Protein modifications giving rise to homo-oligomers.

After the structures of numerous proteins have been established at the atomic level and after a multitude of proteins can be produced with almost no restrictions, the time seems ripe to apply this knowledge for engineering purposes. An apparently simple task is the designed association of protein molecules to form homo-oligomers. A number of worked examples are presented. The associations split into flexible versus rigid designs and also into fixed versus switchable ones. It is shown that the practical work is tightly governed by the multiplicity concept, which in turn is interwoven with symmetry. The available symmetries and multiplicities are explained. Unfortunately, the most desirable contacts with a multiplicity of one, which lead to asymmetric assemblies with 5-50 nm spacings, are most difficult to achieve. Emerging rules for the required surface properties are put forward. Suitable mutations for changing such surfaces are discussed.

Validation of the detergent micelle classification for membrane protein crystals and explanation of the Matthews Graph for soluble proteins.

Protein crystals are of wide-spread interest because many of them allow structure analyses at atomic resolution. For soluble proteins, the packing density of such crystals is distributed according to the Matthews Graph. For integral membrane proteins, the respective graph is similar but at lower density and much broader. By visualizing the relative positions and orientations of membrane proteins in crystals, it has been suggested that the detergent micelles surrounding these proteins form sheets, filaments, or remain isolated in the crystal giving rise to three distinct packing density distributions that superimpose to form the observed broad distribution. This classification was indirect because detergent is not visible in X-ray crystallography. Given the extensive work involved in analyzing detergent structure directly by neutron diffraction, it seems unlikely that a statistically relevant number of them will be established in the near future. Therefore, the proposed classification is here scrutinized by a simulation in which an average detergent-carrying membrane protein was randomly packed to form crystals. The analysis reproduced the three types of detergent structures together with their packing density distributions and relative frequencies, which validates the previous classification. The simulation program was also run for crystals from soluble proteins using ellipsoids as reference shapes and defining a shape factor that quantifies the deviation from the nearest ellipsoid. This series reproduced and thus explained the Matthews Graph.

A new classification of membrane protein crystals.

Although being much smaller than the number of soluble proteins in the Protein Data Bank, the number of membrane proteins therein now approaches 700, and a statistical analysis becomes meaningful. Such an analysis showed that the conventional subdivision into monotopic, ß-barrel and α-helical membrane proteins is appropriate but should be amended by a classification according to the detergent micelle structure in the crystal, which can be derived from the packing of the membrane-immersed parts of the proteins. The crystal packing density is specific for the three conventional types of membrane proteins and soluble proteins. It is also specific for three observed detergent arrangements that are micelle pockets, micelle filaments and micelle sheets, demonstrating that the detergent structure affects crystallization. The packing density distribution of crystals from integral membrane proteins has approximately the same shape as that of soluble proteins but is by a factor of two broader and shifted to lower density. It seems unlikely that the differences can be explained by a mere solvent expansion due to the required detergent. The crystallized membrane proteins were further analyzed with respect to protein mass, oligomerization and crystallographic asymmetric unit, space group, crystal ordering and symmetry. The results provide a new view on membrane proteins.

The dominance of symmetry in the evolution of homo-oligomeric proteins.

The fact that aggregates of identical protein molecules are usually symmetric has remained an enigma. An idealized model of a soluble monomeric protein was constructed and accompanied through a simulated evolutionary process resulting in dimerization, in order to elucidate this peculiarity. The model showed that the probability of a symmetric association is by a factor of 100 or above higher than the probability of an asymmetric one. Unexpectedly, symmetry prevails in the dimer initiation phase much more than in the dimer improvement phase of evolution. The result is clear-cut and robust against a broad spectrum of model inadequacies. It rationalizes the predominance of symmetric homo-oligomers.

Structure and action of the myxobacterial chondrochloren halogenase CndH: a new variant of FAD-dependent halogenases.

