Crystal Structure and Computational Analyses Provide Insights into the Catalytic Mechanism of 2,4-Diacetylphloroglucinol Hydrolase PhlG from Pseudomonas fluorescens

2,4-Diacetylphloroglucinol hydrolase PhlG from Pseudomonas fluorescens catalyzes hydrolytic carbon-carbon (C–C) bond cleavage of the antibiotic 2,4-diacetylphloroglucinol to form monoacetylphloroglucinol, a rare class of reactions in chemistry and biochemistry. To investigate the catalytic mechanism of this enzyme, we determined the three-dimensional structure of PhlG at 2.0 Å resolution using x-ray crystallography and MAD methods. The overall structure includes a small N-terminal domain mainly involved in dimerization and a C-terminal domain of Bet v1-like fold, which distinguishes PhlG from the classical α/β-fold hydrolases. A dumbbell-shaped substrate access tunnel was identified to connect a narrow interior amphiphilic pocket to the exterior solvent. The tunnel is likely to undergo a significant conformational change upon substrate binding to the active site. Structural analysis coupled with computational docking studies, site-directed mutagenesis, and enzyme activity analysis revealed that cleavage of the 2,4-diacetylphloroglucinol C–C bond proceeds via nucleophilic attack by a water molecule, which is coordinated by a zinc ion. In addition, residues Tyr¹²¹, Tyr²²⁹, and Asn¹³², which are predicted to be hydrogen-bonded to the hydroxyl groups and unhydrolyzed acetyl group, can finely tune and position the bound substrate in a reactive orientation. Taken together, these results revealed the active sites and zinc-dependent hydrolytic mechanism of PhlG and explained its substrate specificity as well.     

Structural and biochemical characterization of the salicylyl-acyltranferase SsfX3 from a tetracycline biosynthetic pathway

SsfX3 is a GDSL family acyltransferase that transfers salicylate to the C-4 hydroxyl of a tetracycline intermediate in the penultimate step during biosynthesis of the anticancer natural product SF2575. The C-4 salicylate takes the place of the more common C-4 dimethylamine functionality, making SsfX3 the first acyltransferase identified to act on a tetracycline substrate. The crystal structure of SsfX3 was determined at 2.5 Å, revealing two distinct domains as follows: an N-terminal β-sandwich domain that resembles a carbohydrate-binding module, and a C-terminal catalytic domain that contains the atypical α/β-hydrolase fold found in the GDSL hydrolase family of enzymes. The active site lies at one end of a large open binding pocket, which is spatially defined by structural elements from both the N- and C-terminal domains. Mutational analysis in the putative substrate binding pocket identified residues from both domains that are important for binding the acyl donor and acceptor. Furthermore, removal of the N-terminal carbohydrate-binding module-like domain rendered the stand-alone α/β-hydrolase domain inactive. The additional noncatalytic module is therefore proposed to be required to define the binding pocket and provide sufficient interactions with the spatially extended tetracyclic substrate. SsfX3 was also demonstrated to accept a variety of non-native acyl groups. This relaxed substrate specificity toward the acyl donor allowed the chemoenzymatic biosynthesis of C-4-modified analogs of the immediate precursor to the bioactive SF2575; these were used to assay the structure activity relationships at the C-4 position.     

The Crystal Structure of the Adenylation Enzyme VinN Reveals a Unique β-Amino Acid Recognition Mechanism

