Structure of protease-cleaved Escherichia coli α-2-macroglobulin reveals a putative mechanism of conformational activation for protease entrapment
aInstitute of Infection, Immunity and Inflammation, College of Medical, Veterinary and Life Sciences, University of Glasgow, Glasgow G12 8QQ, Scotland, bWestCHEM, School of Chemistry, College of Science and Engineering, University of Glasgow, Glasgow G12 8QQ, Scotland, cInstitute of Molecular Cell and Systems Biology, College of Medical, Veterinary and Life Sciences, University of Glasgow, Glasgow G12 8QQ, Scotland, and dSchool of Life Sciences, College of Medical, Veterinary and Life Sciences, University of Glasgow, Glasgow G12 8QQ, Scotland
*Correspondence e-mail: email@example.com
Bacterial α-2-macroglobulins have been suggested to function in defence as broad-spectrum inhibitors of host proteases that breach the outer membrane. Here, the X-ray structure of protease-cleaved Escherichia coli α-2-macroglobulin is described, which reveals a putative mechanism of activation and conformational change essential for protease inhibition. In this competitive mechanism, protease cleavage of the bait-region domain results in the untethering of an intrinsically disordered region of this domain which disrupts native interdomain interactions that maintain E. coli α-2-macroglobulin in the inactivated form. The resulting global conformational change results in entrapment of the protease and activation of the thioester bond that covalently links to the attacking protease. Owing to the similarity in structure and domain architecture of Escherichia coli α-2-macroglobulin and human α-2-macroglobulin, this protease-activation mechanism is likely to operate across the diverse members of this group.
α-2-Macroglobulins (α2Ms) are found in eukaryotic blood, invertebrate haemolymph, the eggs of birds and reptiles, and the bacterial periplasm, where they are thought to play a role in the restriction of proteolytic cleavage (Sottrup-Jensen, 1989; Lin et al., 2002; Budd et al., 2004; Li et al., 2004; Doan & Gettins, 2008). Eukaryotic α2Ms have been shown to play important roles in regulating the proteolytic cleavage of a wide range of proteases and are involved in processes such as fibrinolysis and coagulation (De Boer et al., 1993). Bacterial α2Ms (BA2Ms) are produced by a wide range of Gram-negative bacteria ranging from human pathogenic and commensal strains to plant pathogens and marine bacteria (Budd et al., 2004). Escherichia coli α2M (ECAM) contains the conserved thioester bond that is characteristic of the α2Ms and which is essential for covalent binding to cleaving proteases (Budd et al., 2004; Doan & Gettins, 2008; Neves et al., 2012).
The gene encoding ECAM, yfhM, is frequently found in an operon with pbp1C, which encodes penicillin-binding protein 1C (Pbp1C; Budd et al., 2004; Doan & Gettins, 2008). Pbp1C is predicted to be a bifunctional transpeptidase and transglycosylase owing to its homology to Pbp1A and Pbp1B, which are both essential for the synthesis of the peptidoglycan layer (Schiffer & Höltje, 1999; Budd et al., 2004). Both ECAM and Pbp1C are periplasmic proteins that are anchored to the inner membrane and have been proposed to function together in defence and repair against proteases that damage the bacterial cell wall (Budd et al., 2004). Specifically, it has been postulated that the host proteases produced in defence against bacterial infection, which gain access to the periplasmic space, are inhibited by ECAM, with Pbp1C acting to repair damage (Budd et al., 2004).
