Interaction of E. coli outer-membrane protein A with sugars on the receptors of the brain microvascular endothelial cells

Esherichia coli, the most common gram-negative bacteria, can penetrate the brain microvascular endothelial cells (BMECs) during the neonatal period to cause meningitis with significant morbidity and mortality. Experimental studies have shown that outer-membrane protein A (OmpA) of E. coli plays a key role in the initial steps of the invasion process by binding to specific sugar moieties present on the glycoproteins of BMEC. These experiments also show that polymers of chitobiose (GlcNAcbeta1-4GlcNAc) block the invasion, while epitopes substituted with the L-fucosyl group do not. We used HierDock computational technique that consists of a hierarchy of coarse grain docking method with molecular dynamics (MD) to predict the binding sites and energies of interactions of GlcNAcbeta1-4GlcNAc and other sugars with OmpA. The results suggest two important binding sites for the interaction of carbohydrate epitopes of BMEC glycoproteins to OmpA. We identify one site as the binding pocket for chitobiose (GlcNAcbeta1-4GlcNAc) in OmpA, while the second region (including loops 1 and 2) may be important for recognition of specific sugars. We find that the site involving loops 1 and 2 has relative binding energies that correlate well with experimental observations. This theoretical study elucidates the interaction sites of chitobiose with OmpA and the binding site predictions made in this article are testable either by mutation studies or invasion assays. These results can be further extended in suggesting possible peptide antagonists and drug design for therapeutic strategies.     

Site-specific labeling of cytoplasmic peptide:N-glycanase by N,N'-diacetylchitobiose-related compounds

Peptide:N-glycanase (PNGase) is the deglycosylating enzyme, which releases N-linked glycan chains from N-linked glycopeptides and glycoproteins. Recent studies have revealed that the cytoplasmic PNGase is involved in the degradation of misfolded/unassembled glycoproteins. This enzyme has a Cys, His, and Asp catalytic triad, which is required for its enzymatic activity and can be inhibited by "free" N-linked glycans. These observations prompted us to investigate the possible use of haloacetamidyl derivatives of N-glycans as potent inhibitors and labeling reagents of this enzyme. Using a cytoplasmic PNGase from budding yeast (Png1), Man9GlcNAc2-iodoacetoamide was shown to be a strong inhibitor of this enzyme. The inhibition was found to be through covalent binding of the carbohydrate to a single Cys residue on Png1, and the binding was highly selective. The mutant enzyme in which Cys191 of the catalytic triad was changed to Ala did not bind to the carbohydrate probe, suggesting that the catalytic Cys is the binding site for this compound. Precise determination of the carbohydrate attachment site by mass spectrometry clearly identified Cys191 as the site of covalent attachment. Molecular modeling of N,N'-diacetylchitobiose (chitobiose) binding to the protein suggests that the carbohydrate binding site is distinct from but adjacent to that of Z-VAD-fmk, a peptide-based inhibitor of this enzyme. These results suggest that cytoplasmic PNGase has a separate binding site for chitobiose and other carbohydrates, and haloacetamide derivatives can irreversibly inhibit that catalytic Cys in a highly specific manner.     

Primary structure of beta-galactoside alpha 2,6-sialyltransferase. Conversion of membrane-bound enzyme to soluble forms by cleavage of the NH2-terminal signal anchor

This report describes the primary structure of a rat liver beta-galactoside alpha 2,6-sialyltransferase (EC 2.4.99.1), a Golgi apparatus enzyme involved in the terminal sialylation of N-linked carbohydrate groups of glycoproteins. The complete amino acid sequence was deduced from the nucleotide sequence of cDNA clones of the enzyme. The primary structure suggests that the topology of the enzyme in the Golgi apparatus consists of a short NH2-terminal cytoplasmic domain, a 17-residue hydrophobic sequence which serves as the membrane anchor and signal sequence, and a large lumenal, catalytic domain. NH2-terminal sequence analysis of a truncated form of the enzyme, obtained by purification from tissue homogenates, reveals that it is missing a 63-residue NH2-terminal peptide which includes the membrane binding domain. These and supporting results show that soluble forms of the sialyltransferase can be generated by proteolytic cleavage between the NH2-terminal signal-anchor and the catalytic domain.     

