Crystal structures of MMP-1 and -13 reveal the structural basis for selectivity of collagenase inhibitors.

The X-ray crystal structures of the catalytic domain of human collagenase-3 (MMP-13) and collagenase-1 (MMP-1) with bound inhibitors provides a basis for understanding the selectivity profile of a novel series of matrix metalloprotease (MMP) inhibitors. Differences in the relative size and shape of the MMP S1' pockets suggest that this pocket is a critical determinant of MMP inhibitor selectivity. The collagenase-3 S1' pocket is long and open, easily accommodating large P1' groups, such as diphenylether. In contrast, the collagenase-1 S1' pocket must undergo a conformational change to accommodate comparable P1' groups. The selectivity of the diphenylether series of inhibitors for collagenase-3 is largely determined by their affinity for the preformed S1' pocket of collagenase-3, as compared to the induced fit in collagenase-1.

The structure of mammalian 15-lipoxygenase reveals similarity to the lipases and the determinants of substrate specificity.

Here we report the first structure of a mammalian 15-lipoxygenase. The protein is composed of two domains; a catalytic domain and a previously unrecognized beta-barrel domain. The N-terminal beta-barrel domain has topological and sequence identify to a domain in the mammalian lipases, suggesting that these domains may have similar functions in vivo. Within the C-terminal domain, the lipoxygenase substrate binding site is a hydrophobic pocket defined by a bound inhibitor. Arachidonic acid can be docked into this deep hydrophobic pocket with the methyl end extending down into the bottom of the pocket and the acid end tethered by a conserved basic residue on the surface of the enzyme. This structure provides a unifying hypothesis for the positional specificity of mammalian lipoxygenases.

Partial activation of muscle phosphorylase by replacement of serine 14 with acidic residues at the site of regulatory phosphorylation.

Phosphorylation of glycogen phosphorylase at residue Ser14 triggers a conformational transition that activates the enzyme. The N-terminus of the protein, in response to phosphorylation, folds into a 310 helix and moves from its location near a cluster of acidic residues on the protein surface to a site at the dimer interface where a pair of arginine residues form charged hydrogen bonds with the phosphoserine. Site-directed mutagenesis was used to replace Ser14 with Asp and Glu residues, analogs of the phosphoserine, that might be expected to participate in ionic interactions with the arginine side chains at the dimer interface. Kinetic analysis of the mutants indicates that substitution of an acidic residue in place of Ser14 at the site of regulatory phosphorylation partially activates the enzyme. The S14D mutant shows a 1.6-fold increase in Vmax, a 10-fold decrease in the apparent dissociation constant for AMP, and a 3-fold decrease in the S0.5 for glucose 1-phosphate. The S14E mutant behaves similarly, showing a 2.2-fold increase in Vmax, a 6-fold decrease in the apparent dissociation constant for AMP, and a 2-fold decrease in the S0.5 for glucose 1-phosphate. The ability of the mutations to enhance binding of AMP and glucose 1-phosphate and to raise catalytic activity suggests that the introduction of a carboxylate side chain at position 14 promotes docking of the N-terminus at the subunit interface and concomitant stabilization of the activated conformation of the enzyme. Like the native enzyme, both mutants show significant activity only in the presence of the activator, AMP. Full activation, analogous to that provided by covalent phosphorylation of the enzyme, likely is not achieved because of differences in the charge and the geometry of ionic interactions at the phosphorylation site.

Flexibility of the NSAID binding site in the structure of human cyclooxygenase-2.

The first crystal structure of human cyclooxygenase-2, in the presence of a selective inhibitor, is similar to that of cyclooxygenase-1. The structure of the NSAID binding site is also well conserved, although there are differences in its overall size and shape which may be exploited for the further development of selective COX-2 inhibitors. A second COX-2 structure with a different bound inhibitor displays a new, open conformation at the bottom of the NSAID binding site, without significant changes in other regions of the COX-2 structure. These two COX-2 structures provide evidence for the flexible nature of cyclooxygenase, revealing details about how substrate and inhibitor may gain access to the cyclooxygenase active site from within the membrane.

Defining the arachidonic acid binding site of human 15-lipoxygenase. Molecular modeling and mutagenesis.