Adenylation enzymes play important roles in the biosynthesis and degradation of primary and secondary metabolites. Mechanistic insights into the recognition of α-amino acid substrates have been obtained for α-amino acid adenylation enzymes. The Asp residue is invariant and is essential for the stabilization of the α-amino group of the substrate. In contrast, the β-amino acid recognition mechanism of adenylation enzymes is still unclear despite the importance of β-amino acid activation for the biosynthesis of various natural products. Herein, we report the crystal structure of the stand-alone adenylation enzyme VinN, which specifically activates (2S,3S)-3-methylaspartate (3-MeAsp) in vicenistatin biosynthesis. VinN has an overall structure similar to that of other adenylation enzymes. The structure of the complex with 3-MeAsp revealed that a conserved Asp²³⁰ residue is used in the recognition of the β-amino group of 3-MeAsp similar to α-amino acid adenylation enzymes. A mutational analysis and structural comparison with α-amino acid adenylation enzymes showed that the substrate-binding pocket of VinN has a unique architecture to accommodate 3-MeAsp as a β-amino acid substrate. Thus, the VinN structure allows the first visualization of the interaction of an adenylation enzyme with a β-amino acid and provides new mechanistic insights into the selective recognition of β-amino acids in this family of enzymes.     

Characterization of Streptococcus pyogenes ��-NAD+ Glycohydrolase: RE-EVALUATION OF ENZYMATIC PROPERTIES ASSOCIATED WITH PATHOGENESIS*

The Gram-positive pathogen Streptococcus pyogenes injects a β-NAD⁺ glycohydrolase (SPN) into the cytosol of an infected host cell using the cytolysin-mediated translocation pathway. In this compartment, SPN accelerates the death of the host cell by an unknown mechanism that may involve its β-NAD⁺-dependent enzyme activities. SPN has been reported to possess the unique characteristic of not only catalyzing hydrolysis of β-NAD⁺, but also carrying out ADP-ribosyl cyclase and ADP-ribosyltransferase activities, making SPN the only β-NAD⁺ glycohydrolase that can catalyze all of these reactions. With the long term goal of understanding how these activities may contribute to pathogenesis, we have further characterized the enzymatic activity of SPN using highly purified recombinant protein. Kinetic studies of the multiple activities of SPN revealed that SPN possessed only β-NAD⁺ hydrolytic activity and lacked detectable ADP-ribosyl cyclase and ADP-ribosyltransferase activities. Similarly, SPN was unable to catalyze cyclic ADPR hydrolysis, and could not catalyze methanolysis or transglycosidation. Kinetic analysis of product inhibition by recombinant SPN demonstrated an ordered uni-bi mechanism, with ADP-ribose being released as a second product. SPN was unaffected by product inhibition using nicotinamide, suggesting that this moiety contributes little to the binding energy of the substrate. Upon transformation, SPN was toxic to Saccharomyces cerevisiae, whereas a glycohydrolase-inactive SPN allowed for viability. Taken together, these data suggest that SPN functions exclusively as a strict β-NAD⁺ glycohydrolase during pathogenesis.     

Crystal structure and inhibition studies of transglutaminase from Streptomyces mobaraense

The crystal structure of the microbial transglutaminase (MTGase) zymogen from Streptomyces mobaraense has been determined at 1.9-Å resolution using the molecular replacement method based on the crystal structure of the mature MTGase. The overall structure of this zymogen is similar to that of the mature form, consisting of a single disk-like domain with a deep active cleft at the edge of the molecule. A major portion of the prosequence (45 additional amino acid residues at the N terminus of the mature transglutaminase) folds into an L-shaped structure, consisting of an extended N-terminal segment linked with a one-turn short helix and a long α-helix. Two key residues in the short helix of the prosequence, Tyr-12 and Tyr-16, are located on top of the catalytic triad (Cys-110, Asp-301, and His-320) to block access of the substrate acyl donors and acceptors. Biochemical characterization of the mature MTGase, using N-α-benzyloxycarbonyl-l-glutaminylglycine as a substrate, revealed apparent K_m and k_cat/K_m values of 52.66 mm and 40.42 mm⁻¹ min⁻¹, respectively. Inhibition studies using the partial prosequence SYAETYR and homologous sequence SQAETYR showed a noncompetitive inhibition mechanism with IC₅₀ values of 0.75 and 0.65 mm, respectively, but no cross-linking product formation. Nevertheless, the prosequence homologous oligopeptide SQAETQR, with Tyr-12 and Tyr-16 each replaced with Gln, exhibited inhibitory activity with the formation of the SQAETQR-monodansylcadaverine fluorophore cross-linking product (SQAETQR-C-DNS). MALDI-TOF tandem MS analysis of SQAETQR-C-DNS revealed molecular masses corresponding to those of ^NSQAETQ^C-C-DNS and C-DNS-^NQR^C sequences, suggesting the incorporation of C-DNS onto the C-terminal Gln residue of the prosequence homologous oligopeptide. These results support the putative functional roles of both Tyr residues in substrate binding and inhibition.     

