Programmed delivery of novel functional groups to the alpha class glutathione transferases

Here we describe a new route to site- and class-specific protein modification that will allow us to create novel functional proteins with artificial chemical groups. Glutathione transferases from the alpha but not the mu, pi, omega, or theta classes can be rapidly and site-specifically acylated with thioesters of glutathione (GS-thioesters) that are similar to compounds that have been demonstrated to occur in vivo. The human isoforms A1-1, A2-2, A3-3, and A4-4 from the alpha class all react with the reagent at a conserved tyrosine residue (Y9) that is crucial in catalysis of detoxication reactions. The yield of modified protein is virtually quantitative in less than 30 min under optimized conditions. The acylated product is stable for more than 24 h at pH 7 and 25 degrees C. The modification is reversible in the presence of excess glutathione, but the labeled protein can be protected by adding S-methylglutathione. The stability of the ester with respect to added glutathione depends on the acyl moiety. The reaction can also take place in Escherichia coli lysates doped with alpha class glutathione transferases. A control substance that lacks the peptidyl backbone required for binding to the glutathione transferases acylates surface-exposed lysines. There is some acyl group specificity since one out of the three different GS-thioesters that we tried was not able to acylate Y9.     

Orthogonal protein purification--expanding the repertoire of GST fusion systems

We have previously developed a labeling scheme that can be used to site-specifically link human glutathione transferases (hGSTs) from the alpha class to chemical entities such as fluorophores and aldehydes. The reagents are in-house synthesized derivatives of glutathione (GS-derivatives). We have focused on a lysine mutant of hGST A1:A216K. In this study, we wanted to utilize these findings and improve on protein purification schemes that are using GSTs as fusion partners. We have used random mutagenesis to scramble the hydrophobic binding site of A216K through mutations at position M208 and isolated a library of 11 A216K/M208X mutants. All mutants were easily expressed and purified and retained all or parts of the catalytic properties of the parent GST. The mutants were stable over several days at room temperature. The A216K/M208X mutants could be site-specifically labeled using our designed fluorescent reagents. Furthermore, reaction with an aldehyde-containing reagent termed GS-Al results in site-specific introduction of an orthogonal handle for subsequent conjugation with aldehyde-reactive probes. Labeling with coumarin results in a fluorescent protein-conjugate that can bind glutathione (GSH) derivatives for subsequent affinity purification. The K(d) for S-hexyl-GSH of coumarin-labeled A216K was measured to be 2.5 microM. The candidate proteins A216K and A216K/M208F could be purified in high yield in a one-step procedure through affinity chromatography (Glutathione Sepharose 4B). The proteins can readily be perceived as improved GST fusion partners.     

Surface-assisted delivery of fluorescent groups to hGST A1-1 and a lysine mutant

Human glutathione transferase (hGST) A1-1 and a lysine mutant (A216K) can both be rapidly and site-specifically acylated on Y9 and K216, respectively, using a range of thiolesters of glutathione (GS-thiolesters) as modifying reagents. The present investigation was aimed at developing a method with which to deliver a fluorescent acyl group from a solid support under conditions compatible with standard protein purification schemes. A number of fluorescent GS-thiolesters with modified peptide backbones were therefore prepared and tested for reactivity toward hGST A1-1 and the A216K mutant. Substitutions at the alpha-NH2 part of the glutathione backbone were not tolerated by the proteins. However, two fluorescent reagents that carry a biotin moiety at the C-terminal part of glutathione were found through MALDI-MS experiments to react in solution with Y9 of the wild-type protein and one reagent with K216 of A216K. The reaction can take place in the presence of glutathione and even in a crude E. coli lysate of cells expressing A216K. Delivery of the fluorescent group to Y9 or K216 was possible using NeutrAvidin (NA) beads that had been preincubated with biotinylated reagent. Alternatively, excess reagent can be removed by a brief incubation with NA beads. We have thus now developed a system for protein labeling with easy removal of excess and used up low-molecular weight reagent. This strategy can conceivably be utilized in future protein purification and labeling experiments.     