Sortase A‐mediated synthesis of ligand‐grafted cyclized peptides for modulating a model protein‐protein interaction

Specific ligand‐grafted cyclic peptides are promising drug candidates that can modulate protein‐protein interactions (PPIs) with increased proteolytic stability. In this study, we aimed to demonstrate that Sortase A (SrtA)‐mediated peptide transpeptidation can be applied to produce bioactive sequence‐grafted, stable, cyclic peptides. A naturally occurring cyclic peptide, sunflower trypsin inhibitor 1 (SFTI‐1), was selected as the scaffold, and a tetrapeptide motif, Glu‐Ser‐Asp‐Val (ESDV), was grafted into the scaffold as a model ligand. The linear precursor of the grafted peptide with SrtA‐recognition motifs at the N‐ and C‐termini was cyclized in good yield simply by co‐incubation with SrtA. The ESDV‐grafted cyclic SFTI‐1 obtained was confirmed to have high stability against proteolysis by human serum and bound to the target PDZ2 domain of postsynaptic density‐95 protein. An optimized sequence‐grafted cyclic SFTI‐1 could competitively suppress the interaction of PDZ2 with its natural ligand, the C‐terminal peptide of the NR2B subunit of the N‐methyl‐D‐aspartate receptor. These results show that a strategy combining peptide grafting into the SFTI‐1 scaffold with SrtA‐catalyzed cyclization can be a simple and effective method for producing stable peptide drugs.     

Activity and stability of human kallikrein-2-specific linear and cyclic peptide inhibitors.

Human glandular kallikrein (KLK2) is a highly prostate-specific serine protease, which is mainly excreted into the seminal fluid, but part of which is also secreted into circulation from prostatic tumors. Since the expression level of KLK2 is elevated in aggressive tumors and it has been suggested to mediate the metastasis of prostate cancer, inhibition of the proteolytic activity of KLK2 is of potential therapeutic value. We have previously identified several KLK2-specific linear peptides by phage display technology. Two of its synthetic analogs, A R R P A P A P G (KLK2a) and G A A R F K V W W A A G (KLK2b), show specific inhibition of KLK2 but their sensitivity to proteolysis in vivo may restrict their potential use as therapeutic agents. In order to improve the stability of the linear peptides for in vivo use, we have prepared cyclic analogs and compared their biological activity and their structural stability. A series of cyclic variants with cysteine bridges were synthesized. Cyclization inactivated one peptide (KLK2a) and its derivatives, while the other peptide (KLK2b) and its derivatives remained active. Furthermore, backbone cyclization of KLK2b improved significantly the resistance against proteolysis by trypsin and human plasma. Nuclear magnetic resonance studies showed that cyclization of the KLK2b peptides does not make the structures more rigid. In conclusion, we have shown that backbone cyclization of KLK2 inhibitory peptides can be used to increase stability without losing biological activity. This should render the peptides more useful for in vivo applications, such as tumor imaging and prostate cancer targeting.     

Structural and mutational studies on the importance of oligosaccharide binding for the activity of yeast PNGase