The overall structure of α2Ms comprises a series of β-sheet sandwich macroglobulin (MG) domains forming a `keyring' shape, a bait-region domain (BRD) spanning the body of the ring, a mostly helical thioester domain (TED) connected to the `keyring' by a complement protein subcomponent (CUB) domain, and a C-terminal MG (CTMG) domain (Janssen et al., 2005; Marrero et al., 2012; Wong & Dessen, 2014). Within eukaryotic α2Ms there are eight MG domains including CTMG, which is also known as the receptor-binding domain (MG1–MG7, CTMG), while BA2Ms contain ten MG domains including two N-terminal MG domains anchoring the protein to the inner membrane within the periplasm (MG1–MG9, CTMG) (Huang et al., 1998; Doan & Gettins, 2007; Marrero et al., 2012; Wong & Dessen, 2014). The TED contains a conserved CXEQ motif where a thioester bond is formed, which on activation can covalently link the α2M to lysine residues on the surface of the attacking protease (Sottrup-Jensen et al., 1980, 1989; Osterberg & Malmensten, 1984; Jacobsen & Sottrup-Jensen, 1993; Janssen et al., 2005; Abdul Ajees et al., 2006; Marrero et al., 2012; Wong & Dessen, 2014). α2Ms are activated through the protease cleavage of a largely disordered bait region, which results in a conformational change that both traps the protease in a cage-like structure and exposes the highly reactive thioester bond (Travis & Salvesen, 1983; Sottrup-Jensen, 1989; Sottrup-Jensen et al., 1989; Doan & Gettins, 2008).
In human α2M, chemical cleavage of the thioester bond by methylamine results in a large conformational change that has been interpreted by electron microscopy as similar to that observed on protease cleavage (Sottrup-Jensen et al., 1980; Tapon-Bretaudiére et al., 1985; Dodds et al., 1996; Dodds & Law, 1998; Marrero et al., 2012). In contrast, the recently elucidated structures of the BA2M from Salmonella enterica serovar Typhimurium (SaA2M) in its unactivated and methylamine-activated forms show that although the overall domain structure of BA2Ms is highly similar to that of human α2M, there is no major conformational change of the bacterial form on chemical cleavage of the thioester bond (Doan & Gettins, 2008; Neves et al., 2012; Wong & Dessen, 2014). In the structures of both bacterial and eukaryotic α2Ms the thioester bond lies close to the surface of the TED, but is protected from hydrolysis by a hydrophobic pocket at the interface between the TED and the CTMG domain (Janssen et al., 2005; Le et al., 2012; Wong & Dessen, 2014).
Owing to a lack of detailed structural information on protease-cleaved forms of α2M, the mechanism through which cleavage of the BRD activates α2M is not known. However, it has been suggested that upon cleavage within human α2M the cleaved BRD interacts with MG2 (MG4 in BA2M), resulting in conformational activation (Marrero et al., 2012). Alternatively, within human α2M it has been suggested that the bait region interacts with the TED, the MG6 (MG8 in BA2M) and the CUB domains (Marrero et al., 2012). However, in the absence of a crystal structure of a protease-cleaved form of α2M the mechanism of protease-induced activation remains speculative.
To elucidate the mechanism of protease-induced α2M activation, we crystallized and solved the X-ray structure of a porcine elastase-cleaved form of ECAM, a close homologue of SaA2M (81% amino-acid sequence identity) for which the structure of the unactivated form was recently solved (Wong & Dessen, 2014). Interestingly, the structure of protease-activated ECAM is highly similar to that of chemically activated human α2M (12% amino-acid sequence identity) and reveals a clear mechanism of how conformational rearrangement is triggered on protease cleavage. Key to activation is the untethering of the intrinsically disordered bait region on cleavage, allowing this disordered region of polypeptide to outcompete the domain–domain interactions that normally maintain the thioester bond in its unactivated form. This suggests a general mechanism through which members of the large and important α2M superfamily are activated.
All chemicals were purchased from Sigma unless mentioned otherwise.
