research papers
A correction has been published for this article. To view the correction, click here.
Structures of collagen IV globular domains: insight into associated pathologies, folding and network assembly
aDepartment of Biochemistry and Molecular Biology/ERI BIOTECMED, Universitat de València, Dr Moliner 50, Burjassot, 46100 Valencia, Spain, bInstituto de Biomedicina de Valencia, Consejo Superior de Investigaciones Científicas (IBV–CSIC), Jaume Roig 11, 46010 Valencia, Spain, cCIBER de Enfermedades Raras (CIBERER–ISCIII), Spain, dLaboratorio de Reconocimiento Molecular, Centro de Investigación Príncipe Felipe, Eduardo Primo Yúfera 3, 46012 Valencia, Spain, eDepartamento de Microbiología, Facultad de Medicina at Universitat de València, Blasco Ibáñez 17, 46010 Valencia, Spain, fDepartment of Defense, Center for Prostate Disease Research, Bethesda, Maryland, USA, gBioStratum Inc., Durham, North Carolina, USA, hDepartment of Medicine at Division of Nephrology and Hypertension, Vanderbilt University Medical Center, Nashville, TN 37232, USA, iCenter for Matrix Biology, Vanderbilt University Medical Center, Nashville, TN 37232, USA, jDepartment of Biochemistry, Vanderbilt University Medical Center, Nashville, TN 37232, USA, kDepartment of Pathology, Microbiology and Immunology, Vanderbilt University Medical Center, Nashville, TN 37232, USA, lDepartment of Cell and Developmental Biology, Vanderbilt University Medical Center, Nashville, TN 37232, USA, mVanderbilt Ingram Cancer Center, Vanderbilt University Medical Center, Nashville, TN 37232, USA, nVanderbilt Institute of Chemical Biology, Vanderbilt University Medical Center, Nashville, TN 37232, USA, and oDepartamento de Bioquímica y Biología Molecular at Facultad de Medicina y Odontología, Universitat de València, Blasco Ibáñez 15-17, 46010 Valencia, Spain
*Correspondence e-mail: patricia.casino@uv.es, amarina@ibv.csic.es
Basement membranes are extracellular structures of epithelia and endothelia that have collagen IV scaffolds of triple α-chain helical protomers that associate end-to-end, forming networks. The molecular mechanisms by which the noncollagenous C-terminal domains of α-chains direct the selection and assembly of the α1α2α1 and α3α4α5 hetero-oligomers found in vivo remain obscure. Autoantibodies against the noncollagenous domains of the α3α4α5 hexamer or mutations therein cause Goodpasture's or Alport's syndromes, respectively. To gain further insight into oligomer-assembly mechanisms as well as into Goodpasture's and Alport's syndromes, crystal structures of noncollagenous domains produced by recombinant methods were determined. The spontaneous formation of canonical homohexamers (dimers of trimers) of these domains of the α1, α3 and α5 chains was shown and the components of the Goodpasture's disease epitopes were viewed. Crystal structures of the α2 and α4 noncollagenous domains generated by recombinant methods were also determined. These domains spontaneously form homo-oligomers that deviate from the canonical architectures since they have a higher number of subunits (dimers of tetramers and of hexamers, respectively). Six flexible structural motifs largely explain the architectural variations. These findings provide insight into noncollagenous domain folding, while supporting the in vivo operation of extrinsic mechanisms for restricting the self-assembly of noncollagenous domains. Intriguingly, Alport's syndrome missense mutations concentrate within the core that nucleates the folding of the noncollagenous domain, suggesting that this syndrome, when owing to missense changes, is a folding disorder that is potentially amenable to pharmacochaperone therapy.
Keywords: collagen type IV; network assembly; (IV)NC1 hexamers; Goodpasture's disease; Alport's syndrome.
PDB references: collagen type IV α1α2α1NC1, 5nax; α1NC1, 5nay; α2NC1, 5nb2; α3NC1, 5nb0; α4NC1, 5nb1; α5NC1, 5naz
1. Introduction
The extracellular microenvironment plays a pivotal role in tissue genesis, architecture and function. A defining morphological feature of these microenvironments is the basement membrane (BM), an ancient and specialized form of extracellular matrix that is conserved from cnidarians to humans. BMs underlie epithelia and endothelia (Hagios et al., 1998; Rhodes & Simons, 2007; Fidler et al., 2017) and ensheath muscle, fat and Schwann cells (Campbell & Stull, 2003; Sanes, 2003; Sillat et al., 2012; Court et al., 2006). BMs function as supramolecular scaffolds that compartmentalize and provide structural integrity to tissues, guide cell migration and adhesion, delineate apical–basal polarity and modulate cell differentiation during development (Hynes, 2009; Pastor-Pareja & Xu, 2011; Daley & Yamada, 2013).
Type IV collagen is a major constituent of BMs. This evolutionarily conserved protein forms networks and plays a crucial structural role in the maintenance of BM architecture (Brown et al., 2017; Fidler et al., 2018). In addition, it serves as a ligand for cell-surface integrins, thus influencing cell adhesion, migration and differentiation (Wang et al., 2008). Not surprisingly, mutations in collagen IV cause BM destabilization and tissue dysfunction, and are associated with inborn defects such as Alport's syndrome and cerebral haemorrhage (Hudson et al., 2003; Kuo et al., 2012; Gould et al., 2005). Furthermore, Goodpasture's disease results from a selective antibody-mediated aggression to collagen IV, specifically to epitopes localized in the globular noncollagenous domains (NC1s) of collagen IV α-chains 3 and 5 (Hudson et al., 2003; Cui et al., 2016). These are two of the six types of collagen IV chains that exist in humans [designated α1–α6(IV)NC1s; in the following the (IV) will be omitted for simplicity]. The Goodpasture epitopes remain to be structurally characterized, but they are considered to be cryptic, becoming immuno-visible as a consequence of solvent exposure or viral infection (Pedchenko et al., 2010). These collagen IV-related disorders can be deadly as a result of impaired BM function.
Unveiling how collagen IV is organized in BMs seems to be crucial in order to understand the molecular architectures of BMs and to clarify the pathogenesis of BM-related diseases. As previously mentioned, six different types of homologous polypeptide chains of similar lengths (1669–1712 residues) can be found in collagen type IV. These chains (α1–α6), encoded by paralogous genes, are composed of a short N-terminal domain termed the 7S domain (∼25 amino acids), a long central collagenous domain (∼1400 amino acids) and a noncollagenous globular C-terminal domain of around 230 amino acids [abbreviated as (IV)NC1; the (IV) will be omitted in the following; Timpl et al., 1981; Borza et al., 2001]. Three α-chains associate into a protomer with a long parallel triple-helical shaft and nonhelical ends. In the collagen IV network found in BMs, different protomers interconnect by end-to-end interactions. At the C-termini, interactions occur between the trimeric NC1 domains of two protomers, forming a hexamer of these domains reinforced by sulfilimine cross-links (Vanacore et al., 2009). At the N-termini, four protomers interact through their 7S domains, forming a 7S dodecamer reinforced by aldehyde-derived cross-links (Anazco et al., 2016). In this way, a combination of C-terminal hexameric and N-terminal dodecameric interactions generates the three-dimensional collagen IV scaffold to which other components of the BMs bind (Hudson et al., 2003; Brown et al., 2017). Intriguingly, only three distinct protomers, α1α2α1, α3α4α5 and α5α6α5 (abbreviated in the following as α121, α345 and α565, respectively), forming three distinct networks, α121, α345 and α121–α565, are known to occur in BMs (Boutaud et al., 2000; Borza et al., 2001; Gunwar et al., 1998). Previously unrecognized collagen IV α125 chain combinations have recently been found in various cancer cell lines, which form networks of protomers of α1α2 chains and α5-chain homoprotomers (Revert et al., 2018). The formation of these novel collagen IV networks is dependent on an exportable protein kinase, the Goodpasture antigen-binding protein (Raya et al., 1999; Revert et al., 2008, 2018), pointing to the involvement of the cellular machinery in regulating the organization of collagen IV networks.
