Interaction between the Z-type DNA duplex and 1,3-propanediamine: crystal structure of d(CACGTG)2 at 1.2 A resolution

The crystal structure of a hexamer duplex d(CACGTG)(2) has been determined and refined to an R-factor of 18.3% using X-ray data up to 1.2 A resolution. The sequence crystallizes as a left-handed Z-form double helix with Watson-Crick base pairing. There is one hexamer duplex, a spermine molecule, 71 water molecules, and an unexpected diamine (Z-5, 1,3-propanediamine, C(3)H(10)N(2)) in the asymmetric unit. This is the high-resolution non-disordered structure of a Z-DNA hexamer containing two AT base pairs in the interior of a duplex with no modifications such as bromination or methylation on cytosine bases. This structure does not possess multivalent cations such as cobalt hexaammine that are known to stabilize Z-DNA. The overall duplex structure and its crystal interactions are similar to those of the pure-spermine form of the d(CGCGCG)(2) structure. The spine of hydration in the minor groove is intact except in the vicinity of the T5A8 base pair. The binding of the Z-5 molecule in the minor grove of the d(CACGTG)(2) duplex appears to have a profound effect in conferring stability to a Z-DNA conformation via electrostatic complementarity and hydrogen bonding interactions. The successive base stacking geometry in d(CACGTG)(2) is similar to the corresponding steps in d(CG)(3). These results suggest that specific polyamines such as Z-5 could serve as powerful inducers of Z-type conformation in unmodified DNA sequences with AT base pairs. This structure provides a molecular basis for stabilizing AT base pairs incorporated into an alternating d(CG) sequence.     

DNA-drug interactions. The crystal structure of d(CGATCG) complexed with daunomycin

The structure of a d(CGATCG)-daunomycin complex has been determined by single crystal X-ray diffraction techniques. Refinement, with the location of 40 solvent molecules, using data up to 1.5 A, converged with a final crystallographic residual, R = 0.25 (RW = 0.22). The tetragonal crystals are in space group P4(1)2(1)2, with cell dimensions of a = 27.98 A and c = 52.87 A. The self-complementary d(CGATCG) forms a distorted right-handed helix with a daunomycin molecule intercalated at each d(CpG) step. The daunomycin aglycon chromophore is oriented at right-angles to the long axis of the DNA base-pairs. This head-on intercalation is stabilized by direct hydrogen bonds and indirectly via solvent-mediated, hydrogen-bonding interactions between the chromophore and its intercalation site base-pairs. The cyclohexene ring and amino sugar substituent lie in the minor groove. The amino sugar N-3' forms a hydrogen bond with O-2 of the next neighbouring thymine. This electrostatic interaction helps position the sugar in a way that results in extensive van der Waals contacts between the drug and the DNA. There is no interaction between daunosamine and the DNA sugar-phosphate backbone. We present full experimental details and all relevant conformational parameters, and use the comparison with a d(CGTACG)-daunomycin complex to rationalize some neighbouring sequence effects involved in daunomycin binding.     

Crystal structure of triosephosphate isomerase complexed with 2-phosphoglycolate at 0.83-A resolution

The atomic resolution structure of Leishmania mexicana triosephosphate isomerase complexed with 2-phosphoglycolate shows that this transition state analogue is bound in two conformations. Also for the side chain of the catalytic glutamate, Glu(167), two conformations are observed. In both conformations, a very short hydrogen bond exists between the carboxylate group of the ligand and the catalytic glutamate. The distance between O11 of PGA and Oepsilon2 of Glu(167) is 2.61 and 2.55 A for the major and minor conformations, respectively. In either conformation, Oepsilon1 of Glu(167) is hydrogen-bonded to a water network connecting the side chain with bulk solvent. This network also occurs in two mutually exclusive arrangements. Despite the structural disorder in the active site, the C termini of the beta strands that construct the active site display the least anisotropy compared with the rest of the protein. The loops following these beta strands display various degrees of anisotropy, with the tip of the dimer interface loop 3 having very low anisotropy and the C-terminal region of the active site loop 6 having the highest anisotropy. The pyrrolidine ring of Pro(168) at the N-terminal region of loop 6 is in a strained planar conformation to facilitate loop opening and product release.     

