Molecular dynamics, crystallography and mutagenesis studies on the substrate gating mechanism of prolyl oligopeptidase

Altered prolyl oligopeptidase (PREP) activity is found in many common neurological and other genetic disorders, and in some cases PREP inhibition may be a promising treatment. The active site of PREP resides in an internal cavity; in addition to the direct interaction between active site and substrate or inhibitor, the pathway to reach the active site (the gating mechanism) must be understood for more rational inhibitor design and understanding PREP function. The gating mechanism of PREP has been investigated through molecular dynamics (MD) simulation combined with crystallographic and mutagenesis studies. The MD results indicate the inter-domain loop structure, comprised of 3 loops at residues, 189-209 (loop A), 577-608 (loop B), and 636-646 (loop C) (porcine PREP numbering), are important components of the gating mechanism. The results from enzyme kinetics of PREP variants also support this hypothesis: When loop A is (1) locked to loop B through a disulphide bridge, all enzyme activity is halted, (2) nicked, enzyme activity is increased, and (3) removed, enzyme activity is only reduced. Limited proteolysis study also supports the hypothesis of a loop A driven gating mechanism. The MD results show a stable network of H-bonds that hold the two protein domains together. Crystallographic study indicates that a set of known PREP inhibitors inhabit a common binding conformation, and this H-bond network is not significantly altered. Thus the domain separation, seen to occur in lower taxa, is not involved in the gating mechanism for mammalian PREP. In two of the MD simulations we observed a conformational change that involved the breaking of the H-bond network holding loops A and B together. We also found that this network was more stable when the active site was occupied, thus decreasing the likelihood of this transition.     

Thermodynamic, enzymatic and structural effects of removing a salt bridge at the base of loop 4 in (pro)caspase-3

Interactions between loops 2, 2′ and 4, known as the loop bundle, stabilize the active site of caspase-3. Loop 4 (L4) is of particular interest due to its location between the active site and the dimer interface. We have disrupted a salt bridge between K242 and E246 at the base of L4 to determine its role in overall conformational stability and in maintaining the active site environment. Stability measurements show that only the K242A single mutant decreases stability of the dimer, whereas both single mutants and the double mutant demonstrate much lower activity compared to wild-type caspase-3. Structural studies of the caspase-3 variants show the involvement of K242 in hydrophobic interactions that stabilize helix 5, near the dimer interface, and the role of E246 appears to be to neutralize the positive charge of K242 within the hydrophobic cluster. Overall, the results suggest E246 and K242 are important in procaspase-3 for their interaction with neighboring residues, not with one another. Conversely, formation of the K242–E246 salt bridge in caspase-3 is needed for an accurate, stable conformation of loop L4 and proper active site formation in the mature enzyme.     

Crystal structure determination of cholesterol oxidase from Streptomyces and structural characterization of key active site mutants

Cholesterol oxidase is a monomeric flavoenzyme which catalyzes the oxidation and isomerization of cholesterol to cholest-4-en-3-one. The enzyme interacts with lipid bilayers in order to bind its steroid substrate. The X-ray structure of the enzyme from Brevibacterium sterolicum revealed two loops, comprising residues 78-87 and residues 433-436, which act as a lid over the active site and facilitate binding of the substrate [Vrielink et al. (1991) J. Mol. Biol. 219, 533-554; Li et al. (1993) Biochemistry 32, 11507-11515]. It was postulated that these loops must open, forming a hydrophobic channel between the membrane and the active site of the protein and thus sequestering the cholesterol substrate from the aqueous environment. Here we describe the three-dimensional structure of the homologous enzyme from Streptomyces refined to 1.5 A resolution. Structural comparisons to the enzyme from B. sterolicum reveal significant conformational differences in these loop regions; in particular, a region of the loop comprising residues 78-87 adopts a small amphipathic helical turn with hydrophobic residues directed toward the active site cavity and hydrophilic residues directed toward the external surface of the molecule. It seems reasonable that this increased rigidity reduces the entropy loss that occurs upon binding substrate. Consequently, the Streptomyces enzyme is a more efficient catalyst. In addition, we have determined the structures of three active site mutants which have significantly reduced activity for either the oxidation (His447Asn and His447Gln) or the isomerization (Glu361Gln). Our structural and kinetic data indicate that His447 and Glu361 act as general base catalysts in association with conserved water H2O541 and Asn485. The His447, Glu361, H2O541, and Asn485 hydrogen bond network is conserved among other oxidoreductases. This catalytic tetrad appears to be a structural motif that occurs in flavoenzymes that catalyze the oxidation of unactivated alcohols.     