The assembly of the collagen IV network involves two distinct stages of α-chain Three α-chains first associate intracellularly, forming protomers, and the protomers then associate end-to-end outside the cell, generating extracellular networks. Hints on the assembly mechanisms were inferred (i) from the of the α121NC1 hexamer isolated from natural sources (Sundaramoorthy et al., 2002; Than et al., 2002; Vanacore et al., 2004), (ii) from the ability of recombinant α1NC1 chains to selectively assemble into hexamers (Boutaud et al., 2000) and (iii) from refolding experiments with NC1 hexamers (Dölz et al., 1988). Recent studies in one of our laboratories (Cummings et al., 2016) revealed that the NC1 domains function as recognition modules, directing the selection and assembly of α-chains into protomers and networks. In turn, chloride ions activate a molecular switch present in NC1 domains that triggers the of protomers into networks. However, how NC1 recognition modules direct α-chain has so far been ignored.
To obtain further insight into collagen IV network assembly and to advance the characterization of the molecular bases of Goodpasture's and Alport's syndromes, we generated various NC1 domains via recombinant technology and used protein crystallography to identify and structurally characterize the oligomers produced by these domains. Using this approach, while attempting to produce crystals of the α345NC1 heterohexamer that predominates in the BM of the kidney, we solved the three-dimensional structure of the recently reported homohexamer of α5NC1 domains (Revert et al., 2018). In addition, we crystallized and determined the structures of the homohexamers spontaneously formed by recombinantly produced α1NC1 and α3NC1 domains. Interestingly, when carrying out similar studies with crystals of the α2NC1 and α4NC1 domains, oligomers were spontaneously formed that were architecturally similar to the hexamers observed for the other chains except that they were dimers of tetramers (octamers) or of hexamers, respectively, suggesting the existence of restrictions that limit the oligomers produced in vivo. We also validated the use of self-assembled recombinantly produced NC1 domains for structural studies by reproducing in vitro the α121 heterohexamer previously obtained from natural sources (Sundaramoorthy et al., 2002; Than et al., 2002; Vanacore et al., 2004). The information gathered provided insights into the structural features involved in the folding, selection and of collagen IV chains to form a protomer. Furthermore, we provide the first snapshots of the components of the Goodpasture autoantigen. We also provide insight into Alport's syndrome, helping to rationalize the structural bases of the effects of the mutations reported in this disease that map to NC1 domains, and concluding that of these mutations, those that are amino-acid substitutions may cause disease by inducing misfolding, thus opening the way to attempting therapy using pharmacochaperones.
2. Methods
2.1. Cloning, protein expression and purification
We obtained a cDNA library from HEK293 cells (a commercial cell line derived from human embryonic kidney) by retrotranscription with SuperScript reverse transcriptase (Invitrogen) of α1NC1, α2NC1, α4NC1 and α5NC1 domains using a high-fidelity thermophilic DNA polymerase (Pfu polymerase, Stratagene) and the primers listed in Supplementary Table S1. The PCR-amplified coding sequences for these NC1 domains were used in a second round of PCR amplification with additional primers (Supplementary Table S1) to introduce an N-terminal BM40 secretion peptide followed by a FLAG tag preceding the sequences of the indicated NC1 domains (the residues forming each of these domains are indicated in Supplementary Table S2). These BM40-FLAG-tagged domains were then subcloned into pFastBac1 using BamHI and SacI sites for the cloning of α2NC1 and α5NC1, and XhoI and KpnI sites for the cloning of α1NC1 and α4NC1. In the case of α3NC1, the BM40-FLAG-tagged α3NC1 coding sequence was extracted by SacI digestion from the previously reported Fα3ANU-pRC-CMV vector (Gozalbo-Rovira et al., 2013; Netzer et al., 1999) and was then subcloned into the pFastBac1 vector.
extracted from these cells using TRIzol (Invitrogen). This cDNA library was used as template for PCR amplification of theThe resulting pFastBac1 vectors with the indicated inserts were transformed into DH10Bac competent cells and the recombinant baculoviruses were obtained following the instructions provided by the supplier of the baculovirus expression system (Bac-to-Bac Baculovirus Expression System, Invitrogen). The recombinant baculoviruses were used to produce all of the NC1 domains in Sf9 insect cells, as previously described for α2NC1 and α3NC1 (Gozalbo-Rovira et al., 2013). All purification steps were carried out at 4°C. Supernatants from 1 l cultures underwent ultracentrifugation (1 h, 160 000g) and the equivalent of 2 ml column-packed ANTI-FLAG M2 affinity gel (Sigma–Aldrich/Merck) was added. After packing the gel into a column, it was washed with 50 ml 50 mM Tris–HCl pH 7.4, 0.15 M NaCl. The NC1s were eluted with 10 ml of the same solution supplemented with 0.1 mg ml−1 soluble FLAG peptide (DYKDDDDK). The eluted proteins were concentrated and the FLAG peptide was removed by repeated runs of concentration and dilution in elution buffer without peptide using Amicon Ultra-4 10K Centrifugal Filter Devices (Merck–Millipore). Around 1 mg of soluble recombinant protein was usually obtained per litre of culture.
In addition to being produced in the baculovirus/insect-cell expression system, α2NC1 was also obtained in one of our laboratories as an N-terminally FLAG-tagged fusion protein using HEK293 cells for expression and purification, as described elsewhere (Sado et al., 1998).
2.2. Protein crystallization and structure determination
Crystals of α2NC1 (expressed in HEK293 cells) and of α1NC1, α3NC1, α4NC1, α5NC1 and α121NC1 (expressed in insect cells) were obtained by the sitting-drop vapour-diffusion technique at 21°C using 0.4 µl protein solution (7.4 mg ml−1 protein in 50 mM Tris–HCl pH 7.4, 0.15 M NaCl, except for α3NC1, for which the NaCl concentration was 0.5 M) mixed with 0.4 µl of the crystallization solution indicated in Table 1. For the α121NC1 crystals a 2:1 mixture (in terms of mass) of α1NC1 and α2NC1 chains was used, whereas for the α5NC1 crystals a 1:1:1 mixture of α3NC1, α4NC1 and α5NC1 chains was used. The crystal-harvesting solutions used for cryopreservation are listed in Table 1. X-ray diffraction was carried out at 100 K using the indicated beamlines and wavelengths (Table 1) at the European Synchrotron Facility (ESRF), Grenoble, France or the ALBA synchrotron, Cerdanyola, Barcelona, Spain. Crystallographic data sets were processed and scaled using either MOSFLM and SCALA (CCP4 suite; Winn et al., 2011) or the XDS program package (Kabsch, 2010). Data-collection details and unit-cell parameters are given in Table 1. Phases were obtained for all of the crystals by with Phaser (McCoy et al., 2007) using the structure of one α1NC1 domain of the bovine α121 as a search model (PDB entry 1t60; Vanacore et al., 2004). For the α3NC1 and α121NC1 crystals, initial data processing with SCALA indicated 39–40% in both cases. By alternating cycles of with REFMAC5 (Murshudov et al., 2011) with manual model building using Coot (Emsley et al., 2010) and, in the case of the α3NC1 and α121NC1 crystals by applying as implemented in REFMAC5, final models at 1.8–2.8 Å resolution, depending on the crystal, were obtained (see Table 1). The crystals exhibited good quality-control parameters and excellent stereochemistry (the Ramachandran plot distributions of favoured/allowed/disallowed residues for α1NC1, α2NC1, α3NC1, α4NC1, α5NC1 and α121NC1 were 97.7/2.3/0, 97.9/2.1/0, 97.7/2.3/0, 95.3/4.5/0.2, 97.8/2.2/0 and 96.3/3.7/0%, respectively). The structures and structure factors have been deposited in the PDB as entries 5nay, 5nb2, 5nb0, 5nb1, 5naz and 5nax for the α1NC1, α2NC1, α3NC1, α4NC1, α5NC1 and α121NC1 crystals, respectively.