The crystal structure of the FAD-dependent chondrochloren halogenase CndH has been established at 2.1 A resolution. The enzyme contains the characteristic FAD-binding scaffold of the glutathione reductase superfamily. Except for its C-terminal domain, the chainfold of CndH is virtually identical with those of FAD-dependent aromatic hydroxylases. When compared to the structurally known FAD-dependent halogenases PrnA and RebH, CndH lacks a 45 residue segment near position 100 and deviates in the C-terminal domain. Both variations are near the active center and appear to reflect substrate differences. Whereas PrnA and RebH modify free tryptophan, CndH halogenates the tyrosyl group of a chondrochloren precursor that is most likely bound to a carrier protein. In contrast to PrnA and RebH, which enclose their small substrate completely, CndH has a large non-polar surface patch that may accommodate the putative carrier. Apart from the substrate binding site, the active center of CndH corresponds to those of PrnA and RebH. At the halogenation site, CndH has the characteristic lysine (Lys76) but lacks the required base Glu346 (PrnA). This base may be supplied by a residue of its C-terminal domain or by the carrier. These differences were corroborated by an overall sequence comparison between the known FAD-dependent halogenases, which revealed a split into a PrnA-RebH group and a CndH group. The two functionally established members of the CndH group use carrier-bound substrates, whereas three members of PrnA-RebH group are known to accept a free amino acid. Given the structural and functional distinction, we classify CndH as a new variant B of the FAD-dependent halogenases, adding a new feature to the structurally established variant A enzymes PrnA and RebH.

Structural base for enzymatic cyclodextrin hydrolysis.

Cyclodextrins resist hydrolysis by burying all bridge oxygens at their interior. Still, the rings can be opened by a small group of specialized enzymes, the cyclomaltodextrinases. Among them, the enzyme from Flavobacterium sp. no. 92 was mutated, crystallized and soaked with cyclodextrins, giving rise to four complex structures. One of them showed an alpha-cyclodextrin at the outer rim of the active center pocket. In the other complexes, alpha-, beta-and gamma-cyclodextrins were bound in a competent mode in the active center. The structures suggest that Arg464 functions as a chaperone guiding the substrates from the solvent into the active center. Over the last part of this pathway, the cyclodextrins bump on Phe274, which rotates the glucosyl group at subsite (+1) by about 120 degrees and fixes it in the new conformation. This induced fit was observed with all three major cyclodextrins. It makes the bridging oxygen between subsites (+1) and (-1) available for protonation by Glu340, which starts the hydrolysis. The mechanism resembles a spring-lock. The structural data were supplemented by activity measurements, quantifying the initial ring opening reaction for the major cyclodextrins and the transglucosylation activity for maltotetraose. Further activity data were collected for mutants splitting the tetrameric enzyme into dimers and for active center mutants.

Structure and mode of action of a mosquitocidal holotoxin.

The crystal structure of the full mosquitocidal toxin from Bacillus sphaericus (MTX(holo)) has been determined at 2.5 A resolution by the molecular replacement method. The resulting structure revealed essentially the complete chain consisting of four ricin B-type domains curling around the catalytic domain in a hedgehog-like assembly. As the structure was virtually identical in three different crystal packings, it is probably not affected by packing contacts. The structure of MTX(holo) explains earlier autoinhibition data. An analysis of published complexes comprising ricin B-type lectin domains and sugar molecules shows that the general construction principle applies to all four lectin domains of MTX(holo), indicating 12 putative sugar-binding sites. These sites are sequence-related to those of the cytotoxin pierisin from cabbage butterfly, which are known to bind glycolipids. It seems therefore likely that MTX(holo) also binds glycolipids. The seven contact interfaces between the five domains are predominantly polar and not stronger than common crystal contacts so that in an appropriate environment, the multidomain structure would likely uncurl into a string of single domains. The structure of the isolated catalytic domain plus an extended linker was established earlier in three crystal packings, two of which showed a peculiar association around a 7-fold axis. The catalytic domain of the reported MTX(holo) closely resembles all three published structures, except one with an appreciable deviation of the 40 N-terminal residues. A comparison of all structures suggests a possible scenario for the translocation of the toxin into the cytosol.

Inhibition of the glucosyltransferase activity of clostridial Rho/Ras-glucosylating toxins by castanospermine.

Castanospermine was identified as an inhibitor of the Rho/Ras-glucosylating Clostridium sordellii lethal toxin and Clostridium difficile toxin B. Microinjection of castanospermine into embryonic bovine lung cells prevented the cytotoxic effects of toxins. The crystal structure of the glucosyltransferase domain of C. sordellii lethal toxin in complex with castanospermine, UDP and a calcium ion was solved at a resolution of 2.3A. The inhibitor binds in a conformation that brings its four hydroxyl groups and its N-atom almost exactly in the positions of the four hydroxyls and of the ring oxygen of the glucosyl moiety of UDP-glucose, respectively.

A putative alpha-helical porin from Corynebacterium glutamicum.

