Single Peptide Bond Research Articles

The internal thioester and the covalent binding properties of the complement proteins C3 and C4.

The covalent binding of complement components C3 and C4 is critical for their activities. This reaction is made possible by the presence of an internal thioester in the native protein. Upon activation, which involves a conformational change initiated by the cleavage of a single peptide bond, the thioester becomes available to react with molecules with nucleophilic groups. This description is probably sufficient to account for the binding of the C4A isotype of human C4 to amino nucleophiles. The binding of the C4B isotype, and most likely C3, to hydroxyl nucleophiles, however, involves a histidine residue, which attacks the thioester to form an intramolecular acyl-imidazole bond. The released thiolate anion then acts as a base to catalyze the binding of hydroxyl nucleophiles, including water, to the acyl function. This mechanism allows the complement proteins to bind to the hydroxyl groups of carbohydrates found on all biological surfaces, including the components of bacterial cell walls. In addition, the fast hydrolysis of the thioester provides a means to contain this very damaging reaction to the immediate proximity of the site of activation.

Identification of a Hydrophobic Exosite on Tissue Type Plasminogen Activator That Modulates Specificity for Plasminogen

A wide variety of important biological processes, including both the formation and dissolution of blood clots, depend on specific cleavage of individual target proteins by serine proteases. For example, tissue type plasminogen activator (t-PA), a trypsin-like enzyme that catalyzes the rate-limiting step of the endogenous fibrinolytic cascade, has only one known substrate in vivo, a single peptide bond (Arg561-Val562) in the proenzyme plasminogen. We have previously suggested that the specificity of t-PA for plasminogen is mediated in part by direct protein-protein interactions between the protease domain of t-PA and plasminogen that are distinct from those occurring within t-PA's active site. We demonstrate in this study that residues 420-423 of t-PA, which form a fully solvent-exposed, hydrophobic region of a surface loop mapping near one edge of the active site of t-PA, form, or are essential for the integrity of, an important, secondary site of interaction between t-PA and plasminogen that significantly modulates the rate of plasminogen activation in the absence, but not the presence, of fibrin. Identification of this secondary site of interaction between t-PA and plasminogen provides new insight into molecular details of the evolution of stringent substrate specificity by t-PA and suggests a novel strategy to enhance the fibrin dependence of plasminogen activation by t-PA. While the activity of wild type t-PA is stimulated by fibrin by a factor of approximately 650, the activity of two variants characterized in this study, t-PA/R275E,P422G and t-PA/R275E,P422E, is stimulated by a factor of approximately 39,000 or 61,000, respectively. It is therefore possible that, compared with wild type t-PA, the two variants would display enhanced "clot selectivity" in vivo due to reduced activity in the circulation but full activity at a site of fibrin deposition.

A Sensitive and Rapid Fluorescence-Based Assay for Determination of Tetanus Toxin Peptidase Activity

The light chain of tetanus toxin (TeNT-L chain), endowed with a zinc metalloendopeptidase activity, cleaves specifically the vesicle-associated membrane protein (VAMP), also called synaptobrevin, at a single peptide bond (Gln76–Phe77), resulting in the blockade of neuroexocytosis. The 50-mer synaptobrevin peptide S 39–88, synthesized by solid-phase peptide synthesis, was determined to be the minimum substrate of TeNT still notably hydrolyzed by TeNT-L chain. In this peptide, Tyr88was substituted by the highly fluorescent amino acid (L) pyrenylalanine (Pya) which was synthesized in good yields by an enantioselective method. The fluorescent substrate [Pya88] S 39–88 was cleaved four times more rapidly by TeNT-L chain than S 39–88 (kcat/Km= 9635 and 2455M−1·min−1, respectively). One of the two metabolites formed by the action of TeNT L chain, [Pya88] S 77–88, was easily separated from the substrate in one step using Sep-Pak Vac C18cartridges and its concentration quantified by fluorescence. This novel enzymatic assay, which could be easily extended to other clostridial neurotoxins, is a major improvement in term of sensitivity and time saving, compared to currently used methods (SDS–PAGE, HPLC). It lends itself readily to automation for large-scale screening of selective and potent inhibitors of these neurotoxins which remain to be developed.

Specific cleavage of immunoglobulin G by copper ions.