The gene for ECAM, yfhM from E. coli K-12, was amplified by PCR and cloned into pET-21a vector using NdeI and XhoI restriction sites. The first 22 residues from the N-terminus of the gene, containing a signal sequence identified using SignalP, were excluded from the construct. The stop codon was also excluded, resulting in a protein consisting of residues 23–1631 and a C-terminal 6×His tag (LEHHHHHH). ECAM was initially overexpressed in E. coli BL21(DE3) cells and subsequently in T7 Express Crystal Competent E. coli cells (methionine-auxotrophic strain, New England Biosciences) using an inducible T7 promoter with 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) as the inducer. Bacteria expressing ECAM were grown in lysogeny broth; selenomethionine-labelled ECAM was obtained using M9 minimal medium supplemented with 50 mg l−1 selenomethionine and 20 mg l−1 of each of nine essential amino acids (excluding methionine). Cells were grown at 37°C to an OD600 of 0.6, protein production was induced by the addition of 1 mM IPTG and the cells were grown for a further 6 h. The cell pellet was collected by centrifugation at 4400g for 15 min and the cells were resuspended in binding buffer [20 mM Tris, 10 mM imidazole, 500 mM sodium chloride, 5 mM tris(2-carboxyethyl)phosphine (TCEP) pH 7.5] and lysed by sonication with 1 mg ml−1 lysozyme in the presence of protease inhibitors (Complete Mini, Roche). Cell debris was removed by centrifugation at 46 000g for 30 min at 4°C. The cell supernatant was then loaded onto a HisTrap HP column (GE Healthcare) and the bound protein was eluted with elution buffer (20 mM Tris, 500 mM imidazole, 500 mM sodium chloride, 5 mM TCEP pH 7.5) using a linear gradient increasing from 10 to 500 mM. Fractions containing ECAM were pooled and dialysed overnight at 4°C into 50 mM Tris, 200 mM sodium chloride pH 7.5 and run on a Superdex S200 gel-filtration column (GE Healthcare). Central fractions from the peak were combined and concentrated using a 100 kDa molecular-weight cutoff centrifugal concentrator.
Purified ECAM was reacted in a 1:1 molar ratio with porcine elastase (MP Biomedicals) in 50 mM Tris, 200 mM NaCl pH 7.5 on ice for 5 min before being loaded onto a Superdex S200 gel-filtration column (GE Healthcare). The two major peaks from gel filtration were concentrated to 16 mg ml−1 separately using 100 kDa molecular-weight cutoff centrifugal concentrators and used in crystallization trials. Several hundred crystallization conditions were tested, including the JCSG-plus, MIDAS and Morpheus screens (Molecular Dimensions), for both concentrated peaks. A Cartesian Honeybee 8+1 (Harvard Bioscience) robot was used with 96-well plates, dispensing 0.5 µl reservoir solution and 0.5 µl protein sample. Subsequent scaled-up crystal growth was performed using 2.5 µl reservoir solution and 2.5 µl protein sample. The initial crystal was grown in conditions consisting of 0.1 M potassium chloride, 0.1 M HEPES, 25% Sokalan CP 7 pH 7.0, and upon optimization the pH was adjusted to 7.5 for larger crystal growth. Crystals were grown using equal volumes of protease-cleaved ECAM and reservoir solution using sitting-drop vapour diffusion, with crystals appearing after 2 d at 16°C for the second fraction and after two weeks for the first fraction. Cryoprotection was optimized with a 3:2 ratio of xylitol-saturated reservoir solution to reservoir solution. Crystals were briefly soaked and flash-cooled in liquid nitrogen for data collection. The best diffraction resolution obtained was 3.8 Å, and molecular replacement with methylamine-activated α2M (PDB entry 4acq ) was unsuccessful, most likely as the sequence identity with the human homologue was low (12%) and owing to the difference in domain orientation between the structural models (Marrero et al., 2012). Further expression was performed using a methionine-auxotrophic strain of E. coli BL21 (T7 Express Crystal Competent E. coli, New England Bioscience) and the purification and crystallization screens were repeated using selenomethionine-labelled protein. As repeating the previous protocol with selenomethionine-labelled protein was unsuccessful, in situ proteolytic cleavage screening was performed using porcine elastase. Successful crystallization was achieved using a 1:100 ratio of porcine elastase to selenomethionine-labelled ECAM. Crystallization was successful in the same condition as used previously, with the crystal having a similar appearance and the same space group as previous unlabelled crystals. These crystals diffracted to 3.65 Å resolution and phases were obtained using single-wavelength anomalous diffraction (SAD).