The cell wall of Corynebacterium glutamicum contains a mycolic acid layer, which is a protective nonpolar barrier similar to the outer membrane of Gram-negative bacteria. The exchange of material across this barrier requires porins. Porin B (PorB) is one of them. Recombinant PorB has been produced in Escherichia coli, purified, crystallized and analyzed by X-ray diffraction, yielding 16 independent molecular structures in four different crystal forms at resolutions up to 1.8 A. All 16 molecules have the same globular core, which consists of 70 residues forming four alpha-helices tied together by a disulfide bridge. The 16 structures vary greatly with respect to the 29 residues in the N- and C-terminal extensions. Since corynebacteria belong to the group of mycolata that includes some prominent human pathogens, the observed structure may be of medical relevance. Due to the clearly established solid structure of the core, the native porin has to be oligomeric, and the reported structure is one of the subunits. An alpha-helical porin in a bacterial outer envelope is surprising because all presently known structures of such porins consist of beta-barrels. Since none of the four crystal packing arrangements was compatible with an oligomeric membrane channel, we constructed a model of such an oligomer that was consistent with all available data of native PorB. The proposed model is based on the required polar interior and nonpolar exterior of the porin, on a recurring crystal packing contact around a 2-fold axis, on the assumption of a simple C(n) symmetry (a symmetric arrangement around an n-fold axis), on the experimentally established electric conductivity and anion selectivity and on the generally observed shape of porin channels.

Structure of 2,6-dihydroxypyridine 3-hydroxylase from a nicotine-degrading pathway.

The enzyme 2,6-dihydroxypyridine-3-hydroxylase catalyzes the sixth step of the nicotine degradation pathway in Arthrobacter nicotinovorans. The enzyme was produced in Escherichia coli, purified and crystallized. The crystal structure was solved at 2.6 A resolution, revealing a significant structural relationship with the family of FAD-dependent aromatic hydroxylases, but essentially no sequence homology. The structure was aligned with those of the established family members, showing that the FAD molecules are bound at virtually identical locations. The reported enzyme is a dimer like most other family members, but its dimerization contact differs from the others. The binding position of NAD(P)H to this enzyme family is not clear. Since the reported enzyme accepts only NADH for flavin reduction in contrast to the other established members using NADPH, we searched through the structural alignment and found an indication for the position of the 2'-phosphate of NADPH that is in general agreement with mutational studies on a related enzyme, but contradicts a crystal soaking experiment. Using a bound glycerol molecule and the known substrate positions of three related enzymes as a guide, the substrate 2,6-dihydroxypyridine was placed into the active center. The access to the binding site is discussed. The new active center geometry introduces constraints that render some reaction scenarios more likely than others. It suggests that flavin is reduced at its out-position and then drawn into its in-position, where it binds molecular oxygen. The geometry is consistent with the proposal that peroxy-flavin is protonated by the solvent to yield the electrophilic hydroperoxy-flavin. The substrate is activated by two buried histidines but there is no appropriate base to store the surplus proton of the hydroxylated carbon atom. The implications of this problem are discussed.

Conformational changes and reaction of clostridial glycosylating toxins.

The crystal structures of the catalytic fragments of 'lethal toxin' from Clostridium sordellii and of 'alpha-toxin' from Clostridium novyi have been established. Almost half of the residues follow the chain fold of the glycosyl-transferase type A family of enzymes; the other half forms large alpha-helical protrusions that are likely to confer specificity for the respective targeted subgroup of Rho proteins in the cell. In the crystal, the active center of alpha-toxin contained no substrates and was disassembled, whereas that of lethal toxin, which was ligated with the donor substrate UDP-glucose and cofactor Mn2+, was catalytically competent. Surprisingly, the structure of lethal toxin with Ca2+ (instead of Mn2+) at the cofactor position showed a bound donor substrate with a disassembled active center, indicating that the strictly octahedral coordination sphere of Mn2+ is indispensable to the integrity of the enzyme. The homologous structures of alpha-toxin without substrate, distorted lethal toxin with Ca2+ plus donor, active lethal toxin with Mn2+ plus donor and the homologous Clostridium difficile toxin B with a hydrolyzed donor have been lined up to show the geometry of several reaction steps. Interestingly, the structural refinement of one of the three crystallographically independent molecules of Ca2+-ligated lethal toxin resulted in the glucosyl half-chair conformation expected for glycosyl-transferases that retain the anomeric configuration at the C1'' atom. A superposition of six acceptor substrates bound to homologous enzymes yielded the position of the nucleophilic acceptor atom with a deviation of <1 A. The resulting donor-acceptor geometry suggests that the reaction runs as a circular electron transfer in a six-membered ring, which involves the deprotonation of the nucleophile by the beta-phosphoryl group of the donor substrate UDP-glucose.