Global Conformational Change Associated with the Two-step Reaction Catalyzed by Escherichia coli Lipoate-Protein Ligase A

Lipoate-protein ligase A (LplA) catalyzes the attachment of lipoic acid to lipoate-dependent enzymes by a two-step reaction: first the lipoate adenylation reaction and, second, the lipoate transfer reaction. We previously determined the crystal structure of Escherichia coli LplA in its unliganded form and a binary complex with lipoic acid (Fujiwara, K., Toma, S., Okamura-Ikeda, K., Motokawa, Y., Nakagawa, A., and Taniguchi, H. (2005) J Biol. Chem. 280, 33645–33651). Here, we report two new LplA structures, LplA·lipoyl-5′-AMP and LplA·octyl-5′-AMP·apoH-protein complexes, which represent the post-lipoate adenylation intermediate state and the pre-lipoate transfer intermediate state, respectively. These structures demonstrate three large scale conformational changes upon completion of the lipoate adenylation reaction: movements of the adenylate-binding and lipoate-binding loops to maintain the lipoyl-5′-AMP reaction intermediate and rotation of the C-terminal domain by about 180°. These changes are prerequisites for LplA to accommodate apoprotein for the second reaction. The Lys¹³³ residue plays essential roles in both lipoate adenylation and lipoate transfer reactions. Based on structural and kinetic data, we propose a reaction mechanism driven by conformational changes.     

Discovery of a new peptide natural product by Streptomyces coelicolor genome mining

Analyses of microbial genome sequences reveal numerous examples of gene clusters encoding proteins typically involved in complex natural product biosynthesis but not associated with the production of known natural products. In Streptomyces coelicolor M145 there are several gene clusters encoding new nonribosomal peptide synthetase (NRPS) systems not associated with known metabolites. Application of structure-based models for substrate recognition by NRPS adenylation domains predicts the amino acids incorporated into the putative peptide products of these systems, but the accuracy of these predictions is untested. Here we report the isolation and structure determination of the new tris-hydroxamate tetrapeptide iron chelator coelichelin from S. coelicolor using a genome mining approach guided by substrate predictions for the trimodular NRPS CchH, and we show that this enzyme, which lacks a C-terminal thioesterase domain, together with a homolog of enterobactin esterase (CchJ), are required for coelichelin biosynthesis. These results demonstrate that accurate prediction of adenylation domain substrate selectivity is possible and raise intriguing mechanistic questions regarding the assembly of a tetrapeptide by a trimodular NRPS.     

��-Actin Association with Endothelial Nitric-oxide Synthase Modulates Nitric Oxide and Superoxide Generation from the Enzyme

