α1‐Antitrypsin (α1AT) deficiency is a disease with multiple manifestations, including cirrhosis and emphysema, caused by the accumulation of stable polymers of mutant protein in the endoplasmic reticulum of hepatocytes. However, the molecular basis of misfolding and polymerization remain unknown. We produced and crystallized a trimeric form of α1AT that is recognized by an antibody specific for the pathological polymer. Unexpectedly, this structure reveals a polymeric linkage mediated by domain swapping the carboxy‐terminal 34 residues. Disulphide‐trapping and antibody‐binding studies further demonstrate that runaway C‐terminal domain swapping, rather than the s4A/s5A domain swap previously proposed, underlies polymerization of the common Z‐mutant of α1AT in vivo.
There is a Hot off the Press (October 2011) associated with this Scientific Report.
Serine protease inhibitors (serpins), such as α1‐antitrypsin (α1AT), undergo a marked conformational change to inhibit proteases (Huntington et al, 2000). This involves the incorporation of the 20‐residue reactive‐centre loop (RCL) into the middle of the β‐sheet A (as strand 4A; supplementary Fig S1A,B online). The conformational change can also occur spontaneously and is driven by a large free‐energy term (Im et al, 2000). The folding pathway of a native serpin must therefore bypass the global free‐energy minimum in favour of the active metastable state. For this reason, serpins are susceptible to point mutations (positions of mutations shown in supplementary Fig S1A online) that lead to misfolding and the accumulation of stable polymers in the endoplasmic reticulum (ER) of secretory cells (for reviews see Davies & Lomas, 2008; Gooptu & Lomas, 2009; Knaupp & Bottomley, 2009). Polymerization of the serpins antithrombin, C1‐inhibitor and neuroserpin lead to thrombosis, angioedema and dementia, respectively. Mutations in α1AT, including the Z‐mutation (Glu342Lys) carried by 4% of northern Europeans, result in emphysema due to the lack of functional α1AT in the lung, and in liver disease (including cirrhosis) due to the accumulation of polymers in the ER of hepatocytes. Recent structural studies on antithrombin revealed that an unexpected domain swap of more than 50 residues mediated formation of a dimer in vitro (Yamasaki et al, 2008). The intermolecular contact included the insertion of two long β‐strands (s4A and s5A) of one molecule into the β‐sheet A of another (supplementary Fig S1C online). We further demonstrated, using disulphide‐trapping experiments, that α1AT polymers produced by incubation with guanidine hydrochloride (GndHCl) also formed via an analogous s4A/s5A domain‐swapping event. We therefore speculated that a similar s4A/s5A domain swap mediated α1AT polymerization in vivo. Recent work, however, showed that a monoclonal antibody (2C1) specific for hepatocellular inclusions of α1AT (Miranda et al, 2010; that is, the pathological polymer) reacted strongly with heat‐induced polymers but not those induced by GndHCl (Ekeowa et al, 2010). The mechanism of pathological polymer formation of α1AT is therefore likely to be different from the previously described s4A/s5A domain swap.
In vitro polymerization
Incubation of native α1AT either with GndHCl or at elevated temperature resulted in the appearance of smeared bands or interlaced ladders by non‐denaturing polyacrylamide gel electrophoresis (PAGE; Fig 1A, left panel), suggesting multiple types of polymers. In accordance with previous studies (Ekeowa et al, 2010), western blotting the same gel with the 2C1 antibody revealed poor reactivity with GndHCl‐induced polymers (Fig 1A, right panel), and strong reactivity for heat‐induced polymers (Fig 1A, right panel). These results suggest that α1AT can polymerize by at least two different mechanisms in vitro, with the s4A/s5A domain swap predominating in GndHCl, and an alternative mechanism favoured by heat treatment. On the basis of the strong reactivity with the 2C1 antibody, we conclude that the heat‐induced polymers share important structural features with pathological polymers of α1AT formed in vivo.
Formation of α1AT trimer
To preclude formation of the s4A/s5A polymer, we engineered a disulphide bond (Cys 292–Cys 339) in α1AT to link strand 5A to strand 6A (supplementary Fig S1D, left panel online). The s5A–s6A disulphide variant was resistant to polymerization by GndHCl (supplementary Fig S2 online), but polymers could be formed by incubation at 60°C (supplementary Fig S2 online, and Fig 1B), and the resulting ladder on native PAGE reacted strongly with the 2C1 antibody (Fig 1B, right panel). By incubating the s5A–s6A disulphide variant of α1AT at 60°C for 18 h, we obtained short polymers that were resistant to further elongation, consistent with the formation of self‐terminating ‘necklace’‐type polymers previously seen by electron microscopy (Lomas et al, 1993b). The most abundant species was purified to homogeneity by multiple passes over a size‐exclusion column (Fig 1B, lane 3).
