Publications

Domains known as von Willebrand factor type D (VWD) are found in extracellular and cell-surface proteins including von Willebrand factor, mucins, and various signaling molecules and receptors. Many VWD domains have a glycine-aspartate-proline-histidine (GDPH) amino-acid sequence motif, which is hydrolytically cleaved post-translationally between the aspartate (Asp) and proline (Pro). The Fc IgG binding protein (FCGBP), found in intestinal mucus secretions and other extracellular environments, contains 13 VWD domains, 11 of which have a GDPH cleavage site. In this study, we investigated the structural and biophysical consequences of Asp-Pro peptide cleavage in a representative FCGBP VWD domain. We found that endogenous Asp-Pro cleavage increases the resistance of the domain to exogenous proteolytic degradation. Tertiary structural interactions made by the newly generated chain termini, as revealed by a crystal structure of an FCGBP segment containing the VWD domain, may explain this observation. Notably, the Gly-Asp peptide bond, upstream of the cleavage site, assumed the cis configuration in the structure. In addition to these local features of the cleavage site, a global organizational difference was seen when comparing the FCGBP segment structure with the numerous other structures containing the same set of domains. Together, these data illuminate the outcome of GDPH cleavage and demonstrate the plasticity of proteins with VWD domains, which may contribute to their evolution for function in a dynamic extracellular environment.

Aberrant expression of transmem-brane mucins promotes tumor progression and interferes with im-munological and medicinal elimina-tion of cancer cells. In a recent article, Pedram et al. directed an at-tenuated bacterial mucin-specific protease to HER2-positive tumor cells and observed decreased tumor growth rates and extended survival of mice bearing HER2-positive tumors.

The von Willebrand factor (VWF) glycoprotein is stored in tubular form in Weibel-Palade bodies (WPBs) prior to secretion from endothelial cells into the bloodstream. The organization of VWF in the tubules promotes formation of covalently linked VWF polymers and enables orderly secretion without polymer tangling. Recent studies have described the high-resolution structure of helical tubular cores formed in vitro by the D1D2 and D´D3 amino-terminal protein segments of VWF. Here we show that formation of tubules with the helical geometry observed for VWF in intracellular WPBs requires also the VWA1 (A1) domain. We reconstituted VWF tubules from segments containing the A1 domain and discovered it to be inserted between helical turns of the tubule, altering helical parameters and explaining the increased robustness of tubule formation when A1 is present. The conclusion from this observation is that the A1 domain has a direct role in VWF assembly, along with its known activity in hemostasis post-secretion.

In the parasite Trypanosoma brucei, the causative agent of human African sleeping sickness, all mRNAs are trans-spliced to generate a common 59 exon derived from the spliced leader (SL) RNA. Perturbations of protein translocation across the endoplasmic reticulum (ER) induce the spliced leader RNA silencing (SLS) pathway. SLS activation is mediated by a serine-threonine kinase, PK3, which translocates from the cytosolic face of the ER to the nucleus, where it phosphorylates the TATA-binding protein TRF4, leading to the shutoff of SL RNA transcription, followed by induction of programmed cell death. Here, we demonstrate that SLS is also induced by depletion of the essential ER-resident chaperones BiP and calreticulin, ER oxidoreductin 1 (ERO1), and the Golgi complex-localized quiescin sulfhydryl oxidase (QSOX). Most strikingly, silencing of Rhomboid-like 1 (TIMRHOM1), involved in mitochondrial protein import, also induces SLS. The PK3 kinase, which integrates SLS signals, is modified by phosphorylation on multiple sites. To determine which of the phosphorylation events activate PK3, several individual mutations or their combination were generated. These mutations failed to completely eliminate the phosphorylation or translocation of the kinase to the nucleus. The structures of PK3 kinase and its ATP binding domain were therefore modeled. A conserved phenylalanine at position 771 was proposed to interact with ATP, and the PK3F771L mutation completely eliminated phosphorylation under SLS, suggesting that the activation involves most if not all of the phosphorylation sites. The study suggests that the SLS occurs broadly in response to failures in protein sorting, folding, or modification across multiple compartments.