Peptide:N-glycanase (PNGase) is an important component of the endoplasmic reticulum-associated protein degradation pathway in which it de-glycosylates misfolded glycoproteins, thus facilitating their proteasomal degradation. PNGase belongs to the transglutaminase superfamily and features a Cys, His, and Asp catalytic triad, which is essential for its enzymatic activity. An elongated substrate-binding groove centered on the active site Cys191 was visualized in the crystal structure of apo-PNGase, whereas its complex with Z-VAD-fmk, a peptide-based inhibitor of PNGase, revealed that the inhibitor occupied one end of the substrate-binding groove while being covalently linked to the active site Cys. Recently, haloacetamidyl-containing carbohydrate-based inhibitors of PNGase were developed and shown to specifically label the active site Cys. In this study, we describe the crystal structure of yeast PNGase in complex with N,N'-diacetylchitobiose (chitobiose). We found that the chitobiose binds on the side opposite to the peptide binding site with the active site Cys191 being located approximately midway between the carbohydrate and peptide binding sites. Mutagenesis studies confirm the critical role of the chitobiose-interacting residues in substrate binding and suggest that efficient oligosaccharide binding is required for PNGase activity. In addition, the N-terminus of a symmetry-related PNGase was found to bind to the proposed peptide-binding site of PNGase. Together with the bound chitobiose, this enables us to propose a model for glycoprotein binding to PNGase. Finally, deleting the C-terminal residues of yeast PNGase, which are disordered in all structures of this enzyme, results in a significant reduction in enzyme activity, indicating that these residues might be involved in binding of the mannose residues of the glycan chain.     

Approaches to the study of N-linked glycoproteins in human plasma using lectin affinity chromatography and nano-HPLC coupled to electrospray linear ion trap--Fourier transform mass spectrometry

In this publication, we will describe the combination of lectin affinity chromatography with nano high performance liquid chromatography (HPLC) coupled to a linear ion trap Fourier transform mass spectrometer (capillary LC-LTQ/FTMS) to characterize N-linked glycosylation structures in human plasma proteins. We used a well-characterized glycoprotein, tissue plasminogen activator (rt-PA), which is present at low levels in blood, as a standard to determine the dynamic range of this approach. N-linked glycopeptides derived from rt-PA could be characterized at a ratio of 1:200 in human plasma (rtPA: total plasma protein, w/w) by accurate mass measurement in the FTMS and fragmentation (MS(n)) in the linear ion trap. We demonstrated that this platform has the potential to characterize the general N-linked glycosylation structures of abundant glycoproteins present in human plasma without the requirement for antibody-based purification, or additional carbohydrate analytical protocols. This conclusion was supported by the determination of carbohydrate structures for three glycoproteins, IgG, haptoglobin, and alpha-1-acid glycoprotein, at their natural levels in a human plasma sample, but only after the lectin enrichment step.     

Stapled peptide-based membrane fusion inhibitors of hepatitis C virus

The strategy of peptide stapling was used to develop new molecules to inhibit the hepatitis C virus infection via disrupting the binding of HCV envelope glycoprotein E2 with human cell surface protein CD81. The peptide sequence was designed based on the large extra-cellular loop of CD81 with known importance in the HCV E2 binding interaction. Our results showed that the stapled peptides exhibited significantly higher α-helicity and proteolytic stability as compared to their linear peptide counterpart. The optimal compound was found to have an EC₅₀ value of ca. 17–39 μM against different HCV subtypes and represented a new HCV membrane fusion inhibitor.     

Structural basis of sugar-recognizing ubiquitin ligase

SCF(Fbs1) is a ubiquitin ligase that functions in the endoplasmic reticulum (ER)-associated degradation pathway. Fbs1/Fbx2, a member of the F-box proteins, recognizes high-mannose oligosaccharides. Efficient binding to an N-glycan requires di-N-acetylchitobiose (chitobiose). Here we report the crystal structures of the sugar-binding domain (SBD) of Fbs1 alone and in complex with chitobiose. The SBD is composed of a ten-stranded antiparallel beta-sandwich. The structure of the SBD-chitobiose complex includes hydrogen bonds between Fbs1 and chitobiose and insertion of the methyl group of chitobiose into a small hydrophobic pocket of Fbs1. Moreover, NMR spectroscopy has demonstrated that the amino acid residues adjoining the chitobiose-binding site interact with the outer branches of the carbohydrate moiety. Considering that the innermost chitobiose moieties in N-glycans are usually involved in intramolecular interactions with the polypeptide moieties, we propose that Fbs1 interacts with the chitobiose in unfolded N-glycoprotein, pointing the protein moiety toward E2 for ubiquitination.     