Data were collected for ECAM crystals on the I02, I03 and I24 beamlines at Diamond Light Source, Didcot, England at 100 K at the Se K edge (λ = 0.97939 Å) using a PILATUS 6M detector. A high-redundancy SAD data set was processed and scaled using XDS and AIMLESS from the CCP4 suite of programs (Evans, 2006; Kabsch, 2010; Winn et al., 2011). Selenium sites were located using SHELXC/D, with the best substructure solution consisting of 23 sites (Sheldrick, 2010). These selenium sites were input along with the SAD data set to AutoSol within the PHENIX package to perform phasing and density modification (Terwilliger et al., 2009). Density modification in AutoSol was sufficient to break the phase ambiguity owing to the high solvent content of the crystal (69%). This yielded interpretable, low-resolution maps in which density corresponding to secondary-structure elements and larger amino-acid side chains was visible. Initially, the α-helical TED domain was built in Coot using idealized α-helical sections (Emsley et al., 2010). The loops between these sections were connected where density was available. The six selenomethionine sites in this domain provided the starting sites for the building of amino-acid side chains in the TED domain, although owing to the resolution initially only larger side chains and those where continuous sequence could be determined were built. In addition to the building of the TED domain, a number of α2M-derived polyalanine MG domains were rigid-body fitted into their corresponding density and manually real-space refined in Coot (Emsley et al., 2010). As with the TED domain, where possible side chains were fitted using the positions of selenomethionine Se atoms as starting sites. This initial building yielded a partial model, which was then used in conjunction with the selenomethionine substructure to rephase the experimental data using MR-SAD phasing in Phaser (McCoy et al., 2007). This process led to phase improvement and the appearance of new features in the map, which were modelled, and the process was repeated iteratively. Partway through the building process the atomic coordinates for SaA2M were published (PDB entry 4u48 ; Wong & Dessen, 2014), and the domains from this model provided validation of the MG-domain placement and side-chain modelling in our experimentally phased map. The SaA2M structure also provided a template for building the more difficult sections of the model. At this point restrained TLS refinement using REFMAC5 was found to stably improve both Rwork and Rfree, and refinement was performed and the model was improved and finished manually in Coot (Murshudov et al., 2011; Emsley et al., 2010). Electron density for the thioester bond indicated that the deaminated glutamine forms no covalent bond to the cysteine. Before submission of the final model, the quality of the structure was assessed using the MolProbity webserver (Chen et al., 2010). The atomic coordinates and structure factors were deposited in the Protein Data Bank (PDB entry 4rtd ). Statistics for data collection, experimental phasing and refinement are presented in Table 1. For mass-spectrometric analysis of protease-cleaved ECAM, crystals were washed in reservoir solution before being dissolved in deionized water and heated to 96°C in bromophenol blue sample buffer for 5 min. The sample was then run on a NuPAGE Novex 4–12% bis-tris gel (Invitrogen) and visible bands were cut for proteomic analysis. Samples were digested by trypsin and analysed by LC-MS/MS (Orbitrap XL) performed at the Fingerprints Proteomics Facility at the University of Dundee.
‡Percentages of residues in favoured/allowed regions calculated by MolProbity (Chen et al., 2010).
To determine the structural changes that occur on protease cleavage of ECAM, we performed protease digestion with porcine elastase and used the major products from gel filtration of cleaved ECAM to perform crystallization trials. This yielded diffracting crystals in 0.1 M potassium chloride, 0.1 M HEPES, 25% Sokalan CP 7 pH 7.0 with cleaved ECAM. In order to obtain phase information, we attempted to repeat this process with selenomethionine-labelled ECAM, but this failed to yield crystals. However, an alternative strategy of in situ proteolysis and crystallization with selenomethionine-labelled ECAM was successful. This method yielded crystals that diffracted to 3.65 Å resolution. Upon completion and validation of the model of protease-activated ECAM, the number of Ramachandran outliers remaining was appropriate for a crystal structure with a resolution of 3.65 Å. Although the Rmerge and Rp.i.m. values were high, this can be explained by the highly redundant data set used; with a CC1/2 of 0.675 in the highest shell, these data used were judged to be acceptable (Karplus & Diederichs, 2012; Diederichs & Karplus, 2013).
