A molecular characterization of spontaneous frameshift mutagenesis within the trpA gene of Escherichia coli

Spontaneous frameshift mutations are an important source of genetic variation in all species and cause a large number of genetic disorders in humans. To enhance our understanding of the molecular mechanisms of frameshift mutagenesis, 583 spontaneous Trp+ revertants of two trpA frameshift alleles in Escherichia coli were isolated and DNA sequenced. In order to measure the contribution of methyl-directed mismatch repair to frameshift production, mutational spectra were constructed for both mismatch repair-proficient and repair-defective strains. The molecular origins of practically all of the frameshifts analyzed could be explained by one of six simple models based upon misalignment of the template or nascent DNA strands with or without misincorporation of primer nucleotides during DNA replication. Most frameshifts occurred within mononucleotide runs as has been shown often in previous studies but the location of the 76 frameshift sites was usually outside of runs. Mismatch repair generally was most effective in preventing the occurrence of frameshifts within runs but there was much variation from site to site. Most frameshift sites outside of runs appear to be refractory to mismatch repair although the small number of occurrences at most of these sites make firm conclusions impossible. There was a dense pattern of reversion sites within the trpA DNA region where reversion events could occur, suggesting that, in general, most DNA sequences are capable of undergoing spontaneous mutational events during replication that can lead to small deletions and insertions. Many of these errors are likely to occur at low frequencies and be tolerated as events too costly to prevent or repair. These studies also revealed an unpredicted flexibility in the primary amino acid sequence of the trpA product, the alpha subunit of tryptophan synthase.     

Sequence composition and context effects on the generation and repair of frameshift intermediates in mononucleotide runs in Saccharomyces cerevisiae

DNA polymerase slippage occurs frequently in tracts of a tandemly repeated nucleotide, and such slippage events can be genetically detected as frameshift mutations. In long mononucleotide runs, most frameshift intermediates are repaired by the postreplicative mismatch repair (MMR) machinery, rather than by the exonucleolytic proofreading activity of DNA polymerase. Although mononucleotide runs are hotspots for polymerase slippage events, it is not known whether the composition of a run and the surrounding context affect the frequency of slippage or the efficiency of MMR. To address these issues, 10-nucleotide (10N) runs were inserted into the yeast LYS2 gene to create +1 frameshift alleles. Slippage events within these runs were detected as Lys(+) revertants. 10G or 10C runs were found to be more unstable than 10A or 10T runs, but neither the frequency of polymerase slippage nor the overall efficiency of MMR was greatly influenced by sequence context. Although complete elimination of MMR activity (msh2 mutants) affected all runs similarly, analyses of reversion rates in msh3 and msh6 mutants revealed distinct specificities of the yeast Msh2p-Msh3p and Msh2p-Msh6p mismatch binding complexes in the repair of frameshift intermediates in different sequence contexts.     

Spectra of spontaneous frameshift mutations at the hisD3052 allele of Salmonella typhimurium in four DNA repair backgrounds.

To characterize the hisD3052 -1 frameshift allele of Salmonella typhimurium, we analyzed approximately 6000 spontaneous revertants (rev) for a 2-base deletion hotspot within the sequence (CG)4, and we sequenced approximately 500 nonhotspot rev. The reversion target is a minimum of 76 bases (nucleotides 843-918) that code for amino acids within a nonconserved region of the histidinol dehydrogenase protein. Only 0.4-3.9% were true rev. Of the following classes, 182 unique second-site mutations were identified: hotspot, complex frameshifts requiring DeltauvrB + pKM101 (TA98-specific) or not (concerted), 1-base insertions, duplications, and nonhotspot deletions. The percentages of hotspot mutations were 13.8% in TA1978 (wild type), 24.5% in UTH8413 (pKM101), 31.6% in TA1538 (DeltauvrB), and 41.0% in TA98 (DeltauvrB, pKM101). The DeltauvrB allele decreased by three times the mutant frequency (MF, rev/10(8) survivors) of duplications and increased by about two times the MF of deletions. Separately, the DeltauvrB allele or pKM101 plasmid increased by two to three times the MF of hotspot mutations; combined, they increased this MF by five times. The percentage of 1-base insertions was not influenced by either DeltauvrB or pKM101. Hotspot deletions and TA98-specific complex frameshifts are inducible by some mutagens; concerted complex frameshifts and 1-base insertions are not; and there is little evidence for mutagen-induced duplications and nonhotspot deletions. Except for the base substitutions in TA98-specific complex frameshifts, all spontaneous mutations of the hisD3052 allele are likely templated. The mechanisms may involve (1) the potential of direct and inverted repeats to undergo slippage and misalignment and to form quasi-palindromes and (2) the interaction of these sequences with DNA replication and repair proteins.     