Interactions of quinoxaline antibiotic and DNA: the molecular structure of a triostin A-d(GCGTACGC) complex

The crystal structure of a DNA octamer d(GCGTACGC) complexed to an antitumor antibiotic, triostin A, has been solved and refined to 2.2 A resolution by x-ray diffraction analysis. The antibiotic molecule acts as a true bis intercalator surrounding the d(CpG) sequence at either end of the unwound right-handed DNA double helix. As previously observed in the structure of triostin A-d(CGTACG) complex (A.H.-J. Wang, et. al., Science, 225, 1115-1121 (1984)), the alanine amino acid residues of the drug molecule form sequence-specific hydrogen bonds to guanines in the minor groove. The two central A.T base pairs are in Hoogsteen configuration with adenine in the syn conformation. In addition, the two terminal G.C base pairs flanking the quinoxaline rings are also held together by Hoogsteen base pairing. This is the first observation in an oligonucleotide of. Hoogsteen G.C base pairs where the cytosine is protonated. The principal functional components of a bis-intercalative compound are discussed.     

DNA-drug interactions. The crystal structures of d(TGTACA) and d(TGATCA) complexed with daunomycin

The anticancer drug daunomycin has been co-crystallized with the hexanucleotide duplex sequences d(TGTACA) and d(TGATCA) and single crystal X-ray diffraction studies of these two complexes have been carried out. Structure solution of the d(TGTACA) and d(TGATCA) complexes to 1.6 and 1.7 Angstrom resolution, respectively, shows two daunomycin molecules bound to the DNA hexamer. Binding occurs via intercalation of the drug chromophore at the d(TpG) step, and hydrogen bonding interactions involving the drug, DNA and solvent molecules. The daunomycin sugar is located in the minor groove of the DNA hexamer and is stabilized by hydrogen bonds between the amino group of the sugar and functional groups on the floor of the groove. The amino sugar of the d(TGATCA) duplex interacts directly with the DNA sequence, while in the d(TGTACA) duplex, the interaction is via solvent molecules. Two other complexes d(CGTACG)-daunomycin and d(CGATCG)-daunomycin have previously been structurally characterized. Comparison of the four structures with daunomycin bound to the triplet sequences 5'TGT, 5'TGA, 5'CGT and 5'CGA reveals changes in the conformation of both the DNA hexamer and the daunomycin upon complexation, as well as the hydrogen bonding and van der Waals' interactions.     

Crystal structure at 1.5-A resolution of d(CGCICICG), an octanucleotide containing inosine, and its comparison with d(CGCG) and d(CGCGCG) structures.

The octadeoxyribonucleotide d(CGCICICG) has been crystallized in space group P(6)5(22) with unit cell dimensions of a = b = 31.0 A and c = 43.7 A, and X-ray diffraction data have been collected to 1.5-A resolution. Precession photographs and the self-Patterson function indicate that 12 base pairs of Z-conformation DNA stack along the c-axis, and the double helices pack in a hexagonal array similar to that seen in other crystals of Z-DNA. The structure has been solved by both Patterson deconvolution and molecular replacement methods and refined in space group P(6)5 to an R factor of 0.225 using 2503 unique reflections greater than 3.0 sigma (F). Comparison of the molecules within the hexagonal lattice with highly refined crystal structures of other Z-DNA reveals only minor conformational differences, most notably in the pucker of the deoxyribose of the purine residues. The DNA has multiple occupancy of C:I and C:G base pairs, and C:I base pairs adopt a conformation similar to that of C:G base pairs.     

Molecular structure of an apolipoprotein determined at 2.5-A resolution

The three-dimensional structure of an apolipoprotein isolated from the African migratory locust Locusta migratoria has been determined by X-ray analysis to a resolution of 2.5 A. The overall molecular architecture of this protein consists of five long alpha-helices connected by short loops. As predicted from amino acid sequence analyses, these helices are distinctly amphiphilic with the hydrophobic residues pointing in toward the interior of the protein and the hydrophilic side chains facing outward. The molecule falls into the general category of up-and-down alpha-helical bundles as previously observed, for example, in cytochrome c'. Although the structure shows the presence of five long amphiphilic alpha-helices, the alpha-helical moment and hydrophobicity of the entire molecule fall into the range found for normal globular proteins. Thus, in order for the amphiphilic helices to play a role in the binding of the protein to a lipid surface, there must be a structural reorganization of the protein which exposes the hydrophobic interior to the lipid surface. The three-dimensional motif of this apolipoprotein is compatible with a model in which the molecule binds to the lipid surface via a relatively nonpolar end and then spreads on the surface in such a way as to cause the hydrophobic side chains of the helices to come in contact with the lipid surface, the charged and polar residues to remain in contact with water, and the overall helical motif of the protein to be maintained.     