Protein-protein interactions represent an important post-translational mechanism for endothelial nitric-oxide synthase (eNOS) regulation. We have previously reported that β-actin is associated with eNOS oxygenase domain and that association of eNOS with β-actin increases eNOS activity and nitric oxide (NO) production. In the present study, we found that β-actin-induced increase in NO production was accompanied by decrease in superoxide formation. A synthetic actin-binding sequence (ABS) peptide 326 with amino acid sequence corresponding to residues 326–333 of human eNOS, one of the putative ABSs, specifically bound to β-actin and prevented eNOS association with β-actin in vitro. Peptide 326 also prevented β-actin-induced decrease in superoxide formation and increase in NO and l-citrulline production. A modified peptide 326 replacing hydrophobic amino acids leucine and tryptophan with neutral alanine was unable to interfere with eNOS-β-actin binding and to prevent β-actin-induced changes in NO and superoxide formation. Site-directed mutagenesis of the actin-binding domain of eNOS replacing leucine and tryptophan with alanine yielded an eNOS mutant that exhibited reduced eNOS-β-actin association, decreased NO production, and increased superoxide formation in COS-7 cells. Disruption of eNOS-β-actin interaction in endothelial cells using ABS peptide 326 resulted in decreased NO production, increased superoxide formation, and decreased endothelial monolayer wound repair, which was prevented by PEG-SOD and NO donor NOC-18. Taken together, this novel finding indicates that β-actin binding to eNOS through residues 326–333 in the eNOS protein results in shifting the enzymatic activity from superoxide formation toward NO production. Modulation of NO and superoxide formation from eNOS by β-actin plays an important role in endothelial function.     

Structural basis for the substrate specificity of a novel β-N-acetylhexosaminidase StrH protein from Streptococcus pneumoniae R6

The β-N-acetylhexosaminidase (EC 3.2.1.52) from glycoside hydrolase family 20 (GH20) catalyzes the hydrolysis of the β-N-acetylglucosamine (NAG) group from the nonreducing end of various glycoconjugates. The putative surface-exposed N-acetylhexosaminidase StrH/Spr0057 from Streptococcus pneumoniae R6 was proved to contribute to the virulence by removal of β(1,2)-linked NAG on host defense molecules following the cleavage of sialic acid and galactose by neuraminidase and β-galactosidase, respectively. StrH is the only reported GH20 enzyme that contains a tandem repeat of two 53% sequence-identical catalytic domains (designated as GH20-1 and GH20-2, respectively). Here, we present the 2.1 Å crystal structure of the N-terminal domain of StrH (residues Glu-175 to Lys-642) complexed with NAG. It adopts an overall structure similar to other GH20 enzymes: a (β/α)₈ TIM barrel with the active site residing at the center of the β-barrel convex side. The kinetic investigation using 4-nitrophenyl N-acetyl-β-d-glucosaminide as the substrate demonstrated that GH20-1 had an enzymatic activity (k_cat/K_m) of one-fourth compared with GH20-2. The lower activity of GH20-1 could be attributed to the substitution of active site Cys-469 of GH20-1 to the counterpart Tyr-903 of GH20-2. A complex model of NAGβ(1,2)Man at the active site of GH20-1 combined with activity assays of the corresponding site-directed mutants characterized two key residues Trp-443 and Tyr-482 at subsite +1 of GH20-1 (Trp-876 and Tyr-914 of GH20-2) that might determine the β(1,2) substrate specificity. Taken together, these findings shed light on the mechanism of catalytic specificity toward the β(1,2)-linked β-N-acetylglucosides.     

Exploitation of the selectivity-conferring code of nonribosomal peptide synthetases for the rational design of novel peptide antibiotics