Mammalian lipoxygenases have been implicated in the pathogenesis of several inflammatory disorders and are, therefore, important targets for drug discovery. Both plant and mammalian lipoxygenases catalyze the dioxygenation of polyunsaturated fatty acids, which contain one or more 1,4-cis,cis-pentadiene units to yield hydroperoxide products. At the time this study was initiated, soybean lipoxygenase-1 was the only lipoxygenase for which an atomic resolution structure had been determined. No structure of lipoxygenase with substrate or inhibitor bound is currently available. A model of arachidonic acid docked into the proposed substrate binding site in the soybean structure is presented here. Analysis of this model suggested two residues, an aromatic residue and a positively charged residue, could be critical for substrate binding. Validation of this model is provided by site-directed mutagenesis of human 15-lipoxygenase, despite the low amino acid sequence identity between the soybean and mammalian enzymes. Both a positively charged amino acid residue (Arg402) and an aromatic amino acid residue (Phe414) are identified as critical for the binding of fatty acid substrates in human 15-lipoxygenase. Thus, binding determinants shown to be characteristic of non-enzymatic fatty acid-binding proteins are now implicated in the substrate binding pocket of lipoxygenases.

Understanding the P1' specificity of the matrix metalloproteinases: effect of S1' pocket mutations in matrilysin and stromelysin-1.

Matrilysin (MAT) prefers leucine over residues that have aromatic side chains at the P1' position of peptide and protein substrates, while stromelysin (HFS) has a broader specificity. The X-ray structures of these enzymes show that their respective S1' subsites differ primarily due to the amino acids present at positions 214 and 215. To examine the role that these residues play in determining P1' specificity, the amino acids at these positions in matrilysin have been replaced by those found in stromelysin (MAT: Y214L, MAT:A215V, and MAT:Y214L/A215V). The specificity and activity of MAT:A215V are similar to those of wild type matrilysin. Both MAT:Y214L and MAT:Y214L/A215V, however, have P1' specificities that are more similar to stromelysin than matrilysin. Specifically, these enzymes exhibit an 8- to 9-fold reduction in kcat/KM toward a peptide substrate with Leu in subsite P1' relative to wild type matrilysin. This is predominantly the result of an approximate 5-fold decrease in kcat. The KM values only partially increase toward the value observed for stromelysin. Studies of the pre-steady-state reaction of wild type and mutant matrilysin with substrates with Leu and Tyr residues in the P1' position confirm that the KM values for these reactions reflect KD values for substrate binding. Thus, replacement of a single tyrosine residue in the S1' pocket of matrilysin by leucine alters its P1' specificity to resemble that of stromelysin. In contrast, alteration of the S1' subsite of stromelysin (HFS:L214Y/V215A) to resemble matrilysin increases activity (i.e., higher kcat/KM) toward peptide substrates with both leucine and residues with aromatic side chains in the P1' position with only a partial increase in specificity for Leu. These increases in activity are the result of decreases in the KM values for these reactions.

Purification of human matrilysin produced in Escherichia coli and characterization using a new optimized fluorogenic peptide substrate.

Human promatrilysin (matrix metalloproteinase-7) has been produced in Escherichia coli as an N-terminal fusion protein with ubiquitin. The insoluble product was solubilized, refolded, and activated with amino-phenylmercuric acetate. Activation of the fusion protein demonstrated kinetics and intermediates that were very similar to those observed during activation of promatrilysin produced in Chinese Hamster Ovary (CHO) cells. Following activation, matrilysin was purified to > 95% homogeneity using a Sepharose-Pro-Leu-Gly-NHOH affinity column. The matrilysin purified by this procedure is indistinguishable from the enzyme purified from CHO cells with respect to the kinetic parameters for hydrolysis of a peptide substrate and the ability to obtain diffraction quality crystals in the presence of an inhibitor of the enzyme. Additionally, to facilitate detailed kinetic analyses of matrilysin, a new fluorogenic peptide substrate with the optimized sequence Dnp-Arg-Pro-Leu-Ala-Leu-Trp-Arg-Ser (Dnp, dinitrophenyl) has been synthesized. This peptide is the best substrate developed for matrilysin thus far with Km and kcat values of 26 microM and 5.0 s-1, respectively.

Matrilysin-inhibitor complexes: common themes among metalloproteases.