The complex environment of biological cells and tissues has motivated development of three-dimensional (3D) imaging in both light and electron microscopies. To this end, one of the primary tools in fluorescence microscopy is that of computational deconvolution. Wide-field fluorescence images are often corrupted by haze due to out-of-focus light, i.e., to cross-talk between different object planes as represented in the 3D image. Using prior understanding of the image formation mechanism, it is possible to suppress the cross-talk and reassign the unfocused light to its proper source post facto. Electron tomography based on tilted projections also exhibits a cross-talk between distant planes due to the discrete angular sampling and limited tilt range. By use of a suitably synthesized 3D point spread function, we show here that deconvolution leads to similar improvements in volume data reconstructed from cryoscanning transmission electron tomography (CSTET), namely a dramatic in-plane noise reduction and improved representation of features in the axial dimension. Contrast enhancement is demonstrated first with colloidal gold particles and then in representative cryotomograms of intact cells. Deconvolution of CSTET data collected from the periphery of an intact nucleus revealed partially condensed, extended structures in interphase chromatin.

The mucin 2 glycoprotein assembles into a complex hydrogel that protects intestinal epithelia and houses the gut microbiome. A major step in mucin 2 assembly is further multimerization of preformed mucin dimers, thought to produce a honeycomb-like arrangement upon hydrogel expansion. Important open questions are how multiple mucin 2 dimers become covalently linked to one another and how mucin 2 multimerization compares with analogous processes in related polymers such as respiratory tract mucins and the hemostasis protein von Willebrand factor. Here we report the x-ray crystal structure of the mucin 2 multimerization module, found to form a dimer linked by two intersubunit disulfide bonds. The dimer structure calls into question the current model for intestinal mucin assembly, which proposes disulfide-mediated trimerization of the same module. Key residues making interactions across the dimer interface are highly conserved in intestinal mucin orthologs, supporting the physiological relevance of the observed quaternary structure. With knowledge of the interface residues, it can be demonstrated that many of these amino acids are also present in other mucins and in von Willebrand factor, further indicating that the stable dimer arrangement reported herein is likely to be shared across this functionally broad protein family. The mucin 2 module structure thus reveals the manner by which both mucins and von Willebrand factor polymerize, drawing deep structural parallels between macromolecular assemblies critical to mucosal epithelia and the vasculature.

Many mutations that cause familial hypercholesterolemia localize to ligand-binding domain 5 (LA5) of the low-density lipoprotein receptor, motivating investigation of the folding and misfolding of this small, disulfide-rich, calcium-binding domain. LA5 folding is known to involve non-native disulfide isomers, yet these folding intermediates have not been structurally characterized. To provide insight into these intermediates, we used nuclear magnetic resonance (NMR) to follow LA5 folding in real time. We demonstrate that misfolded or partially folded disulfide intermediates are indistinguishable from the unfolded state when focusing on the backbone NMR signals, which provide information on the formation of only the final, native state. However,
¹³C labeling of cysteine side chains differentiated transient intermediates from the unfolded and native states and reported on disulfide bond formation in real time. The cysteine pairings in a dominant intermediate were identified using
¹³C-edited three-dimensional NMR, and coarse-grained molecular dynamics simulations were used to investigate the preference of this disulfide set over other non-native arrangements. The transient population of LA5 species with particular non-native cysteine connectitivies during folding supports the conclusion that cysteine pairing is not random and that there is a bias toward certain structural ensembles during the folding process, even prior to the binding of calcium.

Laminin, an ∼800-kDa heterotrimeric protein, is a major functional component of the extracellular matrix, contributing to tissue development and maintenance. The unique architecture of laminin is not currently amenable to determination at high resolution, as its flexible and narrowsegments complicate both crystallization and single-particle reconstruction by electronmicroscopy. Therefore, we used cross-linking and MS, evaluated using computational methods, to address key questions regarding laminin quaternary structure. This approach was particularly well suited to the ∼750-Å coiled coil that mediates trimer assembly, and our results support revision of the subunit order typically presented in laminin schematics. Furthermore, information on the subunit register in the coiled coil and cross-links to downstream domains provide insights into the self-assembly required for interaction with other extracellular matrix and cell surface proteins.

The ability to query enzyme molecules individually is transforming our view of catalytic mechanisms. Quiescin sulfhydryl oxidase (QSOX) is a multidomain catalyst of disulfide-bond formation that relays electrons from substrate cysteines through two redox-active sites to molecular oxygen. The chemical steps in electron transfer have been delineated, but the conformational changes accompanying these steps are poorly characterized. Here we use single-molecule Förster resonance energy transfer (smFRET) to probe QSOX conformation in resting and cycling enzyme populations. We report the discovery of unanticipated roles for conformational changes in QSOX beyond mediating electron transfer between redox-active sites. In particular, a state of the enzyme not previously postulated or experimentally detected is shown to gate, via a conformational transition, the entrance into a sub-cycle within an expanded QSOX kinetic scheme. By tightly constraining mechanistic models, smFRET data can reveal the coupling between conformational and chemical transitions in complex enzymatic cycles.