Crystal structure of α1AT trimer
We obtained crystals of the purified short polymer that diffracted to 3.9 Å (supplementary Fig S3 online), and determined the structure by molecular replacement (Table 1). A single molecule of α1AT was found in the asymmetric unit in a conformation resembling RCL‐cleaved α1AT (Fig 2A). Close examination revealed clear electron density linking the monomer of the asymmetric unit to a crystallographically related monomer at the C terminus of the RCL (Fig 2B). Three molecules of α1AT were thus linked to form a crystallographic trimer through a 34‐residue C‐terminal domain swap (residues 361–394), including strand 1 from β‐sheet C and strands 4 and 5 from β‐sheet B (Fig 2C). The domain swap we observed could similarly link protomers into linear polymers with no unfolded regions and significant subunit flexibility (Fig 2D), consistent with the ‘beads‐on‐a‐string’ morphology observed on electron micrographs (Lomas et al, 1993b).
Domain swapping in vitro
To determine whether C‐terminal domain swapping is the principal mechanism of α1AT polymerization, we engineered a disulphide bond between the C terminus (residue 392) and an adjacent residue on β‐sheet C (residue 216; supplementary Fig S1D online, centre panel). The C‐terminally constrained variant was able to form polymers when heated, but, similarly to polymers of α1AT formed using GndHCl, they did not react with the 2C1 antibody (supplementary Fig S4 online). The C‐terminal disulphide thus prevents the formation of the pathological polymer, and might exclusively allow the s4A/s5A mechanism. However, when reduced with dithiothreitol (DTT) before heating, the variant polymerized in a manner indistinguishable from control (supplementary Fig S5A online). Furthermore, after reoxidation, the polymers were stable on SDS–PAGE (Fig 3A), indicative of intermolecular disulphide bond formation (Fig 3D). Similar experiments with the s5A–s6A disulphide variant produced a lower proportion of SDS‐stable polymers (Fig 3A, supplementary Fig S5A online and Fig 3E), indicating that the s4A–s5A swap is a minor mode of polymerization of α1AT in vitro, particularly when induced by heat.
Domain swapping in vivo
Polymerization in vivo was tested by expressing the disulphide variants described above coupled to the polymerigenic Z–α1AT mutation in two well‐characterized model cell types, Pichia pastoris (Levina et al, 2009) and COS‐7 (Miranda et al, 2010). Using a His tag, we were able to purify polymers from the cytosol of P. pastoris that reacted strongly with the 2C1 antibody (supplementary Fig S4 online). Furthermore, purified polymers of Z–α1AT with the C‐terminal disulphide (Cys216–Cys392) were SDS stable (Fig 3B, lane 3), indicating the formation of intermolecular disulphide bonds after polymerization (Fig 3D). However, when the Z‐mutation was coupled with the s5A–s6A disulphide variant (Cys 292–Cys 339) or a control (with Cys residues at remote positions 232 and 292; supplementary Fig S1D online, right panel), SDS‐stable polymers were not observed (Fig 3B, lanes 2 and 4). Similar data were obtained when these variants were expressed in mammalian (COS‐7) cells (Fig 3C).
Taken together, these results demonstrate that α1AT is capable of at least two different types of domain swap in vitro. Using a protein engineering approach, we were able to purify and crystallize a single polymer species that reacted strongly with the 2C1 antibody. The resulting structure of a closed α1AT trimer revealed an unexpected C‐terminal domain swap. This linkage satisfyingly explains several known properties of serpin polymers, including a high degree of flexibility between protomers (Lomas et al, 1993b), polymer hyperstability due to full insertion of the RCL (Mast et al, 1992), and persistent donor and acceptor ends that allow facile elongation (‘infectivity’; Zhou & Carrell, 2008) and circularization (Lomas et al, 1993b). In addition, it has been repeatedly demonstrated that accumulation of α1AT polymers in cells does not invoke the unfolded protein response (Graham et al, 1990; Hidvegi et al, 2005). This is consistent with the crystal structure of the trimer, in which there is no exposure of unfolded or improperly folded regions. Linear polymers, on the other hand, would have an exposed hydrophobic C terminus at the donor end (Fig 2D), which might mediate interaction with the ER quality control machinery, as previously observed (Schmidt & Perlmutter, 2005). In addition, the C‐terminal tail of linear polymers would be predicted to be susceptible to proteolysis (cleavage site indicated in Fig 2D). This might explain the previously perplexing observation that polymers of the Mmalton variant purified from blood contained a fraction of RCL‐cleaved α1AT (Lomas et al, 1995).