Ligand-directed labeling of a single lysine residue in hGST A1-1 mutants

Previously, we discovered that human glutathione transferase (hGST) A1-1 could be site-specifically acylated on a tyrosine residue (Y9) to form ester products using thiolesters of glutathione (GS-thiolesters) as acylating reagents. Out of a total of 20 GS-thiolester reagents tested, 15 (75%) are accepted by hGST A1-1 and thus this is a very versatile reaction. The present investigation was aimed at obtaining a more stable product, an amide bond, between the acyl group and the protein, in order to further increase the value of the reaction. Three lysine mutants (Y9K, A216K, and Y9F/A216K) were therefore prepared and screened against a panel of 18 GS-thiolesters. The Y9K mutant did not react with any of the reagents. The double mutant Y9F/A216K reacted with only one reagent, but in contrast, the A216K mutant could be acylated at the introduced lysine 216 with eight (44%) of the GS-thiolesters. The reaction can take place in the presence of glutathione and even in a crude cell lysate for five (28%) of the reagents. Through the screening process we obtained some basic rules relating to reagent requirements. We have thus produced a mutant (A216K) that can be rapidly and site-specifically modified at a lysine residue to form a stable amide linkage with a range of acyl groups. One of the successful reagents is a fluorophore that potentially can be used in downstream protein purification and protein fusion applications.     

Evidence that glutathione participates in the induction of a stress protein

A step in the induction of a 30- to 35-kD stress protein may be the reaction of chemical inducers with glutathione. Effective inducers are sulfhydryl reagents. Further, a comparison of three reagents, 1-chloro-2,4-dinitrobenzene, diethylmaleate, and N-ethylmaleimide, indicates that the first two, which have considerable selectivity for glutathione, are strong inducers of the stress protein but the third, which is much more reactive with protein sulfhydryls, is either a poor or ineffective inducer. A decrease in cellular glutathione does not appear to be inductive, however. An increase in modified glutathione remains as a possible signal for the induction of this stress protein.     

Chemical modification of rat liver microsomal glutathione transferase defines residues of importance for catalytic function.

Amino acid residues that are essential for the activity of rat liver microsomal glutathione transferase have been identified using chemical modification with various group-selective reagents. The enzyme reconstituted into phosphatidylcholine liposomes does not require stabilization with glutathione for activity (in contrast with the purified enzyme in detergent) and can thus be used for modification of active-site residues. Protection by the product analogue and inhibitor S-hexylglutathione was used as a criterion for specificity. It was shown that the histidine-selective reagent diethylpyrocarbonate inactivated the enzyme and that S-hexylglutathione partially protected against this inactivation. All three histidine residues in microsomal glutathione transferase could be modified, albeit at different rates. Inactivation of 90% of enzyme activity was achieved within the time period required for modification of the most reactive histidine, indicating the functional importance of this residue in catalysis. The arginine-selective reagents phenylglyoxal and 2,3-butanedione inhibited the enzyme, but the latter with very low efficiency; therefore no definitive assignment of arginine as essential for the activity of microsomal glutathione transferase can be made. The amino-group-selective reagents 2,4,6-trinitrobenzenesulphonate and pyridoxal 5'-phosphate inactivated the enzyme. Thus histidine residues and amino groups are suggested to be present in the active site of the microsomal glutathione transferase.     

Acceleration of disulfide-coupled protein folding using glutathione derivatives

Protein folding occurs simultaneously with disulfide bond formation. In general, the in vitro folding of proteins containing disulfide bond(s) is carried out in the presence of redox reagents, such as glutathione, to permit native disulfide pairing to occur. It is well known that the formation of a disulfide bond and the correct tertiary structure of a target protein are strongly affected by the redox reagent used. However, little is known concerning the role of each amino acid residue of the redox reagent, such as glutathione. Therefore, we prepared glutathione derivatives - glutamyl-cysteinyl-arginine (ECR) and arginyl-cysteinyl-glycine (RCG) - and examined their ability to facilitate protein folding using lysozyme and prouroguanylin as model proteins. When the reduced and oxidized forms of RCG were used, folding recovery was greater than that for a typical glutathione redox system. This was particularly true when high protein concentrations were employed, whereas folding recovery using ECR was similar to that of the glutathione redox system. Kinetic analyses of the oxidative folding of prouroguanylin revealed that the folding velocity (K(RCG) = 3.69 × 10(-3) s(-1)) using reduced RCG/oxidized RCG was approximately threefold higher than that using reduced glutathione/oxidized glutathione. In addition, folding experiments using only the oxidized form of RCG or glutathione indicated that prouroguanylin was converted to the native conformation more efficiently in the case of RCG, compared with glutathione. The findings indicate that a positively charged redox molecule is preferred to accelerate disulfide-exchange reactions and that the RCG system is effective in mediating the formation of native disulfide bonds in proteins.     