The widespread thioredoxin superfamily enzymes typically share the following features: a characteristic α-β fold, the presence of a Cys-X-X-Cys (or Cys-X-X-Ser) redox-active motif, and a proline in the cis configuration abutting the redox-active site in the tertiary structure. The Cys-X-X-Cys motif is at the solvent-exposed amino terminus of an a-helix, allowing the first cysteine to engage in nucleophilic attack on substrates, or substrates to attack the Cys-X-X-Cys disulfide, depending on whether the enzyme functions to reduce, isomerize, or oxidize its targets. We report here the X-ray crystal structure of an enzyme that breaks many of our assumptions regarding the sequence-structure relationship of thioredoxin superfamily proteins. The yeast Protein Disulfide Isomerase family member Eps1p has Cys-X-X-Cys motifs and proline residues at the appropriate primary structural positions in its first two predicted thioredoxin-fold domains. However, crystal structures show that the Cys-X-X-Cys of the second domain is buried and that the adjacent proline is in the trans, rather than the cis isomer. In these configurations, neither the "active-site" disulfide nor the backbone carbonyl preceding the proline is available to interact with substrate. The Eps1p structures thus expand the documented diversity of the PDI oxidoreductase family and demonstrate that conserved sequence motifs in common folds do not guarantee structural or functional conservation.

Quiescin sulfhydryl oxidase 1 (QSOX1) is a catalyst of disulfide bond formation that undergoes regulated secretion from fibroblasts and is over-produced in adenocarcinomas and other cancers. We have recently shown that QSOX1 is required for incorporation of particular laminin isoforms into the extracellular matrix (ECM) of cultured fibroblasts and, as a consequence, for tumor cell adhesion to and penetration of the ECM. The known role of laminins in integrin-mediated cell survival and motility suggests that controlling QSOX1 activity may provide a novel means of combating metastatic disease. With this motivation, we developed a monoclonal antibody that inhibits the activity of human QSOX1. Here, we present the biochemical and structural characterization of this antibody and demonstrate that it is a tight-binding inhibitor that blocks one of the redox-active sites in the enzyme, but not the site at which de novo disulfides are generated catalytically. Sulfhydryl oxidase activity is thus prevented without direct binding of the sulfhydryl oxidase domain, confirming the model for the interdomain QSOX1 electron transfer mechanism originally surmised based on mutagenesis and protein dissection. In addition, we developed a single-chain variant of the antibody and show that it is a potent QSOX1 inhibitor. The QSOX1 inhibitory antibody will be a valuable tool in studying the role of ECM composition and architecture in cell migration, and the recombinant version may be further developed for potential therapeutic applications based on manipulation of the tumor microenvironment.

The mimivirus genome contains many genes that lack homologs in the sequence database and are thus known as ORFans. In addition, mimivirus genes that encode proteins belonging to known fold families are in some cases fused to domain-sized segments that cannot be classified. One such ORFan region is present in the mimivirus enzyme R596, a member of the Erv family of sulfhydryl oxidases. We determined the structure of a variant of full-length R596 and observed that the carboxy-terminal region of R596 assumes a folded, compact domain, demonstrating that these ORFan segments can be stable structural units. Moreover, the R596 ORFan domain fold is novel, hinting at the potential wealth of protein structural innovation yet to be discovered in large double-stranded DNA viruses. In the context of the R596 dimer, the ORFan domain contributes to formation of a broad cleft enriched with exposed aromatic groups and basic side chains, which may function in binding target proteins or localization of the enzyme within the virus factory or virions. Finally, we find evidence for an intermolecular dithiol/disulfide relay within the mimivirus R596 dimer, the first such extended, intersubunit redox-active site identified in a viral sulfhydryl oxidase.