Source and target enzyme signature in serine protease inhibitor active site sequences

Amino acid sequences of proteinaceous proteinase inhibitors have been extensively analysed for deriving information regarding the molecular evolution and functional relationship of these proteins. These sequences have been grouped into several well defined families. It was found that the phylogeny constructed with the sequences corresponding to the exposed loop responsible for inhibition has several branches that resemble those obtained from comparisons using the entire sequence. The major branches of the unrooted tree corresponded to the families to which the inhibitors belonged. Further branching is related to the enzyme specificity of the inhibitor. Examination of the active site loop sequences of trypsin inhibitors revealed that here are strong preferences for specific amino acids at different positions of the loop. These preferences are inhibitor class specific. Inhibitors active against more than one enzyme occur within a class and confirm to class specific sequence in their loops. Hence, only a few positions in the loop seem to determine the specificity. The ability to inhibit the same enzyme by inhibitors that belong to different classes appears to be a result of convergent evolution.     

Coupling effects of distal loops on structural stability and enzymatic activity of Escherichia coli dihydrofolate reductase revealed by deletion mutants

Residues distal from the active site in dihydrofolate reductase (DHFR) have regulatory roles in catalytic reaction and also folding stability. The couplings of the distal residues to the ones in the active site have been analyzed using site-directed mutants. To expand our understanding of the structural and functional influences of distal residue mutation, we explored the structural stability and enzymatic activity of deletion mutants. Deletion has greater structural and dynamical impacts on the corresponding part than site-directed mutation does. Thus, deletion amplifies the effects caused by distal mutations, which should make the mutual couplings among the distant residues more apparent. We focused on residues 52, 67, 121, and 145 in the four distinct loops of DHFR. All the single-residue deletion mutants showed marked reduction in stability, except for Δ52 in an αC–βC loop. Double deletion mutants showed that the loop αC–βC has nonadditive couplings with the βF–βG and βG–βH loops regarding stability. Single deletion to the loops αC–βC or βC–βD resulted in considerable activity reduction, demonstrating that the loops couple to the residues near the active site. The four loops were shown to be functionally interdependent from the double deletion experiments.     

Introduction of foreign peptides in surface loops of alkaline phosphatase

The introduction of foreign peptides into alkaline phosphatase that have a minimal influence on the enzymatic activity of a protein is described in the work. Calculated based on the methods of molecular dynamics the longest surface loops of alkaline phosphatase, which are not adjacent to both active center of a dimeric enzyme and contact surface between its monomers, were used as sites of the introduction of foreign peptides. The length of loops was estimated by several methods. Two loops were used to genetically engineer the introduction of peptides. Experiments demonstrated that the introduction of several peptides after the Ala218 residue of alkaline phosphatase insignificantly influences on the enzyme activity. According to experimental data, the selection of the loop for introducing the foreign peptide can be determined using the methods of molecular dynamics based on calculating the mobility of dihedral angles. In order to create macromolecules with new features by introducing foreign polypeptides in enzymatically active proteins, it is possible in practice to use an estimation of the loop length by means of the methods of molecular dynamics.     

Saccharomyces cerevisiae mutants selected for increased production of Trichoderma reesei cellulases

 Trichoderma reesei endoglucanase I (EGI) was used as a reporter enzyme for screening mutagenized yeast strains for increased ability to produce protein. Sixteen haploid Saccharomyces cerevisiae strains, transformed with a yeast multicopy vector pALK222, containing the EGI cDNA under the ADH1 promoter, produced EGI activity of 10^-5–10^-4 g/l. On the average 93% of the total activity was secreted into the culture medium. Two strains with opposite mating types were mutagenized, and several mutants were isolated possessing up to 45-fold higher EGI activity. The best mutants were remutagenized and a second-generation mutant, strain 2804, with an additional twofold increase in EGI activity was selected. The mutant strain 2804 grew more slowly and reached a lower final cell density than the parental strain. In the selective minimal medium, the 2804 strain produced 40 mg/l immunoreactive EGI protein, but only 2% was active enzyme. In the rich medium the secreted EGI enzyme stayed active, but without selection pressure the EGI production ceased after 2 days of cultivation, when the strain 2804 had produced 10 mg/l of EGI. A sevenfold difference was found between the parental and the 2804 strain in their total EGI production relative to cell density. The difference in favour of the mutant strain was also detected on the mRNA level. The 2804 mutant was found to be more active than the parental strain also in the production of T. reesei cellulases, cellobiohydrolase I, and cellobiohydrolase II. Received: 22 December 1995/Received revision: 26 February 1996/Accepted: 17 March 1996     