|
2.3. Residue numbering for NC1 domains
We have followed the convention used for the previously reported structures of α121NC1 heterohexamers obtained from BMs (PDB entries 1li1, 1t60 and 1t61; Than et al., 2002; Vanacore et al., 2004). Correspondences with the residue numbering for each complete α-chain are provided in Supplementary Table S2.
2.4. Other methods
Structure-based alignments, structural superimpositions and analysis of protein contacts were performed with MUSTANG (Konagurthu et al., 2006; Papadopoulos & Agarwala, 2007), SUPERPOSE and NCONT (CCP4 suite, Winn et al., 2011), respectively. Figures were produced using PyMOL (https://www.pymol.org).
Size-exclusion M Tris–HCl pH 7.4, 0.15 M NaCl at a flow rate of 0.5 ml min−1 as a running buffer. Individual NC1 chains were incubated overnight at 20°C to allow before injection, continuously monitoring the absorbance at 280 nm in the effluent.
was performed at 20°C using a Superdex S200 (10/300) column fitted to an ÅKTA 900 FPLC system (column and instrument from GE Healthcare, Barcelona, Spain) using 50 mSurface plasmon resonance (SPR) assays were performed using a Biacore T100 instrument (GE Healthcare). α3NC1 or α5NC1 was immobilized on CM5 sensor chips (GE Healthcare) using amine-coupling chemistry by passing 1 µg ml−1 solutions of either NC1 chain in 10 mM sodium acetate pH 3 to attain a level of attachment of ∼1000 resonance units (RU). The assays were performed at 25°C by passing a solution consisting of 10 mM Tris–HCl pH 7.4, 0.15 M NaCl, 3 mM ethylenediaminetetraacetic acid, 0.005% P20 (GE Healthcare) including the indicated concentrations of α3NC1 or α5NC1 at a flow rate of 10 µl min−1 for 300 s, followed by a 300 s dissociation time with protein-free fluid. Between runs, the chip was regenerated by washing with 20 mM NaOH at a flow rate of 75 µl min−1 for 20 s followed by re-equilibration in running buffer. BIAevaluation 2.0.3 (GE Healthcare) was used to extract the kinetic data and to estimate Kd values.
SDS–PAGE was performed according to Laemmli (1970). Coomassie Blue (Bradford, 1976) was used for protein determination using a commercial reagent (Bio-Rad, California, USA) with bovine serum albumin as a standard.
3. Results
3.1. In vitro self-assembly of recombinantly produced chains generates the α5NC1 homohexamer
In an attempt to visualize the Goodpasture antigens, we tried to crystallize the α345NC1 heterohexamer by using an equimolecular mixture of pure recombinant α3NC1, α4NC1 and α5NC1 domains (Supplementary Fig. S1). We obtained crystals that diffracted X-rays to 1.85 Å resolution but that only revealed homohexamers of α5NC1 chains (Table 1 and Fig. 1a), most likely representing the novel α5NC1 homohexamer recently reported in cancer cell lines (Revert et al., 2018). experiments confirmed a strong tendency of α5NC1 to hexamerize: ∼80% of this domain was self-organized as hexamers in solution, while isolated α3NC1 or α4NC1 eluted as monomers (Supplementary Fig. S2a).
The architecture of the α5NC1 homohexamer (hereafter referred to as α5NC1homo) is essentially the same as that of the α121NC1 hexamer obtained from natural sources (Sundaramoorthy et al., 2002; Than et al., 2002; Vanacore et al., 2009). α5NC1homo was shaped as a prolate ellipsoid of ∼92 × ∼63 Å (Figs. 1a and 2, Supplementary Fig. S3a and Table S3). It was composed of two trimeric protomers meeting front-to-front at the equatorial plane of the ellipsoid in a nearly planar interaction surface. The subunit fold (Fig. 1b) showed a tandem arrangement of two similarly folded subdomains (hereafter referred to as N-sub and C-sub), as expected for the collagen IV NC1-domain fold (Pfam PF01413; https://pfam.xfam.org/family/PF01413; Supplementary Table S4). Thus, each subdomain of the α5NC1 subunit contains one α-helix and ten β-strands forming two antiparallel β-sheets (sheets I and II in N-sub and I′ and II′ in C-sub; a prime identifies the elements of C-sub; Fig. 1b). In each protomer, the I and I′ sheets (the topologies of these sheets are β1β10β2β5 and β1′β10′β2′β5′, respectively) of the three subunits alternate, forming the polar layers of the prolate ellipsoid (the regions that link the ellipsoid to the collagenous domains at both narrow ends of the ellipsoid), whereas the equatorial layers that connect both protomers front-to-front are formed by the six alternating II and II′ sheets of the three chains (Fig. 1b, left). The II and II′ sheets provide a clipping mechanism that glues the three subunits of each protomer together (Fig. 1b). Each II or II′ sheet is composed of four strands of one subdomain and of two strands forming a β-hairpin projected from the corresponding sheet of the adjacent subdomain or subunit (the topologies of the II and II′ sheets are β4β3β8β6′β7′β9 and β4′β3′β8′β6β7β9′, respectively; Fig. 1b). Thus, the β6′β7′ hairpin of sheet II corresponds to C-sub of the same subunit and the β6β7 hairpin of sheet II′ to N-sub of the adjacent subunit.
3.2. Structures of the α1NC1 homohexamer and of an unreported α3NC1 homohexamer
In vitro studies have shown that α1NC1 can assemble into homohexamers (Khoshnoodi, Sigmundsson et al., 2006). We grew crystals of self-assembled recombinant α1NC1 subunits (Supplementary Fig. S1) that diffracted X-rays to 1.8 Å resolution (Table 1). The contained a homohexamer (α1NC1homo) with the same `canonical' architecture as α121NC1 and α5NC1homo (Fig. 2, Supplementary Fig. S3b). We also obtained crystals of an α3NC1 homohexamer that diffracted X-rays to 2.7 Å resolution (Table 1), revealing canonical homohexamers (α3NC1homo; Fig. 2, Supplementary Fig. S3b). Although α3NC1 homohexamers were not observed by (Supplementary Fig. S2a), plasmon resonance studies (Supplementary Fig. S2b) showed self-association of α3NC1 with a similar affinity as α5NC1 (Kd values of 1.71 and 4.92 µM, respectively; Supplementary Fig. S2b) or α1NC1 (Khoshnoodi, Sigmundsson et al., 2006). These self-affinities, the canonical architectures of α1NC1homo, α3NC1homo and α5NC1homo, and the recent discovery of biologically formed α5NC1homo (Revert et al., 2018) indicate that either α1NC1homo and α3NC1homo hexamers exist in vivo and have not been discovered to date or there are biological mechanisms that prevent their formation in vivo.
