The structure of tumor necrosis factor-alpha at 2.6 A resolution. Implications for receptor binding

The three-dimensional structure of tumor necrosis factor (TNF-alpha), a protein hormone secreted by macrophages, has been determined at 2.6 A resolution by x-ray crystallography. Phases were determined by multiple isomorphous replacement using data collected from five heavy atom derivatives. The multiple isomorphous replacement phases were further improved by real space symmetry averaging, exploiting the noncrystallographic 3-fold symmetry of the TNF-alpha trimer. An atomic model corresponding to the known amino acid sequence of TNF-alpha was readily built into the electron density map calculated with these improved phases. The 17,350-dalton monomer forms an elongated, antiparallel beta-pleated sheet sandwich with a "jelly-roll" topology. Three monomers associate intimately about a 3-fold axis of symmetry to form a compact bell-shaped trimer. Examination of the model and comparison to known protein structures reveals striking structural homology to several viral coat proteins, particularly satellite tobacco necrosis virus. Locations of residues conserved between TNF-alpha and lymphotoxin (TNF-beta, a related cytokine known to bind to the same receptors as TNF-alpha) suggest that lymphotoxin, like TNF-alpha, binds to the receptor as a trimer and that the general site of interaction with the receptor is at the "base" of the trimer.     

Aspartic acid 50 and tyrosine 108 are essential for receptor binding and cytotoxic activity of tumour necrosis factor beta (lymphotoxin).

Single amino acid substitutions were generated in predicted hydrophilic loop regions of the human tumour necrosis factor beta (TNF-beta) molecule, and the mutant proteins were expressed in Escherichia coli and purified. Mutants with single amino acid changes at either of two distinct loop regions, at positions aspartic acid 50 or tyrosine 108, were found to have greatly reduced receptor binding and cytotoxic activity. These two regions in TNF-beta correspond to known loop regions where mutations also result in loss of biological activity of TNF-alpha, a related cytokine which shares the same cellular receptors with TNF-beta. The two distinct loops at positions 31-34 and 84-89 in the known three-dimensional structure of TNF-alpha (equivalent to positions 46-50 and 105-110 respectively in TNF-beta), lie on opposite sides of the TNF-alpha monomer. When the TNF-alpha monomer forms a trimer, the two loops, each from a different subunit of the trimer, come together and lie in a cleft between adjacent subunits. Together, these findings suggest that a TNF receptor binds to a cleft between subunits via surface loops at amino acid residues 31-34 and 84-89 in TNF-alpha, and similarly via surface loops including amino acids aspartic acid 50 and tyrosine 108 in TNF-beta.     

Characterization of recombinant human lymphotoxin (tumor necrosis factor-beta) produced by a mammalian cell line

Lymphotoxin (LT) was purified from serum-free conditioned media of a recombinant mammalian cell line transfected with human lymphotoxin cDNA. The purification scheme consisted of controlled pore glass chromatography, Q-Sepharose ion-exchange chromatography, and concanavalin A-Sepharose chromatography. The purified protein was found to be homogeneous by reverse-phase high-performance liquid chromatography and had an approximate specific activity of 130 X 10(6) units per milligram protein as determined by the L-929 cytotoxicity assay. Purified LT had an isoelectric point of approximately 6.85 and an apparent molecular weight of 50,000 by gel permeation high-pressure liquid chromatography. However, when analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, two distinct bands at approximate molecular sizes of 25 and 20 kDa were observed. Both the bands were immunoreactive by Western blot analysis and found to be associated with biological activity. The two forms of lymphotoxin differed from each other with respect to protein structure. Amino-terminal amino acid sequence analysis revealed that the 25-kDa LT sequence starts with Leu-Pro-Gly-residues whereas that of the 20-kDa LT begins with His-Leu-Ala; thus the latter form is truncated by 20 amino acid residues from the amino terminal. Two species of LT also differed from each other with respect to carbohydrate structure. Enzymatic removal of sialic acid reduced the molecular weight of 25 kDa by approximately 5 kDa whereas that of the 20-kDa LT was unchanged. A reduction in an apparent molecular size by approximately 4 kDa of both species of LT was observed on removal of N-linked oligosaccharides. Treatment with O-Glycanase had minimal effect on either form of LT. The recombinant lymphotoxin described here was found superior in its solubility behavior as compared to bacterial cell derived LT. Overall, mammalian cell line derived recombinant LT appears closer in its properties to natural LT than does bacterial cell derived recombinant LT.     