Perturbation of DNA hairpins containing the EcoRI recognition site by hairpin loops of varying size and composition: physical (NMR and UV) and enzymatic (EcoRI) studies.

We have investigated loop-induced structural perturbation of the stem structure in hairpins d(GAATTCXnGAATTC) (X = A, T and n = 3, 4, 5 and 6) that contain an EcoRI restriction site in close proximity to the hairpin loop. Oligonucleotides containing either a T3 or a A3 loop were not hydrolyzed by the restriction enzyme and also showed only weak binding to EcoRI in the absence of the cofactor Mg2+. In contrast, hairpins with larger loops are hydrolyzed by the enzyme at the scission site next to the loop although the substrate with a A4 loop is significantly more resistant than the oligonucleotide containing a T4 loop. The hairpin structures with 3 loop residues were found to be thermally most stable while larger hairpin loops resulted in structures with lower melting temperatures. The T-loop hairpins are thermally more stable than the hairpins containing the same number of A residues in the loop. As judged from proton NMR spectroscopy and the thermodynamic data, the base pair closest to the hairpin loop did form in all cases studied. The hairpin loops did, however, affect the conformation of the stem structure of the hairpins. From 31P and 1H NMR spectroscopy we conclude that the perturbation of the stem structure is stronger for smaller hairpin loops and that the extent of the perturbation is limited to 2-3 base pairs for hairpins with T3 or A4 loops. Our results demonstrate that hairpin loops modulate the conformation of the stem residues close to the loop and that this in turn reduces the substrate activity for DNA sequence specific proteins.     

Monoclonal antibodies against core and cellulose-binding domains of Trichoderma reesei cellobiohydrolases I and II and endoglucanase I.

Cellulases from Trichoderma reesei form an enzyme group with a common structural organization. Each cellulase enzyme is composed of two functional domains, the core region containing the active site and the cellulose-binding domain (CBD). To facilitate the specific detection of each domain, monoclonal antibodies (mAb) against cellobiohydrolase I (CBHI), cellobiohydrolase II (CBHII) and endoglucanase I (EGI) were produced. Five mAb were obtained against CBHI, ten against CBHII and eight against EGI. The location of the antigenic epitope for each antibody was mapped by allowing the antibodies to react with truncated cellulases, synthesized from deleted cDNA in Saccharomyces cerevisiae. Proteolytic fragments of Trichoderma cellulases, obtained by papain digestion, were used to confirm the results. Specific antibodies were detected against the core and the CBD epitopes for all three cellulases. Using the truncated enzymes, it was possible to locate the epitopes to a reasonably short region within the protein. To obtain a quantitative assay for each enzyme, a specific mAb against each antigen was chosen, based on the affinity to the corresponding antigen on Western-blot staining and on filter blots of the cellulolytic yeasts. The mAb were used to quantitative the corresponding enzymes in T. reesei culture medium. Specific quantitation of each cellulase enzyme has not been possible by biochemical assays or using polyclonal antibodies, due to their cross-reactions. Now, these mAb can be specifically used to recognize and quantitate different domains of these three important cellulolytic enzymes.     

Exon grafting yields a "two active-site" lysozyme

The design of enzymes with enhanced stability and activity has long been a goal in protein engineering. We report a strategy to engineer an additional active site for human lysozyme, grafted the entire human lysozyme exon 2, which encodes the catalytically competent domain, into the gene at a position corresponding to an exposed loop region in the translated protein. Exon 2 grafting created a novel lysozyme with twice the activity of the wild type enzyme, equal activity came from each of the two active sites. We dissected the contributions of each active site using site-directed mutagenesis of the catalytic doublets of (E35A/D53A), circular dichroism, fluorescence spectra, and molecular modeling. Temperature and pH stability of the "two active-site" enzyme were similar to those of wild-type lysozyme. Thus, we provide a novel strategy for engineering the active site of enzymes.     