The hinge region of a recombinant-DNA-produced human IgG1 (Campath 1H) is specifically cleavable at a single copper-sensitive peptide bond, yielding a distinct fragment resolved by size-exclusion high-performance liquid chromatography. This novel metal ion-catalysed cleavage at slightly alkaline pH is inhibited by EDTA and its rate is reduced at slightly acidic conditions (pH 5-6) and accelerated by increasing concentrations of cupric ion and higher temperature. Complete cleavage was observed after incubation at pH 8 for 24 h with 1 mM CuCl2. Sequence analysis determined the cleavage site to be the Lys226-Thr227 bond in the hinge-region sequence DKTHT. Cleavage of other IgGs was observed to varying degrees, and specific cleavage of synthetic peptides containing this pentapeptide sequence was also observed.

Thermodynamic Stability Effects of Single Peptide Bond Hydrolysis in Protein Inhibitors of Serine Proteinases

Kunitz type soybean trypsin inhibitor (STI) and basic pancreatic trypsin inhibitor (BPTI) were used as model proteins to measure the thermodynamic consequences of single peptide bond hydrolysis. For each inhibitor the reactive site cleaved form was prepared and, additionally, each inhibitor was selectively cleaved at a Met residue. Selective cleavage generally led to reducing of the T den value from 7 K up to 75 K. For STI cleaved at Met84 a slight stabilization (increase of T den by 1.0 K) was observed. In terms of Δ G den , the difference between the most extreme cleavage effect was 11.44 kcal/mole, much larger than resulting from the theoretical effects of crosslinks. It was found that hydrolysis of a single peptide bond affects not only entropy, but also enthalpy and heat capacity parameters. Moreover, the sign of change is opposite for two inhibitors: Δ H den and Δ S den increase for both cleaved forms of STI, while they decrease for the two nicked forms of BPTI. To understand the stability effects, a thermodynamic cycle analysis was applied based on comparison of stabilities of intact and cleaved protein with peptide bond hydrolysis equilibria in native and denatured states. The cycle revealed a good agreement of the theoretical effect of crosslink elimination with a free energy difference for hydrolysis of a single peptide bond in a denatured protein. It appears that hydrolysis constants for single peptide bonds in a native protein span over at least 20 orders of magnitude. They are very low for peptide bonds placed in α-helices and very high if cleavage reaction leads to a formation of a new secondary structure element.

Proteolytic cleavage and activation of pro-macrophage-stimulating protein by resident peritoneal macrophage membrane proteases.

Macrophage stimulating protein (MSP), which is secreted as biologically inactive pro-MSP, is activated to MSP by cleavage at a single peptide bond. Our objectives were to determine the form of MSP in circulating blood and to study proteolytic activation of pro-MSP by its target cell. Western blot of immunoaffinity-purified serum MSP showed that all the protein was pro-MSP, without detectable MSP. The circulating form of the protein is therefore pro-MSP, and conversion to MSP does not occur when blood is shed. Incubation of radiolabeled pro-MSP with murine peritoneal macrophages caused proteolytic cleavage to predominantly inactive fragments. Among several protease inhibitors, soybean trypsin inhibitor was one of two that inhibited nonspecific cleavage and revealed a macrophage proteolysis of pro-MSP, and certain concentrations enhanced cleavage to mature MSP. Macrophage membranes had nonspecific and specific pro-MSP proteolytic activity, which was not present in macrophage culture fluids. The results suggest that control of MSP activity can occur at the level of the target cell by proteolytic cleavage of pro-MSP to mature MSP or to inactive fragments.

Novel neuropeptide Y processing in human cerebrospinal fluid from depressed patients.

In vitro processing of neuropeptide Y (NPY) in cerebrospinal fluid (CSF) of patients with depression was monitored by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS). Single peptide bonds in NPY were cleaved to yield N- or C-terminal fragments. Multiple cleavage to form internal peptides was unimportant. Degradation rates varied between individuals, whereas the product distributions were fairly constant. Other peptides did not evidence such proteolysis. MALDI-TOF MS will facilitate extensive investigations of NPY processing that could provide the basis for clinical assays and illuminate the pathophysiology related to depression.