Matrix metalloproteases are a family of enzymes that play critical roles in the physiological and pathological degradation of the extracellular matrix. These enzymes may be important therapeutic targets for the treatment of various diseases where tissue degradation is part of the pathology, such as cancer and arthritis. Matrilysin is the smallest member of this family of enzymes, all of which require zinc for catalytic activity. The first X-ray crystal structures of human matrilysin are presented. Inhibitors of metalloproteases are often characterized by the chemical group that interacts with the active site zinc of the protein. The structures of matrilysin complexed with hydroxamate (maximum resolution 1.9 A), carboxylate (maximum resolution 2.4 A), and sulfodiimine (maximum resolution 2.3 A) inhibitors are presented here and provide detailed information about how each functional group interacts with the catalytic zinc. Only the zinc-coordination group is variable in this series of inhibitors. Examination of these inhibitor-matrilysin complexes emphasizes the dominant role the zinc-coordinating group plays in determining the relative potencies of the inhibitors. The structures of these matrilysin-inhibitor complexes also provide a basis for comparing the catalytic mechanism of MMPs and other metalloproteins.

Identification of the molecular trigger for allosteric activation in glycogen phosphorylase.

Activation of protein function through phosphorylation can be mimicked by the engineering of specific metal binding sites. The addition of two histidine residues to glycogen phosphorylase allows enzymatic activation by transition metals in a cooperative and allosteric manner. Crystal structures of the metallo-enzyme have been determined and show that the structural transition induced upon metal binding (Ni2+) is, in part, analogous to the mode of activation of the native enzyme. The designed metal activation site allows assignment of the structural changes which trigger activation in this allosteric enzyme and, further, provide insight into the evolutionary development of multiple activation sites.

Purification of glycogen phosphorylase isozymes by metal-affinity chromatography.

Mammalian phosphorylase isozymes from muscle, brain and liver were expressed in Escherichia coli and purified from the crude bacterial cell extracts in one step using a copper-loaded, metal-affinity matrix. Good chromatographic behavior, enzyme activity and protein stability were maintained by judicious choice of pH and buffer which contained 250 mM sodium chloride and 25 mM beta-glycerophosphate at pH 7.0. Small amounts of beta-mercaptoethanol and EDTA in the buffers further stabilized the enzymes, but stripped some of the metal from the column which, nonetheless, retained good chromatographic characteristics. Owing to the presence of multiple surface histidine residues in the phosphorylase dimers, good enzyme purities (90-98%) and recoveries (>90%) were routinely obtained from crude bacterial lysates after two passes through the copper column. Of the various metal ions which were investigated, Cu2+ gave the best chromatographic results. Imidazole gradients at constant pH were used to selectively desorb the phosphorylase from the metal column whose capacity for phosphorylase binding in the presence of bacterial proteins exceeded 30 mg enzyme per milliliter of matrix.

Cooperative binding is not required for activation of muscle phosphorylase.

Muscle and liver glycogen phosphorylase isozymes differ in their responsiveness to the activating ligand AMP. The muscle enzyme, which supplies glucose in response to strenuous activity, binds AMP cooperatively, and its enzymatic activity becomes greatly enhanced. The liver isozyme regulates the level of blood glucose, and AMP is not the primary activator. In muscle glycogen phosphorylase, the residue proline 48 links two secondary structural elements that bind AMP. This amino acid residue is replaced with a threonine in the liver isozyme; unlike the muscle enzyme, liver binds AMP noncooperatively, and the enzymatic activity is not greatly increased. We have substituted proline 48 in the muscle enzyme with threonine, alanine, and glycine and characterized the recombinant enzymes kinetically and structurally to determine if proline at this position is critical for cooperative AMP binding and activation. Importantly, all of the engineered enzymes were fully activated by phosphorylation, indicating that enzymatic activity was not compromised. Only the mutant enzyme with alanine at position 48 responds like the wild-type enzyme to the presence of AMP, indicating that proline is not absolutely required for full cooperative activation. The substitution of either threonine or glycine at this position, however, creates enzymes that no longer bind AMP cooperatively. The enzyme with threonine at position 48 further mimics the liver enzyme, in that the maximal enzymatic activity is also reduced. Significantly, the glycine substitution caused the enzyme to be fully activated by AMP, although binding was not cooperative. The hyperactivation of the glycine mutant by AMP suggests that the total free energy of activation has decreased.(ABSTRACT TRUNCATED AT 250 WORDS)

Tracking conformational states in allosteric transitions of phosphorylase.