Secondary binding site of trypsin: revealed by crystal structure of trypsin-peptide complex

Designed synthetic heterochiral peptides, when added to porcine trypsin, resulted in reduction of enzyme activity. The crystal structure of a complex formed between porcine trypsin and a heterochiral hepta peptide Boc-Pro-DAsp-Aib-Leu-Aib-Leu-Ala-NHMe has been determined at 1.9 A resolution. The hepta peptide does not bind at the active site, but is located in the interstitial region, and interacts with the calcium-binding loop (residues 60-80). The bound peptide interacts with the active site residue Ser195 through an acetate ion, and with Lys 60 mediated by water molecules. The structure, when compared with the other trypsin-peptide complexes, suggests that the flexibility of surface loops, concerted movement of the loops towards the active site, and the interaction of the bound peptide with Lys 60, may be responsible for the reduction in enzyme activity. This study provides a structural evidence for the earlier biochemical observation regarding the role of surface loops in the catalysis of the enzyme.     

Participation of S-S loops in inhibitory activity of peanut protease inhibitor B-III

A peanut Bowman-Birk (BBI) type protease inhibitors B-III has two regions, 1 and 2, homologous with each other. Each region contains three S-S loops and a reactive site in its outermost loop. The inhibitor was used to investigate the contribution of the S-S loops of BBI-type inhibitors to their inhibitory activity. Two steps of Edman degradation of the native inhibitor cleaved loop III (the innermost S-S loop) of region 1 of B-III, and the antichymotryptic activity of the first reactive site decreased to about 1/4 of that of native B-III. A third step of Edman degradation split loop II and the inhibitory activity at that site became extremely low (about 1/200 of the original value). These results suggest that protease inhibitor B-III maintains its active conformation by means of the three S-S loops and that the conformation is markedly changed by the splitting of loop II.     

Loop‐loop interactions govern multiple steps in indole‐3‐glycerol phosphate synthase catalysis

Substrate binding, product release, and likely chemical catalysis in the tryptophan biosynthetic enzyme indole‐3‐glycerol phosphate synthase (IGPS) are dependent on the structural dynamics of the β1α1 active‐site loop. Statistical coupling analysis and molecular dynamic simulations had previously indicated that covarying residues in the β1α1 and β2α2 loops, corresponding to Arg54 and Asn90, respectively, in the Sulfolobus sulfataricus enzyme (ssIGPS), are likely important for coordinating functional motions of these loops. To test this hypothesis, we characterized site mutants at these positions for changes in catalytic function, protein stability and structural dynamics for the thermophilic ssIGPS enzyme. Although there were only modest changes in the overall steady‐state kinetic parameters, solvent viscosity and solvent deuterium kinetic isotope effects indicated that these amino acid substitutions change the identity of the rate‐determining step across multiple temperatures. Surprisingly, the N90A substitution had a dramatic effect on the general acid/base catalysis of the dehydration step, as indicated by the loss of the descending limb in the pH rate profile, which we had previously assigned to Lys53 on the β1α1 loop. These changes in enzyme function are accompanied with a quenching of ps‐ns and µs‐ms timescale motions in the β1α1 loop as measured by nuclear magnetic resonance studies. Altogether, our studies provide structural, dynamic and functional rationales for the coevolution of residues on the β1α1 and β2α2 loops, and highlight the multiple roles that the β1α1 loop plays in IGPS catalysis. Thus, substitution of covarying residues in the active‐site β1α1 and β2α2 loops of indole‐3‐glycerol phosphate synthase results in functional, structural, and dynamic changes, highlighting the multiple roles that the β1α1 loop plays in enzyme catalysis and the importance of regulating the structural dynamics of this loop through noncovalent interactions with nearby structural elements.     