3.3. Recombinant α1NC1 and α2NC1 chain mixtures reproduce the canonical naturally existing α121NC1 hexamer
We validated the use of self-assembled recombinant NC1 domains to reflect natural structures by generating the same α121NC1 hexamer as previously isolated from BMs (Sundaramoorthy et al., 2002; Than et al., 2002; Vanacore et al., 2009) from a 2:1 mixture of pure α1NC1 and α2NC1 chains (Supplementary Fig. S1). Crystals of recombinant α121NC1 diffracted X-rays to 2.8 Å resolution (Table 1) and had the same architecture and folding (Fig. 2, Supplementary Fig. S3c) as the α121NC1 hexamer isolated from BMs (the r.m.s.d. for the superimposition of all Cα atoms of the hexamer is 0.48 Å; Supplementary Table S3). The recombinant α121NC1 hexamer was composed of two identical protomers formed by two α1NC1 domains and one α2NC1 domain (Supplementary Figs. S4a, S4b and S4c). Each individual chain (designated α1NC1α121 and α2NC1α121) had an identical fold and relations to its natural counterpart (Supplementary Figs. S3c and S4; Supplementary Table S4; Than et al., 2002; Sundaramoorthy et al., 2002; Vanacore et al., 2004). Thus, recombinant NC1 chains seem to be appropriate for the formation and structural characterization of collagen IV NC1-domain assemblies. Therefore, the covalent modifications and/or cross-links found in natural α121NC1 hexamers or potential directive roles of the collagenous domains or other macromolecular components do not seem to be essential for the correct self-assembly of α121NC1.
3.4. Structures of self-assembled α4NC1 and α2NC1 oligomers reveal possibly informative noncanonical homo-oligomers
Given the lack of structural information on α4NC1, we grew crystals of this chain alone that diffracted X-rays to 2.8 Å resolution (Table 1). Surprisingly, they revealed a noncanonical hexameric protomer that generated a dodecamer (α4NC1homo) with an almost spherical shape upon application of the crystal symmetry (Fig. 2, Supplementary Fig. S5). The protomer resembled a canonical trimeric protomer, except that it had six subunits surrounding the molecular symmetry axis (Fig. 2), while the dodecamer resembled a canonical hexamer except for the increased number of subunits, resulting in a wider equatorial circumference (∼100 Å diameter) and a widened central pole-to-pole tunnel (∼40 Å diameter versus ∼17 Å in canonical homohexamers) (Fig. 2, Supplementary Fig. S5).
We also crystallized and determined the structure of the isolated α2NC1 subunit at 2.5 Å resolution (Table 1). The included two subunits. Following application of the crystal symmetry, each subunit formed a protomer with fourfold molecular symmetry. One of these protomers was isolated and the other generated a homo-octamer (α2NC1homo; Fig. 2, Supplementary Fig. S5). α2NC1homo resembled a canonical hexamer, except for an increase in the number of subunits to four per protomer, enlarging the equatorial circumference somewhat (∼70 Å diameter) and widening the central tunnel (∼30 Å).
These two noncanonical protomers could be artifactual owing to the lack of the restrictions imposed by the triple-helical collagenous domain that is present in the complete collagen IV protomer. However, the protomers observed here in α1NC1homo, α3NC1homo, α5NC1homo and α121NC1 were trimeric despite the absence of a collagenous triple helix. Therefore, the two self-assembled noncanonical structures of α2NC1 or α4NC1 subunits may lack some structural elements that are present in the canonical trimeric protomers that could intrinsically restrict the formation of protomers with more than three subunits. The comparison between the structure of the α2NC1 domain in its noncanonical homohexamer and the same domain in α121NC1 (α2NC1α121) revealed that there are conformational differences that might explain the greater number of subunits (Fig. 2, Supplementary Fig. S6). Similar differences were observed when α4NC1 was compared with the NC1 domains in the canonical hexamers (Fig. 3, Supplementary Fig. S6). These differences affected β-sheets II and II′ exclusively. These two sheets ensure the clipping of the protomer into a closed trimeric structure and form the equatorial layer that connects both protomers front-to-front. Here, we define three structural motifs in sheet II (Fig. 3), SM1 (hairpin β3–β4), SM2 (hairpin β6–β7) and SM3 (β9 and the preceding loop), and their homologous counterparts in sheet II′, SM1′ (hairpin β3′–β4′), SM2′ (hairpin β6′–β7′) and SM3′ (β9′ and the preceding loop) (Figs. 3a and 3b).
As can be seen in the canonical structures, SM1 and SM1′ (Fig. 3b) face the central tunnel, creating a closed `barrel'-like arrangement of six alternating β4 and β4′ strands, which allows compact intertwining of three monomers (Than et al., 2002). These hairpins (residues 37–45 in the case of SM1 and residues 144–154 in the case of SM1′) were not visible in the structure of α2NC1homo (Fig. 3), as would be expected if they were disordered in this noncanonical homo-octamer. On the other hand, they were visible in α4NC1homo but with a different conformation to that in α2NC1α121 (Fig. 3). These changes are expected to decrease the strength of the association of the subunits into the protomer.
As described above (Fig. 1), SM2 and SM2′ are inserted into the equatorial layers of the adjacent subunit or subdomain, respectively. In α2NC1homo SM2 underwent a rigid-body rotation of 22.5° (calculated by DYNDOM; Hayward & Berendsen, 1998) relative to its position in α2NC1α121, while in α4NC1homo SM2 adopted a random-coil conformation (Fig. 3, Supplementary Fig. S6). These alterations could weaken the closing belt that would have contributed to restricting the number of subunits to three in a canonical protomer.
The SM3′ motif (Figs. 3a and 3b) flanks SM2 peripherally (relative to the molecular symmetry axis of the protomer). SM3′ was disordered and was not visible in either α2NC1homo or α4NC1homo (Fig. 3), again weakening the insertion of SM2 into the II′ sheet and thus the clipping together of the oligomer. In fact, the loop connecting SM2 to β8 (Lβ7β8; residues 62–77), as well as the homologous loop connecting strands β7′ and β8′ (Lβ7′β8′; residues 185–189) in sheet II′, also adopted alternative arrangements relative to the canonical structures (Fig. 3c) in α2NC1homo and α4NC1homo. Given the closeness of the Lβ7β8 and Lβ7′β8′ loops to the SM2′ and SM2 motifs, which are inserted between strands β8 and β9 of the two homologous sheets II′ and II, these conformational changes should weaken the restricting belt that helps compel the protomer to be trimeric.