Analysis of functionalization of methoxy-PEG as maleimide-PEG

The design of the extension arm-facilitated PEGylation (EAFP) of proteins takes advantage of the high selective and quantitative aspects of the thiol-maleimide reaction. However, the efficiency of EAFP with hemoglobin varied with the batches of maleimide-PEG. The low level of functionalization of monomethoxy-PEG (mPEG) as maleimide-PEG has been now investigated as the potential source of this variation. New chemical approaches for the estimation of the functionalization of mPEG using the reaction of the thiol groups of glutathione, dithiothreitol, and hemoglobin with maleimide-PEG have been developed. The single-step modular approach to the synthesis of maleimidophenyl-PEG (MPPEG) that involved the condensation of p-maleimidophenyl isocyanate with mPEG has been optimized to generate a product with an overall purity of 80%. The NMR approach correlates well with the estimates made by the new chemical approaches. Commercial maleimide-PEG reagents synthesized using multiple steps exhibited a lower level of functionalization as reflected by these chemical estimations. The better functionalization of MPPEG increases the efficiency of EAFP as reflected by the generation of hexaPEGylated Hb and the masking of the D antigen of RBCs. This new EAFP protocol is expected to improve the cost effectiveness of the generation of hexaPEGylated Hb, PEGylated albumin, and PEGylated RBCs as new PEGylated therapeutics.     

New photocleavable linker: α-Thioacetophenone-type linker

Photocleavable linkers are advantageous over the common linkers because they could be cleaved without using reagents. A novel photocleavable linker with an α-thioacetophenone moiety has been developed. This linker, which can be cleaved upon irradiation at 365 nm via the Norrish type II reaction, is applicable to a protein affinity purification system, allowing target proteins to be effectively isolated. This novel linker would serve as an effective tool in chemical biology.     

Quantitative determination of SH groups in low- and high-molecular-weight compounds by an electron spin resonance method

To quantitatively determine SH groups in high- and low-molecular-weight compounds, a disulfide biradical (RS-SR), where R is imidazoline residue, has been used. The biradical is shown to participate in a thiol-disulfide exchange reaction with compounds containing SH groups. In this case the ESR spectra of the biradical RS-SR and the resulting monoradical R-SH are different. The reaction of the biradical with cysteine, glutathione, and human serum albumin has been studied using the ESR method and the rate constants kf of this reaction have been calculated. Studies of the pH dependence of kf indicate that the thiol-disulfide exchange occurs by reaction with mercaptidione. Protein human serum albumin and hemoglobin have been modified by RS-SR. It has been shown that the treatment of modified proteins with reduced glutathione leads to removal of the radical from the protein; such modifications are thus reversible. The method proposed has been used to quantitatively determine the SH groups of cysteine and glutathione in mouse and rat blood. The method is shown to coincide within experimental error with the determination of glutathione and cysteine by titration with p-chloromercuribenzoate or reaction with Ellman's reagent. This method allows detection of 10(-6)-10(-7) M SH compounds even in colored and highly absorbing samples. The kinetics of the SH group modification can also be determined, leading to deduction about accessibility of the SH group in protein.     

The use of amphipathic maleimides to study membrane-associated proteins

A series of amphiphilic polymethylenecarboxymaleimides has been synthesized for use as sulfhydryl reagents applicable to membrane proteins. Physical properties of the compounds which are relevant to their proposed mode of action have been determined. By comparing rates of reaction in aqueous and aprotic solvents, the compounds have been shown to react exclusively with the thiolate ion. The effects of the reagents on three membrane-associated proteins are reported, and in two cases a comparative study has been made of the effects on the proteins in the absence of membranes. A mechanism is proposed whereby the reagents are anchored at the lipid/water interface by the negatively charged carboxyl group, thus siting the reactive maleimide in a plane whose depth is defined by the length of the reagent. Supporting evidence for this model is provided by the inability of the reagents to traverse membranes, and variation of their inhibitory potency with chain length when the proteins are embedded in the membrane, but not when extracted into solution. As examples of general use of the reagents to probe sulfhydryl groups in membrane proteins, the reagents have been used to (a) determine the depths in the membrane at which two populations of sulfhydryl groups occur in the mitochondrial phosphate transporter; (b) locate a single sulfhydryl associated with the active site ofD--hydroxybutyrate dehydrogenase in the inner mitochondrial membrane; (c) examine sulfhydryl groups in theD-3-glyceraldehyde phosphate dehydrogenase associated with the human red blood cell membrane.     