Amino acids outside of the loops that define the agonist binding site are important for ligand binding to insect nicotinic acetylcholine receptors

Nicotinic acetylcholine (ACh) receptors (nAChRs) are the targets of several kinds of insecticides. Based on the mutagenesis studies of Torpedo californica nAChRs and solved structure of a molluscan, glial-derived soluble ACh-binding protein, a model of the agonist site was constructed with contributing amino acids from three distinct loops (A, B, and C) of the α subunits and another three loops (D, E, and F) of the non-α subunits. According to this model, most insect nAChR subunits can form the functional heteromeric or homomeric receptors. Actually, insect subunits themselves did not form any functional receptor at various combinations as yet, and only part of them can form the functional receptors with vertebrate non-α subunits. These findings suggested that the agonist binding for insect nAChRs was not only contributed by those key amino acids in six loops, but also some unidentified amino acids from other regions. In our previous studies on nAChRs for Nilaparvata lugens , a target-site mutation (Y151S) was found within two α subunits (Nlα1 and Nlα3). In Drosophila S2 cells and Xenopus oocytes, Nlα1 can form functional receptors with rat β2 subunit. However, the same thing was not observed in Nlα3. In the present paper, by exchanging the corresponding regions between Nlα1 and Nlα3 to generate different chimeras, amino acid residues or residue clusters in the regions outside the six loops were found to play essential roles in agonist binding, especially for the amino acid clusters between loop B and C. This result indicated that the residues in the six loops could be necessary, but not enough for the activity of agonist binding.     

Lymphotoxin is expressed as a heteromeric complex with a distinct 33-kDa glycoprotein on the surface of an activated human T cell hybridoma.

We characterized the membrane-associated form of lymphotoxin (surface LT) on the activated II-23.D7 T cell hybridoma. Antibodies to rLT precipitated both surface LT and a distinct 33-kDa glycoprotein (p33). Because p33 and surface LT were antigenically unrelated, their coprecipitation suggested a physical association of p33 and surface LT on the membrane. Pulse-chase analysis indicated that LT and p33 associate with each other early in the LT biosynthetic pathway, precluding the possibility that LT is secreted and bound to p33 or a surface receptor. Furthermore, no p33 was associated with the secreted form of LT. Isoelectric focusing of surface LT and p33 under nondenaturing and denaturing conditions confirmed that surface LT and p33 existed as a complex. Treatment of cells with a high concentration of salt or with acid indicated that surface LT is a peripheral membrane protein. Although secreted LT is a homologous trimer, protein cross-linking studies revealed that surface LT existed as a monomer associated with a dimer of p33. Together the results demonstrate a novel mechanism for stable membrane expression of LT by activated T cells.     

Structure of a flavivirus envelope glycoprotein in its low-pH-induced membrane fusion conformation

Enveloped viruses enter cells via a membrane fusion reaction driven by conformational changes of specific viral envelope proteins. We report here the structure of the ectodomain of the tick-borne encephalitis virus envelope glycoprotein, E, a prototypical class II fusion protein, in its trimeric low-pH-induced conformation. We show that, in the conformational transition, the three domains of the neutral-pH form are maintained but their relative orientation is altered. Similar to the postfusion class I proteins, the subunits rearrange such that the fusion peptide loops cluster at one end of an elongated molecule and the C-terminal segments, connecting to the viral transmembrane region, run along the sides of the trimer pointing toward the fusion peptide loops. Comparison with the low-pH-induced form of the alphavirus class II fusion protein reveals striking differences at the end of the molecule bearing the fusion peptides, suggesting an important conformational effect of the missing membrane connecting segment.     