Base composition of mononucleotide runs affects DNA polymerase slippage and removal of frameshift intermediates by mismatch repair in Saccharomyces cerevisiae

The postreplicative mismatch repair (MMR) system is important for removing mutational intermediates that are generated during DNA replication, especially those that arise as a result of DNA polymerase slippage in simple repeats. Here, we use a forward mutation assay to systematically examine the accumulation of frameshift mutations within mononucleotide runs of variable composition in wild-type and MMR-defective yeast strains. These studies demonstrate that (i) DNA polymerase slippage occurs more often in 10-cytosine/10-guanine (10C/10G) runs than in 10-adenine/10-thymine (10A/10T) runs, (ii) the MMR system removes frameshift intermediates in 10A/10T runs more efficiently than in 10C/10G runs, (iii) the MMR system removes -1 frameshift intermediates more efficiently than +1 intermediates in all 10-nucleotide runs, and (iv) the repair specificities of the Msh2p-Msh3p and Msh2p-Msh6p mismatch recognition complexes with respect to 1-nucleotide insertion/deletion loops vary dramatically as a function of run composition. These observations are relevant to issues of genome stability, with both the rates and types of mutations that accumulate in mononucleotide runs being influenced by the primary sequence of the run as well as by the status of the MMR system.     

High frequencies of short frameshifts in poly-CA/TG tandem repeats borne by bacteriophage M13 in Escherichia coli K-12.

Slipped-strand mispairing (SSM) may play an major role in repetitive DNA sequence evolution by generating large numbers of short frameshift mutations within simple tandem repeats. Here we examine the frequency and size spectrum of frameshifts generated within poly-CA/TG sequences inserted into bacteriophage M13 in Escherichia coli hosts. The frequency of detectable frameshifts within a 40 bp tract of poly-CA/TG is greater than one percent and increases more than linearly with length, being lower by a factor of four in a 22 bp target sequence. The frequency increases more than 13-fold in mutL and mutS host cells, suggesting that a high proportion of frameshift events are normally repaired by methyl-directed mismatch repair. Of the 87 sequenced frameshifts in this study, 96% result from deletion or insertion of only or two 2 bp repeat units. The most frequent events are 2 bp deletions, 2 bp insertions, and 4 bp deletions, the relative frequencies of these events being about 18:6:1.     

Frameshift mutagenesis of lambda prophage by 9-aminoacridine, proflavin and ICR-191

Summary The changes in DNA base sequence induced in the lambda cI gene in an E. coli lysogen have been determined following mutagenesis by three acridine derivatives: 9-aminoacridine and proflavin, which bind reversibly to DNA; and ICR-191, which attaches covalently to DNA through a half-mustard group. For all three derivatives, most mutations are +1 and-1 frameshifts in runs of adjacent G:C pairs. The specificity of mutagenesis at various sites is similar for all three compounds. Prophage in mutL host cells, deficient in mismatch repair, are much more susceptible to mutagenesis by 9-aminoacridine. The induced mutations are also frameshifts, and the site specificity is the same as in lysogens of wild type cells. Thus, additions or deletions of single bases can be corrected by the mismatch repair system, but mismatch repair does not play an important role in determining the sequence specificity of the mutational events.     