Genomes of nucleocytoplasmic large DNA viruses (NCLDVs) encode enzymes that catalyze the formation of disulfide bonds between cysteine amino acid residues in proteins, a function essential for the proper assembly and propagation of NCLDV virions. Recently, a catalyst of disulfide formation was identified in baculoviruses, a group of large double-stranded DNA viruses considered phylogenetically distinct from NCLDVs. The NCLDV and baculovirus disulfide catalysts are flavin adenine dinucleotide (FAD)-binding sulfhydryl oxidases related to the cellular Erv enzyme family, but the baculovirus enzyme, the product of the Ac92 gene in Autographa californica multiple nucleopolyhedrovirus (AcMNPV), is highly divergent at the amino acid sequence level. The crystal structure of the Ac92 protein presented here shows a configuration of the active-site cysteine residues and bound cofactor similar to that observed in other Erv sulfhydryl oxidases. However, Ac92 has a complex quaternary structural arrangement not previously seen in cellular or viral enzymes of this family. This novel assembly comprises a dimer of pseudodimers with a striking 40-degree kink in the interface helix between subunits. The diversification of the Erv sulfhydryl oxidase enzymes in large double-stranded DNA viruses exemplifies the extreme degree to which these viruses can push the boundaries of protein family folds.

In this issue of Molecular Cell, Ron and colleagues (Zito et al., 2010b) show that an enzyme responsible for cleaning up hydrogen peroxide in the endoplasmic reticulum can contribute productively to disulfide bond formation.

Large double-stranded DNA viruses, including poxviruses and mimiviruses, encode enzymes to catalyze the formation of disulfide bonds in viral proteins produced in the cell cytosol, an atypical location for oxidative protein folding. These viral disulfide catalysts belong to a family of sulfhydryl oxidases that are dimers of a small five-helix fold containing a Cys-X-X-Cys motif juxtaposed to a flavin adenine dinucleotide cofactor. We report that the sulfhydryl oxidase pB119L from African swine fever virus (ASFV) uses for self-assembly surface different from that observed in homologs from mammals, plants, and fungi. Within a protein family, different packing interfaces for the same oligomerization state are extremely rare. We find that the alternate dimerization mode seen in ASFV pB119L is not characteristic of all viral sulfhydryl oxidases, as the flavin-binding domain from a mimivirus sulfhydryl oxidase assumes the same dimer structure as the known eukaryotic enzymes. ASFV pB119L demonstrates the potential of large double-stranded DNA viruses, which have faster mutation rates than their hosts and the tendency to incorporate host genes, to pioneer new protein folds and self-assembly modes.

Oxidoreductases belonging to the protein disulfide isomerase (PDI) family promote proper disulfide bond formation in substrate proteins in the endoplasmic reticulum. In plants and metazoans, new family members continue to be identified and assigned to various functional niches. PDI-like proteins typically contain tandem thioredoxin-fold domains. The limited information available suggested that the relative orientations of these domains may be quite uniform across the family, and structural models based on this assumption are appearing. However, the X-ray crystal structure of the yeast PDI family protein Mpd1p, described here, demonstrates the radically different domain orientations and surface properties achievable with multiple copies of the thioredoxin fold. A comparison of Mpd1p with yeast Pdi1p expands our perspective on the contexts in which redox-active motifs are presented in the PDI family.

The interaction of membrane-embedded voltage-activated potassium channels (Kv) with intracellular scaffold proteins, such as the postsynaptic density 95 (PSD-95) protein, is mediated by the channel C-terminal segment. This interaction underlies Kv channel clustering at unique membrane sites and is important for the proper assembly and functioning of the synapse. In the current study, we address the molecular mechanism underlying Kv/PSD-95 interaction. We provide experimental evidence, based on hydrodynamic and spectroscopic analyses, indicating that the isolated C-terminal segment of the archetypical Shaker Kv channel (ShB-C) is a random coil, suggesting that ShB-C belongs to the recently defined class of intrinsically disordered proteins. We show that isolated ShB-C is still able to bind its scaffold protein partner and support protein clustering in vivo, indicating that unfoldedness is compatible with ShB-C activity. Pulldown experiments involving C-terminal chains differing in flexibility or length further demonstrate that intrinsic disorder in the C-terminal segment of the Shaker channel modulates its interaction with the PSD-95 protein. Our results thus suggest that the C-terminal domain of the Shaker Kv channel behaves as an entropic chain and support a "fishing rod" molecular mechanism for Kv channel binding to scaffold proteins. The importance of intrinsically disordered protein segments to the complex processes of synapse assembly, maintenance, and function is discussed.