Crystal structure of ribonuclease T1 complexed with adenosine 2'-monophosphate at 1.8-A resolution.

Ribonuclease T1 was purified from an Escherichia coli overproducing strain and co-crystallized with adenosine 2'-monophosphate (2'-AMP) by microdialysis against 50% (v/v) 2-methyl-2,4-pentanediol in 20 mM sodium acetate, 2 mM calcium acetate, pH 4.2. The crystals have orthorhombic space group P2(1)2(1)2(1), with cell dimensions a = 48.93(1), b = 46.57(4), c = 41.04(2) A; Z = 4 and V = 93520 A3. The crystal structure was determined on the basis of the isomorphous structure of uncomplexed RNase T1 (Martinez-Oyanedel et al. (1991) submitted for publication) and refined by least squares methods using stereochemical restraints. The refinement was based on Fhkl of 7,445 reflections with Fo greater than or equal to 1 sigma (Fo) in the resolution range of 10-1.8 A, and converged at a crystallographic R factor of 0.149. The phosphate group of 2'-AMP is tightly hydrogen-bonded to the side chains of the active site residues Tyr38, His40, Glu58, Arg77, and His92, comparable with vanadate binding in the respective complex (Kostrewa, D., Choe, H.-W., Heinemann, U., and Saenger, W. (1989) Biochemistry 28, 7592-7600) and different from the complex with guanosine 2'-monophosphate (Arni, R., Heinemann, U., Tokuoka, R., and Saenger, W. (1988) J. Biol. Chem. 263, 15358-15368) where the phosphate does not interact with Arg77 and His92. The adenosine moiety is not located in the guanosine recognition site but stacked on Gly74 carbonyl and His92 imidazole, which serve as a subsite, as shown previously (Lenz, A., Cordes, F., Heinemann, U., and Saenger, W. (1991) J. Biol. Chem. 266, 7661-7667); in addition, there are hydrogen bonds adenine N6H . . . O Gly74 (minor component of three-center hydrogen bond) and adenosine O5' . . . O delta Asn36. These binding interactions readily explain why RNase T1 has some affinity for 2'-AMP. The molecular structure of RNase T1 is only marginally affected by 2'-AMP binding. Its "empty" guanosine-binding site features a flipped Asn43-Asn44 peptide bond and the side chains of Tyr45, Glu46 adopt conformations typical for RNase T1 not involved in guanosine binding. The side chains of amino acids Leu26, Ser35, Asp49, Val78 are disordered. The disorder of Val78 is of interest since this amino acid is located in a hydrophobic cavity, and the disorder appears to be correlated with an "empty" guanosine-binding site. The two Asp15 carboxylate oxygens and six water molecules coordinate a Ca2+ ion 8-fold in the form of a square antiprism.     

Refinement of the structure of recombinant rat intestinal fatty acid-binding apoprotein at 1.2-A resolution.

The three-dimensional structure of the 131-residue rat intestinal fatty acid-binding protein, without bound ligand (apoI-FABP), has been refined with x-ray diffraction data to a nominal resolution of 1.19 A. The final model has a conventional crystallographic R-factor of 16.9% for 34,290 unique reflections [a root mean square (r.m.s.) deviation for bond length of 0.012 A and a r.m.s. deviation of 2.368 degrees for bond angles]. Ninety-two residues are present as components of the protein's 10 anti-parallel beta-strands while 14 residues are part of its two short alpha-helices. The beta-strands and alpha-helices are organized into two nearly orthogonal beta-sheets. Particular attention has been placed in defining solvent structure and the structures of discretely disordered groups in this protein. Two hundred thirty-seven solvent molecules have been identified; 24 are located within apoI-FABP. The refined model includes alternate conformers for 228 protein atoms (109 main-chain, 119 side-chain) and 63 solvent molecules. We have found several aromatic side-chains with multiple conformations located near, or in, the protein's ligand binding site. This observation, along with the fact that these side-chains have a temperature factor that is relatively higher than that of other aromatic residues, suggests that they may be involved in the process of noncovalent binding of fatty acid. The absence of a true hydrophobic core in I-FABP suggests that its structural integrity may be maintained primarily by a hydrogen bonding network involving protein and solvent atoms.     