Adaptive mutations in Salmonella typhimurium phenotypic of purR super-repression

Under non-lethal selective conditions, a non-dividing or very slowly dividing microbial population gives rise to mutations that relieve selective pressures. This process is described as adaptive mutation. Salmonella typhimurium strain 5-28 has been used as a system for studying adaptive mutations in the chromosomal regulatory gene purR and its target, the purD operator. When this strain is plated on a minimal lactose medium, no apparent growth of parent lawn is observed, yet the revertant colonies accumulate over a period of time. Analysis of the purR mutational spectra showed that the frequencies of transitions and transversions were not significantly different among the growth-dependent and adaptive mutations. But the frequencies for five kinds of -1 frameshifts were significantly different between the growth-dependent and adaptive types. Among the growth-dependent mutations, most one-base deletions occurred in non-iterated bases and were distributed randomly. Among adaptive mutations, the frequency of one-base deletions in small mononucleotide repeats was higher and mutations were concentrated at three hotspots. One-base deletion in small mononucleotide repeats are generally believed to result from DNA polymerase slippage errors, which are not corrected by DNA repair machinery. We further investigated the role of DNA repair on adaptive mutation. Our results showed that the mismatch repair (MMR) might function less efficiently during adaptive mutation. However, DNA oxidative damage repair seemed no less effective in correcting errors under selective pressures than during non-selective growth.     

Escherichia coli mutator (Delta)polA is defective in base mismatch correction: the nature of in vivo DNA replication errors

We constructed a set of Escherichia coli strains containing deletions in genes encoding three SOS polymerases, and defective in MutS and DNA polymerase I (PolI) mismatch repair, and estimated the rate and specificity of spontaneous endogenous tonB(+)-->tonB- mutations. The rate and specificity of mutations in strains proficient or deficient in three SOS polymerases was compared and found that there was no contribution of SOS polymerases to the chromosomal tonB mutations. MutS-deficient strains displayed elevated spontaneous mutation rates, consisting of dominantly minus frameshifts and transitions. Minus frameshifts are dominated by warm spots at run-bases. Among 57 transitions (both G:C-->A:T and A:T-->G:C), 35 occurred at two hotspot sites. PolI-deficient strains possessed an increased rate of deletions and frameshifts, because of a deficiency in postreplicative deletion and frameshift mismatch corrections. Frameshifts in PolI-deficient strains occurred within the entire tonB gene at non-run and run sequences. MutS and PolI double deficiency indicated a synergistic increase in the rate of deletions, frameshifts and transitions. In this case, mutS-specific hotspots for frameshifts and transitions disappeared. The results suggested that, unlike the case previously known pertaining to postreplicative MutS mismatch repair for frameshifts and transitions and PolI mismatch repair for frameshifts and deletions, PolI can recognize and correct transition mismatches. Possible mechanisms for distinct MutS and PolI mismatch repair are discussed. A strain containing deficiencies in three SOS polymerases, MutS mismatch repair and PolI mismatch repair was also constructed. The spectrum of spontaneous mutations in this strain is considered to represent the spectrum of in vivo DNA polymerase III replication errors. The mutation rate of this strain was 219x10(-8), about a 100-fold increase relative to the wild-type strain. Uncorrected polymerase III replication errors were predominantly frameshifts and base substitutions followed by deletions.     

A Set of Lacz Mutations in Escherichia Coli That Allow Rapid Detection of Specific Frameshift Mutations

We have used site-directed mutagenesis to alter bases in lacZ near the region encoding essential residues in the active site of beta-galactosidase. The altered sequences generate runs of six or seven identical base pairs which create a frameshift, resulting in a Lac- phenotype. Reversion to Lac+ in each strain can occur only by a specific frameshift at these sequences. Monotonous runs of A's (or of T's on the opposite strand) and G's (or C's) have been constructed, as has an alternating -C-G- sequence. These specific frameshift indicator strains complement a set of six previously described strains which detect each of the base substitutions. We have examined a variety of mutagens and mutators for their ability to cause reversion to Lac+. Surprisingly, frameshifts are well stimulated at many of these runs by ethyl methanesulfonate, N-methyl-N'-nitro-N-nitrosoguanidine and 2-amino-purine, mutagens not widely known to induce frameshifts. A comparison of ethyl methanesulfonate, N-methyl-N'-nitro-N-nitrosoguanidine and 2-aminopurine frameshift specificity with that found with a mutH strain suggests that these mutagens partially or fully saturate or inactivate the methylation-directed mismatch repair system and allow replication errors leading to frameshifts to escape repair. This results in a form of indirect mutagenesis, which can be detected at certain sites.     