Designed protein-protein association.

The analysis of natural contact interfaces between protein subunits and between proteins has disclosed some general rules governing their association. We have applied these rules to produce a number of novel assemblies, demonstrating that a given protein can be engineered to form contacts at various points of its surface. Symmetry plays an important role because it defines the multiplicity of a designed contact and therefore the number of required mutations. Some of the proteins needed only a single side-chain alteration in order to associate to a higher-order complex. The mobility of the buried side chains has to be taken into account. Four assemblies have been structurally elucidated. Comparisons between the designed contacts and the results will provide useful guidelines for the development of future architectures.

Antenna domain mobility and enzymatic reaction of L-rhamnulose-1-phosphate aldolase.

The enzyme l-rhamnulose-1-phosphate aldolase from Escherichia coli participates in the degradation pathway of l-rhamnose, a ubiquitous deoxy-hexose. It is a homotetramer of the rare C4-symmetric type with N-terminal domains protruding like antennas from the main body. A mobility analysis of the enzyme gave rise to the hypothesis that an anisotropic thermal antenna motion may support the catalysis (Kroemer et al., Biochemistry 42, 10560, 2003). We checked this hypothesis by generating four single mutants and one disulfide bridge that were designed to reduce the mobility of the antenna domain without disturbing the chain-fold or the active center. The catalytic rates of the mutants revealed activity reductions that correlated well with the expected antenna fixation. Among these mutants, K15W was crystallized, structurally elucidated, and used as a guide for modeling the others. The structure confirmed the design because the mutation introduced a tight nonpolar contact to a neighboring subunit that fixed the antenna but did not affect the main chain. The fixation was confirmed by a comparison of the anisotropic B-factors describing the mobility of the domains. It turned out that the distinctly anisotropic mobility of the wild-type antenna domain has become isotropic in K15W, in agreement with the design. We suggest that, like K15W, the other mutations also followed the design, validating the correlation between antenna mobility and activity. This correlation suggests that the domain mobility facilitates the reaction.

Structure and action of the N-oxygenase AurF from Streptomyces thioluteus.

Nitro groups are found in a number of bioactive compounds. Most of them arise by a stepwise mono-oxygenation of amino groups. One of the involved enzymes is AurF participating in the biosynthesis of aureothin. Its structure was established at 2.1 A resolution showing a homodimer with a binuclear manganese cluster. The enzyme preparation, which yielded the analyzed crystals, showed activity using in vitro and in vivo assays. Chain fold and cluster are homologous with ribonucleotide reductase subunit R2 and related enzymes. The two manganese ions and an iron content of about 15% were established by anomalous X-ray diffraction. A comparison of the cluster with more common di-iron clusters suggested an additional histidine in the coordination sphere to cause the preference for manganese over iron. There is no oxo-bridge. The substrate p-amino-benzoate was modeled into the active center. The model is supported by mutant activity measurements. It shows the geometry of the reaction and explains the established substrate spectrum.

High resolution structure and catalysis of O-acetylserine sulfhydrylase isozyme B from Escherichia coli.

The crystal structure of the dimeric O-acetylserine sulfhydrylase isozyme B from Escherichia coli (CysM), complexed with the substrate analog citrate, has been determined at 1.33 A resolution by X-ray diffraction analysis. The C1-carboxylate of citrate was bound at the carboxylate position of O-acetylserine, whereas the C6-carboxylate adopted two conformations. The activity of the enzyme and of several active center mutants was determined using an assay based on O-acetylserine and thio-nitrobenzoate (TNB). The unnatural substrate TNB was modeled into the reported structure. The substrate model and the observed mutant activities may facilitate future protein engineering attempts designed to broaden the substrate spectrum of the enzyme. A comparison of the reported structure with previously published CysM structures revealed large conformational changes. One of the crystal forms contained two dimers, each of which comprised one subunit in a closed and one in an open conformation. Although the homodimer asymmetry was most probably caused by crystal packing, it indicates that the enzyme can adopt such a state in solution, which may be relevant for the catalytic reaction.

Structure and action of the C-C bond-forming glycosyltransferase UrdGT2 involved in the biosynthesis of the antibiotic urdamycin.

