(Received for publication, December 2, 1996, and in revised form, April 6, 1997)
From the Radiobiology Program, Cross Cancer Institute, Edmonton, Alberta T6G 1Z2, Canada
Several oxidative DNA-damaging agents, including
ionizing radiation, can generate multiply damaged sites in DNA. Among
the postulated lesions are those with abasic sites located in close proximity on opposite strands. The repair of an abasic site requires strand scission by a repair endonuclease such as human
apurinic/apyrimidinic endonuclease (Ape) or exonuclease III in
Escherichia coli. Therefore, a potential consequence of the
"repair" of bistranded abasic sites is the formation of
double-strand breaks. To test this possibility and to investigate the
influence of the relative distance between the two abasic sites and
their orientation to each other, we prepared a series of
oligonucleotide duplexes containing abasic sites at defined positions
either directly opposite each other or separated by 1, 3, or 5 base
pairs in the 5- or 3
-direction. Analysis following Ape and
exonuclease III treatment of these substrates indicated a variety of
responses. In general, cleavage at abasic sites was slower in duplexes
with paired lesions than in control duplexes with single lesions.
Double-strand breaks were, however, readily generated in duplexes with
abasic sites positioned 3
to each other. With the duplex containing
abasic sites set 1 base pair apart, 5
to each other, both Ape and
exonuclease III slowly cleaved the abasic site on one strand only and
were unable to incise the other strand. With the duplex containing
abasic sites set 3 base pairs apart, 5
to each other, Ape protein was
unable to cleave either strand. These data suggest that closely
positioned abasic sites could have several deleterious consequences in
the cell. In addition, this approach has allowed us to map bases that make significant contact with the enzymes when acting on an abasic site
on the opposite strand.
The cytotoxic effects of many agents are believed to be the result of damage to DNA. In addition to producing isolated DNA base and deoxyribose lesions, several important antineoplastic agents, including ionizing radiation and drugs such as bleomycin and neocarzinostatin, can generate clustered lesions or locally multiply damaged sites (LMDS)1 located within one or two helical turns (1-5). Such LMDS may be critical lesions in the DNA as they present an additional challenge to the repair machinery of the cell.
One type of LMDS, the DNA double-strand break, has been the subject of
considerable study and is considered to contribute significantly to
cytotoxicity (6, 7). However, the biological consequences of other
types of LMDS, such as 2 modified bases or two abasic sites close to
each other, require further investigation. A great deal will depend on
if and how these lesions are recognized by the initial damage
recognition enzymes. In the case of abasic sites, the enzymes that
recognize these lesions and cleave the DNA have been classified
according to mechanism. The major apurinic/apyrimidinic (AP) DNA repair
endonuclease in human cells is Ape (8) or HAP1 (9). The major AP
endonuclease of Escherichia coli is the multifunctional enzyme exonuclease III (10). Both enzymes, which share several structural and functional similarities, cleave the phosphodiester bond
in duplex DNA immediately 5 to the abasic site and are thus classified
as class II AP endonucleases (11). E. coli endonuclease III,
on the other hand, cleaves DNA 3
to an abasic site via a
-elimination mechanism and has therefore been classified as a class
I AP endonuclease or AP lyase. An important potential outcome of
cleavage of closely opposed abasic sites would be the production of
double-strand breaks.
Data regarding the enzymatic fate of abasic sites as components of more complex lesions are relatively sparse. Recent physicochemical measurements of an oligonucleotide duplex with a bistranded abasic site were suggestive of a structure in which the lesion forms an extrahelical loop (12). This may explain the earlier observation that abasic sites on opposite strands reduce the efficiency of strand cleavage by Ape (13) and exonuclease III (14) and that bistranded abasic sites site-specifically placed in shuttle vectors and transfected into E. coli and COS-7 simian cells appear to be highly mutagenic (15). All these studies, however, were restricted to stable synthetic analogues of deoxyribose, such as tetrahydrofuranyl groups, placed directly opposite each other. Povirk and Houlgrave (16), on the other hand, examined the enzymatic cleavage of abasic sites generated within the complex lesions induced by bleomycin and neocarzinostatin. They observed that considerably higher concentrations of E. coli exonuclease III (but not endonuclease III) were required to effect DNA cleavage at these lesions in comparison to the more spatially isolated abasic sites generated by heat depurination.