In addition to playing a key role in the formation of the protomer, all six SMs, as well as loops Lβ7β8 and Lβ7′β8′, play crucial roles in the front-to-front interactions through the equatorial plane that glue two protomers together to generate the final particle (see the top row in Fig. 3b). Loop Lβ7β8 is the chloride-binding motif (ClA motif) that assists protomer–protomer interaction across the equatorial plane following chloride binding (Cummings et al., 2016; Fig. 4a). The rearrangements and structural alterations in these elements observed in α2NC1homo and α4NC1homo (Fig. 3) modify the equatorial plane of the tetrameric and hexameric protomers, abolishing most of the protomer–protomer interactions that are observed in the canonical hexamer. However, the persistence of some interactions through the equatorial plane can explain the association of the two noncanonical protomers of α2NC1 and α4NC1 into octamers and dodecamers, respectively, which overall follow the general plan of the canonical hexameric assembly. The presence of an isolated tetrameric protomer that is not associated into an octamer in the α2NC1homo crystal is structural evidence of weakened interactions across the equatorial plane.
Overall, a comparison of noncanonical and canonical oligomers (Figs. 3a and 3b) revealed that the SM1, SM1′ and SM2 flexible regions are key determinants in protomer formation, while the loops connecting these motifs, Lβ7β8 (motif ClA), Lβ7′β8′ and SM3′, are needed to establish the protomer–protomer interactions that generate the hexamer. It is interesting that among these structurally more variable regions (Fig. 3c), the loops that connect β8 and β9 (Lβ8β9), β6′ and β7′ (Lβ6′β7′), and β8′ and β9′ (Lβ8′β9′) present length differences of one or two residues among the various types of NC1 chains (see below; Fig. 5b), suggesting that these regions provide structural variability to the oligomers. Indeed, previous work with the natural α121NC1 heterohexamer has reported structural inter-chain variability in the conformations of these loops (Sundaramoorthy et al., 2002; Than et al., 2002). Here, these loops show variations in the structures of our canonical homohexamers (Supplementary Table S4 and Supplementary Fig. S7).
In summary, mobility and conformational changes are concentrated in the equatorial layer formed by β-sheets II and II′, a layer that is crucial for clipping the structure of the protomer and that connects both protomers front-to-front, whereas sheets I and I′ of the polar layer exhibit high structural conservation in canonical and noncanonical protomers (Fig. 3c). In the polar layer only the loop between β1 and β2 exhibited substantial movement (translated into higher values for the r.m.s.d. between odd-numbered and even-numbered NC1 chains; Supplementary Fig. S7), caused by a short flapping movement (∼30°) with no alteration of the actual structure of this loop, leading only to changes in its position. Given their high structural stability, sheets I and I′ may form the initial structural core of the subunit, folding first, while the flexible regions of β-sheets II and II′ could be involved in the termination of the folding process, ensuring proper protomer assembly. At this point, we are unable to identify the reasons for the insertion of just one extra subunit or of three extra subunits in the α2NC1 and α4NC1 protomers, respectively. However, it seems clear that the disorder or loss of the SM1 and SM1′ secondary structure observed in the noncanonical α2NC1homo and α4NC1homo protomers, which leads to loss of the β-barrel-like organization around the central tunnel, could affect SM2 and the ClA motif of the same subunit, resulting in the displacement of these two elements away from the central tunnel. The structural disruption of SM1′ could also affect the adjacent β8′ strand (which loses its secondary structure in the noncanonical protomers) and the insertion of SM2 in the next subunit (Fig. 3). SM2 is a key protomer-clipping element that would become relaxed in the two noncanonical protomers, allowing the insertion of extra subunits. In turn, the resulting abnormal placement of SM2 could prevent interactions with SM3′, which would avoid the generation of a narrow canonical protomer.
3.5. Chloride-mediated protomer–protomer stabilization and potential for sulfilimine cross-linking in canonical but not in noncanonical self-assembled oligomers of recombinant NC1 chains
The α121NC1 structures obtained from natural sources showed the presence of ions and covalent sulfilimine bonds associated with stabilization of the hexamers (Vanacore et al., 2004; Than et al., 2002; Robertson et al., 2014). Chloride ions were also present in our α1NC1homo, α3NC1homo and α5NC1homo structures at two different positions in the protomer–protomer interface. Six chloride ions (ClA), one per monomer, were modelled at the position previously observed in the natural α121NC1 structure (Figs. 4a and 4c). ClA interacted with the amino groups of the ClA motif. In the structure of the α121NC1 hexamer obtained from the BM of bovine lens capsule (Cummings et al., 2016), this motif was one of the flexible regions shown to be crucial for triggering the assembly of the two α121NC1 protomers into a hexamer.
Six additional chloride ions (ClB), again one per monomer, were also observed in the α1NC1homo, α3NC1homo, α5NC1homo and α121NC1 hexamers (Figs. 4b and 4c). In the α121NC1 structure from the BM of bovine lens the nonprotein electron densities at these positions were interpreted as potassium ions, while in the equivalent α121NC1 structures from the BMs of bovine and human placenta they were modelled as bromide ions and acetate, respectively (Vanacore et al., 2004). We modelled the densities at these positions as Cl− ions since the nature and geometry of the contacts mediated by them (with a hydroxyl group and with amino groups) and the electron densities fitted best with the presence of this anion, which is abundant in the crystallization drop. ClB was located at the interprotomer interface and was involved in an interaction network between conserved residues that are present in Lβ7′β8′ (hereafter referred to as the ClB motif) and SM2 from three different monomers of opposing protomers (Fig. 4c). In contrast, in the noncanonical homoligomers α2NC1homo and α4NC1homo all chloride ions were missing mainly owing to the rearrangement of the ClA and ClB motifs as well as of SM2 and β8′ (Fig. 3), which prevented the coordination of these chlorides. Since all recombinant proteins used were prepared in NaCl buffers, and since hexamer assembly in the BM from bovine lens has recently been demonstrated to be chloride-dependent (Cummings et al., 2016), the conformational changes observed in α2NC1homo and α4NC1homo that prevent chloride binding may hinder the final protomer–protomer assembly.
As expected, in the final stage of network assembly the natural NC1 hexamers of collagen IV cross-linked by covalent sulfilimine bonds between Met93 from SM3 of one monomer and Lys211 from SM3′ of another monomer facing across the equatorial plane were formed (α1NC1 residue numbering; Than et al., 2002; Vanacore et al., 2009). The formation of the sulfilimine bonds is catalyzed by peroxidasin, an enzyme found in BMs (Bhave et al., 2012). No sulfilimine bonds were seen in the structures presented in this work, possibly because all of the proteins were produced in a recombinant system in which either the conditions were inadequate for the of peroxidasin or this enzyme was absent. However, in the α1NC1homo, α3NC1homo, α5NC1homo and α121NC1 structures Lys211 interacts via a hydrogen bond with the SM3 containing Met93 of an opposing monomer (Figs. 4c and 4d). An alternative conformation of the Lys211 side chain could allow sulfilimine-bond formation with Met93 (Fig. 4d). This was not the case for α2NC1homo and α4NC1homo (Fig. 3), where the inappropriate location of these residues for sulfilimine-bond formation reflects the high flexibility of β9′ in SM3′, strongly suggesting that the structural elements involved in protomer–protomer assembly acquire their final conformation at the end the NC1 folding process. Thus, NC1 folding is correlated with the proposed monomer–protomer–hexamer sequential assembly model for collagen IV network formation (Boutaud et al., 2000; Kalluri, 2003; Khoshnoodi, Cartailler et al., 2006; Cummings et al., 2016).