Molecular Cloning and Expression of Rat Brain Endopeptidase 3.4.24.16

We have isolated by immunological screening of a lambda ZAPII cDNA library constructed from rat brain mRNAs a cDNA clone encoding endopeptidase 3.4.24.16. The longest open reading frame encodes a 704-amino acid protein with a theoretical molecular mass of 80,202 daltons and bears the consensus sequence of the zinc metalloprotease family. The sequence exhibits a 60.2% homology with those of another zinc metallopeptidase, endopeptidase 3.4.24.15. Northern blot analysis reveals two mRNA species of about 3 and 5 kilobases in rat brain, ileum, kidney, and testis. We have transiently transfected COS-7 cells with pcDNA3 containing the cloned cDNA and established the overexpression of a 70-75-kDa immunoreactive protein. This protein hydrolyzes QFS, a quenched fluorimetric substrate of endopeptidase 3.4.24.16, and cleaves neurotensin at a single peptide bond, leading to the formation of neurotensin (1-10) and neurotensin (11-13). QFS and neurotensin hydrolysis are potently inhibited by the selective endopeptidase 3.4.24.16 dipeptide blocker Pro-Ile and by dithiothreitol, while the enzymatic activity remains unaffected by phosphoramidon and captopril, the specific inhibitors of endopeptidase 3.4.24.11 and angiotensin-converting enzyme, respectively. Altogether, these physicochemical, biochemical, and immunological properties unambiguously identify endopeptidase 3.4.24.16 as the protein encoded by the isolated cDNA clone.

Structure and function of tetanus and botulinum neurotoxins.

Tetanus and botulinum neurotoxins are produced by Clostridia and cause the neuroparalytic syndromes of tetanus and botulism. Tetanus neurotoxin acts mainly at the CNS synapse, while the seven botulinum neurotoxins act peripherally. Clostridial neurotoxins share a similar mechanism of cell intoxication: they block the release of neurotransmitters. They are composed of two disulfide-linked polypeptide chains. The larger subunit is responsible for neurospecific binding and cell penetration. Reduction releases the smaller chain in the neuronal cytosol, where it displays its zinc-endopeptidase activity specific for protein components of the neuroexocytosis apparatus. Tetanus neurotoxin and botulinum neurotoxins B, D, F and G recognize specifically VAMP/ synaptobrevin. This integral protein of the synaptic vesicle membrane is cleaved at single peptide bonds, which differ for each neurotoxin. Botulinum A, and E neurotoxins recognize and cleave specifically SNAP-25, a protein of the presynaptic membrane, at two different sites within the carboxyl-terminus. Botulinum neurotoxin type C cleaves syntaxin, another protein of the nerve plasmalemma. These results indicate that VAMP, SNAP-25 and syntaxin play a central role in neuroexocytosis. These three proteins are conserved from yeast to humans and are essential in a variety of docking and fusion events in every cell. Tetanus and botulinum neurotoxins form a new group of zinc-endopeptidases with characteristic sequence, mode of zinc coordination, mechanism of activation and target recognition. They will be of great value in the unravelling of the mechanisms of exocytosis and endocytosis, as they are in the clinical treatment of dystonias.

Enhancement of the Enzymatic Activity of Single-chain Urokinase Plasminogen Activator by Soluble Urokinase Receptor

Single-chain urokinase (scuPA), the unique form of urokinase secreted by cells, is converted to an active two-chain molecule through the cleavage of a single peptide bond by plasmin and other specific proteinases. Although scuPA may express limited enzymatic activity, its contribution to plasminogen activation on cell surfaces remains uncertain. Further, although it is well known that scuPA binds to a specific extracellular urokinase-type plasminogen activator receptor, the effect of this interaction on the enzymatic activity of scuPA has not been described. In the present paper we report that the binding of scuPA to cellular an to recombinant soluble urokinase-type plasminogen activator receptors (suPAR) increases its catalytic activity as measured by the cleavage of a urokinase-specific chromogenic substrate. suPAR increased the Vmax of scuPA 5-fold with little change in its Km. suPAR also stimulated the plasminogen activator activity of scuPA by decreasing its Km for Glu-plasminogen from 1.15 microM to 0.022 microM and by increasing the kcat of this reaction from 0.0015 to 0.022 s-1. Preincubation of scuPA with suPAR also enhances its susceptibility to inhibition by plasminogen activator inhibitor type 1, consistent with exposure of its catalytic site. The activity of scuPA bound to suPAR is not accompanied by cleavage of scuPA, which continues to migrate as a single band in SDS-polyacrylamide gel electrophoresis under reducing conditions. Moreover, suPAR increases the plasminogen activator activity of a plasmin-insensitive variant, scuPA (scuPA-Glu158), as well. Enhancement of scuPA activity by suPAR is both prevented and reversed by its aminoterminal fragment (amino acids 1-135), which competes for receptor binding, suggesting that continued binding to the receptor is required for expression of scuPA's enzymatic activity. Thus, our data suggest that scuPA may undergo a reversible transformation between a latent and an active state. The urokinase receptor may induce or stabilize scuPA in its active conformation, thereby contributing to the initiation of plasminogen activation on cell surfaces.