Retroviral aspartyl proteases are homodimeric, whereas eukaryotic aspartyl proteases tend to be large, monomeric enzymes with 2-fold internal symmetry. It has been proposed that contemporary monomeric aspartyl proteases evolved by gene duplication and fusion from a primordial homodimeric enzyme. Recent sequence analyses have suggested that such "fossil" dimeric aspartyl proteases are still encoded in the eukaryotic genome. We present evidence for retention of a dimeric aspartyl protease in eukaryotes. The X-ray crystal structure of a domain of the Saccharomyces cerevisiae protein Ddi1 shows that it is a dimer with a fold similar to that of the retroviral proteases. Furthermore, the double Asp-Thr-Gly-Ala amino acid sequence motif at the active site of HIV protease is found with identical geometry in the Ddi1 structure. However, the putative substrate binding groove is wider in Ddi1 than in the retroviral proteases, suggesting that Ddi1 accommodates bulkier substrates. Ddi1 belongs to a family of proteins known as the ubiquitin receptors, which have in common the ability to bind ubiquitinated substrates and the proteasome. Ubiquitin receptors contain an amino-terminal ubiquitin-like (UBL) domain and a carboxy-terminal ubiquitin-associated (UBA) domain, but Ddi1 is the only representative in which the UBL and UBA domains flank an aspartyl protease-like domain. The remarkable structural similarity between the central domain of Ddi1 and the retroviral proteases, in the global fold and in active-site detail, suggests that Ddi1 functions proteolytically during regulated protein turnover in the cell. (c) 2006 Elsevier Ltd. All rights reserved.

Three different classes of thiol-oxidoreductases that facilitate the formation of protein disulfide bonds have been identified. They are the Ero1 and SOX/ALR family members in eukaryotic cells, and the DsbB family members in prokaryotic cells. These enzymes transfer oxidizing potential to the proteins PDI or DsbA, which are responsible for directly introducing disulfide bonds into substrate proteins during oxidative protein folding in eukaryotes and prokaryotes, respectively. A comparison of the recent X-ray crystal structure of Ero1 with the previously solved structure of the SOX/ALR family member Erv2 reveals that, despite a lack of primary sequence homology between Ero1 and Erv2, the core catalytic domains of these two proteins share a remarkable structural similarity. Our search of the DsbB protein sequence for features found in the Ero1 and Erv2 structures leads us to propose that, in a fascinating example of structural convergence, the catalytic core of this integral membrane protein may resemble the soluble catalytic domain of Ero1 and Erv2. Our analysis of DsbB also identified two new groups of DsbB proteins that, based on sequence homology, may also possess a catalytic core similar in structure to the catalytic domains of Ero1 and Erv2.

Infection of T lymphocytes by the cytopathic retrovirus feline leukemia virus subgroup T (FeLV-T) requires FeLIX, a cellular coreceptor that is encoded by an endogenous provirus and closely resembles the receptor-binding domain (RBD) of feline leukemia virus subgroup B (FeLV-B). We determined the structure of FeLV-B RBD, which has FeLIX activity, to a 2.5-Å resolution by X-ray crystallography. The structure of the receptor-specific subdomain of this glycoprotein differs dramatically from that of Friend murine leukemia virus (Fr-MLV), which binds a different cell surface receptor. Remarkably, we find that Fr-MLV RBD also activates FeLV-T infection of cells expressing the Fr-MLV receptor and that FeLV-B RBD is a competitive inhibitor of infection under these conditions. These studies suggest that FeLV-T infection relies on the following property of mammalian leukemia virus RBDs: the ability to couple interaction with one of a variety of receptors to the activation of a conserved membrane fusion mechanism. A comparison of the FeLV-B and Fr-MLV RBD structures illustrates how receptor-specific regions are linked to conserved elements critical for postbinding events in virus entry.

Yeast Sec18p and its mammalian orthologue N-ethylmaleimide-sensitive fusion protein (NSF) are hexameric ATPases with a central role in vesicle trafficking. Aided by soluble adapter factors (SNAPs), Sec18p/NSF induces ATP-dependent disassembly of a complex of integral membrane proteins from the vesicle and target membranes (SNAP receptors). During the ATP hydrolysis cycle, the Sec18p/NSF homohexamer undergoes a large-scale conformational change involving repositioning of the most N terminal of the three domains of each protomer, a domain that is required for SNAP-mediated interaction with SNAP receptors. Whether an internal conformational change in the N-terminal domains accompanies their reorientation with respect to the rest of the hexamer remains to be addressed. We have determined the structure of the N- terminal domain from Sec18p by x-ray crystallography. The Sec18p N-terminal domain consists of two β-sheet-rich subdomains connected by a short linker. A conserved basic cleft opposite the linker may constitute a SNAP-binding site. Despite structural variability in the linker region and in an adjacent loop, all three independent molecules in the crystal asymmetric unit have the identical subdomain interface, supporting the notion that this interface is a preferred packing arrangement. However, the linker flexibility allows for the possibility that other subdomain orientations may be sampled.