3.6. Insight into Alport's and Goodpasture's syndromes
The α345 collagen IV network is a major component of the glomerular BM and its alteration underlies the pathogenesis of Alport's and Goodpasture's syndromes (Hudson et al., 2003; Pedchenko et al., 2010). In Alport's syndrome, mutations in these chains, including α5NC1, lead to a defective network assembly, causing the multi-laminar splitting of the glomerular BM. In Goodpasture's disease autoantibodies directed to the α3NC1 and α5NC1 domains are generated. Our structures of α3NC1homo, α4NC1homo and α5NC1homo show that two structurally equivalent regions of α3NC1, EA and EB, as well as the homologous regions of α5NC1, encompass the Goodpasture epitopes (Figs. 5a and 5b; Cui et al., 2016). Furthermore, the Alport missense mutations in the NC1 domains (Crockett et al., 2010; see Supplementary Table S5) predominantly map to the more structurally stable regions of the NC1 domain (∼90% of the mutations), mainly β-sheets I/I′, with few mutations mapping to those regions that showed high flexibility in α2NC1homo and α4NC1homo (∼10% of the mutations; Figs. 5b and 5c). Since the structural core of the NC1 domains should acquire its conformation in the initial steps of NC1 folding (see Fig. 6 and Discussion) to nucleate protomer assembly, these mutations could strongly compromise collagen IV network formation. Supporting this idea, most of the Alport mutations (68%) are found in residues with high structural relevance, such as cysteine, glycine or proline (Supplementary Table S5).
To our knowledge, the structures presented here represent the first visualization of the Goodpasture autoantigen. The regions that encompass the EA and EB epitopes lie in β-sheets I/I′, a more structurally constant and fixed part of the NC1 structural core (Figs. 1b and 5c). The comparison of EA and EB of α3NC1 and α5NC1 also showed that they are identical in structure and also to the corresponding part of α1NC1, although no antigenicity has been determined in the latter chain in patients with Goodpasture's disease (Supplementary Figs. S8 and S9). Solvent-accessibility analysis of residues at the EA and EB epitopes in the α3NC1 and α5NC1 structures, in comparison to the corresponding residues in α1NC1, showed similar exposed surface areas in either monomers or hexamers. Very small differences in solvent accessibility were observed for Leu27 and Tyr28 in EA and for Pro131 and Trp134 in EB (Supplementary Fig. S10). These residues are highly conserved among NC1 chains (Supplementary Figs. S8a and S9a). They form a hydrophobic patch on the exposed surface of the protomer (Supplementary Fig. S11a). Further comparisons with the corresponding regions of α2NC1homo and α4NC1homo revealed conservation of the fold in this region even in these noncanonical oligomers (Supplementary Figs. S11b and S11c). Therefore, the differential antigenicity observed between chains must be owing to sequence determinants, as has recently been proposed (Cui et al., 2016).
4. Discussion
The very recent report of the existence of the α5NC1 homohexamer in vivo (Revert et al., 2018) gives credence to the view that the canonical α1NC1 and α3NC1 homohexamers presented here may also exist in vivo, particularly given their structural stability, their proper binding of chloride ions for hexamer stabilization and their correct disposition of the residues that mediate the sulfilimine cross-linking bonds that stabilize the hexamer. Their spontaneous self-assembly in solutions of individual components, and even in mixtures of different components, attests to their thermodynamic stability and the absence of important kinetic barriers against their formation.
The finding that only a few combinations of collagen IV chain types are present in BMs, among the >1000 possible combinations of the six collagen IV chains, is difficult to explain based only on spontaneous thermodynamically driven assembly. Although the contribution of the collagenous parts to chain selection has not yet been clarified, our findings with isolated NC1 domains without the collagenous parts suggest the existence of regulatory mechanisms that could allow or guide the selection of a given chain combination. Spontaneous assembly has been proposed for α121NC1, where the α2NC1 chain shows a higher affinity to interact with α1NC1 than with itself (Khoshnoodi, Sigmundsson et al., 2006). In fact, we show here that 2:1 mixtures of α1NC1 and α2NC1 chains spontaneously generate the natural α121NC1 heterohexamer. Conversely, our experimental results have shown that equimolecular mixtures of α3NC1, α4NC1 and α5NC1 only produce α5NC1homo crystals. Therefore, further studies will be required to clarify the mechanism underlying the alternative formation of α5 and α345 networks in vivo.
Our findings of noncanonical assemblies for α2NC1 and α4NC1 homo-oligomers and the structural changes observed in the chains forming these noncanonical protomers in comparison to α2NC1α121 and to the chains in α1NC1homo, α3NC1homo and α5NC1homo could imply that is closely related to protomer assembly. It is unlikely that α2NC1homo and α4NC1homo could represent stable physiological assemblies, but the structures of the individual chains in these noncanonical oligomers might provide a frozen glimpse of transient conformational states in the process of NC1 folding and hexamer building. This is supported by the identical folding of many parts of these noncanonical oligomers and canonical structures, particularly in polar layers (β-sheets I and I′), and the restriction of flexible regions in both noncanonical structures to essentially the same sequence stretches belonging to the equatorial layer (mainly β-sheets II and II′).
As already mentioned, β-sheets I/I′ could act as a stable structural core nucleating the NC1 domains, which could prevent their final folding until integrated into a trimer, since the SM2 would have to be inserted into β-sheet II′ of an adjacent subunit. A plausible scenario for NC1 equatorial layer folding that is compatible with our observations with noncanonical oligomers would involve first the formation of the central β-barrel-like structure by SM1/SM1′, with the immediate insertion of SM2′ into the forming β-sheet II of the same subunit. This would make SM2 ready for insertion into the SM3′ of an adjacent folding protomer. Following this insertion, β-sheet II′ would acquire its final structure. The process of a central β-barrel-like organization reduces the size of the central tunnel, restricting the number of monomers in the protomer to three, which is further stabilized by the presence of the triple-helical collagenous domain. Subsequently, flexible loops (ClA and ClB motifs) at the equatorial plane of the protomer could acquire a competent conformation for chloride binding, leading to the final assembly of the hexamer involving the interaction of two protomers through their flat surfaces in a chloride-rich medium, i.e. the extracellular fluid, with further hexamer stabilization by sulfilimine cross-linkings (Fig. 6). In the previous description, the potential participation of an uncharacterized cellular machinery in assisting this process of chain sorting, folding and assembly has not been taken into account. It has recently been reported that the production of an α5 network in cancer cell lines is dependent on the expression and activity of the Goodpasture antigen-binding protein (GPBP). This strongly suggests that this machinery exists and that GPBP forms an important part of it (Revert et al., 2018).
Our present work helps to clarify why missense mutations affecting NC1 domains in Alport's syndrome are more frequently found in the structural core of the NC1 domain. Given the key role of β-sheets I/I′ in subunit folding that is suggested by our findings, mutations in this region are likely to have increased visibility (`eloquent mutations') since they will have a higher impact on the final structure in comparison to mutations in structurally less relevant parts. Thus, the genetic variants of Alport's syndrome owing to missense mutations in this domain can be considered to cause a folding disease, as is supported by the occurrence of the unfolded protein response and of endoplasmic reticulum stress in Alport's syndrome (Gould et al., 2007; Pieri et al., 2014). With this in mind, chemical chaperones could be used to help to restore the proper conformation and folding in cases with certain missense mutants (Murray et al., 2014; Wang et al., 2017).