Methylation of the EcoRI recognition site does not alter DNA conformation: the crystal structure of d(CGCGAm6ATTCGCG) at 2.0-A resolution

Methylation of nucleic acid bases is known to prevent the cleavage of DNA by restriction endonucleases. The effect on the conformation of the DNA molecule itself and hence its interactions with other DNA binding proteins has been a subject of general interest. To help address this question, we have solved the crystal structure at 2.0 A of the methylated dodecamer, d(CGCGAm6ATTCGCG), which contains the EcoRI recognition sequence and have compared the conformation of the methylated molecule with that of its nonmethylated counterpart. This methylation produces a bulky hydrophobic patch on the floor of the major groove of B-DNA which plays an important role in the mechanism of inhibition of EcoRI restriction activity. However, with the exception of small perturbations in the immediate vicinity of the methyl groups, the structure is virtually unchanged. Given the lack of a conformational change upon methylation, we have extended this thesis of the recognition process to other types of restriction systems and found that different restriction enzymes seem to have their own characteristic protein-DNA interactions. The relative spatial orientations of methylation sites and cleavage sites must play a major role in ordering protein secondary structure elements as well as subunit-subunit interactions along the DNA strand.     

Crystal and molecular structure of a DNA fragment: d(CGTGAATTCACG)

The crystal structure of the dodecanucleotide d(CGTGAATTCACG) has been determined to a resolution of 2.7 A and refined to an R factor of 17.0% for 1532 reflections. The sequence crystallizes as a B-form double helix, with Watson-Crick base pairing. This sequence contains the EcoRI restriction endonuclease recognition site, GAATTC, and is flanked by CGT on the 5'-end and ACG on the 3'-end, in contrast to the CGC on the 5'-end and GCG on the 3'-end in the parent dodecamer d(CGCGAATTCGCG). A comparison with the isomorphous parent compound shows that any changes in the structure induced by the change in the sequence in the flanking region are highly localized. The global conformation of the duplex is conserved. The overall bend in the helix is 10 degrees. The average helical twist values for the present and the parent structures are 36.5 degrees and 36.4 degrees, respectively, corresponding to 10 base pairs per turn. The buckle at the substituted sites are significantly different from those seen at the corresponding positions in the parent dodecamer. Step 2 (GpT) is underwound with respect to the parent structure (27 degrees vs 36 degrees) and step 3 (TpG) is overwound (34 degrees vs 27 degrees). There is a spine of hydration in the narrow minor groove. The N3 atom of adenine on the substituted A10 and A22 bases are involved in the formation of hydrogen bonds with other duplexes or with water; the N3 atom of guanine on G10 and G22 bases in the parent structure does not form hydrogen bonds.     

The structure of human cytochrome P450 2C9 complexed with flurbiprofen at 2.0-A resolution

The structure of human P450 2C9 complexed with flurbiprofen was determined to 2.0 A by x-ray crystallography. In contrast to other structurally characterized P450 2C enzymes, 2C5, 2C8, and a 2C9 chimera, the native catalytic domain of P450 2C9 differs significantly in the conformation of the helix F to helix G region and exhibits an extra turn at the N terminus of helix A. In addition, a distinct conformation of the helix B to helix C region allows Arg-108 to hydrogen bond with Asp-293 and Asn-289 on helix I and to interact directly with the carboxylate of flurbiprofen. These interactions position the substrate for regioselective oxidation in a relatively large active site cavity and are likely to account for the high catalytic efficiency exhibited by P450 2C9 for the regioselective oxidation of several anionic non-steroidal anti-inflammatory drugs. The structure provides a basis for interpretation of a number of observations regarding the substrate selectivity of P450 2C9 and the observed effects of mutations on catalysis.     

Anthracycline binding to DNA. High-resolution structure of d(TGTACA) complexed with 4'-epiadriamycin.

Crystallographic methods have been applied to determine the high-resolution structure of the complex formed between the self-complementary oligonucleotide d(TGTACA) and the anthracycline antibiotic 4'-epiadriamycin. The complex crystallises in the tetragonal system, space group P4(1)2(1)2 with a = 2.802 nm and c = 5.293 nm, and an asymmetric unit consisting of a single DNA strand, one drug molecule and 34 solvent molecules. The refinement converged with an R factor of 0.17 for the 2381 reflections with F greater than or equal to 3 sigma F in the resolution range 0.70-0.14 nm. Two asymmetric units associate such that a distorted B-DNA-type hexanucleotide duplex is formed incorporating two drug molecules that are intercalated at the TpG steps. The amino sugar of 4'-epiadriamycin binds in the minor groove of the duplex and displays different interactions from those observed in previously determined structures. Interactions between the hydrophilic groups of the amino sugar and the oligonucleotide are all mediated by solvent molecules. Ultraviolet melting measurements and comparison with other anthracycline-DNA complexes suggest that these indirect interactions have a powerful stabilising effect on the complex.     