Design of cyclic and other templates for potent and selective peptide alpha-MSH analogues

For over three decades, the design of linear peptide ligands often has incorporated cyclic constraints to improve potency, receptor selectivity, proteolytic stability and biodistribution. Its importance has been so well established that modern day schemes for ligand-based drug design often start with cyclization of linear peptides to rigidify peptide structure, to limit its conformational possibilities, and to find key pharmacophore elements in three-dimensional space. In the past several years, cyclic constraints have been used to develop ligands with improved efficacy, binding affinity, biostability and receptor selectivity for alpha-melanocyte-stimulating hormone (alpha-MSH). Furthermore, potent cyclic alpha-MSH analogues, such as MT-II and SHU-9119, have made structure-activity relationship studies and molecular modeling more useful for creating new three-dimensional, topographical pharmacophore templates.     

Design and characterization of a hybrid miniprotein that specifically inhibits porcine pancreatic elastase

Studying protease/peptide inhibitor interactions is a useful tool for understanding molecular recognition in general and is particularly relevant for the rational design of inhibitors with therapeutic potential. An inhibitory peptide (PMTLEYR) derived from the third domain of turkey ovomucoid inhibitor and optimized for specific porcine pancreatic elastase inhibition was introduced into an inhibitor scaffold to increase the proteolytic stability of the peptide. The trypsin-specific squash inhibitor EETI II from Ecballium elaterium was chosen as the scaffold. The resulting hybrid inhibitor HEI-TOE I (hybrid inhibitor from E. elaterium and the optimized binding loop of the third domain of turkey ovomucoid inhibitor) shows a specificity and affinity to porcine pancreatic elastase similar to the free inhibitory peptide but with significantly higher proteolytic stability. Isothermal titration calorimetry revealed that elastase binding of HEI-TOE I occurs with a small unfavorable positive enthalpy contribution, a large favorable positive entropy change, and a large negative heat capacity change. In addition, the inhibitory peptide and the hybrid inhibitor HEI-TOE I protected endothelial cells against degradation following treatment with porcine pancreatic elastase.     

Synthesis of human angiotensinogen (1-17) containing one of the putative glycosylation binding sites and its hydrolysis by human renin and porcine pepsin.

The N-terminal heptadecapeptide of human angiotensinogen (Asp-Arg-Val-Tyr-Ile-His-Pro-Phe-His-Leu-Val-Ile-His-Asn-Glu-Ser-Thr-NH2 ), with the C-terminal carboxyl group amidated, was synthesized in order to study the role of Asn-Glu-Ser, a putative carbohydrate binding site, on the hydrolysis by human renin. The synthesis was performed by fragment condensation using the Honzl and Rudinger azide procedure. In our conditions for azide segment condensation, histidine racemization was demonstrated to be negligible for most of the condensation reactions. Human renin liberates angiotensin I from h-angiotensinogen (1-17)-NH2 with a Km value of 3.4 x 10(-5) M, at pH 7.3 and 37 degrees being similar to h-angiotensinogen (1-13), an analog without the carbohydrate binding site. However, the Vmax value of 4.1 x 10(-9) mol/G.U. min is one order of magnitude higher. Porcine pepsin was demonstrated to cleave preferentially Leu10-Val11 bond and, surprisingly, His9-Leu10 as well.     

Structural fingerprinting of Asn-linked carbohydrates from specific attachment sites in glycoproteins by mass spectrometry: application to tissue plasminogen activator