In Goodpasture's autoimmune disease, the autoepitopes are immunoreactive only when the native α345NC1 hexamers are dissociated into monomers or dimer subunits (Hudson et al., 2003; Pedchenko et al., 2010). The underlying mechanism has been attributed to conformational changes forming epitopes in dissociated α3 and α5 subunits that bind the pathogenic autoantibodies (Calvete et al., 2006; Pedchenko et al., 2010). Our crystal structures of the α3NC1 and α5NC1 homohexamers provide the first view of the arrangement of the EA and EB regions, which are known critical components of two epitopes that reside in the dissociated α3 and α5 subunits of native α345NC1 hexamers. In the homohexamers these regions are exposed on the surface and are presumably unreactive to antibody binding, as they are in the native α345NC1 hexamers. This knowledge of EA and EB structures will be critical in elucidating the structural mechanism that enables presentation only in dissociated subunits. Potentially, the mechanism will shed light on the aetiology of Goodpasture's disease.
It is conceivable that the extreme structural stability of the regions visualized here as pertaining to the EA and EB epitopes could be related to a high antigenicity because of the presentation of a single conformer, while other areas of the protein with higher flexibility could present many different conformations, populated with reduced frequencies, to the antibody-making machinery. The preference of Goodpasture antibodies to recognize the α3NC1 chain could be explained by its sequence composition, although the recent description of Goodpasture's disease with autoantibodies reactive exclusively to α5NC1 (Cui et al., 2016) opens the door to additional factors being implicated in pathogenesis. Anti-α3NC1 autoantibodies are found in most Goodpasture's disease patients. The presence of α5NC1-specific autoantibodies could be considered to be to some extent atypical or exceptional in the context of Goodpasture's disease, and thus α3NC1 must exhibit some singularity that makes it more susceptible to undergoing autoimmune attack in comparison with other chains.
Supporting information
PDB references: collagen type IV α1α2α1NC1, 5nax; α1NC1, 5nay; α2NC1, 5nb2; α3NC1, 5nb0; α4NC1, 5nb1; α5NC1, 5naz
Supplementary Tables and Figures. DOI: https://doi.org/10.1107/S2052252518012459/tj5018sup1.pdf
Footnotes
‡Both authors contributed equally.
Acknowledgements
We thank the IBV–CSIC Crystallogenesis Facility for protein crystallization screening. The experimental results reported in this article are derived from measurements made at the ESRF (Grenoble, France), Diamond Light Source (Didcot, England) and ALBA (Cerdanyola del Valles, Spain) synchrotrons. We thank the staff of the beamlines used for assistance. X-ray diffraction data collection was supported by Diamond Light Source block allocation group (BAG) Proposals MX10121 and MX14739 and Spanish Synchrotron Radiation Facility ALBA Proposals 2015071314 and 2017072262.
Funding information
The following funding is acknowledged: Spanish Government (Ministerio de Economia y Competitividad; grant Nos. BIO2013-42619 and BIO2016-78571-P to Alberto Marina; grant Nos. SAF 2009-10703, IPT-2011-1527-010000 and RTC-2014-2415-1 to Juan Saus; grant No. BFU2016-78606-P to Patricia Casino; grant No. BFU2014-58229-P to Vicente Rubio; contract No. RYC-2014-16490 to Patricia Casino); Valencian Government Prometeo program (grant No. II/2014/029 to Alberto Marina, Vicente Rubio and Juan Saus); National Institutes of Health (grant NIH-DK18381 to Billy G. Hudson); European Community's Seventh Framework Programme FP7/2007-2013 (BioStruct-X; grant No. 283570); and a visiting researcher grant from the program `Atracció de Talent' of the University of Valencia to Billy G. Hudson.
References
Anazco, C., Lopez-Jimenez, A. J., Rafi, M., Vega-Montoto, L., Zhang, M.-Z., Hudson, B. G. & Vanacore, R. M. (2016). J. Biol. Chem. 291, 25999. Web of Science PubMed Google Scholar
Bhave, G., Cummings, C. F., Vanacore, R. M., Kumagai-Cresse, C., Ero-Tolliver, I. A., Rafi, M., Kang, J.-S., Pedchenko, V., Fessler, L. I., Fessler, J. H. & Hudson, B. G. (2012). Nature Chem. Biol. 8, 784–790. Web of Science CrossRef Google Scholar
Borza, D. B., Bondar, O., Ninomiya, Y., Sado, Y., Naito, I., Todd, P. & Hudson, B. G. (2001). J. Biol. Chem. 276, 28532–28540. Web of Science CrossRef PubMed Google Scholar
Boutaud, A., Borza, D.-B., Bondar, O., Gunwar, S., Netzer, K.-O., Singh, N., Ninomiya, Y., Sado, Y., Noelken, M. E. & Hudson, B. G. (2000). J. Biol. Chem. 275, 30716–30724. Web of Science CrossRef PubMed Google Scholar
Bradford, M. M. (1976). Anal. Biochem. 72, 248–254. Web of Science CrossRef CAS PubMed Google Scholar
Brown, K. L., Cummings, C. F., Vanacore, R. M. & Hudson, B. G. (2017). Protein Sci. 26, 2151–2161. Web of Science CrossRef PubMed Google Scholar
Calvete, J. J., Revert, F., Blanco, M., Cervera, J., Tárrega, C., Sanz, L., Revert-Ros, F., Granero, F., Pérez-Payá, E., Hudson, B. G. & Saus, J. (2006). Proteomics, 6, S237–S244. Web of Science CrossRef PubMed Google Scholar
Campbell, K. P. & Stull, J. T. (2003). J. Biol. Chem. 278, 12599–12600. Web of Science CrossRef PubMed Google Scholar
Court, F. A., Wrabetz, L. & Feltri, M. L. (2006). Curr. Opin. Neurobiol. 16, 501–507. Web of Science CrossRef PubMed Google Scholar
Crockett, D. K., Pont-Kingdon, G., Gedge, F., Sumner, K., Seamons, R. & Lyon, E. (2010). Hum. Mutat. 31, E1652–E1657. Web of Science CrossRef PubMed Google Scholar
Cui, Z., Zhao, M.-H., Jia, X.-Y., Wang, M., Hu, S.-Y., Wang, S.-X., Yu, F., Brown, K. L., Hudson, B. G. & Pedchenko, V. (2016). J. Autoimmun. 70, 1–11. Web of Science CrossRef PubMed Google Scholar
Cummings, C. F. et al. (2016). J. Cell Biol. 213, 479–494. Web of Science CrossRef PubMed Google Scholar
Daley, W. P. & Yamada, K. M. (2013). Curr. Opin. Genet. Dev. 23, 408–414. Web of Science CrossRef PubMed Google Scholar
Dölz, R., Engel, J. & Kühn, K. (1988). Eur. J. Biochem. 178, 357–366. PubMed Web of Science 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