Similar to the domain architecture of native SaA2M, uncleaved ECAM consists of ten MG domains, with a largely disordered bait-region domain found within MG8, a TED that houses the reactive thioester bond and a CUB domain (Fig. 1a). Elastase-cleaved ECAM (Fig. 1b) adopts a conformation similar to that of methylamine-activated human α2M (Fig. 1c) but distinct from the unactivated form of SaA2M (Fig. 1d). Despite the low sequence identity between human α2M and ECAM (12%), the r.m.s.d. between Cα atoms for these proteins is 14.1 Å, while that for SaA2M and ECAM, which share 82% sequence identity, is 22.1 Å. Most notably, in the structure of the protease-cleaved ECAM, interactions between the TED and the CTMG domain, which protects the thioester bond in unactivated SaA2M, are not present. Instead, the thioester region of TED is solvent-exposed and faces the expected location of the attacking protease, as it would be positioned when cleaving the bait region of ECAM (Fig. 1b). Electron density for elastase, in addition to that for MG domains 1, 2, 3 and 7, was absent in 2Fo − Fc maps of cleaved ECAM, as was electron density for 20 amino-acid residues (Arg923–Leu942) within the bait region. Analysis of elastase-cleaved ECAM by SDS–PAGE and mass spectrometry confirms that MG domains 1, 2, 3 and 7 are absent from the crystallized protein (Supplementary Fig. S1). Two bands at 90 and 75 kDa on SDS–PAGE have similar peptide coverage, encompassing the same domains (Supplementary Fig. S1). However, the reason for the difference in apparent molecular weights between these two species is not known.
Although it is known that protease cleavage of the largely disordered bait region of both human and bacterial α2Ms gives rise to a large conformational rearrangement and activation of the thioester bond, the mechanism through which this is mediated was unclear (Barrett et al., 1979; Neves et al., 2012). However, the structure of protease-cleaved ECAM provides a clear mechanism for this process. Thioester bond release is achieved by the untethering of a large region of an unstructured polypeptide chain upon cleavage of the BRD that forms new interactions with the CTMG domain, thereby preventing the protective interaction with the TED. Specifically, in the elastase-cleaved form of ECAM additional residues (Phe947–Asn963) of the BRD are observed that are disordered in the uncleaved SaA2M structure. All of these residues in the protease-cleaved ECAM structure are ordered owing to the formation of a new binding interface between the BRD and the CTMG domain upon protease cleavage (Fig. 2a). Critically, the region of the CTMG domain that is involved in BRD binding substantially overlaps with the region of the CTMG domain that forms the binding interface with the TED in the uncleaved form of the protein (Fig. 2b). The buried surface area between the CTMG domain and TED in SaA2M is 1972 Å2, with three hydrogen bonds between these domains, whereas the buried surface area between the elastase-cleaved BRD and the CTMG domain in ECAM is 1438 Å2, with five hydrogen bonds between the domains. In the protease-cleaved ECAM, hydrogen-bond formation between residues of the CTMG domain and the cleaved bait region involves the highly conserved RDDR and EXMY motifs (Fig. 3). Interestingly, it is residues within these motifs that also form hydrogen bonds to the TED in the uncleaved SaA2M protein (Fig. 3; Wong & Dessen, 2014).
We propose that the loss of the interaction of the TED with the CTMG domain is sufficient to enable both the large TED movement observed in the cleaved ECAM structure relative to the unactivated SaA2M structure and the exposure of the thioester bond (Fig. 4a, Supplementary Movie S1). The movement of the TED domain shows an overall shift of 36 Å, with the MG6 domain moving by 50 Å and with both the MG6 and TED arms hugging the position at which the attacking protease would be located to cleave the BRD (Fig. 4a, Supplementary Movie S1). In addition to these global conformational changes, protease cleavage leads to localized changes in the environment of the thioester bond that are likely to lead to its activation. In the unactivated SaA2M structure the conserved methionine and tyrosine side chains (Met1625 and Tyr1626 in SaA2M) from the CTMG domain form part of the hydrophobic pocket at the interface with the TED that has been shown to be important in maintenance of the thioester bond (Wong & Dessen, 2014). The loss of these key side-chain interactions and the movement of the additional side chains which constitute the protective hydrophobic pocket (Tyr1175 and Tyr1177 in SaA2M and Tyr1183 and Tyr1185 in ECAM) from the TED exposes the thioester bond to the solvent, allowing hydrolysis or covalent-bond formation with the attacking protease (Fig. 4b). As there are no contacts present from the CTMG domain providing a hydrophobic pocket to protect the thioester from hydrolysis by water, the thioester bond could be cleaved and the deaminated glutamine (Gln1190) would be converted to a glutamic acid. Within our protease-activated ECAM structure the distance between the S atom of Cys1187 and the Cγ atom of Gln1190 is 4.6 Å, indicating that the thioester bond may not be intact (Supplementary Fig. S2). However, we cannot rule out the possibility that there is a mixed population of molecules, some of which possess an intact thioester bond. Owing to this ambiguity, we have represented the thioester without a covalent bond between Cys1187 and the Cγ atom of Gln1190 and have also omitted the O atom from the deaminated glutamine that would be formed on hydrolysis of the thioester bond.