4-Oxalocrotonate tautomerase, a 41-kDa homohexamer: backbone and side-chain resonance assignments, solution secondary structure, and location of active site residues by heteronuclear NMR spectroscopy.

4-Oxalocrotonate tautomerase (4-OT), a homohexamer consisting of 62 residues per subunit, catalyzes the isomerization of unsaturated alpha-keto acids using Pro-1 as a general base (Stivers et al., 1996a, 1996b). We report the backbone and side-chain 1H, 15N, and 13C NMR assignments and the solution secondary structure for 4-OT using 2D and 3D homonuclear and heteronuclear NMR methods. The subunit secondary structure consists of an alpha-helix (residues 13-30), two beta-strands (beta 1, residues 2-8; beta 2, residues 39-45), a beta-hairpin (residues 50-57), two loops (I, residues 9-12; II, 34-38), and two turns (I, residues 30-33; II, 47-50). The remaining residues form coils. The beta 1 strand is parallel to the beta 2 strand of the same subunit on the basis of cross stand NH(i)-NH(j) NOEs in a 2D 15N-edited 1H-NOESY spectrum of hexameric 4-OT containing two 15N-labeled subunits/hexamer. The beta 1 strand is also antiparallel to another beta 1 strand from an adjacent subunit forming a subunit interface. Because only three such pairwise interactions are possible, the hexamer is a trimer of dimers. The diffusion constant, determined by dynamic light scattering, and the rotational correlation time (14.5 ns) estimated from 15N T1/T2 measurements, are consistent with the hexameric molecular weight of 41 kDa. Residue Phe-50 is in the active site on the basis of transferred NOEs to the bound partial substrate 2-oxo-1,6-hexanedioate. Modification of the general base, Pro-1, with the active site-directed irreversible inhibitor, 3-bromopyruvate, significantly alters the amide 15N and NH chemical shifts of residues in the beta-hairpin and in loop II, providing evidence that these regions change conformation when the active site is occupied.     

Structural refinement and analysis of Mengo virus

The structure of Mengo encephalomyelitis virus was refined at 3 A resolution with a final R-factor of 0.221 and a root-mean-square deviation from idealized bond lengths of 0.019 A for 10 A to 3 A data with F greater than or equal to 3 sigma(F). The Hendrickson-Konnert refinement was restrained by the phases derived from the molecular replacement averaging procedure and constrained by the icosahedral symmetry of the virus. The virus consists of 60 protomers each having three major subunits, VP1, VP2 and VP3, along with one smaller internal protein, VP4. The three major subunits form similar eight-stranded beta-barrel structures. Alterations in the original sequence were found at position 45 in VP1 (Arg to Ala) and at position 58 in VP3 (Met to Val). The residues in loops I and II of VP1 (82 to 102), the "FMDV loop" in VP1 (205 to 213), the flexible loop of VP3 in the putative receptor attachment site (175 to 185) as well as the terminal regions 260 to 268 in VP1, 253 to 256 in VP2 and 13 to 15 in VP4 were built or modified in regions of weak density. The variation in temperature factors at the end of the refinement is over a wide range (from 2 to 80 A2), with the disordered outer and inner regions showing high mobility. Four cis proline residues, 105 in VP1, 85 and 152 in VP2 and 59 in VP3, have been identified. The disulfide bridge Cys86 to Cys88 in VP3 has been characterized. One phosphate ion and 233 water positions were included in the refinement. It is suggested that this phosphate is associated with the receptor attachment site. There are two major hydrogen-bonding networks involving solvent atoms; one involving only the subunits of a protomer, and the other connecting the protomers in a pentamer. The distribution of atom types around the icosahedral symmetry axes shows that the 5-fold channel is more hydrophobic than that along the 3-fold axis and that there are more charged residues around the 2-fold axis. The analysis of contacts between the different subunits supports the assignment of the protomeric unit. The five protomers that form the pentameric unit are held together by interactions involving the smaller VP4 protein and the amino termini of VP1 and VP3. The pentamers are associated by means of the amino-terminal region of the VP2 subunits, the beta F strand of the VP3 subunits, the C terminus of the VP4 subunits and the electrostatic helical (alpha A) interactions of VP2 subunits across the icosahedral 2-fold axes. The superposition of the corresponding subunits of Mengo virus, human rhinovirus 14 and southern bean mosaic virus has provided an improved sequence alignment. The largest structural similarity is between the VP3 subunits of Mengo virus and rhinovirus, while the least similarity is between the VP1 subunits. The various specialized insertions in the different subunits can be associated with specific functional requirements.     