Estradiol metabolites as isoform-specific inhibitors of human glutathione S-transferases

Numerous studies have suggested that the lifetime dose of unopposed estrogen is a significant risk factor for breast and uterine cancer. Estradiol (E2) plays a putative role as a tumor promoter through interaction with estrogen receptors but can also be metabolized to redox active and/or mutagenic semiquinones and quinones. Similarly, equine estrogens (components of certain hormone replacement therapy preparations) are converted to quinone metabolites. The use of hormone replacement therapy has also been associated with increased breast and endometrial cancer risk. Recently, metabolites of certain equine estrogens have been shown to inhibit human glutathione S-transferases (hGSTs). Since E2 and equine estrogens share similarities in other biological interactions, we have investigated the inhibitory capacity of endogenously formed E2 metabolites toward various hGSTs. The quinone metabolite of 2-hydroxy-17-beta-estradiol (2-OH-E2) was synthesized, and inhibition of hGST-mediated biotransformation of model substrates was assessed. Inhibition of purified recombinant hGSTM1-1 and hGSTA1-1 occurred in a concentration-dependent manner with IC50-values of approximately 250 and 350 nM, respectively. hGSTs M2-2, P1-1 and T1-1 were significantly less sensitive to inhibition. Specific glutathione-conjugates of the estrogen quinone also potently inhibited hGSTM1-1 and hGSTA1-1. Mass spectrometry data indicate that the inhibition was not mediated via covalent adduction. Although we have demonstrated hGST inhibition via E2 metabolites, our findings indicate that the isoform specificity and potency of GST inhibition by endogenous E2 metabolites is different than that of equine estrogen metabolites.     

Use of the heterobifunctional cross-linker m-maleimidobenzoyl N-hydroxysuccinimide ester to affinity label cholecystokinin binding proteins on rat pancreatic plasma membranes

The binding of 125I-cholecystokinin-33 (125I-CCK-33) to its receptors on rat pancreatic membranes was decreased by modification of membrane protein sulfhydryl groups. Sulfhydryl modifying reagents also caused an accelerated release of bound 125I-CCK-33 from its receptor. Because of the presence of an essential sulfhydryl group(s) in CCK receptor binding we studied the application of the heterobifunctional (SH,NH2) cross-linker, m-maleimidobenzoyl N-hydroxysuccinimide ester (MBS), to affinity label 125I-CCK-33 binding proteins on rat pancreatic plasma membranes. Analysis of the cross-linked products by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and autoradiography revealed that this heterobifunctional cross-linker affinity labeled a major Mr = 80,000-95,000 protein previously identified as part of the CCK receptor on the basis of affinity labeling using homobifunctional and heterobifunctional photoreactive cross-linkers. Additional proteins of Mr greater than 200,000, and Mr = 130,000-140,000 were affinity labeled using MBS. The efficiency of the cross-linking reaction between 125I-CCK-33 and its membrane binding proteins with MBS was significantly greater than that obtained with NH2-directed homobifunctional reagents such as disuccinimidyl suberate. The efficiency of cross-linking could be dramatically improved by reduction of membrane proteins with low-molecular weight thiols prior to binding and cross-linking. The differential labeling patterns of the CCK binding proteins obtained with chemical cross-linkers of similar length but different chemical reactivity underscores the need for caution in predicting native receptor structure from affinity labeling data alone. Using the same pancreatic plasma membrane preparation and 125I-insulin, the Mr = 125,000 alpha-subunit of the insulin receptor was affinity labeled using MBS as cross-linker, demonstrating its utility in identifying other peptide hormone receptors.     