Serpin folding is thought to proceed via a critical intermediate that has the capacity to form the native state or to polymerize. Our data suggest that the fate of this intermediate is governed by two competing events—the incorporation of the C terminus into the folded serpin core versus the completion of β‐sheet A (Fig 4). We hypothesize that in wild‐type serpins the intermediate is short‐lived, and that the native fold is achieved because the rapid folding of the C‐terminal 34 residues effectively tethers the RCL and prevents it from inserting into β‐sheet A. The mutations that cause polymerization, including the Z‐variant, have been shown to influence the folding pathway by increasing the lifetime of the polymerigenic intermediate (Yu et al, 1995). This would predictably increase the fraction of the molecules that insert the RCL before the C terminus can be incorporated into the serpin body. The result of such a misfolding event would be the formation of the monomer observed in our crystal structure—a molecule with the RCL inserted, a gap in β‐sheet B and the C terminus exposed. The only way for this kinetically trapped monomer to reach its global free‐energy minimum is to polymerize by inserting its C‐terminal tail into another molecule (monomer or polymer; Fig 4). Such a mechanism might also explain why mutations that slow RCL incorporation (such as replacing the hydrophobic P8–P6 region with aspartic acids) result in a significant rescue of the secretion defect of Z–α1AT (Yamasaki et al, 2010). Indeed, this observation provides a proof of principle that interventions that alter the folding pathway of α1AT might be of therapeutic value in treating the loss‐of‐function and gain‐of‐function manifestations of α1AT deficiency.
α1AT variants. Wild‐type α1AT was purified from plasma as described previously (Lomas et al, 1993a). Recombinant variants of α1AT were all based on a template with a reactive‐centre mutation M → R (M358R) and the only Cys residue (232) mutated to Ser or Ala. The s5A–s6A disulphide variant was created by substituting Cys residues at adjacent positions on s5A and s6A (T339C and S292C). The C‐terminal disulphide variant was constructed by introducing Cys residues on s3C and on the C‐terminal loop (V216C and T392C). Mutagenesis was performed using the QuikChange kit and associated protocol (Stratagene). The α1AT variants for crystallographic and biochemical studies were produced recombinantly in Escherichia coli, as described previously (Zhou et al, 2001).
Native PAGE and 2C1 western blots. Different monomer and polymer forms of α1AT were assessed by 8% non‐denaturing PAGE and visualized by Coomassie or silver staining, or by western blotting using the mouse monoclonal antibody specific for α1AT polymers (2C1) as the primary antibody, as described previously (Miranda et al, 2010). All polymers were prepared at a concentration of 0.2 mg/ml in PBS using the following conditions: GndHCl polymer by incubation of plasma α1AT with 1 M GndHCl at 37°C for 16 h; heat polymer by incubation of plasma α1AT at 60°C for 2 h; yeast control monomer and Z‐polymers purified as described below; s5A–s6A disulphide (S292C/T339C) polymer by incubation at 60°C for 8 h; s5A–s6A disulphide trimer purified as described below; C‐terminal disulphide V216C/T392C polymer by incubation at 60°C for 8 h. The disulphide mutants were treated with DTT and iodoacetamide before running on the native gel to block the Cys residues before western blotting.