Although it has been suggested that small-molecule-activated human α2M resembles the conformation of the protease-cleaved form, there have to date been no detailed structural data to confirm this (Tapon-Bretaudiére et al., 1985). The similarity in the overall conformation of protease-cleaved ECAM to that of methylamine-activated human α2M confirms that the protease-activated and chemically activated forms of human α2M are structurally equivalent. In addition, these data suggest that the entrapment of cleaving proteases is likely to occur in a similar manner for BA2Ms as has been proposed for human α2M (Marrero et al., 2012; Meyer et al., 2012). However, there are clearly key differences in the details of the interactions that maintain the inactive conformations of BA2Ms and human α2M, since the recent structures of unactivated and methylamine-activated SaA2M show that chemical cleavage of the thioester bond does not lead to global conformational changes in this case (Marrero et al., 2012; Wong & Dessen, 2014). This difference is likely to be owing to the domain location of the side chains that comprise the thioester-protecting pocket (Supplementary Fig. S3). In BA2Ms this pocket comprises two tyrosine side chains from the TED and a tyrosine and a methionine side chain from the CTMG domain, whereas in eukaryotic α2Ms all four residues (three tyrosine side chains and one methionine side chain) are found in the CTMG domain. When SaA2M is reacted with methylamine no conformational change is seen, but Tyr1175 from the TED is displaced (Wong & Dessen, 2014). The structural counterpart of this side chain (Tyr1307 from TEP1) in eukaryotic α2M family members is, however, found within the CTMG domain (Supplementary Fig. S3; Baxter et al., 2007). It is presumably the rearrangement of this side chain and perhaps other CTMG side chains that comprise the thioester-protecting pocket which leads to the loss of TED–CTMG domain binding and subsequent global conformational changes. The buried surface area seen between the TED and the CTMG domain in native SaA2M decreases by 27% compared with that of the BRD and the CTMG domain in protease-cleaved ECAM; however, the number of hydrogen bonds increases. Although there is a decrease in the buried surface area, it is the interaction between important residues for protecting the thioester that would be expected to trigger conformational change releasing TED and allow subsequent interaction with a cleaving protease.
The role of BA2Ms, which are inner membrane-anchored periplasmic proteins, has been suggested as protease inhibitors that inhibit exogenous proteases that have breached the outer membrane. Comparison of the structures of unactivated SaA2M and protease-cleaved ECAM illustrates how protease-induced conformational changes enable protease entrapment (Fig. 5). The movement of the TED-domain arm and MG6-domain arm around the central pocket above the bait region is similar to the movement seen in methylamine-activated human α2M; when accounting for the domains that are not present in the protease-activated ECAM structure no clashes are seen between the moving arms and the absent domains (Marrero et al., 2012). The entrapment of proteases would limit the proteolysis to smaller substrates, as has been suggested for human α2M, and would prevent the cleavage of important larger substrates such as the peptide component of the peptidoglycan layer or large proteins (Fig. 5; Sottrup-Jensen, 1989). Although we cannot be sure why the covalently bound elastase is not present in our structure, this may be owing to a lack of available and correctly positioned lysine side chains on the surface of the protease, since the thioester bond is preferentially cleaved by this side chain (Marrero et al., 2012). The lack of MG domains 1, 2, 3 and 7 within the crystal lattice is likely to be owing to the MG domains being cleaved by the elastase.