Substrate Specificity of Tissue Type Plasminogen Activator: CHARACTERIZATION OF THE FIBRIN INDEPENDENT SPECIFICITY OF t-PA FOR PLASMINOGEN

Tissue-type plasminogen activator (t-PA) is a remarkably specific protease: the only known substrate of this enzyme in vivo is a single peptide bond (Arg560-Val561) within the proenzyme plasminogen. Part of the substrate specificity of t-PA is due to a ternary interaction between fibrin, t-PA and plasminogen which reduces the Km of t-PA for plasminogen by a factor of 440. However, even in the absence of fibrin, t-PA continues to hydrolyze plasminogen more rapidly than does trypsin, a homologous serine protease. We have measured the extent of the specificity of t-PA for plasminogen by assaying t-PA and trypsin toward substrates modeled after the peptide sequence in plasminogen surrounding Arg560-Val561. Surprisingly, t-PA hydrolyzes these substrates with kcat/Km values which are 28,000-210,000-fold lower than those obtained using trypsin. Both the high activity toward plasminogen and the low activity toward peptides are also exhibited by the isolated protease domain. This suggests that the protease domain, in spite of its high homology to the nonspecific enzyme trypsin, is inherently specific for recognition of one or more structural features displayed by native plasminogen.

Clostridial neurotoxins as tools to investigate the molecular events of neurotransmitter release

The clostridial neurotoxins responsible for tetanus and botulism are eight different proteins, composed of two disulfide-linked polypeptide chains. They bind specifically to the presynaptic membrane via the heavy chain, while the light chain enters the cytosol of the neurons, where it displays a zinc-endopeptidase activity directed to proteins of the neuroexocytosis apparatus. Tetanus neurotoxin and botulinum neurotoxin serotypes B, D, F and G cleave specifically and at single different peptide bonds VAMP/synaptobrevin, a component of small synaptic vesicles. In contrast, the other neurotoxins catalyze the hydrolysis of proteins of the presynaptic membrane. Serotypes A and E of botulinum neurotoxin cleave SNAP-25, at different sites located within the carboxyl-terminus, while the specific target of serotype C is syntaxin.

Mechanism of action of tetanus and botulinum neurotoxins.

The clostridial neurotoxins responsible for tetanus and botulism are metallo-proteases that enter nerve cells and block neurotransmitter release via zinc-dependent cleavage of protein components of the neuroexocytosis apparatus. Tetanus neurotoxin (TeNT) binds to the presynaptic membrane of the neuromuscular junction and is internalized and transported retroaxonally to the spinal cord. Whilst TeNT causes spastic paralysis by acting on the spinal inhibitory interneurons, the seven serotypes of botulinum neurotoxins (BoNT) induce a flaccid paralysis because they intoxicate the neuromuscular junction. TeNT and BoNT serotypes B, D, F and G specifically cleave VAMP/synaptobrevin, a membrane protein of small synaptic vesicles, at different single peptide bonds. Proteins of the presynaptic membrane are specifically attacked by the other BoNTs: serotypes A and E cleave SNAP-25 at two different sites located within the carboxyl terminus, whereas the specific target of serotype C is syntaxin.