Human DNA polymerases lambda and beta show different efficiencies of translesion DNA synthesis past abasic sites and alternative mechanisms for frameshift generation

Human DNA polymerases (pols) beta and lambda could promote template slippage and generate -1 frameshifts on defined heteropolymeric DNA substrates containing a single abasic site. Kinetic data demonstrated that pol lambda was more efficient than pol beta in catalyzing translesion DNA synthesis past an abasic site, particularly in the presence of low nucleotide concentrations. Moreover, pol lambda was found to generate frameshifts in two ways: first, by using a nucleotide-stabilized primer misalignment mechanism, or second, by promoting primer reannealing using microhomology regions between the terminal primer sequence and the template strand. Our results suggest a molecular mechanism for the observed high in vivo rate of frameshifts generation by pol lambda and highlight the remarkable ability of pol lambda to promote microhomology pairing between two DNA strands, further supporting its proposed role in the nonhomologous end joining process.     

Novel PMS1 alleles preferentially affect the repair of primer strand loops during DNA replication

Null mutations in DNA mismatch repair (MMR) genes elevate both base substitutions and insertions/deletions in simple sequence repeats. Data suggest that during replication of simple repeat sequences, polymerase slippage can generate single-strand loops on either the primer or template strand that are subsequently processed by the MMR machinery to prevent insertions and deletions, respectively. In the budding yeast Saccharomyces cerevisiae and mammalian cells, MMR appears to be more efficient at repairing mispairs comprised of loops on the template strand compared to loops on the primer strand. We identified two novel yeast pms1 alleles, pms1-G882E and pms1-H888R, which confer a strong defect in the repair of "primer strand" loops, while maintaining efficient repair of "template strand" loops. Furthermore, these alleles appear to affect equally the repair of 1-nucleotide primer strand loops during both leading- and lagging-strand replication. Interestingly, both pms1 mutants are proficient in the repair of 1-nucleotide loop mispairs in heteroduplex DNA generated during meiotic recombination. Our results suggest that the inherent inefficiency of primer strand loop repair is not simply a mismatch recognition problem but also involves Pms1 and other proteins that are presumed to function downstream of mismatch recognition, such as Mlh1. In addition, the findings reinforce the current view that during mutation avoidance, MMR is associated with the replication apparatus.     

Molecular handles on adaptive mutation

In one experimental system, several handles on the molecular mechanism of apparent adaptive mutation have emerged. The system is reversion of a lac frame-shift mutation in Escherichia coli . The molecular handles include a requirement for homologous recombination; the implication of DNA double-strand breaks as a molecular intermediate; a unique sequence spectrum of −1 deletions in mononucleotide repeats which implies polymerase errors, and also implies a failure of post-synthesis mismatch repair on those errors; and the involvement of sexual functions at some stage of the process. These molecular handles are revealing an unexpected new mechanism of mutagenesis.     

Efficient incorporation of large (>2 kb) heterologies into heteroduplex DNA: Pms1/Msh2-dependent and -independent large loop mismatch repair in Saccharomyces cerevisiae

DNA double-strand break (DSB) repair in yeast is effected primarily by gene conversion. Conversion can conceivably result from gap repair or from mismatch repair of heteroduplex DNA (hDNA) in recombination intermediates. Mismatch repair is normally very efficient, but unrepaired mismatches segregate in the next cell division, producing sectored colonies. Conversion of small heterologies (single-base differences or insertions <15 bp) in meiosis and mitosis involves mismatch repair of hDNA. The repair of larger loop mismatches in plasmid substrates or arising by replication slippage is inefficient and/or independent of Pms1p/Msh2p-dependent mismatch repair. However, large insertions convert readily (without sectoring) during meiotic recombination, raising the question of whether large insertions convert by repair of large loop mismatches or by gap repair. We show that insertions of 2.2 and 2.6 kbp convert efficiently during DSB-induced mitotic recombination, primarily by Msh2p- and Pms1p-dependent repair of large loop mismatches. These results support models in which Rad51p readily incorporates large heterologies into hDNA. We also show that large heterologies convert more frequently than small heterologies located the same distance from an initiating DSB and propose that this reflects Msh2-independent large loop-specific mismatch repair biased toward loop loss.     