An intrinsic molecular property of a protein domain can be determined by calculating its principal axes from the inertia tensor matrix. The mass-weighted principal axes can be used to calculate an ellipsoid representing the shape of the protein domain, providing an easy means of visualizing domain movements. Most importantly, the mass-weighted principal axes provide an intuitive means of characterizing domain relationships within a protein, as well as the disposition of domains in different protein conformers. Thus, this method provides a simple, quantitative description of differences of domain positions within various protein structures. We show the utility of this method by characterizing the quaternary and tertiary differences as observed in eight structures of phosphorylated or dephosphorylated glycogen phosphorylase with different effectors bound. This analysis revealed domain movements which were characteristic of the activated phosphorylase structures. The monomers of the phosphorylase dimer were found to move apart by a 2.5-A translation and to rotate apart, in three orthogonal directions, by a minimum of 3.2 degrees. Analysis of the three domains within the phosphorylase monomer showed that both simple and complex domain movements occur and that multiple domain configurations are energetically stable. We suggest that the C-terminal domain of phosphorylase moves along a simple path in the transition from an inactive to active conformation. The direction of translation and rotation is consistent, but the magnitude is variable. In contrast, this analysis showed that the activation domain did not behave as a rigid body, and therefore, the motion of this domain is not as easily characterized.

Phosphorylase: a biological transducer.

A transducer is a device that receives energy from one system and transmits it, often in a different form, to another. Glycogen phosphorylase receives information from the cell or organism in the form of metabolic signals. The energy associated with the binding of these ligand signals is integrated and transmitted at an atomic level, allowing precise adjustment of the enzymatic activity. Understanding this elegant allosteric control has required several different approaches, but the structural requirements of allostery are being defined.

Expression of the catalytic subunit of phosphorylase phosphatase (protein phosphatase-1) in Escherichia coli.

The catalytic subunit of rabbit skeletal muscle protein phosphatase-1 was expressed in Escherichia coli. Expression of phosphatase-1 in the pET3a vector, which is based on the use of the T7 promoter, resulted in the expression of the enzyme as an insoluble aggregate. The insoluble enzyme could be renatured by high dilutions of the urea-solubilized protein in buffers containing dithiothreitol, Mn2+, and high NaCl concentrations. However, under all conditions tested, only partial (less than 5%) renaturation was achieved. A second attempt was made using a vector with the trp-lac hybrid promoter. In this case it was possible to express the enzyme as a soluble protein at levels of 3-4% of the soluble E. coli protein. The recombinant enzyme was purified by DEAE-Sepharose and heparin-Sepharose chromatography. Approximately 20 mg of purified enzyme was reproducibly obtained from the cells derived from 2 liters of culture. The purified enzyme had a specific activity toward phosphorylase alpha comparable to that reported for the authentic protein and had an Mr of 37,000 by sodium dodecyl sulfate-polyacrylamide gel electrophoresis. The recombinant enzyme displayed similar sensitivities to inhibition by inhibitor-2, okadaic acid, and microcystin-LR as for the protein isolated from rabbit muscle. At all stages of purification the recombinant phosphatase behaved as an essentially inactive enzyme that required the presence of microM Mn2+ for full expression of its activity.

Expression of the protease inhibitor ecotin and its co-crystallization with trypsin.

We have expressed the serine protease inhibitor ecotin to high levels (greater than 400 mg/l of cell culture) in its natural mileau, the Escherichia coli periplasm, using the endogenous signal peptide and the heterologous tac promoter. After induction, functional, soluble ecotin comprises 15% of total cellular protein. This expression system has facilitated initiation of a crystallographic study to determine the structural basis for inhibition of the pancreatic serine proteases by ecotin. Ecotin was co-crystallized with rat trypsin mutant D102N. Preliminary crystallographic analysis of co-crystals showed that they diffract to at least 2.7 A, and indicate that they belong to the monoclinic space group, P21. The cell constants are a = 52.0 A, b = 93.3 A, c = 160.7 A, and beta = 96 degrees. Four molecules each of trypsin and ecotin are found in the asymmetric unit.

An engineered liver glycogen phosphorylase with AMP allosteric activation.