We have previously modeled LMDS using plasmid DNA constructs containing site-specific bistranded dihydrothymine lesions and abasic sites and studied the response of the glycosylase and AP lyase activities of E. coli endonuclease III toward such damage (17). With substrates containing modified bases set 1 and 3 bp apart, the enzyme was able to remove only one of the dihydrothymines and cleave the resulting abasic site, thereby generating a single-strand break. Treatment of the substrates containing two abasic sites, however, readily yielded double-strand breaks. It was thus possible to infer that the glycosylase activity of endonuclease III, but not the AP lyase activity, is inhibited by the presence of a closely positioned break in the opposite strand. In this work, we have extended our modeling of LMDS by preparing oligonucleotide duplexes with bistranded natural abasic sites (i.e. 2-deoxy-D-erythro-pentofuranose) set various distances apart and have examined the reactivity of human Ape protein and E. coli exonuclease III with this type of clustered lesion.
T4 polynucleotide kinase was supplied by U. S.
Biochemical Corp. Uracil-DNA glycosylase was purchased from Life
Technologies, Inc. and Epicentre Technologies. (One unit of uracil-DNA
glycosylase catalyzes the release of 1 nmol of free uracil from
[3H]poly(dU) in 1 h at 37 °C.) Endonuclease III
(fraction IV) was purified from strain N99C1857 carrying
the pHIT1 plasmid (kindly provided by Dr. R. P. Cunningham, State
University of New York, Albany, NY) according to the procedure of
Asahara et al. (18). Exonuclease III was obtained from Life
Technologies, Inc. (One unit of exonuclease III produces 1 nmol of
acid-soluble nucleotide from sonicated DNA in 30 min at 37 °C.)
Recombinant human Ape protein (in the form of the glutathione
S-transferase fusion protein (GST-Ape) and the purified
protein isolated after clipping the fusion protein with Factor Xa) was
generously provided by Drs. David Wilson III and Bruce Demple (Harvard
University). The purification of the enzyme is fully described by
Wilson et al. (13). One unit of enzyme cleaves or releases 1 pmol of damaged sites/min from a synthetic substrate (19) under
standard conditions (20, 21). The concentration of the original stock
of GST fusion protein was 500 µg/ml (100,000 units/ml). For our
purposes, the enzyme was diluted to 7 µg/ml in enzyme reaction buffer
(50 mM Hepes-KOH, pH 7.5, 100 µg/ml bovine serum albumin,
10 mM MgCl2, and 0.05% Triton X-100)
containing 50% glycerol. This protein stock, which was stable at
20 °C for several months, was further diluted to achieve the
desired protein concentrations.
Oligonucleotides (23- or 28-mers) containing
deoxyuridine (dU) at various positions (see Table I) were synthesized
by the DNA Synthesis Core Facility at the University of Alberta.
Oligonucleotides were purified according to Sambrook et al.
(22). Briefly, after the electrophoresis of synthetic oligonucleotides
on a 20% polyacrylamide gel containing 7 M urea, the bands
containing the oligonucleotides were located, using a fluorescent Merck
Silica Gel 60 F254 thin-layer chromatography plate
(Brinkmann Instruments) and UV illumination from above in a Brinkmann
Chromato-Vue (Model CC 20) apparatus, and excised. The gel slices were
crushed, and the oligonucleotides were eluted in a buffer containing
0.1% SDS, 0.5 M ammonium acetate, and 10 mM
magnesium acetate at 37 °C for 12 h in a shaker incubator. The
solution was filtered through a Millex HV filter (0.45 µm Millipore
Corp.), and oligonucleotides were isolated from a Waters reversed-phase
Sep-Pak C18 cartridge. The oligonucleotides (10 pmol) were
labeled at the 5-end with 15 units of T4 polynucleotide kinase in the
reaction buffer provided by the enzyme supplier (U. S. Biochemical
Corp.) and 5 µCi of [
-32P]ATP (specific activity,
3000 Ci/mmol; 10 mCi/ml; Amersham Corp.) at 37 °C for 30 min.
Labeled oligonucleotides were purified by ethanol precipitation (22),
and a sample was run on a 16% polyacrylamide gel to verify their size
and purity.