Site-directed mutational analysis of human tumor necrosis factor-alpha receptor binding site and structure-functional relationship.

In order to define the receptor binding site and the structure-functional relationship of tumor necrosis factor (TNF), single amino acid substitutions were made by site-directed mutagenesis at selected residues of human tumor necrosis factor, using a phagemid mutagenesis/expression vector. The recombinant TNF mutants were compared to the wild type TNF in assays using crude bacterial lysates, for protein yield, solubility, subunit trimerization, receptor binding inhibition activity, and in vitro cytotoxic activity. All mutants which did not form cross-linkable trimer also showed little cytotoxic activity or receptor binding inhibition activity, indicating that trimer formation is obligatory for TNF-alpha activity. Most mutations of internal residues yielded no cross-linkable trimer, while most mutations of surface residues yielded cross-linkable trimer. Mutations at surface residues Leu29, Arg31, and Ala35 yielded cross-linkable trimers with good activities, except proline substitutions which may cause conformational changes in the polypeptide chain. This suggested that these residues are near the receptor binding site. Mutations at other strictly conserved internal residues such as Ser60, His78, and Tyr119 form cross-linkable trimer with little activity. These mutations may indirectly affect the receptor binding site by forming trimers with undetectable abnormalities. Mutants of surface residues Tyr87, Ser95, Ser133, and Ser147 affect receptor binding and cytotoxic activity but not trimer formation, suggesting that these residues are involved directly in receptor binding. The fact that residues Arg31, Ala35, Tyr87, Ser95, and Ser147, located on the opposite sides of a monomer, are clustered at the intersubunit grooves of TNF trimer supports the current notion that TNF receptor binding sites are trivalent and are located at the three intersubunit grooves. However, our finding that Ser133, which is outside the groove, can also be involved directly in receptor binding suggested that the receptor binding sites of TNF may not be confined to the intersubunit grooves, but extended to include additional surface residues.     

Production of transmembrane and secreted forms of tumor necrosis factor (TNF)-alpha by HIV-1-specific CD4+ cytolytic T lymphocyte clones. Evidence for a TNF-alpha-independent cytolytic mechanism.