5-Arylidene-2-phenylimino-4-thiazolidinones as PTP1B and LMW-PTP inhibitors

As part of a project aimed at identifying effective low molecular weight nonphosphorus monoanionic inhibitors of PTPs, we have synthesized 4-[(5-arylidene-4-oxo-2-phenyliminothiazolidin-3-yl)methyl]benzoic acids (4) and evaluated their inhibitory activity against human PTP1B and LMW-PTP enzymes. The introduction of a 2-phenylimino moiety onto the 4-thiazolidinone ring was designed to enhance the inhibitor/enzyme affinity by means of further favourable interactions with residues of the active site and the surrounding loops. Some of the compounds (4a–d, f) showed interesting inhibition levels in the low micromolar range. The 5-arylidene moiety of acids 4 proved to markedly influence the potency of these inhibitors. Molecular modeling experiments inside the binding sites of both enzymes were performed.     

Role of loop dynamics in thermal stability of mesophilic and thermophilic adenylosuccinate synthetase: a molecular dynamics and normal mode analysis study

Enzymes from thermophiles are poorly active at temperatures at which their mesophilic homologs exhibit high activity and attain corresponding active states at high temperatures. In this study, comparative molecular dynamics (MD) simulations, supplemented by normal mode analysis, have been performed on an enzyme Adenylosuccinate synthetase (AdSS) from E. coli (mesophilic) and P. horikoshii (thermophilic) systems to understand the effects of loop dynamics on thermal stability of AdSS. In mesophilic AdSS, both ligand binding and catalysis are facilitated through the coordinated movement of five loops on the protein. The simulation results suggest that thermophilic P. horikoshii preserves structure and catalytic function at high temperatures by using the movement of only a subset of loops (two out of five) for ligand binding and catalysis unlike its mesophilic counterpart in E. coli. The pre-arrangement of the catalytic residues in P. horikoshii is well-preserved and salt bridges remain stable at high temperature (363K). The simulations suggest a general mechanism (including pre-arrangement of catalytic residues, increased polar residue content, stable salt bridges, increased rigidity, and fewer loop movements) used by thermophilic enzymes to preserve structure and be catalytically active at elevated temperatures.     

Primary recovery of a genetically engineered Trichoderma reesei endoglucanase I (Cel 7B) fusion protein in cloud point extraction systems

Here we present data to demonstrate how partitioning of a hydrophilic enzyme can be directed to the hydrophobic detergent-enriched phase of an aqueous two-phase system by addition of short stretches of amino acid residues to the protein molecule. The target enzyme was the industrially important endoglucanase I, EGI (endo-1,4-beta-D-glucan-4-glucanohydrolase, EC 3.2.1.4, Cel7B) of Trichoderma reesei. We investigated the partitioning of three EGI variants containing various C-terminal peptide extensions including Trp-Pro motifs of different lengths and localizations. Additionally, a recently developed system composed of the thermoseparating copolymer HM-EOPO was utilized to study the effects of fusion tags. The addition of peptides containing tryptohan residues enhanced the partitioning of EGI to the HM-EOPO-rich phase. The system composed of a nonionic detergent (Agrimul NRE1205) resulted in the highest partition coefficient (K = 31) and yield (90%) with the construct EGI(core-P5)(WP)(4) containing (Trp-Pro)(4) after a short linker stretch. A recombinant strain of T. reesei Rut-C30 for large-scale production was constructed in which the fusion protein EGI(core-P5)(WP)(4) was expressed from the strong promoter of the cellulase gene cbh1. The fusion protein was successfully expressed and secreted from the fungus during shake-flask cultivations. Cultivation in a 28-L bioreactor however, revealed that the fusion protein is sensitive to proteases. Consequently, only low production levels were obtained in large-scale production trials.     

Molecular insight into the role of the leucine residue on the L2 loop in the catalytic activity of caspases 3 and 7

Various apoptotic signals can activate caspases 3 and 7 by triggering the L2 loop cleavage of their proenzymes. These two enzymes have highly similar structures and functions, and serve as apoptotic executioners. The structures of caspase 7 and procaspase 7 differ significantly in the conformation of the loops constituting the active site, indicating that the enzyme undergoes a large structural change during activation. To define the role of the leucine residue on the L2 loop, which shows the largest movement during enzyme activation but has not yet been studied, Leu168 of caspase 3 and Leu191 of caspase 7 were mutated. Kinetic analysis indicated that the mutation of the leucine residues sometimes improved the Km but also greatly decreased the kcat, resulting in an overall decrease in enzyme activity. The tryptophan fluorescence change at excitation/emission = 280/350 nm upon L2-L2' loop cleavage was found to be higher in catalytically active mutants, including the corresponding wild-type caspase, than in the inactive mutants. The crystal structures of the caspase 3 mutants were solved and compared with that of wild-type. Significant alterations in the conformations of the L1 and L4 loops were found. These results indicate that the leucine residue on the L2 loop has an important role in maintaining the catalytic activity of caspases 3 and 7.     