Single peptide bond hydrolysis/resynthesis in squash inhibitors of serine proteinases. 1. Kinetics and thermodynamics of the interaction between squash inhibitors and bovine .beta.-trypsin

The substrate and inhibitory parameters are described for the interaction between Cucurbita maxima trypsin inhibitor I (CMTI I) and bovine beta-trypsin. The data are fully consistent with the reactive site hypothesis and the standard mechanism proposed for the protein inhibitor-serine proteinase interaction. The second-order association rate constant (k(on)) for the interaction of the intact inhibitor and trypsin is high, above 10(6) M-1 s-1. The same value is only 22-fold lower for the reactive site hydrolyzed inhibitor. This result implicates a very low transition-state barrier for the hydrolysis of the Arg5-Ile6 reactive site peptide bond. The equilibrium constant Ka (= 1/Km,f) and K(assoc) change by 6 orders of magnitude in the pH range 4.0-8.3. The steady-state parameters for the hydrolysis and resynthesis of the reactive site have been determined over the pH range 3.2-8.3. Catalytic rate constants, but not kcat/km, exhibit strong pH dependence. The dependence of the hydrolysis constant (Khyd) on pH fits the simplest form of the Dobry equation, indicating that after the hydrolysis of the reactive site, pK values of any preexistent groups are not perturbed. It is suggested that a major factor leading to high kcat/Km values is the presence of Arg or Lys residues at the P1 position. Low values of Km result from a conservation of the ground-state conformation of the inhibitor binding loop upon the complex formation. The crucial stage of the reactive site hydrolysis seems to be associated with a change of basic side-chain interactions within the S1 binding pocket.

Identification of the nerve terminal targets of botulinum neurotoxin serotypes A, D, and E.

Botulinum neurotoxins are metalloproteins with one zinc atom bound to the zinc binding motif of zinc endopeptidases. Here we show that botulinum neurotoxin serotypes A, D, and E are zinc endoproteases specific for components of the synaptic vesicle docking and fusion complex. Serotypes A and E cleave SNAP-25, a 25-kDa protein of the synaptic terminal, while serotype D is specific for VAMP/synaptobrevin, a membrane protein of synaptic vesicles. Both rat brain VAMP isoforms are cleaved at a single Lys-Leu peptide bond. The proteolytic activity of these neurotoxins is inhibited by EDTA and captopril.

Antibodies Against Rat Brain Vesicle‐Associated Membrane Protein (Synaptobrevin) Prevent Inhibition of Acetylcholine Release by Tetanus Toxin or Botulinum Neurotoxin Type B

Tetanus and botulinum B neurotoxins are zinc endopeptidases that cleave vesicle-associated membrane protein (VAMP or synaptobrevin) at a single peptide bond. To test the possibility that in vivo also the toxin-induced blockade of neurotransmission is due to cleavage of VAMP, rat brain VAMP-specific antibodies were raised in rabbits. IgGs purified from one antiserum, which bind specifically to rat brain VAMP, also specifically recognize proteins from Aplysia californica in immunoblotting. When injected into neurons in the buccal ganglion of Aplysia, these IgGs did not affect the release of acetylcholine but effectively prevented the inhibitory action of both toxins on neurotransmitter release, thus indicating that the block of neurotransmission by these neurotoxins is consequent to the cleavage of VAMP or specific interaction with VAMP.

Botulinum neurotoxin serotype F is a zinc endopeptidase specific for VAMP/synaptobrevin

Botulinum neurotoxin serotype F contains the zinc binding motif of zinc endopeptidases. Atomic adsorption analysis of highly purified toxin preparation revealed the presence of one atom of zinc per molecule of toxin, which could be removed with EDTA or o-phenanthroline. The light chain of the neurotoxin was shown to have a zinc-dependent protease activity specific for VAMP/synaptobrevin, an integral membrane protein of synaptic vesicles. Both isoforms of rat VAMP were cleaved at the same site corresponding to the single Gln-Lys peptide bond present in their sequences. This proteolytic activity was inhibited by EDTA, o-phenanthroline, and captopril as well as by VAMP peptides spanning the cleavage site.

Reconstitution of active catalytic trimer of aspartate transcarbamoylase from proteolytically cleaved polypeptide chains.