Fidler, A. L., Darris, C. E., Chetyrkin, S. V., Pedchenko, V. K., Boudko, S. P., Brown, K. L., Jerome, W. G., Hudson, J. K., Rokas, A. & Hudson, B. G. (2017). Elife, 6, e24176. Web of Science CrossRef PubMed Google Scholar
Fidler, A. L., Boudko, S. P., Rokas, A. & Hudson, B. G. (2018). J. Cell Sci. 131, jcs203950. Web of Science CrossRef PubMed Google Scholar
Gould, D. B., Marchant, J. K., Savinova, O. V., Smith, R. S. & John, S. W. (2007). Hum. Mol. Genet. 16, 798–807. Web of Science CrossRef PubMed Google Scholar
Gould, D. B., Phalan, F. C., Breedveld, G. J., van Mil, S. E., Smith, R. S., Schimenti, J. C., Aguglia, U., van der Knaap, M. S., Heutink, P. & John, S. W. (2005). Science, 308, 1167–1171. Web of Science CrossRef PubMed Google Scholar
Gozalbo-Rovira, R., Rodríguez-Díaz, J., Saus, J. & Cervera, J. (2013). Kidney Int. 83, 438–445. Web of Science PubMed Google Scholar
Gunwar, S., Ballester, F., Noelken, M. E., Sado, Y., Ninomiya, Y. & Hudson, B. G. (1998). J. Biol. Chem. 273, 8767–8775. Web of Science CrossRef PubMed Google Scholar
Hagios, C., Lochter, A. & Bissell, M. J. (1998). Philos. Trans. R. Soc. Lond. B Biol. Sci. 353, 857–870. Web of Science CrossRef PubMed Google Scholar
Hayward, S. & Berendsen, H. J. (1998). Proteins, 30, 144–154. CrossRef CAS PubMed Google Scholar
Hudson, B. G., Tryggvason, K., Sundaramoorthy, M. & Neilson, E. G. (2003). N. Engl. J. Med. 348, 2543–2556. Web of Science CrossRef PubMed Google Scholar
Hynes, R. O. (2009). Science, 326, 1216–1219. Web of Science CrossRef PubMed Google Scholar
Kabsch, W. (2010). Acta Cryst. D66, 125–132. Web of Science CrossRef CAS IUCr Journals Google Scholar
Kalluri, R. (2003). Nature Rev. Cancer, 3, 422–433. Web of Science CrossRef Google Scholar
Khoshnoodi, J., Cartailler, J.-P., Alvares, K., Veis, A. & Hudson, B. G. (2006). J. Biol. Chem. 281, 38117–38121. Web of Science CrossRef PubMed Google Scholar
Khoshnoodi, J., Sigmundsson, K., Cartailler, J.-P., Bondar, O., Sundaramoorthy, M. & Hudson, B. G. (2006). J. Biol. Chem. 281, 6058–6069. Web of Science CrossRef PubMed Google Scholar
Konagurthu, A. S., Whisstock, J. C., Stuckey, P. J. & Lesk, A. M. (2006). Proteins, 64, 559–574. Web of Science CrossRef PubMed CAS Google Scholar
Kuo, D. S., Labelle-Dumais, C. & Gould, D. B. (2012). Hum. Mol. Genet. 21, R97–110. Web of Science CrossRef PubMed Google Scholar
Laemmli, U. K. (1970). Nature (London), 227, 680–685. CrossRef CAS PubMed Web of Science 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
Murray, L. S., Lu, Y., Taggart, A., Van Regemorter, N., Vilain, C., Abramowicz, M., Kadler, K. E. & Van Agtmael, T. (2014). Hum. Mol. Genet. 23, 283–292. Web of Science CrossRef PubMed 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
Netzer, K.-O., Leinonen, A., Boutaud, A., Borza, D.-B., Todd, P., Gunwar, S., Langeveld, J. P. M. & Hudson, B. G. (1999). J. Biol. Chem. 274, 11267–11274. Web of Science CrossRef PubMed Google Scholar
Papadopoulos, J. S. & Agarwala, R. (2007). Bioinformatics, 23, 1073–1079. Web of Science CrossRef PubMed CAS Google Scholar
Pastor-Pareja, J. C. & Xu, T. (2011). Dev. Cell, 21, 245–256. Web of Science PubMed Google Scholar
Pedchenko, V., Bondar, O., Fogo, A. B., Vanacore, R., Voziyan, P., Kitching, A. R., Wieslander, J., Kashtan, C., Borza, D.-B., Neilson, E. G., Wilson, C. B. & Hudson, B. G. (2010). N. Engl. J. Med. 363, 343–354. Web of Science CrossRef PubMed Google Scholar
Pieri, M. et al. (2014). J. Am. Soc. Nephrol. 25, 260–275. Web of Science CrossRef PubMed Google Scholar
Raya, A., Revert, F., Navarro, S. & Saus, J. (1999). J. Biol. Chem. 274, 12642–12649. Web of Science CrossRef PubMed Google Scholar
Revert, F. et al. (2018). Oncotarget, 9, 11020–11045. CrossRef PubMed Google Scholar
Revert, F., Ventura, I., Martínez-Martínez, P., Granero-Moltó, F., Revert-Ros, F., Macías, J. & Saus, J. (2008). J. Biol. Chem. 283, 30246–30255. Web of Science CrossRef PubMed Google Scholar
Rhodes, J. M. & Simons, M. (2007). J. Cell. Mol. Med. 11, 176–205. Web of Science CrossRef PubMed Google Scholar
Robertson, W. E., Rose, K. L., Hudson, B. G. & Vanacore, R. M. (2014). J. Biol. Chem. 289, 25601–25610. Web of Science CrossRef PubMed Google Scholar
Sado, Y., Boutaud, A., Kagawa, M., Naito, I., Ninomiya, Y. & Hudson, B. G. (1998). Kidney Int. 53, 664–671. Web of Science CrossRef PubMed Google Scholar
Sanes, J. R. (2003). J. Biol. Chem. 278, 12601–12604. Web of Science CrossRef PubMed Google Scholar
Sillat, T., Saat, R., Pöllänen, R., Hukkanen, M., Takagi, M. & Konttinen, Y. T. (2012). J. Cell. Mol. Med. 16, 1485–1495. Web of Science CrossRef PubMed Google Scholar
Sundaramoorthy, M., Meiyappan, M., Todd, P. & Hudson, B. G. (2002). J. Biol. Chem. 277, 31142–31153. Web of Science CrossRef PubMed Google Scholar
Than, M. E., Henrich, S., Huber, R., Ries, A., Mann, K., Kühn, K., Timpl, R., Bourenkov, G. P., Bartunik, H. D. & Bode, W. (2002). Proc. Natl Acad. Sci. USA, 99, 6607–6612. Web of Science CrossRef PubMed CAS Google Scholar
Timpl, R., Wiedemann, H., van Delden, V., Furthmayr, H. & Kühn, K. (1981). Eur. J. Biochem. 120, 203–211. CrossRef PubMed Web of Science Google Scholar
Vanacore, R., Ham, A. J., Voehler, M., Sanders, C. R., Conrads, T. P., Veenstra, T. D., Sharpless, K. B., Dawson, P. E. & Hudson, B. G. (2009). Science, 325, 1230–1234. Web of Science CrossRef PubMed Google Scholar
Vanacore, R. M., Shanmugasundararaj, S., Friedman, D. B., Bondar, O., Hudson, B. G. & Sundaramoorthy, M. (2004). J. Biol. Chem. 279, 44723–44730. Web of Science CrossRef PubMed Google Scholar
Wang, D., Mohammad, M., Wang, Y., Tan, R., Murray, L. S., Ricardo, S., Dagher, H., van Agtmael, T. & Savige, J. (2017). Kidney Int. Rep. 2, 739–748. Web of Science CrossRef PubMed Google Scholar
Wang, X., Harris, R. E., Bayston, L. J. & Ashe, H. L. (2008). Nature (London), 455, 72–77. Web of Science CrossRef PubMed Google Scholar
Winn, M. D. et al. (2011). Acta Cryst. D67, 235–242. Web of Science CrossRef CAS IUCr Journals 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.