A sensitive and specific strategy has been developed for determining the sites of attachment of Asn-linked carbohydrates in glycoproteins, and defining the compositions and molecular heterogeneity of carbohydrates at each specific attachment site. In this carbohydrate 'fingerprinting' strategy, potential glycopeptides are identified by comparing the high pressure liquid chromatography (HPLC) chromatograms of proteolytic digests of a glycoprotein obtained before and after digestion with a glycosidase, usually peptide:N-glycosidase F (PNGase F). The glycopeptide-containing HPLC fractions are analyzed by fast atom bombardment mass spectrometry (FAB MS) prior to and after digestion with PNGase F to identify the former glycosylation site peptide and its sequence location (Carr and Roberts, (1986) Anal. Biochem. 157, 396-406). Carbohydrates are extracted from these fractions as the peracetates which are then permethylated and analyzed by FAB MS. The spectra exhibit molecular weight-related ions for each of the parent oligosaccharides present in the fraction which provide composition in terms of hexose, deoxyhexose, N-acetylhexosamine and sialic acid. The relative ratios of these peaks reflect the relative abundances of the various carbohydrate homologs present in the mixture. The derivatives formed are directly amenable to methylation analysis for determination of linkage. This strategy enables the structural classes of carbohydrates at specific attachment sites to be determined using only a few nmol of glycoprotein. The carbohydrate fingerprinting strategy has been applied to a number of glycoproteins including tissue plasminogen activator, the results for which are described herein.     

Inhibition of lymphocyte proliferation by synthetic peptides homologous to human plasma apolipoproteins B and E

Apolipoproteins B and E of the human plasma lipoproteins are known inhibitors of lymphocyte proliferation. In this report, two synthetic peptide amides, apoB3358-3372 and apoE141-155, showed a dose-dependent inhibition of both the murine mixed lymphocyte culture reaction and the anti-T3 induced proliferation of lymphocytes. Their structures and antiproliferative potencies were similar to that of the heptadecapeptide CKS-17, a consensus peptide of a highly conserved region among HTLV-I, -II and C-type human retroviral proteins. SP-9-2-amide, a peptide homologous to the amino-terminal half of CKS-17, also suppressed lymphocyte activation. In contrast, a peptide homologous to the gp41 protein of HTLV-III that is sequence related to CKS-17 (approximately 35% homology) showed little antiproliferative activity. Neurotensin, a control peptide, showed no activity. The results suggest that a basic tetrapeptide sequence common to CKS-17-amide, SP-9-2, apoB3358-3372 and apoE141-155, but not HTLV-III-amide may account, in part, for the antiproliferative activities of these peptides.     

Antigenic properties of peptide mimotopes of HIV-1-associated carbohydrate antigens

The glycan shield of the human immunodeficiency virus (HIV) envelope protein presents many potential epitopes for vaccine development. To augment immune responses to HIV, type 1 (HIV-1), envelope-associated carbohydrate antigens, we are defining peptide mimics of HIV-associated carbohydrate antigens that function as antigen mimotopes that upon immunization will induce antibodies cross-reactive with carbohydrate antigens. We have previously defined peptides with a putative sequence tract RYRY that mimic concanavalin A-binding glycans. To imitate the multivalent binding of carbohydrates, we compared the avidity of a linear (911) and cyclic peptide (D002) reactive with concanavalin A presented in a multiple antigen peptide (MAP) format. The affinity of the MAP-D002 peptide was higher than that of the peptide MAP-911, whereas the avidity of D002 peptide was lower than that of 911. Serum from mice immunized with MAP-911 had lower titer for oligomannose-9 than those elicited by MAP-D002 under the same conditions, but both immunogens elicited antibodies that can block the binding of GP120 to dendritic cells. Antibodies that bind to the studied MAPs were found in a preparation of normal human immunoglobulin for intravenous use. Those that were purified on 911 bound back to 911 and D002, whereas anti-D002 antibodies were specific only for D002. Human antibodies reactive with both mimotopes and with a mannosyl preparation were observed to bind to envelope protein. These results suggested the potential to fine-tune the antibody response to carbohydrate antigens by modifying structural features of peptide mimotope-based immunogens.     