Treatment of the catalytic (C) trimer of Escherichia coli aspartate transcarbamoylase (ATCase) with alpha-chymotrypsin by a procedure similar to that used by Chan and Enns (1978, Can. J. Biochem. 56, 654-658) has been shown to yield an intact, active, proteolytically cleaved trimer containing polypeptide fragments of 26,000 and 8,000 MW. Vmax of the proteolytically cleaved trimer (CPC) is 75% that of the wild-type C trimer, whereas Km for aspartate and Kd for the bisubstrate analog, N-(phosphonacetyl)-L-aspartate, are increased about 7- and 15-fold, respectively. CPC trimer is very stable to heat denaturation as shown by differential scanning microcalorimetry. Amino-terminal sequence analyses as well as results from electrospray ionization mass spectrometry indicate that the limited chymotryptic digestion involves the rupture of only a single peptide bond leading to the production of two fragments corresponding to residues 1-240 and 241-310. This cleavage site involving the bond between Tyr 240 and Ala 241 is in a surface loop known to be involved in intersubunit contacts between the upper and lower C trimers in ATCase when it is in the T conformation. Reconstituted holoenzyme comprising two CPC trimers and three wild-type regulatory (R) dimers was shown by enzyme assays to be devoid of the homotropic and heterotropic allosteric properties characteristic of wild-type ATCase. Moreover, sedimentation velocity experiments demonstrate that the holoenzyme reconstituted from CPC trimers is in the R conformation. These results indicate that the intact flexible loop containing Tyr 240 is essential for stabilizing the T conformation of ATCase. Following denaturation of the CPC trimer in 4.7 M urea and dilution of the solution, the separate proteolytic fragments re-associate to form active trimers in about 60% yield. How this refolding of the fragments, docking, and association to form trimers are achieved is not known.

Importance of the evolutionarily conserved glycine residue in the N-terminal region of human cystatin C (Gly-11) for cysteine endopeptidase inhibition.

Human cystatin C variants in which the evolutionarily conserved Gly-11 residue has been replaced by residues with positively charged (Arg), negatively charged (Glu), bulky hydrophobic (Trp), or small (Ser or Ala) side-chains have been produced by site-directed mutagenesis and expression in Escherichia coli. The five variants were isolated and structurally verified. Their inhibitory properties were compared with those of wild-type recombinant cystatin C by determination of the equilibrium constants for dissociation (Ki) of their complexes with the cysteine endopeptidases papain and human cathepsin B and with the cysteine exopeptidase dipeptidyl peptidase I. The Ser-11 and Ala-11 cystatin C variants displayed Ki values for the two endopeptidases that were approx. 20-fold higher than those of wild-type cystatin C, while the corresponding values for the Trp-11. Arg-11 and Glu-11 variants were increased by a factor of about 2000. In contrast, the Ki values for the interactions of all five variants with the exopeptidase differed from that of wild-type cystatin C by a factor of less than 10. Wild-type cystatin C and the Ser-11, Ala-11 and Glu-11 variants were incubated with neutrophil elastase, which in all cases resulted in the rapid hydrolysis of a single peptide bond, between amino acid residues 10 and 11. The Ki values for the interactions with papain of these three N-terminal-decapeptide-lacking cystatin C variants were 20-50 nM, just one order of magnitude higher than the value for N-terminally truncated wild-type cystatin C, which in turn was similar to the corresponding values for the full-length Glu-11, Arg-11 and Trp-11 variants. These data indicate that the crucial feature of the conserved Gly residue in position 11 of wild-type cystatin C is that this residue, devoid of a side-chain, will allow the N-terminal segment of cystatin C to adopt a conformation suitable for interaction with the substrate-binding pockets of cysteine endopeptidases, resulting in high-affinity binding and efficient inhibition. The functional properties of the remaining part of the proteinase contact area, which is built from more C-terminal inhibitor segments, are not significantly affected even when amino acids with bulky or charged side-chains replace the Gly-11 residue of the N-terminal segment.

Bovine inositol monophosphatase: proteolysis and structural studies

Bovine brain inositol monophosphatase is inactivated when trypsin catalyses the cleavage of a single peptide bond between Lys-36 and Ser-37. This proteolysis is closely followed by cleavage at two other sites in the protein between Lys-78 and Ser-79 and between Lys-156 and Ser-157 suggesting that all of these sites are exposed in the native conformation of the protein. All of these residues are predicted to lie at the ends of α helices. The most susceptible bond (Lys-36-Ser-37) is predicted to lie in a highly flexible region of the protein. Circular dichroism studies suggest that approximately 40% of the secondary structure of this protein is helical which is similar to that predicted by the algorithm of Gamier et al. [(1978) J. Mol. Biol. 120, 97-120].