|
For each oligonucleotide duplex, two separate hybridizations were carried out, either with labeled top or bottom strand (see Table I). The unlabeled strand was always in 2-fold molar excess in the hybridization reaction. The oligonucleotides were taken up in uracil-DNA glycosylase buffer (30 mM Tris-HCl, pH 8.3, 50 mM KCl, and 5 mM MgCl2) and heated for 10 min at 60 °C. The hybridization was allowed to proceed at room temperature for 1 h, followed by the addition of 1-3 units of uracil-DNA glycosylase/20-30 pmol of oligonucleotide. The samples were incubated at 37 °C for 1 h, and the oligonucleotides were precipitated at 0 °C for 30 min with 4 volumes of 5 M ammonium acetate and 12.5 volumes of ice-cold ethanol. The removal of uracil from the DNA was confirmed by monitoring the cleavage at the abasic sites by gel electrophoresis after incubation of the oligonucleotide with 0.5 M NaOH at 37 °C for 1 h or with endonuclease III (6 ng of enzyme/10 pmol of oligonucleotide) in reaction buffer (50 mM Tris-HCl, pH 7.6, 100 mM KCl, 1 mM EDTA, and 0.1 mM dithiothreitol) at 37 °C for 1 h. To minimize nonspecific cleavage at the abasic sites, vortexing, vigorous pipetting, and freezing/thawing of the oligonucleotides were avoided.
Ape Reaction ConditionsOligonucleotide duplex substrates (2.5 × 105 cpm, ~1 pmol) were incubated at 37 °C with 70 pg (unless otherwise indicated) of GST-Ape fusion protein in enzyme reaction buffer (13) in a total volume of 30 µl. Aliquots were removed at various times, and the reaction was stopped on ice by the addition of an equal volume of gel loading buffer containing 90% formamide, 0.02% bromphenol blue, and xylene cyanol in 89 mM Tris borate, pH 8.3, and 2.5 mM EDTA.
Exonuclease III Reaction ConditionsApproximately 5 × 104 cpm (0.2 pmol) of the oligonucleotide duplex were incubated with different amounts of exonuclease III in 66 mM Tris-HCl, pH 8.0, 125 mM NaCl, 5 mM CaCl2, and 10 mM dithiothreitol in a final volume of 10 µl at 37 °C for 10 min. The reaction was stopped on ice by the addition of an equal volume of gel loading buffer.
Analysis of Cleavage ProductsThe reaction products were
analyzed on a 1.5-mm-thick 16% polyacrylamide gel containing 7 M urea. The gel was run at 25 V/cm for 3 h, and after
scanning in a PhosphorImager (Molecular ImagerTM GS-250 system,
Bio-Rad), the cleavage products were quantified with Molecular
AnalystTM software (Bio-Rad). The bands corresponding to the reaction
products and uncleaved material were also excised from the gel after
autoradiography. The radioactivity in each gel band was determined with
a Beckman Model LS 5801 liquid scintillation counter. A typical gel is
shown in Fig. 1. The streak of radioactivity lying
between the full-length (23-mer) oligonucleotides and the labeled
cleaved products is the result of chemical decomposition at the abasic
site occurring during gel electrophoresis. For the purpose of
calculating the percentage cleavage, the radioactivity in the streak
was added to the radioactivity in the band of uncleaved oligonucleotide.
A panel of oligonucleotides (23- or 28-mers) with dU
residues at unique sites were synthesized and annealed to their
complementary strand to form duplexes. Each duplex contained either a
single dU or paired dU nucleosides. The paired dU nucleosides were
either sited directly opposite each other or separated by 1, 3, or 5 bp
in the 5- and 3
-directions. The sequences of these duplexes are given
in Table I. The unlabeled strand in the duplex was always in molar excess compared with the labeled strand to enhance the
hybridization of the labeled strand. The duplexes were treated with
uracil-DNA glycosylase to remove the incorporated uracil and thus
create abasic sites at these positions.
To assess the extent of uracil removal, the oligonucleotide duplexes
were subjected to alkali or endonuclease III treatment to induce
quantitative cleavage at the abasic sites. (An example of NaOH and
endonuclease III analysis for duplex F+A is shown in Fig.