Structural basis for recognition of high mannose type glycoproteins by mammalian transport lectin VIP36

VIP36 functions as a transport lectin for trafficking certain high mannose type glycoproteins in the secretory pathway. Here we report the crystal structure of VIP36 exoplasmic/luminal domain comprising a carbohydrate recognition domain and a stalk domain. The structures of VIP36 in complex with Ca(2+) and mannosyl ligands are also described. The carbohydrate recognition domain is composed of a 17-stranded antiparallel beta-sandwich and binds one Ca(2+) adjoining the carbohydrate-binding site. The structure reveals that a coordinated Ca(2+) ion orients the side chains of Asp(131), Asn(166), and His(190) for carbohydrate binding. This result explains the Ca(2+)-dependent carbohydrate binding of this protein. The Man-alpha-1,2-Man-alpha-1,2-Man, which corresponds to the D1 arm of high mannose type glycan, is recognized by eight residues through extensive hydrogen bonds. The complex structures reveal the structural basis for high mannose type glycoprotein recognition by VIP36 in a Ca(2+)-dependent and D1 arm-specific manner.     

Ca2+ binding protects against gelsolin amyloidosis

Amyloid diseases occur when native or mutant polypeptides misfold and aggregate to form deposits in the extracellular space. There are at least 20 proteins associated with amyloid diseases, including the well-known amyloid-beta peptide that is the causative agent for Alzheimer's disease (AD). This review describes familial amyloidosis of Finnish type (FAF), an amyloid disease caused by mutations in plasma gelsolin, a secreted protein that contains multiple Ca2+-binding domains. The FAF mutations result in a loss of the Ca2+-binding site in domain 2 of plasma gelsolin. The resulting decreased stability gives rise to susceptibility to the protease furin in the Golgi. Furin cleavage generates a secreted fragment that undergoes a second proteolytic event in the extracellular matrix to produce a peptide that self-assembles into amyloid plaques. Thus, Ca2+ binding in native plasma gelsolin protects against amyloid disease.     

Solution structure of the Reps1 EH domain and characterization of its binding to NPF target sequences

The recently described EH domain recognizes proteins containing Asn-Pro-Phe (NPF) sequences. Using nuclear magnetic resonance (NMR) data, we determined the solution structure of the EH domain from the Reps1 protein and characterized its binding to linear and cyclic peptides derived from a novel targeting protein. The structure calculation included 1143 distance restraints and 122 angle restraints and resulted in structures with a root-mean-square deviation of 0.40 +/- 0.05 A for backbone atoms of superimposed secondary structural elements. The structure comprises two helix-loop-helix motifs characteristic of EF-hand domains. Titration data with NPF-containing peptides showed evidence of intermediate exchange on the NMR chemical shift time scale, which required an analysis that includes curve fitting to obtain accurate equilibrium constants and dissociation rate constants. The cyclic and linear peptides bound with similar affinities (Kd = 65 +/- 17 and 46 +/- 14 microM, respectively) and to the same hydrophobic pocket formed between helices B and C. The cyclic peptide formed a complex that dissociated more slowly (k(off) = 440 +/- 110 s(-1)) than the linear peptide (k(off) = 1800 +/- 250 s(-1)), but had little change in affinity because of the slower rate of association of the cyclic peptide. In addition, we characterized binding to a peptide containing a DPF sequence (Kd = 0.5 +/- 0.2 mM). The characterization of binding between the Reps1 EH domain and its target proteins provides information about their role in endocytosis.     

Molecular determinants of tuberoinfundibular peptide of 39 residues (TIP39) selectivity for the parathyroid hormone-2 (PTH2) receptor. N-terminal truncation of TIP39 reverses PTH2 receptor/PTH1 receptor binding selectivity