Length of CTG.CAG repeats determines the influence of mismatch repair on genetic instability

We showed previously that mutations in methyl-directed mismatch repair of Escherichia coli reduced the occurrence of large deletions in (CTG.CAG)(175) repeats contained on plasmids. By contrast, other workers reported that mutations in mismatch repair increase the frequency of small-length changes in the shorter (CTG.CAG)(64). Using plasmids with a variety of lengths and purity of (CTG.CAG) repeats, we have resolved these apparently conflicting observations. We show that all lengths of (CTG.CAG) repeats are subject to small-length changes (eight repeats) in (CTG.CAG)(n) occur more readily in cells with active mismatch repair. The frequency of large deletions is proportional to the tract length; in our assays they become prominent in tracts greater than 100 repeats. Interruptions in repeat purity enhance the occurrence of large deletions. In addition, we observed a high level of incidence of deletions in (CTG.CAG) repeats for cultures passing repeatedly through stationary phase during long-term growth experiments of all strains (i.e. with active or inactive mismatch repair). These results agree with current theories on mismatch repair acting on DNA slippage events that occur in DNA triplet-repeats.     

Frameshift intermediates in homopolymer runs are removed efficiently by yeast mismatch repair proteins.

A change in the number of base pairs within a coding sequence can result in a frameshift mutation, which almost invariably eliminates the function of the encoded protein. A frameshift reversion assay with Saccharomyces cerevisiae that can be used to examine the types of insertions and deletions that are generated during DNA replication, as well as the editing functions that remove such replication errors, has been developed. Reversion spectra have been obtained in a wild-type strain and in strains defective for defined components of the postreplicative mismatch repair system (msh2, msh3, msh6, msh3 msh6, pms1, and mih1 mutants). Comparison of the spectra reveals that yeast mismatch repair proteins preferentially remove frameshift intermediates that arise in homopolymer tracts and indicates that some of the proteins have distinct substrate or context specificities.     

A DNA polymerase epsilon mutant that specifically causes +1 frameshift mutations within homonucleotide runs in yeast

The DNA polymerases delta and epsilon are the major replicative polymerases in the yeast Saccharomyces cerevisiae that possess 3' --> 5' exonuclease proofreading activity. Many errors arising during replication are corrected by these exonuclease activities. We have investigated the contributions of regions of Polepsilon other than the proofreading motifs to replication accuracy. An allele, pol2-C1089Y, was identified in a screen of Polepsilon mutants that in combination with an exonuclease I (exo1) mutation could cause a synergistic increase in mutations within homonucleotide runs. In contrast to other polymerase mutators, this allele specifically results in insertion frameshifts. When pol2-C1089Y was combined with deletions of EXO1 or RAD27 (homologue of human FEN1), mutation rates were increased for +1 frameshifts while there was almost no effect on -1 frameshifts. On the basis of genetic analysis, the pol2-C1089Y mutation did not cause a defect in proofreading. In combination with a deletion of the mismatch repair gene MSH2, the +1 frameshift mutation rate for a short homonucleotide run was increased nearly 100-fold whereas the -1 frameshift rate was unchanged. This suggests that the Pol2-C1089Y protein makes +1 frameshift errors during replication of homonucleotide runs and that these errors can be corrected by either mismatch repair (MMR) or proofreading (in short runs). This is the first report of a +1-specific mutator for homonucleotide runs in vivo. The pol2-C1089Y mutation defines a functionally important residue in Polepsilon.     

Role of proofreading and mismatch repair in maintaining the stability of nucleotide repeats in DNA.

The role of the proofreading exonuclease in maintaining the stability of multiply repeated units in DNA was studied in Escherichia coli. Reversion of plasmids in which the beta-galactosidase alpha complementing sequence was moved +2 out of frame by inserts containing (CA)14, (CA)5, (CA)2 or (TA)6 or +1 by creating a run of 8 C was compared in mutS and mutSdnaQ strains. Proofreading corrects at least half of the frameshift errors for all the plasmids and at least 99% of the errors in the (CA)2 plasmid. The (CA)2 plasmid reverts mostly by +1 frameshifts in the restriction sites flanking the insert. With the (CA)14, (TA)6, (CA)5 and 8C plasmids, reversion is mainly by loss of a repeat unit. The data support the hypothesis that the dnaQgene product recognizes frameshifts close to the DNA growing point. Frameshifts distal to the growing point are mainly corrected by mismatch repair.We speculate that mismatches in mononucleotide repeats are susceptible to proofreading because they can either migrate to a point where they are recognized by the exonuclease or, alternatively, because single nucleotide distortions are more readily detected than dinucleotides.     