Crystal structure of a human tRNA(Gly) microhelix at 1.2A resolution

The tRNA^Gly/glycyl-tRNA synthetase (GlyRS) system belongs to the so-called ‘class II aminoacyl-tRNA synthetase system’ in which tRNA identity elements are assured by rather few and simple determinants mostly located in the tRNA acceptor stem. Regarding evolutionary aspects, the tRNA^Gly/GlyRS system is a special case. There exist two different types of GlyRS, namely an archaebacterial/human type and a eubacterial type reflecting an evolutionary divergence within this system.Here we report the crystal structure of a human tRNA^Gly acceptor stem microhelix at 1.2 Å resolution. The local geometric parameters of the microhelix and the water network surrounding the RNA are presented. The structure complements the previously published Escherichia coli tRNA^Gly aminoacyl stem structure.     

The molecular structure of the left-handed Z-DNA double helix at 1.0-A atomic resolution. Geometry, conformation, and ionic interactions of d(CGCGCG)

The structure of d(CGCGCG) crystallized in the presence of magnesium and sodium ions alone is compared to that of the spermine form of the molecule. The very high resolution nature of these structure determinations allows the first true examination of an oligonucleotide structure in fine detail. The values of bond distances and angles are compared to those derived from small molecule crystal structures. In addition, the interactions of cations and polyamines with the Z-DNA helix are analyzed. In particular, multiple cationic charges appear to offer enhanced stabilization for the Z-DNA conformation. The location of spermine molecules along the edge of the deep groove and also spanning the entrance to the groove emphasizes the importance of polyamines for stabilizing this left-handed structure. On averaging, we obtained very similar structural parameters for the two different structures with standard deviations generally smaller than the deviations of the crystallographic model from ideal values. This indicates a high degree of accuracy of the two structures, which have been refined using different data and different refinement methods. The derived bond lengths and angles may thus be more representative of this polymeric DNA structure than those derived from mono- and dinucleotide structures at a similar accuracy.     

Crystal and molecular structure of human plasminogen kringle 4 refined at 1.9-A resolution.

The crystal structure of human plasminogen kringle 4 (PGK4) has been solved by molecular replacement using the bovine prothrombin kringle 1 (PTK1) structure as a model and refined by restrained least-squares methods to an R factor of 14.2% at 1.9-A resolution. The K4 structure is similar to that of PTK1, and an insertion of one residue at position 59 of the latter has minimal effect on the protein folding. The PGK4 structure is highly stabilized by an internal hydrophobic core and an extensive hydrogen-bonding network. Features new to this kringle include a cis peptide bond at Pro30 and the presence of two alternate, perpendicular, and equally occupied orientations for the Cys75 side chain. The K4 lysine-binding site consists of a hydrophobic trough formed by the Trp62 and Trp72 indole rings, with anionic (Asp55/Asp57) and cationic (Lys35/Arg71) charge pairs at either end. With the adjacent Asp5 and Arg32 residues, these result in triply charged anionic and cationic clusters (pH of crystals at 6.0), which, in addition to the unusually high accessibility of the Trp72 side chain, serve as an obvious marker of the binding site on the K4 surface. A complex intermolecular interaction occurs between the binding sites of symmetry-related molecules involving a highly ordered sulfate anion of solvation in which the Arg32 side chain of a neighboring kringle occupies the binding site.     

The molecular structure of a 4'-epiadriamycin complex with d(TGATCA) at 1.7A resolution: comparison with the structure of 4'-epiadriamycin d(TGTACA) and d(CGATCG) complexes.