5C.) In accordance with the observation of Eftedal et al. (23), the efficiency of enzymatic release of uracil was dependent on sequence context. In duplexes F+A, F+K, G+A, H+K, and M+N,
the uracils were resistant to uracil-DNA glycosylase and therefore
required the use of 3 units of enzyme to effect >90% removal. The
subsequent percent cleavage of abasic sites after incubation of these
substrates with Ape protein and exonuclease III was normalized
according to the available AP sites in each substrate.
Interaction of Ape with Single AP Sites
Oligonucleotide
duplexes B+A, F+K, G+K, H+K, and M+N containing a single AP site at
various positions (Table I) served as controls. These substrates were
incubated with 70 pg of Ape protein for various periods of time, and
the percent cleavage at the AP sites was determined after
electrophoresis of reaction products on a polyacrylamide gel (see
"Experimental Procedures"). The rate of cleavage in these
substrates is shown in Fig. 2. With all substrates except H+K, the enzyme rapidly cleaved the oligonucleotide, resulting in >50% cleavage after 5 min of incubation. The cleavage rate of
duplex H+K was slow, probably due to the close proximity of the AP site
to the 3-terminus of the oligonucleotide. When cleavage of a longer
duplex (M+N; Table I) containing an AP site in the same sequence
context as in duplex H+K but farther from the 3
-terminus was examined,
the rate of scission matched closely that of the other substrates.
Interaction of Ape with Bistranded AP Sites Directly Opposite or 1 and 3 bp Apart, 3
Substrates C+A (AP sites
directly opposite each other), D+A (AP sites 1 bp apart), and E+A (AP
sites 3 bp apart) were incubated with Ape protein, and the degree of
oligonucleotide incision was determined at various time points. Two
reactions were carried out with each duplex, with only one strand
labeled in each reaction, thus permitting the monitoring of cleavage of
both strands. Hydrolysis of the strands of duplex E+A is shown as an
example in Fig. 1, where the labeled strand is marked with an
asterisk. The kinetics of cleavage of these substrates are
presented in Fig. 3A. Substrate B+A, with a
single abasic site, is included for comparison. Although cleavage rates
were universally slower for the abasic sites in duplexes with
bistranded lesions than for the control, there appeared to be no strong
influence of distance between the lesions.
Interaction of Ape with Bistranded AP Sites 1, 3, and 5 bp Apart, 5
The rate of incision of substrates F+A (AP sites
1 bp apart), G+A (3 bp apart), and H+A (5 bp apart) resulting from
incubation with 70 pg of Ape protein is shown in Fig. 3B.
There is clearly a marked difference between the enzyme response to
bistranded lesions in this orientation compared with lesions displaced
3 to each other. In substrate F+A, strand F was not cleaved even after
1 h of incubation, whereas only ~30% of strand A could be cleaved during this incubation period. Even more striking was the
inability of Ape to incise either AP site in substrate G+A. When the
distance between the abasic sites was extended to 5 bp in substrate
H+A, we observed a near-normal rate of cleavage for strand A, but no
scission of strand H. The slow cleavage of strand H in the control
substrate, H+K (Fig. 2), suggested that the lack of strand H cleavage
in duplex H+A was in part attributable to the proximity of the abasic
site in strand H to the 3
-terminus of the oligonucleotide. This was
confirmed when incubation of Ape with the longer duplex, M+L, resulted
in substantial cleavage of strand M, although notably still less than
scission of strand L, or strand M in the control duplex M+N (Fig.
3C).
The refractory nature of the abasic sites in duplexes F+A and G+A was further evaluated by incubation with 210 and 700 pg of the enzyme (data not shown). Incubation of substrate F+A with 210 pg of Ape protein resulted in ~70% cleavage of strand A, but 700 pg of the enzyme were required to cleave strand F to ~40% after 1 h of incubation. Incubation of substrate G+A with 700 pg of the enzyme resulted in 30-35% cleavage of AP sites in each strand.
Comparison of Clipped Ape Versus GST-Ape Fusion ProteinTo
rule out the possibility that the inability of the enzyme to cleave AP
sites in substrates F+A and G+A was an artifact due to interference
from the glutathione S-transferase component of the fusion
protein, the cleavage of these substrates was re-examined using the
clipped (GST-free) protein. The substrates, including the control
substrate B+A, were incubated with an equal number of units (0.014 units) of GST-Ape and clipped Ape protein for 30 min, and the percent
cleavage was determined. The results (Fig. 4) indicate
little difference in response between the two forms of the enzyme.