The glycosyltransferase UrdGT2 from Streptomyces fradiae catalyzes the formation of a glycosidic C-C bond between a polyketide aglycone and D-olivose. The enyzme was expressed in Escherichia coli, purified and crystallized. Its structure was established by X-ray diffraction at 1.9 A resolution. It is the first structure of a C-glycosyltransferase. UrdGT2 belongs to the structural family GT-B of the glycosyltransferases and is likely to form a C(2)-symmetric dimer in solution. The binding structures of donor and acceptor substrates in five structurally homologous enzymes provided a clear and consistent guide for the substrate-binding structure in UrdGT2. The modeled substrate locations suggest the deeply buried Asp137 as the activator for C-C bond formation and explain the reaction. The putative model can be used to design mutations that change the substrate specificity. Such mutants are of great interest in overcoming the increasing danger of antibiotic resistance.

Substrate spectrum of L-rhamnulose kinase related to models derived from two ternary complex structures.

The enzyme L-rhamnulose kinase from Escherichia coli participates in the degradation pathway of L-rhamnose, a common natural deoxy-hexose. The structure of the enzyme in a ternary complex with its substrates ADP and L-rhamnulose has been determined at 1.55A resolution and refined to R(cryst)/R(free) values of 0.179/0.209. The result was compared with the lower resolution structure of a corresponding complex containing L-fructose instead of L-rhamnulose. In light of the two established sugar positions and conformations, a number of rare sugars have been modeled into the active center of L-rhamnulose kinase and the model structures have been compared with the known enzymatic phosphorylation rates. Rare sugars are of rising interest for the synthesis of bioactive compounds.

Structure and action of a C-C bond cleaving alpha/beta-hydrolase involved in nicotine degradation.

The enzyme 2,6-dihydroxy-pseudo-oxynicotine hydrolase from the nicotine-degradation pathway of Arthrobacter nicotinovorans was crystallized and the structure was determined by an X-ray diffraction analysis at 2.1 A resolution. The enzyme belongs to the alpha/beta-hydrolase family as derived from the chain-fold and from the presence of a catalytic triad with its oxyanion hole at the common position. This relationship assigns a pocket lined by the catalytic triad as the active center. The asymmetric unit contains two C(2)-symmetric dimer molecules, each adopting a specific conformation. One dimer forms a more spacious active center pocket and the other a smaller one, suggesting an induced-fit. All of the currently established C-C bond cleaving alpha/beta-hydrolases are from bacterial meta-cleavage pathways for the degradation of aromatic compounds and cover their active center with a 40 residue lid placed between two adjacent strands of the beta-sheet. In contrast, the reported enzyme shields its active center with a 110 residue N-terminal domain, which is absent in the meta-cleavage hydrolases. Since neither the substrate nor an analogue could be bound in the crystals, the substrate was modeled into the active center using the oxyanion hole as a geometric constraint. The model was supported by enzymatic activity data of 11 point mutants and by the two dimer conformations suggesting an induced-fit. Moreover, the model assigned a major role for the large N-terminal domain that is specific to the reported enzyme. The proposal is consistent with the known data for the meta-cleavage hydrolases although it differs in that the reaction does not release alkenes but a hetero-aromatic compound in a retro-Friedel-Crafts acylation. Because the hydrolytic water molecule can be assigned to a geometrically suitable site that can be occupied in the presence of the substrate, the catalytic triad may not form a covalent acyl-enzyme intermediate but merely support a direct hydrolysis.

Structure and reaction geometry of geranylgeranyl diphosphate synthase from Sinapis alba.

The crystal structure of the geranylgeranyl diphosphate synthase from Sinapis alba (mustard) has been solved in two crystal forms at 1.8 and 2.0 A resolutions. In one of these forms, the dimeric enzyme binds one molecule of the final product geranylgeranyl diphosphate in one subunit. The chainfold of the enzyme corresponds to that of other members of the farnesyl diphosphate synthase family. Whereas the binding modes of the two substrates dimethylallyl diphosphate and isopentenyl diphosphate at the allyl and isopentenyl sites, respectively, have been established with other members of the family, the complex structure presented reveals for the first time the binding mode of a reaction product at the isopentenyl site. The binding geometry of substrates and product in conjunction with the protein environment and the established chemistry of the reaction provide a clear picture of the reaction steps and atom displacements. Moreover, a comparison with a ligated homologous structure outlined an appreciable induced fit: helix alpha8 and its environment undergo a large conformational change when either the substrate dimethylallyl diphosphate or an analogue is bound to the allyl site; only a minor conformational change occurs when the other substrate isopentenyl diphosphate or the product is bound to the isopentenyl site.