Recently, the solved crystal structure of a phenylalanine-activating adenylation (A) domain enlightened the structural basis for the specific recognition of the cognate substrate amino acid in nonribosomal peptide synthetases (NRPSs). By adding sequence comparisons and homology modeling, we successfully used this information to decipher the selectivity-conferring code of NRPSs. Each codon combines the 10 amino residues of a NRPS A domain that are presumed to build up the substrate-binding pocket. In this study, the deciphered code was exploited for the first time to rationally alter the substrate specificity of whole NRPS modules in vitro and in vivo. First, the single-residue Lys239 of the L-Glu-activating initiation module C-A(Glu)-PCP of the surfactin synthetase A was mutated to Gln239 to achieve a perfect match to the postulated L-Gln-activating binding pocket. Biochemical characterization of the mutant protein C-A(Glu)-PCP(Lys239 --> Gln) revealed the postulated alteration in substrate specificity from L-Glu to L-Gln without decrease in catalytic efficiency. Second, according to the selectivity-conferring code, the binding pockets of L-Asp and L-Asn-activating A domains differs in three positions: Val299 versus Ile, His322 versus Glu, and Ile330 versus Val, respectively. Thus, the binding pocket of the recombinant A domain AspA, derived from the second module of the surfactin synthetases B, was stepwisely adapted for the recognition of L-Asn. Biochemical characterization of single, double, and triple mutants revealed that His322 represents a key position, whose mutation was sufficient to give rise to the intended selectivity-switch. Subsequently, the gene fragment encoding the single-mutant AspA(His322 --> Glu) was introduced back into the surfactin biosynthetic gene cluster. The resulting Bacillus subtilis strain was found to produce the expected so far unknown lipoheptapeptide [Asn(5)]surfactin. This indicates that site-directed mutagenesis, guided by the selectivity-conferring code of NRPS A domains, represents a powerful alternative for the genetic manipulation of NRPS biosynthetic templates and the rational design of novel peptide antibiotics.     

Biochemical and structural studies of uncharacterized protein PA0743 from Pseudomonas aeruginosa revealed NAD+-dependent L-serine dehydrogenase

The β-hydroxyacid dehydrogenases form a large family of ubiquitous enzymes that catalyze oxidation of various β-hydroxy acid substrates to corresponding semialdehydes. Several known enzymes include β-hydroxyisobutyrate dehydrogenase, 6-phosphogluconate dehydrogenase, 2-(hydroxymethyl)glutarate dehydrogenase, and phenylserine dehydrogenase, but the vast majority of β-hydroxyacid dehydrogenases remain uncharacterized. Here, we demonstrate that the predicted β-hydroxyisobutyrate dehydrogenase PA0743 from Pseudomonas aeruginosa catalyzes an NAD⁺-dependent oxidation of l-serine and methyl-l-serine but exhibits low activity against β-hydroxyisobutyrate. Two crystal structures of PA0743 were solved at 2.2–2.3-Å resolution and revealed an N-terminal Rossmann fold domain connected by a long α-helix to the C-terminal all-α domain. The PA0743 apostructure showed the presence of additional density modeled as HEPES bound in the interdomain cleft close to the predicted catalytic Lys-171, revealing the molecular details of the PA0743 substrate-binding site. The structure of the PA0743-NAD⁺ complex demonstrated that the opposite side of the enzyme active site accommodates the cofactor, which is also bound near Lys-171. Site-directed mutagenesis of PA0743 emphasized the critical role of four amino acid residues in catalysis including the primary catalytic residue Lys-171. Our results provide further insight into the molecular mechanisms of substrate selectivity and activity of β-hydroxyacid dehydrogenases.     

Nuclear Magnetic Resonance Evidence for the Role of the Flexible Regions of the E1 Component of the Pyruvate Dehydrogenase Complex from Gram-negative Bacteria

Most bacterial pyruvate dehydrogenase complexes from either Gram-positive or Gram-negative bacteria have E1 components with an α₂ homodimeric quaternary structure. In a sequel to our previous publications, we present the first NMR study on the flexible regions of the E1 component from Escherichia coli and its biological relevance. We report sequence-specific NMR assignments for 6 residues in the N-terminal 1–55 region and for a glycine in each of the two mobile active center loops of the E1 component, a 200-kDa homodimer. This was accomplished by using site-specific substitutions and appropriate labeling patterns along with a peptide with the sequence corresponding to the N-terminal 1–35 amino acids of the E1 component. To study the functions of these mobile regions, we also examined the spectra in the presence of (a) a reaction intermediate analog known to affect the mobility of the active center loops, (b) an E2 component construct consisting of a lipoyl domain and peripheral subunit binding domain, and (c) a peptide corresponding to the amino acid sequence of the E2 peripheral subunit binding domain. Deductions from the NMR studies are in excellent agreement with our functional finding, providing a clear indication that the N-terminal region of the E1 interacts with the E2 peripheral subunit binding domain and that this interaction precedes reductive acetylation. The results provide the first structural support to the notion that the N-terminal region of the E1 component of this entire class of bacterial pyruvate dehydrogenase complexes is responsible for binding the E2 component.     