Liver and muscle glycogen phosphorylases, which are products of distinct genes, are both activated by covalent phosphorylation, but in the unphosphorylated (b) state, only the muscle isozyme is efficiently activated by the allosteric activator AMP. The different responsiveness of the phosphorylase isozymes to allosteric ligands is important for the maintenance of tissue and whole body glucose homeostasis. In an attempt to understand the structural determinants of differential sensitivity of the muscle and liver isozymes to AMP, we have developed a bacterial expression system for the liver enzyme, allowing native and engineered proteins to be expressed and characterized. Engineering of the single amino acid substitutions Thr48Pro, Met197Thr and the double mutant Thr48Pro, Met197Thr in liver phosphorylase, and Pro48Thr in muscle phosphorylase, did not qualitatively change the response of the two isozymes to AMP. These sites had previously been implicated in the configuration of the AMP binding site. However, when nine amino acids among the first 48 in liver phosphorylase were replaced with the corresponding muscle phosphorylase residues (L1M2-48L49-846), the engineered liver enzyme was activated by AMP to a higher maximal activity than native liver phosphorylase. Interestingly, the homotropic cooperativity of AMP binding was unchanged in the engineered phosphorylase b protein, and heterotropic cooperativity between the glucose-1-phosphate and AMP sites was only slightly enhanced. The native liver, native muscle and L1M2-48L49-846 phosphorylases were converted to the a form by treatment with purified phosphorylase kinase; the maximal activity of the chimeric a enzyme was greater than the native liver a enzyme and approached that of muscle phosphorylase a. From these results we suggest that tissue-specific phosphorylase isozymes have evolved a complex mechanism in which the N-terminal 48 amino acids modulate intrinsic activity (Vmax), probably by affecting subunit interactions, and other, as yet undefined regions specify the allosteric interactions with ligands and substrates.

Temperature-sensitive production of rabbit muscle glycogen phosphorylase in Escherichia coli.

In order to understand how allosteric switches regulate both the catalytic activity and molecular interactions of glycogen phosphorylase, it is necessary to design and analyze variant proteins that test hypotheses about the structural details of the allosteric mechanism. Essential to such an investigation is the ability to obtain large amounts of variant proteins. We developed a system for obtaining milligram amounts (greater than 20 mg/l) of rabbit muscle phosphorylase from bacteria. Phosphorylase aggregates as inactive protein when a strong bacterial promoter is used under full inducing conditions and normal growth conditions. However, when the growth temperature of bacteria expressing phosphorylase is reduced to 22 degrees C we obtain active muscle phosphorylase. The degree to which the induced expression of phosphorylase protein is temperature sensitive depends on the strain of bacteria used. New assay and purification methods were developed to allow rapid purification of engineered phosphorylase proteins from bacterial cultures. The rabbit muscle phosphorylase obtained from the bacterial expression system is enzymatically identical to the enzyme purified from rabbit muscle. The expressed protein crystallizes in the same conditions used for growing crystals of protein from rabbit muscle and the crystal form is isomorphous. Rabbit muscle phosphorylase is one of the largest oligomeric mammalian enzymes successfully expressed in Escherichia coli. Our results indicate that optimization of a combination of growth and induction conditions will be important in the expression of other heterologous proteins in bacteria.

Purification and crystallization of 15-lipoxygenase from rabbit reticulocytes.

We report a new purification of rabbit reticulocyte 15-lipoxygenase that has resulted in the first crystallization of a mammalian lipoxygenase. The enzyme was purified to homogeneity (greater than 98% pure by SDS-PAGE) using high pressure liquid chromatography on hydrophobic-interaction, hydroxyapatite and cation-exchange columns. Crystals were grown by the vapor diffusion method from concentrated solutions of the protein in sodium phosphate buffer, pH 7.0. The hexagonal, rod-shaped crystals were on average 0.09 mm x 0.09 mm x 0.4 mm, with approximate unit cell dimensions of a = b = 260 A, c = 145 A. The crystals diffract to 5 A resolution.

Primary structure of rabbit skeletal muscle glycogen synthase deduced from cDNA clones.

The complete amino acid sequence of rabbit skeletal muscle glycogen synthase was deduced from cDNA clones with a composite length of 3317 bp. An mRNA of 3.6 kb was identified by Northern blot analysis of rabbit skeletal muscle RNA. The mRNA coded for a protein of 734 residues with a molecular weight of 83,480. The deduced NH2-terminal and COOH-terminal sequences corresponded to those reported for the purified protein, indicating the absence of any proteolytic processing. At the nucleotide level, the 5' untranslated and coding regions were 79 and 90% identical for rabbit and human muscle glycogen synthases, whereas the 3' untranslated regions were significantly less similar. The enzymes had 97% amino acid sequence identity. Interestingly, the NH2 and COOH termini of rabbit and human muscle glycogen synthase, the regions of phosphorylation, showed the greatest sequence variation (15 of 19 mismatches and two insertion/deletion events), which may indicate different evolutionary constraints in the regulatory and catalytic regions of the molecule.