Tuberoinfundibular peptide of 39 residues (TIP39) and the parathyroid hormone-2 (PTH2) receptor form part of an extended family of related signaling molecules that includes the PTH1 receptor, which responds to PTH and PTH-related protein. TIP39 does not appreciably activate the PTH1 receptor, but in this study it is shown to bind the receptor with moderate affinity (59 nm). In this study, we investigated the molecular determinants of both ligand and receptor for the PTH2 receptor selectivity of TIP39 and quantitatively evaluated the role of molecular elements in the binding of TIP39 to the PTH2 and PTH1 receptors. A chimeric receptor composed of the N-terminal extracellular domain of the PTH1 receptor and the remainder (juxtamembrane domain) of the PTH2 receptor (P2-NP1) was fully activated by TIP39 (E(max) = 98% of the rPTH-(1-34), E(max), EC(50) = 2.0 nm). This receptor chimera bound TIP39 with an equivalent affinity to the wild-type PTH2 receptor (2. 3 and 2.0 nm, respectively). The reciprocal chimeric receptor (P1-NP2) was not activated by TIP39 and bound the ligand with an affinity equivalent to that of the PTH1 receptor. Thus, the juxtamembrane receptor domain specifies the signaling and binding selectivity of TIP39 for the PTH2 receptor over the PTH1 receptor. Removing six N-terminal residues of TIP39 eliminated activation of the PTH2 receptor and reduced binding affinity 70-fold. In contrast, this truncation increased affinity for the PTH1 receptor 10-fold, reversing the PTH2/PTH1 receptor binding selectivity and resulting in a high affinity interaction of TIP-(7-39) with the PTH1 receptor (6 nm). These findings can be explained by a strong interaction between the N-terminal region of TIP39 and the juxtamembrane domain of the PTH2 receptor, with the corresponding domain of the PTH1 receptor acting as a selectivity barrier against high affinity binding of TIP39. As a result, TIP-(7-39) is a highly potent, selective antagonist for the PTH1 receptor.     

Characterization of guanosine diphospho-D-mannose dehydrogenase from Pseudomonas aeruginosa. Structural analysis by limited proteolysis.

Alginate is believed to be a major virulence factor in the pathogenicity of Pseudomonas aeruginosa in the lungs of patients suffering from cystic fibrosis. Guanosine diphospho-D-mannose dehydrogenase (GDPmannose dehydrogenase, EC 1.1.1.132) is a key enzyme in the alginate biosynthetic pathway which catalyzes the oxidation of guanosine diphospho-D-mannose (GDP-D-mannose) to GDP-D-mannuronic acid. In this paper, we report the structural analysis of GMD by limited proteolysis using three different proteases, trypsin, submaxillary Arg-C protease, and chymotrypsin. Treatment of GMD with these proteases indicated that the amino-terminal part of this enzyme may fold into a structural domain with an apparent molecular mass of 25-26 kDa. Multiple proteolytic cleavage sites existed at the carboxyl-terminal end of this domain, indicating that this segment may represent an exposed region of the protein. Initial proteolysis also generated a carboxyl-terminal fragment with an apparent molecular mass of 16-17 kDa which was further digested into smaller fragments by trypsin and chymotrypsin. The proteolytic cleavage sites were localized by partial amino-terminal sequencing of the peptide fragments. Arg-295 was identified as the initial cleavage site for trypsin and Tyr-278 for chymotrypsin. Catalytic activity of GMD was totally abolished by the initial cleavage. However, binding of the substrate, GDP-D-mannose, increased stability toward proteolysis and inhibited the loss of enzyme activity. GMP and GDP (guanosine 5'-mono- and diphosphates) also blocked the initial cleavage, but NAD and mannose showed no effect. These results suggest that binding of the guanosine moiety at the catalytic site of GMD may induce a conformational change that reduces the accessibility of the cleavage sites to proteases. Binding of [14C]GDP-D-mannose to the amino-terminal domain was not affected by the removal of the carboxyl-terminal 16-kDa fragment. Furthermore, photoaffinity labeling of GMD with [32P]arylazido-beta-alanine-NAD followed by proteolysis demonstrated that the radioactive NAD was covalently linked to the amino-terminal domain. These observations imply that the amino-terminal domain (25-26 kDa) contains both the substrate and cofactor binding sites. However, the carboxyl-terminal fragment (16-17 kDa) may possess amino acid residues essential for catalysis. Thus, proteolysis had little effect on substrate binding, but totally eliminated catalysis. These biochemical data are in complete agreement with amino acid sequence analysis for the existence of substrate and cofactor sites of GMD. A linear peptide map of GMD was constructed for future structure/functional studies.