Structure of a calcium-dependent 11R-lipoxygenase suggests a mechanism for Ca2+ regulation

Lipoxygenases (LOXs) are a key part of several signaling pathways that lead to inflammation and cancer. Yet, the mechanisms of substrate binding and allosteric regulation by the various LOX isoforms remain speculative. Here we report the 2.47-Å resolution crystal structure of the arachidonate 11R-LOX from Gersemia fruticosa, which sheds new light on the mechanism of LOX catalysis. Our crystallographic and mutational studies suggest that the aliphatic tail of the fatty acid is bound in a hydrophobic pocket with two potential entrances. We speculate that LOXs share a common T-shaped substrate channel architecture that gives rise to the varying positional specificities. A general allosteric mechanism is proposed for transmitting the activity-inducing effect of calcium binding from the membrane-targeting PLAT (polycystin-1/lipoxygenase/α-toxin) domain to the active site via a conserved π-cation bridge.     

Genetic manipulation revealing an unusual N-terminal region in a stand-alone non-ribosomal peptide synthetase involved in the biosynthesis of ramoplanins

Ramoplanins produced by Actinoplanes are new structural class of lipopeptide and are currently in phase III clinical trials for the prevention of vancomycin-resistant enterococcal infections. The depsipeptide structures of ramoplanins are synthesized by non-ribosomal peptide synthetases (NRPS). Romo-orf17, a stand-alone NRPS, is responsible for the recruitment of Thr into the linear NRPS pathways for which the corresponding adenylation domain is absent. Here, systematical gene inactivation and complementation have been carried out in a Actinoplanes sp. using homologous recombination and site-specific integration methods. A hybrid gene coding for the N-terminal region of the stand-alone NRPS and the A-PCP domains of a heterologous NRPS restored production of ramoplanins. The results elucidate the unusual N-terminal region which is essential for the biosynthesis of ramoplanins.     

非核糖体肽合成酶基因腺苷酰化结构域序列克隆及分析

微生物许多非核糖体肽类次生代谢产物主要是由非核糖体肽合成酶(NRPS)催化合成。参考Gontang发布的非核糖体肽合成酶(NRPS)通用引物设计扩增NRPS腺苷酰化结构域基因序列的特异引物,从海洋链霉菌L1的基因组DNA中扩增获得一个715 bp的NRPS基因序列。测序结果及比对分析表明该片段属于NRPS腺苷酰化结构域部分序列。对其拟翻译的氨基酸序列组成成分、理化性质进行分析,显示其包含AFD class I超基因家族核心结合区,为NRPS腺苷酰化结构域(A结构域)所在区域。对氨基酸序列的二级结构预测和三级结构模拟,发现与数据库中肠菌素合酶F组分的结构相似。为后续研究A结构域的特异性及完整NRPS基因簇克隆提供了参考。     

Crystal Structure of Enzyme I of the Phosphoenolpyruvate Sugar Phosphotransferase System in the Dephosphorylated State