Determinants of Neutralization Resistance in the Envelope Glycoproteins of a Simian-Human Immunodeficiency Virus Passaged In Vivo

In vivo passage of a simian-human immunodeficiency virus (SHIV-89.6) generated a virus, SHIV-89.6P, that exhibited increased resistance to some neutralizing antibodies (G. B. Karlsson et al., J. Exp. Med. 188:1159-1171, 1998). Here we examine the range of human immunodeficiency virus type 1 (HIV-1) neutralizing antibodies to which the passaged virus became resistant and identify envelope glycoprotein determinants of antibody resistance. Compared with the envelope glycoproteins derived from the parental SHIV-89.6, the envelope glycoproteins of the passaged virus were resistant to antibodies directed against the gp120 V3 variable loop and the CD4 binding site. By contrast, both viral envelope glycoproteins were equally sensitive to neutralization by two antibodies, 2G12 and 2F5, that recognize poorly immunogenic structures on gp120 and gp41, respectively. Changes in the V2 and V3 variable loops of gp120 were necessary and sufficient for full resistance to the IgG1b12 antibody, which is directed against the CD4 binding site. Changes in the V3 loop specified complete resistance to a V3 loop-directed antibody, while changes in the V1/V2 loops conferred partial resistance to this antibody. The epitopes of the neutralizing antibodies were not disrupted by the resistance-associated changes. These results indicate that in vivo selection occurs for HIV-1 envelope glycoproteins with variable loop conformations that restrict the access of antibodies to immunogenic neutralization epitopes.     

Evolution of an antibiotic resistance enzyme constrained by stability and activity trade-offs

Pressured by antibiotic use, resistance enzymes have been evolving new activities. Does such evolution have a cost? To investigate this question at the molecular level, clinically isolated mutants of the beta-lactamase TEM-1 were studied. When purified, mutant enzymes had increased activity against cephalosporin antibiotics but lost both thermodynamic stability and kinetic activity against their ancestral targets, penicillins. The X-ray crystallographic structures of three mutant enzymes were determined. These structures suggest that activity gain and stability loss is related to an enlarged active site cavity in the mutant enzymes. In several clinically isolated mutant enzymes, a secondary substitution is observed far from the active site (Met182-->Thr). This substitution had little effect on enzyme activity but restored stability lost by substitutions near the active site. This regained stability conferred an advantage in vivo. This pattern of stability loss and restoration may be common in the evolution of new enzyme activity.     

Determinants of Human Immunodeficiency Virus Type 1 Envelope Glycoprotein Activation by Soluble CD4 and Monoclonal Antibodies

Infection by some human immunodeficiency virus type 1 (HIV-1) isolates is enhanced by the binding of subneutralizing concentrations of soluble receptor, soluble CD4 (sCD4), or monoclonal antibodies directed against the viral envelope glycoproteins. In this work, we studied the abilities of different antibodies to mediate activation of the envelope glycoproteins of a primary HIV-1 isolate, YU2, and identified the regions of gp120 envelope glycoprotein contributing to activation. Binding of antibodies to a variety of epitopes on gp120, including the CD4 binding site, the third variable (V3) loop, and CD4-induced epitopes, enhanced the entry of viruses containing YU2 envelope glycoproteins. Fab fragments of antibodies directed against either the CD4 binding site or V3 loop also activated YU2 virus infection. The activation phenotype was conferred on the envelope glycoproteins of a laboratory-adapted HIV-1 isolate (HXBc2) by replacing the gp120 V3 loop or V1/V2 and V3 loops with those of the YU2 virus. Infection by the YU2 virus in the presence of activating antibodies remained inhibitable by macrophage inhibitory protein 1β, indicating dependence on the CCR5 coreceptor on the target cells. Thus, antibody enhancement of YU2 entry involves neither Fc receptor binding nor envelope glycoprotein cross-linking, is determined by the same variable loops that dictate enhancement by sCD4, and probably proceeds by a process fundamentally similar to the receptor-activated virus entry pathway.