Interaction of Exonuclease III with Single and Bistranded Abasic Sites
We carried out a similar analysis with E. coli
exonuclease III in which the substrates were treated with increasing
quantities of enzyme. To suppress the exonucleolytic activity of the
enzyme, CaCl2 was substituted for MgCl2 in the
incubation buffer (14). Nonetheless, some exonucleolytic activity was
still evident, especially with the higher concentrations of exonuclease
III. Examples of the responses are presented in Fig. 5
(A and B). Fig. 5A shows a comparison
of cleavage of strand A with three different complementary oligonucleotides, whereas Fig. 5B contrasts the incision of
the abasic sites in duplex F+A. Fig. 5C shows the result of
cleavage of strand F by endonuclease III (En) or NaOH
(Alk) used to measure the induction of abasic sites.
(Endonuclease III cleavage generates 3-termini with
4-hydroxy-2-pentenal phosphate groups, whereas NaOH generates these
termini as well as termini with unmodified 3
-phosphate groups, hence
the doublet of products seen in the Alk lane.)
Analysis of the percent cleavage of the various oligonucleotide
duplexes (Fig. 6) demonstrated the following. (i) The
rates of scission of all the bistranded lesion-containing duplexes were slower than those of the controls, but varied greatly from one duplex
to the next. (ii) Direct opposition of the abasic site (duplex C+A) did
not lead to a sharp reduction in the cleavage of either abasic site.
(iii) Of those constructs with lesions lying 3 to each other, the most
marked effect was seen with duplex D+A, where the abasic sites were 1 bp apart. (iv) Of the constructs with lesions lying 5
to each other,
the most resistant to enzyme-induced scission was duplex F+A (abasic
sites 1 bp apart), with the abasic site in strand F being even more
resistant than the opposing abasic site.
Of the wide variety of DNA lesions generated by ionizing radiation, double-strand breaks are considered to be the predominant lesion responsible for cell lethality (6). However, in addition to the double-strand breaks initially induced by radiation, double-strand breaks can arise after the cessation of irradiation (24-26) and are most likely due to repair of closely opposed damaged or missing bases or damaged/missing bases opposite single-strand breaks. Theoretical calculations (e.g. Refs. 27 and 28) and recent experimentation (e.g. Refs. 1 and 3) have indicated that a single deposition of energy from ionizing radiation can lead to the formation of two or more DNA lesions (single-strand break or base damage/loss) within one to two helical turns.
In an earlier study (17), we examined the capacity of E. coli endonuclease III, a well characterized DNA glycosylase/AP lyase, to generate double-strand breaks in substrates containing bistranded base lesions (dihydrothymine). The present study has focused
on the major AP endonuclease in human and E. coli cells and
their response to bistranded abasic sites. Our data clearly indicate
that, for both enzymes, the oligonucleotide constructs elicited a
marked variation in response, which was dependent on the distance
between the abasic sites and their orientation to one another. If these
in vitro observations reflect the repair responses in human
and bacterial cells, this class of locally multiply damaged site would
have several deleterious consequences. First, the bistranded abasic
sites in most of the configurations that were examined, especially
those oriented 3 to each, gave rise to double-strand breaks when
treated with either of the AP endonucleases. Thus, in the cell, such
lesions would be anticipated to contribute to lethality caused by
double-strand breaks. Second, the reaction of duplex F+A with Ape and
exonuclease III generated a single-strand break opposite an
endonuclease-resistant abasic site. Third, abasic sites with the
configuration seen in duplex G+A are both resistant to the human AP
endonuclease. Since unrepaired AP sites represent a potent class of
premutagenic lesions (29), the latter configuration of bistranded
abasic sites must be considered to have a high mutagenic potential
unless cells can respond to such endonuclease-resistant abasic sites by
an alternative mechanism (e.g. the SOS response in E. coli).
In addition to contributing to our understanding of the possible
biological consequences of bistranded lesions, this investigation has
afforded information regarding the physical interactions between the
endonucleases and DNA. The strategy of changing the location of one
abasic site while keeping the other constant (strand A or L) in each
duplex has allowed us to map at least some of the bases on the opposite
strand that strongly interact with the enzymes when acting on the
constantly positioned abasic site (Fig. 7). This is a
similar approach to the "missing thymine site interference assay"
used by Devchand et al. (30) to map thymines essential for
binding of the lac repressor to the lac operator.