The bacterial phosphoenolpyruvate (PEP) sugar phosphotransferase system mediates sugar uptake and controls the carbon metabolism in response to carbohydrate availability. Enzyme I (EI), the first component of the phosphotransferase system, consists of an N-terminal protein binding domain (EIN) and a C-terminal PEP binding domain (EIC). EI transfers phosphate from PEP by double displacement via a histidine residue on EIN to the general phosphoryl carrier protein HPr. Here we report the 2.4 Å crystal structure of the homodimeric EI from Staphylococcus aureus. EIN consists of the helical hairpin HPr binding subdomain and the phosphorylatable βα phospho-histidine (P-His) domain. EIC folds into an (βα)₈ barrel. The dimer interface of EIC buries 1833 Ų of accessible surface per monomer and contains two Ca²⁺ binding sites per dimer. The structures of the S. aureus and Escherichia coli EI domains (Teplyakov, A., Lim, K., Zhu, P. P., Kapadia, G., Chen, C. C., Schwartz, J., Howard, A., Reddy, P. T., Peterkofsky, A., and Herzberg, O. (2006) Proc. Natl. Acad. Sci. U.S.A. 103, 16218–16223) are very similar. The orientation of the domains relative to each other, however, is different. In the present structure the P-His domain is docked to the HPr binding domain in an orientation appropriate for in-line transfer of the phosphate to the active site histidine of the acceptor HPr. In the E. coli structure the phospho-His of the P-His domain projects into the PEP binding site of EIC. In the S. aureus structure the crystallographic temperature factors are lower for the HPr binding domain in contact with the P-His domain and higher for EIC. In the E. coli structure it is the reverse.     

Substrate selection of adenylation domains for nonribosomal peptide synthetase (NRPS) in bacillamide C biosynthesis by marine <Emphasis Type="Italic">Bacillus atrophaeus</Emphasis> C89

Nonribosomal peptide synthetases (NRPSs) are multi-modular enzymes involved in the biosynthesis of natural products. Bacillamide C was synthesized by Bacillus atrophaeus C89. A nonribosomal peptide synthetase (NRPS) cluster found in the genome of B. atrophaeus C89 was hypothesized to be responsible for the biosynthesis of bacillamide C using alanine and cysteine as substrates. Here, the structure analysis of adenylation domains based on homologous proteins with known crystal structures indicated locations of the substrate-binding pockets. Molecular docking suggested alanine and cysteine as the potential substrates for the two adenylation domains in the NRPS cluster. Furthermore, biochemical characterization of the purified recombinant adenylation domains proved that alanine and cysteine were the optimum substrates for the two adenylation domains. The results provided the in vitro evidence for the hypothesis that the two adenylation domains in the NRPS of B. atrophaeus C89 preferentially select alanine and cysteine, respectively, as a substrate to synthesize bacillamide C. Furthermore, this study on substrates selectivity of adenylation domains provided basis for rational design of bacillamide analogs.     

Predicting substrate specificity of adenylation domains of nonribosomal peptide synthetases and other protein properties by latent semantic indexing

Successful genome mining is dependent on accurate prediction of protein function from sequence. This often involves dividing protein families into functional subtypes (e.g., with different substrates). In many cases, there are only a small number of known functional subtypes, but in the case of the adenylation domains of nonribosomal peptide synthetases (NRPS), there are >500 known substrates. Latent semantic indexing (LSI) was originally developed for text processing but has also been used to assign proteins to families. Proteins are treated as ‘‘documents’’ and it is necessary to encode properties of the amino acid sequence as ‘‘terms’’ in order to construct a term-document matrix, which counts the terms in each document. This matrix is then processed to produce a document-concept matrix, where each protein is represented as a row vector. A standard measure of the closeness of vectors to each other (cosines of the angle between them) provides a measure of protein similarity. Previous work encoded proteins as oligopeptide terms, i.e. counted oligopeptides, but used no information regarding location of oligopeptides in the proteins. A novel tokenization method was developed to analyze information from multiple alignments. LSI successfully distinguished between two functional subtypes in five well-characterized families. Visualization of different ‘‘concept’’ dimensions allows exploration of the structure of protein families. LSI was also used to predict the amino acid substrate of adenylation domains of NRPS. Better results were obtained when selected residues from multiple alignments were used rather than the total sequence of the adenylation domains. Using ten residues from the substrate binding pocket performed better than using 34 residues within 8 Å of the active site. Prediction efficiency was somewhat better than that of the best published method using a support vector machine.