Candidate AIDS vaccines consisting of recombinant forms of the HIV-1 envelope glycoprotein induce, in seronegative human volunteers, an env-specific T cell response that includes CD4+, MHC class II-restricted CTL capable of lysing HIV-1-infected target cells. In this study, we have analyzed the production of the cytokines TNF-alpha and lymphotoxin (LT) by a set of env-specific CD4+ human CTL clones. TNF-alpha and LT are of interest because of their potential role in target cell destruction by CD4+ CTL. Our studies focused on the possibility that a cell surface form of TNF-alpha expressed by CTL after physiologic activation with target APC might participate in the cytolytic reactions mediated by these clones. We found that, upon interaction with target cells expressing env epitopes in the context of the appropriate MHC class II molecules, CD4+ CTL released TNF-alpha with kinetics that were rapid, compared with other cytokines, and that were generally similar to the kinetics of target cell destruction. LT secretion was not detected during the time course of the cytolytic reactions. A novel flow cytometric assay was used to show that physiologic activation of CD4+ CTL with target APC induced expression by the CTL of cell surface forms of TNF-alpha. Immunoprecipitations from activated, surface-iodinated CTL clones revealed two forms of surface TNF-alpha, a 26-kDa form, representing the transmembrane precursor of secreted TNF-alpha, as well as the 17-kDa secreted form bound to the cell surface. For a subset of CD4+ CTL, we found that treatment of CTL with cyclosporin A inhibited Ag-induced production of both transmembrane and secreted forms of TNF-alpha but had no effect on cytolysis. Thus, although transmembrane and secreted TNF-alpha produced by HIV-1-specific CD4+ CTL may have important effects in vivo, the rapid destruction of target APC by the set of CD4+ CTL clones described here occurs through a TNF-alpha-independent mechanism.     

Structural studies of the streptavidin binding loop.

The streptavidin-biotin complex provides the basis for many important biotechnological applications and is an interesting model system for studying high-affinity protein-ligand interactions. We report here crystallographic studies elucidating the conformation of the flexible binding loop of streptavidin (residues 45 to 52) in the unbound and bound forms. The crystal structures of unbound streptavidin have been determined in two monoclinic crystal forms. The binding loop generally adopts an open conformation in the unbound species. In one subunit of one crystal form, the flexible loop adopts the closed conformation and an analysis of packing interactions suggests that protein-protein contacts stabilize the closed loop conformation. In the other crystal form all loops adopt an open conformation. Co-crystallization of streptavidin and biotin resulted in two additional, different crystal forms, with ligand bound in all four binding sites of the first crystal form and biotin bound in only two subunits in a second. The major change associated with binding of biotin is the closure of the surface loop incorporating residues 45 to 52. Residues 49 to 52 display a 3(10) helical conformation in unbound subunits of our structures as opposed to the disordered loops observed in other structure determinations of streptavidin. In addition, the open conformation is stabilized by a beta-sheet hydrogen bond between residues 45 and 52, which cannot occur in the closed conformation. The 3(10) helix is observed in nearly all unbound subunits of both the co-crystallized and ligand-free structures. An analysis of the temperature factors of the binding loop regions suggests that the mobility of the closed loops in the complexed structures is lower than in the open loops of the ligand-free structures. The two biotin bound subunits in the tetramer found in the MONO-b1 crystal form are those that contribute Trp 120 across their respective binding pockets, suggesting a structural link between these binding sites in the tetramer. However, there are no obvious signatures of binding site communication observed upon ligand binding, such as quaternary structure changes or shifts in the region of Trp 120. These studies demonstrate that while crystallographic packing interactions can stabilize both the open and closed forms of the flexible loop, in their absence the loop is open in the unbound state and closed in the presence of biotin. If present in solution, the helical structure in the open loop conformation could moderate the entropic penalty associated with biotin binding by contributing an order-to-disorder component to the loop closure.     

Increase in permeability of human endothelial cell monolayer by recombinant human lymphotoxin

The effect of lymphotoxin (LT) on transendothelial permeability was examined using in vitro human endothelial cell (EC) culture system. To assess permeability, we measured the movement of human IgG labeled with horseradish peroxidase (HRP-huIgG) across an EC monolayer cultured on a fibronectin and collagen-coated membrane. LT increases the permeability dose dependently within 16 hours. Similar results were obtained with tumor necrosis factor (TNF)-alpha. The results suggest that increase in vascular permeability by LT and TNF-alpha plays an important role in initiating hemorrhagic necrosis of tumors in vivo.     