Trimer preparation, crystallization and structure determination. The s5A–s6A disulphide variant was polymerized by incubation at 0.1 mg/ml in PBS at 60°C for 16 h. The most abundant polymer species was purified using a Hiload 16/60 Superdex 200 prep‐grade size‐exclusion column. The material was subjected to Asp‐N (Sigma‐Aldrich) digestion (1:750 ratio by weight) to remove the 25 N‐terminal residues that precede the serpin core. This material was subsequently concentrated to 10 mg/ml in 5 mM Tris–HCl (pH 7.4) and 50 mM NaCl for crystallization. Crystals were grown in 18% (w/w) ethanol and 0.1 M Tris–HCl (pH 8.5) in 96‐well sitting‐drop vapour diffusion plates using our robotic crystallization facility. Crystals were transferred into 42% (w/w) 2‐methyl‐2,4‐pentanediol and 0.1 M Tris‐HCl (pH 8.5), and flash‐cooled in liquid N2. Diffraction data were collected on Diamond Light Source beamline I04 (wavelength 0.9763 Å) and processed using Mosflm, Scala and Truncate (Leslie, 1992). Data were of sufficient quality from 20 to 3.9 Å for molecular replacement and refinement (Table 1 and supplementary Fig S3 online). The best molecular replacement solution was a single copy of RCL‐cleaved α1AT (PDB: 3NDD), found using the program Phaser (McCoy et al, 2005). The structure was refined using jelly‐body restrains with the program Refmac (Murshudov et al, 1997; version 5.6) and rebuilt using sharpened maps (produced by Refmac) with the program COOT (Emsley & Cowtan, 2004). Data processing and refinement statistics are given in Table 1. A total of 89.2% of the residues in the final model were in the favoured regions and 99.2% were in the allowed regions when Ramachandran analysis was carried out using the MolProbity server (Lovell et al, 2003). Structural figures were made using the program PyMOL (DeLano, 2002).
Disulphide‐trapping studies. The disulphide variants and control Cys variant (0.2 mg/ml) were subjected to 2 mM DTT at room temperature for 30 min and incubated for 8 h at 4°C or 53°C, or at 37°C in the presence of 0.75 M GndHCl. After incubation, the samples were dialysed into PBS for 3 h to remove DTT and GndHCl. The disulphide bond was reformed by incubation with 0.5 mM tetramethylazodicarboxamide (Fuluka) for 1 h at room temperature in the dark. Formation of intermolecular disulphide bonds was assessed by non‐reducing 10% SDS–PAGE (NuPAGE Bis‐Tris Gel, Invitrogen).
α1AT expression in P. pastoris. α1AT variants were expressed in the cytosol of the methylotrophic yeast species P. pastoris, as previously described (Levina et al, 2009). The Z‐mutation (E342K) and Cys variants (S292C/T339C, V216C/T392C and A232C/S292C) were introduced by site‐directed mutagenesis using the QuikChange kit. The template was C232A. After cell disruption, protein was purified by His‐tag affinity chromatography using a HiTrap IMAC HP column (GE Healthcare). Monomer and polymer fractions were separated by salt gradient using a HiTrap Q HP column (GE Healthcare), essentially as described before (Levina et al, 2009).
α1AT expression in COS‐7 cells. The α1AT DNA was cloned from pQE30 into the pCEP4 vector (Invitrogen) using NotI and KpnI sites and extended to add the signal sequence derived from α1AT. Mutations were introduced by site‐directed mutagenesis, as described before. Cells were maintained, transfected and lysed as described previously (Miranda et al, 2010). A measure of 30 μg of protein (as determined by Bradford assay) was loaded onto an SDS gel (NuPAGE 10% Bis‐Tris Gel, Invitrogen) or an 8% non‐denaturing gel and visualized by western blotting using an anti‐α1AT rabbit polyclonal primary antibody (AbD Serotec) and an anti‐rabbit horseradish peroxidase‐conjugated secondary antibody (Sigma) and exposed to film.
Accession codes. Protein Data Bank: coordinates and structure factors have been deposited with accession code 3T1P.
Supplementary information is available at EMBO reports online (http://www.emboreports.org).
Conflict of Interest
The authors declare that they have no conflict of interest.
We thank E. Miranda and D. Lomas for providing the 2C1 antibody, and S. Bottomley and R. Carrell for comments on the manuscript. This work was supported by the Medical Research Council (MRC, J.A.H.), and J.A.H. is a Senior MRC Non‐clinical Fellow. J.C.W. is an Australian Research Council Federation Fellow and honorary National Health and Medical Research Council of Australia Principal Research Fellow. This work was carried out with the support of the Diamond Light Source.
Author Contributions: M.Y. produced and crystallized the α1AT trimer, and collected the diffraction data. M.Y. and J.A.H. processed the data, and solved and refined the structure. M.Y. and T.J.S. conducted biochemical analysis of in vitro polymerization. T.J.S. and M.C.P. conducted the studies on polymers formed in cells. J.A.H., M.Y. and J.C.W. analysed the structure. J.A.H. and J.C.W. wrote the paper. J.A.H. directed the project.
- Copyright © 2011 European Molecular Biology Organization