Exonuclease III and Ape cleave abasic sites in double-stranded DNA far
more efficiently than in single-stranded substrates (31, 32); thus, reduction of activity toward a substrate with abasic sites could be due
to either partial denaturation of the substrate or loss of interaction
with a specific base. Data on denaturation caused by paired abasic
sites, although limited, do suggest that there is sufficient
denaturation to alter enzyme activity. For example, paired abasic sites
set 1 and 3 base pairs apart cause sufficient destabilization to allow
the single strand-specific nuclease, S1 nuclease, to induce
double-strand breaks (33). The analysis of a duplex with abasic sites
situated directly opposite each other demonstrated a melting
temperature 12 °C lower than the duplex lacking an abasic site (12).
Partial denaturation may therefore be responsible for the modest
decrease in the rate of incision by both enzymes of duplexes C+A, D+A,
and E+A (Figs. 3A and 6). In the case of duplexes F+A and
G+A with Ape and duplex F+A with exonuclease III, the resistance of the
substrates to cleavage is much more severe, suggesting a strong
interaction with the missing base. Inspection of the B-form of the DNA
helix reveals that the bases missing in duplexes F+A and G+A are
situated close to each other across the minor groove. (Those in
duplexes D+A and E+A face each other across the major groove.) In their model of exonuclease III/DNA interaction based on x-ray
crystallographic data, Mol et al. (34) proposed two minor
groove binding domains that would make significant contact with the
DNA: the first involving Arg-90 and Lys-36 interacting with the
phosphate backbone while Tyr-63 stacks with the sugar ring of the
deoxyribose and the second being the V-
VI
loop. While no x-ray crystallographic model has been published for Ape,
we would predict a similar interaction with possibly even more
extensive contact with bases in the minor groove given the greater
resistance to Ape than to exonuclease III demonstrated by duplex G+A. A
recent methylation interference analysis of Ape interaction with a
synthetic oligonucleotide duplex carrying an abasic site on one strand
displayed a substantial interference signal from an adenine two
nucleotides 5
to the abasic site (equivalent to the guanine lying
between the two thymines marked in Fig. 7) (35). If this is due to
alkylation of N-3 of adenine, it would also be strongly indicative of
binding within the minor groove.
The x-ray crystallography of exonuclease III (34) also indicated that the enzyme has a relatively large positively charged DNA-binding face (55 Å long). It is therefore possible that subtle differences in DNA charge and hydrophobicity due to sequence context close to the abasic site could influence DNA binding and may account for the unexpected preferential cleavage of strand A in duplex F+A (Fig. 6). Alternatively, sequence context may influence the degree to which the two abasic sites can protrude into the enzyme's active site for cleavage. Again, we anticipate that similar properties would account for the same strand-specific cleavage of this duplex shown by Ape (Figs. 3B and 4).
It is worth noting that many of the more potent chemicals that produce
complex lesions bind to the minor groove and abstract hydrogen atoms
from deoxyribose carbon atoms on opposite strands that protrude into
the minor groove. Neocarzinostatin, for example, generates a lesion
very similar to the product of duplex F+A after Ape or exonuclease III
cleavage, i.e. a single-strand break opposite an abasic site
situated 2 bases 5 to the strand break (36). In this case, the abasic
site is oxidized to a 2-deoxyribonolactone moiety (37). It remains to
be seen if the resistance of the abasic site to exonuclease III (38) is
due to its oxidation or to the presence of a strand break as seen with
the product of duplex F+A.
Finally, the relatively slow rate of Ape-catalyzed scission of strand H
in duplex H+K, where the abasic site lies 7 bases from the 3-terminus,
in comparison to strand M in duplex M+N (Fig. 2), where the abasic site
lies 12 bases from the 3
-terminus, implies that the enzyme interacts
with at least the 7 bases 3
to the abasic site on the same strand.
Wilson et al. (13) have previously shown an absolute
requirement for 3 base pairs on the 3
-side of the lesion.
We thank Drs. David Wilson III and Bruce Demple for generosity in supplying purified recombinant Ape protein and Drs. Lynn Harrison and Susan Wallace (University of Vermont) and Dr. John Ward (University of California, San Diego, CA) for thoughtful discussions and advice.