Reactivity and specificity of alpha-dicarbonyl compounds towards essential aminoacyl residues of D-beta-hydroxybutyrate dehydrogenase from the inner mitochondrial membrane

The effects of a alpha-dicarbonyl chromophoric reagent: 4-hydroxy-3-nitrophenylglyoxal on the D-beta-hydroxybutyrate dehydrogenase have been compared to those of phenylglyoxal, a specific arginyl reagent in proteins. Both reagents inactivate irreversibly the enzyme. Kinetic experiments show that only one molecule of these reagents per molecule of enzyme is sufficient to inactivate the enzyme. The second order inactivation rate constant is more than 500 times higher with the chromophoric reagent than with phenylglyoxal. A pseudosubstrate (methylmalonate) in presence of coenzyme (NAD) strongly protects enzyme against inactivation by both reagents. Coenzyme alone has no effect on inactivation by phenylglyoxal while it protects whether inhibitor is the chromophoric reagent or N-ethylmaleimide: a thiol specific reagent. These results indicate: 1. That one arginyl residue is essential for D-beta-hydroxybutyrate dehydrogenase activity (experiments with phenylglyoxal). 2. That the presence of a nitro group on position 3 and a hydroxyl-group on position 4 strongly increase the reactivity of the alpha-dicarbonyl groups, but the specificity of the chemical reaction with arginyl residues seems to be lost for the benefit of cysteyl residues.     

Comparison of 4-plex to 8-plex iTRAQ quantitative measurements of proteins in human plasma samples

Methods for isobaric tagging of peptides, iTRAQ or TMT, are commonly used platforms in mass spectrometry based quantitative proteomics. These two methods are very often used to quantitate proteins in complex samples, e.g., serum/plasma or CSF supporting biomarker discovery studies. The success of these studies depends on multiple factors, including the accuracy of ratios of reporter ions reflecting quantitative changes of proteins. Because reporter ions are generated during peptide fragmentation, the differences of chemical structure of iTRAQ balance groups may have an effect on how efficiently these groups are fragmented and thus how differences in protein expression will be measured. Because 4-plex and 8-plex iTRAQ reagents do have different structures of balanced groups, it has been postulated that indeed differences in protein identification and quantitation exist between these two reagents. In this study we controlled the ratios of tagged samples and compared quantitation of proteins using 4-plex versus 8-plex reagents in the context of a highly complex sample of human plasma using ABSciex 4800 MALDI-TOF/TOF mass spectrometer and ProteinPilot 4.0 software. We observed that 8-plex tagging provides more consistent ratios than 4-plex without compromising protein identification, thus allowing investigation of eight experimental conditions in one analytical experiment.     

Reactive sulfhydryl groups in Saccharomyces cerevisiae phosphoenolpyruvate carboxykinase

Saccharomyces cerevisiae phosphoenolpyruvate carboxykinase (ATP:oxaloacetate carboxy-lyase (transphosphorylating), EC 4.1.1.49) is inactivated by several thiol- and vicinal dithiol-specific reagents. Titration experiments of the enzyme with 5,5'-dithiobis(2-nitrobenzoate) (DTNB) show the presence of reactive monothiol and vicinal dithiol groups, whose modifications lead to enzyme inactivation. The enzyme is also inactivated by N-(1-pyrenyl)iodoacetamide (PyrIAM), with a binding stoichiometry of approx. 2 mol per mol of enzyme subunit. A high level of pyrene excimer fluorescence is detected on the labeled enzyme, thus implying the reaction of the reagent with two spatially close sulfhydryl groups in the protein. The carboxykinase is not completely inactivated by different vicinal dithiol-specific reagents, thus implying a catalytically non-essential character for these groups. From substrate protection experiments of the enzyme inactivation by DTNB, PyrIAM and vicinal dithiol-specific reagents, it is concluded that the loss of enzyme activity is caused by the modification of both thiol and vicinal dithiol groups in the substrate binding region.     

