyperstable mutant recombinant antithrombin was in the latent conformation, we evaluated the accessibility of the reactive centre loop to thrombin cleavage. The migration on SDS gel under non-reducing conditions demonstrated that the mutant hyperstable protein was cleaved by thrombin (Fig. S3). Moreover, N-terminal sequencing confirmed cleavage by thrombin at R393 (P1 of the RCL).
Structure of the hyperstable conformation of variant antithrombin
In order to fully understand the structure induced by this mutation, and particularly to know how the insertion was incorporated in the structure of antithrombin and how antithrombin obtained a hyperstable conformation without the incorporation of the RCL into β-sheet A, we determined the crystal structure of the recombinant protein. This was achieved by forming a hetero-dimer with α-wild-type latent antithrombin. This strategy was followed after failing to crystallize the mutant molecule on its own, and once we had experimental evidence of an accessible RCL that might interact with a latent wild-type form. Crystals appeared after 2 days and data were collected to 2.8 Å from a single crystal. The structure was solved by molecular replacement using 1E04 (Table 1).
We point out two relevant features of this molecule. First, the mutant molecule had a new strand in the central β-sheet (s4A) formed by the whole insertion plus some residues that normally connect the helix F to the strand 3A by a loop. It had therefore a six-stranded conformation that explains the hyperstability (Fig. 5).
Description of new mutations in serpins in general and antithrombin in particular has increased our knowledge of function and dysfunction, and the identification of new mechanisms involved in different diseases, including thrombosis [13,18–20]. In general, mutations associated with type II deficiencies do not affect protein folding, and the variant protein is secreted at high rates. These variant antithrombin molecules have a native conformation with impaired anticoagulant activity, because the mutation affects more or less severely relevant functional domains (heparin binding domain, RCL, or both) , or increases the conformational sensitivity of the molecule to transform to the latent inactive conformation . In contrast, there are mutations in antithrombin associated with null or low secretion rates.
Mutations in antithrombin and other serpins have helped in the development of our understanding of serpin folding and misfolding. The archetype of these mutations is E342K in α1-antitrypsin , but some relevant examples have also been described in serpins that regulate blood coagulation, such as the homologous mutation in heparin cofactor II (E428K)  and different mutations in antithrombin (P80S and G424R, C95R and F229L) [11,24,25]. Recent evidence suggests that these mutations influence the intermediate folding state immediately prior to the folding of the metastable native conformation, facilitating the domain swap between two mutant molecules, in a process that leads to the formation of intracellular polymers, which are not secreted . The apparent conformational sensitivity of serpins means that it would be expected that small deletions or insertions would never allow secretion of the variant serpin. The finding in this study is therefore an unexpected exception identified in a patient with antithrombin deficiency and thrombosis.
The presence of a relatively large in-frame insertion affecting one relevant and conserved region of serpins does not lead to a substantial inappropriate folding and the subsequent degradation by the proteasome quality control machinery  or to the formation of polymers. The mutation allows a correct native-like folding of the molecule and its proper glycosylation, avoiding intracellular accumulation, although a reduced level of protein is detected (Fig. 3A). An increased intracellular retention is not observed in the cells when recombinant protein is expressed, so a reduced RNA stability or transcription rate, an enhanced clearance or an inefficient or incorrect folding of some of the variant molecules should be considered to explain the reduction of levels of this variant in plasma of carriers. The single distortion of the standard intracellular pathway of antithrombin is proteolytic cleavage in the N-terminal region by a still unknown intracellular protease. The presence in this proteolyzed peptide of two cysteines involved in the formation of two of the three disulphide bonds established in the protein  explains the electrophoretic mobility and mass results (Fig. 3A and Fig. S1), as the proteolyzed peptide remains bound to the protein. This proteolytic attack, located close to the heparin binding site, might also help explain the reduced heparin affinity of the variant .
The insertion must have a minor influence on the native folding, resulting in a stressed conformation. However, the metastable native conformation of this variant is extremely sensitive, because purification or a few freeze/thaw cycles transform this antithrombin mutant into a new relaxed form, with six strands in the central β-sheet that do not include the RCL. This conformational sensitivity is close to that of PAI-I, in which the transition to the hyperstable conformation takes place spontaneously . Our study has been able to characterize the structure of this hyperstable conformation, which has a native RCL, including the hinge inserted normally at the top of the central β sheet (Fig. 5), similar to that of peptide complexed serpins . Obligatory substrate behavior of the hyperstable mutant would be predicted given that the novel strand occupies the same position that the RCL would fruitlessly attempt to occupy. This suggests that the hyperstable form could be an endpoint in the misfolding pathway, as it occurs with the δ-antichymotrypsin conformation provoked by the L55P mutation , which has been recently identified at physiological temperature and pH, suggesting that it may have a role in both health and disease .
Unfortunately, these features have made it impossible to obtain the variant in its native conformation from plasma or cell cultures, but the substrate behavior of the variant suggests the partial insertion of the mutation between strands 3 and 5 of β-sheet A.
Finally, it should be emphasized that this mutation has been identified in a family with antithrombin deficiency that displays a severe thrombotic history. The mutation causes a loss of function by at least two mechanisms: the impaired heparin affinity does not allow the activation of the molecule  and more importantly, the insertion transforms the serpin into a substrate for the target proteases. The difference with other variants that also behave as substrates is that the cleaved RCL of this variant could not be inserted because the s4A position is occupied by the insertion.
IM-M is a researcher from Fundación para la Formación e Investigación Sanitarias. JN-F holds post-doctoral contract Sara Borrell from ISCIII. The authors want to thank N. Gómez from Servicio de Hematología, H. Universitario de la Princesa, Madrid, Spain, for providing the samples from the patient, and T. Seara Sevivas from Centro Hospitalar de Coimbra (Portugal) for her help.
J. Corral was the principal investigator and takes primary responsibility for the paper. I. Martínez-Martínez, D. J. D. Johnson, M. Yamasaki, J. Navarro-Fernández and A. Ordóñez performed the laboratory work for this study. D. J. D. Johnson and J. A. Huntington solved and refined the crystal structure. V. Vicente, J. A. Huntington and J. Corral coordinated the research. I. Martínez-Martínez, J. A. Huntington and J. Corral and wrote the paper.
This work was supported by SAF2009-08993 (MCYT & FEDER), RETICS RECAVA RD06/0014/0039 (ISCIII & FEDER), Fundación Séneca (04515/GERM/06), and Fundación Mutua Madrileña. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Disclosure of Conflict of Interests
The authors state that they have no conflict of interest.