Biochemical and functional analyses of the human Toll-like receptor 3 ectodomain

The structure of the human Toll-like receptor 3 (TLR3) ectodomain (ECD) was recently solved by x-ray crystallography, leading to a number of models concerning TLR3 function (Choe, J., Kelker, M. S., and Wilson, I. A. (2005) Science 309, 581-585; Bell, J. K., Botos, I., Hall, P. R., Askins, J., Shiloach, J., Segal, D. M., and Davies, D. R. (2005) Proc. Natl. Acad. Sci. U. S. A. 102, 10976-10980) The structure revealed four pairs of cysteines that are putatively involved in disulfide bond formation, several residues that are predicted to be involved in dimerization between ECD subunits, and surfaces that could bind to poly(I:C). In addition, there are two loops that protrude from the central solenoid structure of the protein. We examined the recombinant TLR3 ECD for disulfide bond formation, poly(I:C) binding, and protein-protein interaction. We also made over 80 mutations in the residues that could affect these features in the full-length TLR3 and examined their effects in TLR3-mediated NF-kappaB activation. A number of mutations that affected TLR3 activity also affected the ability to act as dominant negative inhibitors of wild type TLR3. Loss of putative RNA binding did not necessarily affect dominant negative activity. All of the results support a model where a dimer of TLR3 is the form that binds RNA and activates signal transduction.     

Crystal structure of osmoporin OmpC from E. coli at 2.0 A

Porins form transmembrane pores in the outer membrane of Gram-negative bacteria with matrix porin OmpF and osmoporin OmpC from Escherichia coli being differentially expressed depending on environmental conditions. The three-dimensional structure of OmpC has been determined to 2.0 A resolution by X-ray crystallography. As expected from the high sequence similarity, OmpC adopts the OmpF-like 16-stranded hollow beta-barrel fold with three beta-barrels associated to form a tight trimer. Unlike in OmpF, the extracellular loops form a continuous wall at the perimeter of the vestibule common to the three pores, due to a 14-residues insertion in loop L4. The pore constriction and the periplasmic outlet are very similar to OmpF with 74% of the pore lining residues being conserved. Overall, only few ionizable residues are exchanged at the pore lining. The OmpC structure suggests that not pore size, but electrostatic pore potential and particular atomic details of the pore linings are the critical parameters that physiologically distinguish OmpC from OmpF.     

Effect of local secondary structure of mRNA on expression of recombinant human lymphotoxin inEscherichia coli

We substituted a few alternative codons close to the N-terminal coding region of human lymphotoxin (LT) gene for natural ones to avoid stable local secondary structures of mRNA, predicted in the 5-terminal 130 nucleotides of mRNA. With this modification, we have overexpressed a recombinant lymphotoxin (rLT) inEscherichia coli under the control of trc promoter. Most of the produced rLT was a soluble and active form.     

Crystal structure of the catalytic trimer of Methanococcus jannaschii aspartate transcarbamoylase

The catalytic trimer of Methanococcus jannaschii aspartate transcarbamoylase is extremely heat stable, maintaining 75% of its activity after heat treatment for 60 min at 75 degrees C. We undertook its structural analysis in order to understand the molecular basis of its thermostability and gain insight on how its catalytic function adapts to high temperature. Several structural elements potentially contributing to thermostability were identified. These include: (i) changes in the amino acid composition such as a decrease in the thermolabile residues Gln and Asn, an increase in the charged residues Lys and Glu, an increase in Tyr and a decrease in Ala residues; (ii) a larger number of salt bridges, in particular, the improvement of ion-pair networks; (iii) shortening of the N-terminus and shortening of three loops. An interesting feature of the crystal structure is the association of two crystallographically independent catalytic subunits into a staggered complex with an intertrimer distance of 33.8 A. The active site appears similar to Escherichia coli. Upon substrate binding, smaller changes in the global orientation of domains and larger conformational changes of the active site residues are expected as compared to E. coli.