Kinetic analysis of the slow ionization of glutathione by microsomal glutathione transferase MGST1

An important aspect of the catalytic mechanism of microsomal glutathione transferase (MGST1) is the activation of the thiol of bound glutathione (GSH). GSH binding to MGST1 as measured by thiolate anion formation, proton release, and Meisenheimer complex formation is a slow process that can be described by a rapid binding step (K(GSH)d = 47 +/- 7 mM) of the peptide followed by slow deprotonation (k2 = 0.42 +/- 0.03 s(-1). Release of the GSH thiolate anion is very slow (apparent first-order rate k(-2) = 0.0006 +/- 0.00002 s(-)(1)) and thus explains the overall tight binding of GSH. It has been known for some time that the turnover (kcat) of MGST1 does not correlate well with the chemical reactivity of the electrophilic substrate. The steady-state kinetic parameters determined for GSH and 1-chloro-2,4-dinitrobenzene (CDNB) are consistent with thiolate anion formation (k2) being largely rate-determining in enzyme turnover (kcat = 0.26 +/- 0.07 s(-1). Thus, the chemical step of thiolate addition is not rate-limiting and can be studied as a burst of product formation on reaction of halo-nitroarene electrophiles with the E.GS- complex. The saturation behavior of the concentration dependence of the product burst with CDNB indicates that the reaction occurs in a two-step process that is characterized by rapid equilibrium binding ( = 0.53 +/- 0.08 mM) to the E.GS- complex and a relatively fast chemical reaction with the thiolate (k3 = 500 +/- 40 s(-1). In a series of substrate analogues, it is observed that log k3 is linearly related (rho value 3.5 +/- 0.3) to second substrate reactivity as described by Hammett sigma- values demonstrating a strong dependence on chemical reactivity that is similar to the nonenzymatic reaction (rho = 3.4). Microsomal glutathione transferase 1 displays the unusual property of being activated by sulfhydryl reagents. When the enzyme is activated by N-ethylmaleimide, the rate of thiolate anion formation is greatly enhanced, demonstrating for the first time the specific step that is activated. This result explains earlier observations that the enzyme is activated only with more reactive substrates. Taken together, the observations show that the kinetic mechanism of MGST1 can be described by slow GSH binding/thiolate formation followed by a chemical step that depends on the reactivity of the electrophilic substrate. As the chemical reactivity of the electrophile becomes lower the rate-determining step shifts from thiolate formation to the chemical reaction.     

Topological information content and chemical reactions

The difference of the topological information content of two reacting molecules and that of their reaction products is calculated for several topological types of chemical reactions, illustrating the influence of the structure of the reagents and of the reaction product. It is shown that the change in the topological information content in a chemical reaction can be positive as well as negative, depending on the way the reagents approach each other and thus on the reaction product formed. A quantitative measure of structural specificity is introduced.     

The role of human glutathione S-transferases (hGSTs) in the detoxification of the food-derived carcinogen metabolite N-acetoxy-PhIP, and the effect of a polymorphism in hGSTA1 on colorectal cancer risk

Food-derived heterocyclic amines (HCAs), particularly 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP), are implicated in the etiology of human colorectal cancer (CRC) via a process of N-oxidation followed by O-acetylation or O-sulfation to form electrophilic metabolites that react with DNA. Glutathione S-transferases (GSTs) detoxify activated carcinogen metabolites by catalysis of their reaction with GSH. However, among HCAs, only N-acetoxy-PhIP has been shown to be a substrate for the GSTs. By using a competitive DNA-binding assay, we confirm that hGSTA1-1 is an efficient catalyst of the detoxification of N-acetoxy-PhIP. Further, we show that hGSTs A2-2, P1-1, M1-1, T1-1 and T2-2 appear to have low activity towards N-acetoxy-PhIP, and that hGSTs A4-4, M2-2, M4-4 and Z1-1 appear to have no activity towards N-acetoxy-PhIP. A genetic polymorphism in the 5'-regulatory sequence of hGSTA1 has been shown to correlate with the relative and absolute levels of expression of GSTA1/GSTA2 in human liver. Examination of hGSTA1 allele frequency in 100 Caucasian CRC patients and 226 Caucasian controls demonstrated a significant over-representation of the homozygous hGSTA1*B genotype among cases compared to controls (24.0 and 13.7%, respectively, P=0.04). This corresponds to an odds ratio for risk of CRC of 2.0 (95% CI 1.0-3.7) when comparing homozygous hGSTA1*B individuals with all other genotypes. Thus, individuals who are homozygous hGSTA1*B, and who would be predicted to have the lowest levels of hGSTA1 expression in their livers, appear to be at risk of developing CRC, possibly as a result of inefficient hepatic detoxification of N-acetoxy-PhIP.