The structure of the complex between d(TGATCA) and the anthracycline 4'-epiadriamycin has been determined by crystallographic methods. The crystals are tetragonal, space group P4(1)2(1)2 with unit cell dimensions of a = 28.01, c = 52.95A. The asymmetric unit consists of one strand of hexanucleotide, one molecule of 4'-epiadriamycin and 34 waters. The R-factor is 20.2% for 1694 reflections with F greater than or equal to 2 sigma F to 1.7A. Two asymmetric units associate to generate a duplex complexed with two drug molecules at the d(TpG) steps of the duplex. The chromophore intercalates between these base pairs with the anthracycline amino-sugar positioned in the minor groove. The double helix is a distorted B-DNA type structure. Our structure determination of d(TGATCA) complexed to 4'-epiadriamycin allows for comparison with the previously reported structures of 4'-epiadriamycin bound to d(TGTACA) and to d(CGATCG). The three complexes are similar in gross features and the intercalation geometry is the same irrespective of whether a d(CpG) or d(TpG) sequence is involved. However, the orientation of the amino-sugar displays a dependence on the sequence adjacent to the intercalation site. The flexibility of this amino-sugar may help explain why this class of antibiotics displays a relative insensitivity to base sequence when they bind to DNA.     

Covalent modification of guanine bases in double-stranded DNA. The 1.2-A Z-DNA structure of d(CGCGCG) in the presence of CuCl2

We have solved the single crystal structure to 1.2-A resolution of the Z-DNA sequence d(CGCGCG) soaked with copper(II) chloride. This structure allows us to elucidate the structural properties of copper in a model that mimics a physiologically relevant environment. A copper(II) cation was observed to form a covalent coordinate bond to N-7 of each guanine base along the hexamer duplex. The occurrence of copper bound at each site was dependent on the exposure of the bases and the packing of the hexamers in the crystal. The copper at the highest occupied site was observed to form a regular octahedral complex, with four water ligands in the equatorial plane and a fifth water along with N-7 of the purine base at the axial positions. All other copper complexes appear to be variations of this structure. By using the octahedral complex as the prototype for copper(II) binding to guanine bases in the Z-DNA crystal, model structures were built showing that duplex B-DNA can accommodate octahedral copper(II) complexes at the guanine bases as well as copper complexes bridged at adjacent guanine residues by a reactive dioxygen species. The increased susceptibility to oxidative DNA cleavage induced by copper(II) ions in solution of the bases located 5' to one or more adjacent guanine residues can thus be explained in terms of the cation and DNA structures described by these models.     

Crystal structure of AlgQ2, a macromolecule (alginate)-binding protein of Sphingomonas sp. A1, complexed with an alginate tetrasaccharide at 1.6-A resolution

Sphingomonas sp. A1 possesses a high molecular weight (HMW) alginate uptake system composed of a novel pit formed on the cell surface and a pit-dependent ATP-binding cassette (ABC) transporter in the inner membrane. The transportation of HMW alginate from the pit to the ABC transporter is mediated by the periplasmic HMW alginate-binding proteins AlgQ1 and AlgQ2. We determined the crystal structure of AlgQ2 complexed with an alginate tetrasaccharide using an alginate-free (apo) form as a search model and refined it at 1.6-A resolution. One tetrasaccharide was found between the N and C-terminal domains, which are connected by three extended hinge loops. The tetrasaccharide complex took on a closed domain form, in contrast to the open domain form of the apo form. The tetrasaccharide was bound in the cleft between the domains through van der Waals interactions and the formation of hydrogen bonds. Among the four sugar residues, the nonreducing end residue was located at the bottom of the cleft and exhibited the largest number of interactions with the surrounding amino acid residues, suggesting that AlgQ2 mainly recognizes and binds to the nonreducing part of a HMW alginate and delivers the polymer to the ABC transporter through conformational changes (open and closed forms) of the two domains.     

The structure of human lymphotoxin (tumor necrosis factor-beta) at 1.9-A resolution.

The three-dimensional structure of recombinant human lymphotoxin (residues 24-171 of the mature protein) has been determined by x-ray crystallography at 1.9-A resolution (Rcryst = 0.215 for I greater than 3 sigma (I)). Phases were derived by molecular replacement using tumor necrosis factor (TNF-alpha) as a search model. Like TNF-alpha, lymphotoxin (LT) folds to form a "jellyroll" beta-sheet sandwich. Three-fold related LT subunits form a trimer stabilized primarily by hydrophobic interactions. A cluster of 6 basic residues around the 3-fold axis may account for the acid lability of the trimer. Although the structural cores of TNF-alpha and LT are similar, insertions and deletions relative to TNF-alpha occur in loops at the "top" of the LT trimer and significantly alter the local structure and the overall shape trimer is highly conserved. The sites of two mutations (Asp-50 and Tyr-108) that abolish the cytotoxicity of LT are contained within poorly ordered loops of polypeptide chain that flank the cleft between neighboring subunits at the base of the molecule, suggesting that the receptor recognizes an intersubunit binding site.