The effect of sequence context on spontaneous Polzeta-dependent mutagenesis in Saccharomyces cerevisiae

The Polζ translesion synthesis (TLS) DNA polymerase is responsible for over 50% of spontaneous mutagenesis and virtually all damage-induced mutagenesis in yeast. We previously demonstrated that reversion of the lys2ΔA746 −1 frameshift allele detects a novel type of +1 frameshift that is accompanied by one or more base substitutions and depends completely on the activity of Polζ. These ‘complex’ frameshifts accumulate at two discrete hotspots (HS1 and HS2) in the absence of nucleotide excision repair, and accumulate at a third location (HS3) in the additional absence of the translesion polymerase Polη. The current study investigates the sequence requirements for accumulation of Polζ-dependent complex frameshifts at these hotspots. We observed that transposing 13 bp of identity from HS1 or HS3 to a new location within LYS2 was sufficient to recapitulate these hotspots. In addition, altering the sequence immediately upstream of HS2 had no effect on the activity of the hotspot. These data support a model in which misincorporation opposite a lesion precedes and facilitates the selected slippage event. Finally, analysis of nonsense mutation revertants indicates that Polζ can simultaneously introduce multiple base substitutions in the absence of an accompanying frameshift event.     

Cell cycle and genetic requirements of two pathways of nonhomologous end-joining repair of double-strand breaks in Saccharomyces cerevisiae.

In Saccharomyces cerevisiae, an HO endonuclease-induced double-strand break can be repaired by at least two pathways of nonhomologous end joining (NHEJ) that closely resemble events in mammalian cells. In one pathway the chromosome ends are degraded to yield deletions with different sizes whose endpoints have 1 to 6 bp of homology. Alternatively, the 4-bp overhanging 3' ends of HO-cut DNA (5'-AACA-3') are not degraded but can be base paired in misalignment to produce +CA and +ACA insertions. When HO was expressed throughout the cell cycle, the efficiency of NHEJ repair was 30 times higher than when HO was expressed only in G1. The types of repair events were also very different when HO was expressed throughout the cell cycle; 78% of survivors had small insertions, while almost none had large deletions. When HO expression was confined to the G1 phase, only 21% were insertions and 38% had large deletions. These results suggest that there are distinct mechanisms of NHEJ repair producing either insertions or deletions and that these two pathways are differently affected by the time in the cell cycle when HO is expressed. The frequency of NHEJ is unaltered in strains from which RAD1, RAD2, RAD51, RAD52, RAD54, or RAD57 is deleted; however, deletions of RAD50, XRS2, or MRE11 reduced NHEJ by more than 70-fold when HO was not cell cycle regulated. Moreover, mutations in these three genes markedly reduced +CA insertions, while significantly increasing the proportion of both small (-ACA) and larger deletion events. In contrast, the rad5O mutation had little effect on the viability of G1-induced cells but significantly reduced the frequency of both +CA insertions and -ACA deletions in favor of larger deletions. Thus, RAD50 (and by extension XRS2 and MRE11) exerts a much more important role in the insertion-producing pathway of NHEJ repair found in S and/or G2 than in the less frequent deletion events that predominate when HO is expressed only in G1.     

8-oxoguanine incorporation into DNA repeats in vitro and mismatch recognition by MutSalpha

DNA 8-oxoguanine (8-oxoG) causes transversions and is also implicated in frameshifts. We previously identified the dNTP pool as a likely source of mutagenic DNA 8-oxoG and demonstrated that DNA mismatch repair prevented oxidation-related frameshifts in mononucleotide repeats. Here, we show that both Klenow fragment and DNA polymerase α can utilize 8-oxodGTP and incorporate the oxidized purine into model frameshift targets. Both polymerases incorporated 8-oxodGMP opposite C and A in repetitive DNA sequences and efficiently extended a terminal 8-oxoG. The human MutSα mismatch repair factor recognized DNA 8-oxoG efficiently in some contexts that resembled frameshift intermediates in the same C or A repeats. DNA 8-oxoG in other slipped/mispaired structures in the same repeats adopted configurations that prevented recognition by MutSα and by the OGG1 DNA glycosylase thereby rendering it invisible to DNA repair. These findings are consistent with a contribution of oxidative DNA damage to frameshifts. They also suggest how mismatch repair might reduce the burden of DNA 8-oxoG and prevent frameshift formation.