In summary, the structure of protease-activated ECAM shows a competitive mechanism of activation in which cleavage of the BRD allows the normally intrinsically disordered region of this domain to outcompete the TED for CTMG-domain binding. Loss of the TED–CTMG domain interaction leads to a large conformational rearrangement of ECAM and exposure of the reactive thioester bond. The structural similarity between methylamine-activated human α2M and protease-cleaved ECAM suggests that similar mechanisms are likely to operate across the diverse members of the α2M family.
We would like to acknowledge the Diamond Light Source for access to beamlines I02, I03 and I24 (proposal No. MX5689). CDF is supported by a four-year studentship from the BBSRC/EPSRC doctoral training centre in cell and proteomic engineering (Grant No. EP/F500424/1). IJ is supported by a four-year studentship from the Wellcome Trust (Grant No. 093597/Z/10/Z). RG is supported by a Kelvin Smith Scholarship from the University of Glasgow.
Abdul Ajees, A., Gunasekaran, K., Volanakis, J. E., Narayana, S. V. L., Kotwal, G. J. & Krishna Murthy, H. M. (2006). Nature (London), 444, 221–225. CrossRef PubMed CAS Google Scholar
Barrett, A. J., Brown, M. A. & Sayers, C. A. (1979). Biochem. J. 181, 401–418. CAS PubMed Google Scholar
Baxter, R. H. G., Chang, C.-I., Chelliah, Y., Blandin, S., Levashina, E. A. & Deisenhofer, J. (2007). Proc. Natl Acad. Sci. USA, 104, 11615–11620. CrossRef PubMed CAS Google Scholar
Boer, J. P. de, Creasey, A. A., Chang, A., Abbink, J. J., Roem, D., Eerenberg, A. J., Hack, C. E. & Taylor, F. B. (1993). Infect. Immun. 61, 5035–5043. PubMed Google Scholar
Budd, A., Blandin, S., Levashina, E. A. & Gibson, T. J. (2004). Genome Biol. 5, R38. CrossRef PubMed Google Scholar
Chen, V. B., Arendall, W. B., Headd, J. J., Keedy, D. A., Immormino, R. M., Kapral, G. J., Murray, L. W., Richardson, J. S. & Richardson, D. C. (2010). Acta Cryst. D66, 12–21. Web of Science CrossRef CAS IUCr Journals Google Scholar
Diederichs, K. & Karplus, P. A. (2013). Acta Cryst. D69, 1215–1222. Web of Science CrossRef CAS IUCr Journals Google Scholar
Doan, N. & Gettins, P. G. W. (2007). Biochem. J. 407, 23–30. CrossRef PubMed CAS Google Scholar
Doan, N. & Gettins, P. G. W. (2008). J. Biol. Chem. 283, 28747–28756. CrossRef PubMed CAS Google Scholar
Dodds, A. W. & Law, S. K. A. (1998). Immunol. Rev. 166, 15–26. CrossRef CAS PubMed Google Scholar
Dodds, A. W., Ren, X.-D., Willis, A. C. & Law, S. K. A. (1996). Nature (London), 379, 177–179. CrossRef CAS PubMed Google Scholar
Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. (2010). Acta Cryst. D66, 486–501. Web of Science CrossRef CAS IUCr Journals Google Scholar
Evans, P. (2006). Acta Cryst. D62, 72–82. Web of Science CrossRef CAS IUCr Journals Google Scholar
Huang, W., Dolmer, K., Liao, X. & Gettins, P. G. W. (1998). Protein Sci. 7, 2602–2612. CrossRef CAS PubMed Google Scholar
Jacobsen, L. & Sottrup-Jensen, L. (1993). Biochemistry, 32, 120–126. CrossRef CAS PubMed Google Scholar
Janssen, B. J. C., Huizinga, E. G., Raaijmakers, H. C. A., Roos, A., Daha, M. R., Nilsson-Ekdahl, K., Nilsson, B. & Gros, P. (2005). Nature (London), 437, 505–511. Web of Science CrossRef PubMed CAS Google Scholar
Kabsch, W. (2010). Acta Cryst. D66, 125–132. Web of Science CrossRef CAS IUCr Journals Google Scholar
Karplus, P. A. & Diederichs, K. (2012). Science, 336, 1030–1033. Web of Science CrossRef CAS PubMed Google Scholar
Le, B. V., Williams, M., Logarajah, S. & Baxter, R. H. G. (2012). PLoS Pathog. 8, e1002958. CrossRef PubMed Google Scholar
Li, Z.-F., Wu, X.-H. & Engvall, E. (2004). Genomics, 83, 1083–1093. CrossRef PubMed CAS Google Scholar
Lin, M., Sutherland, D. R., Horsfall, W., Totty, N., Yeo, E., Nayar, R., Wu, X.-F. & Schuh, A. C. (2002). Blood, 99, 1683–1691. CrossRef PubMed CAS Google Scholar
Marrero, A., Duquerroy, S., Trapani, S., Goulas, T., Guevara, T., Andersen, G. R., Navaza, J., Sottrup-Jensen, L. & Gomis-Rüth, F. X. (2012). Angew. Chem. Int. Ed. 51, 3340–3344. CrossRef CAS Google Scholar
McCoy, A. J., Grosse-Kunstleve, R. W., Adams, P. D., Winn, M. D., Storoni, L. C. & Read, R. J. (2007). J. Appl. Cryst. 40, 658–674. Web of Science CrossRef CAS IUCr Journals Google Scholar
Meyer, C., Hinrichs, W. & Hahn, U. (2012). Angew. Chem. Int. Ed. 51, 5045–5047. CrossRef CAS Google Scholar
Murshudov, G. N., Skubák, P., Lebedev, A. A., Pannu, N. S., Steiner, R. A., Nicholls, R. A., Winn, M. D., Long, F. & Vagin, A. A. (2011). Acta Cryst. D67, 355–367. Web of Science CrossRef CAS IUCr Journals Google Scholar
Neves, D., Estrozi, L. F., Job, V., Gabel, F., Schoehn, G. & Dessen, A. (2012). PLoS One, 7, e35384. Web of Science CrossRef PubMed Google Scholar
Osterberg, R. & Malmensten, B. (1984). Eur. J. Biochem. 143, 541–544. CrossRef CAS PubMed Google Scholar
Schiffer, G. & Höltje, J. V. (1999). J. Biol. Chem. 274, 32031–32039. CrossRef PubMed CAS Google Scholar
Sheldrick, G. M. (2010). Acta Cryst. D66, 479–485. Web of Science CrossRef CAS IUCr Journals Google Scholar
Sottrup-Jensen, L. (1989). J. Biol. Chem. 264, 11539–11542. CAS PubMed Google Scholar
Sottrup-Jensen, L., Petersen, T. E. & Magnusson, S. (1980). FEBS Lett. 121, 275–279. CAS PubMed Google Scholar
Sottrup-Jensen, L., Sand, O., Kristensen, L. & Fey, G. H. (1989). J. Biol. Chem. 264, 15781–15789. CAS PubMed Google Scholar
Tapon-Bretaudiére, J., Bros, A., Couture-Tosi, E. & Delain, E. (1985). EMBO J. 4, 85–89. PubMed Google Scholar
Terwilliger, T. C., Adams, P. D., Read, R. J., McCoy, A. J., Moriarty, N. W., Grosse-Kunstleve, R. W., Afonine, P. V., Zwart, P. H. & Hung, L.-W. (2009). Acta Cryst. D65, 582–601. Web of Science CrossRef CAS IUCr Journals Google Scholar
Travis, J. & Salvesen, G. S. (1983). Annu. Rev. Biochem. 52, 655–709. CrossRef CAS PubMed Google Scholar
Winn, M. D. et al. (2011). Acta Cryst. D67, 235–242. Web of Science CrossRef CAS IUCr Journals Google Scholar
Wong, S. G. & Dessen, A. (2014). Nature Commun. 5, 4917. CrossRef Google Scholar
This is an open-access article distributed under the terms of the Creative Commons Attribution (CC-BY) Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original authors and source are cited.