ABSTRACT
Repair of UV-induced cyclobutane pyrimidine dimers (CPDs) was examined at single
nucleotide resolution in the yeast
Saccharomyces cerevisiae,
using an improved protocol for genomic end-labelling. To obtain the sensitivity required for adduct detection in
yeast, an oligonucleotide-directed enrichment step was introduced into the current methodology
developed for adduct detection in
Escherichia coli
. With this method, heterogeneous repair of CPDs within the
RPB2
locus is observed. Individual CPDs positioned in the transcribed strand are
removed very efficiently with identical kinetics. This fast repair starts
within 23 bases downstream of the transcription initiation site. The non-transcribed strand of the active gene exhibits slow repair without
detectable repair variations between individual lesions. In contrast, CPDs positioned in the promoter region show profound repair heterogeneity. Here, CPDs at
specific sites are removed very quickly, with comparable rates to CPDs
positioned in the transcribed strand, while at other positions lesions are not
repaired at all during the period studied. Interestingly, the fast repair in
the promoter region is dependent on the
RAD7
and
RAD16
genes, as are the slowly repaired CPDs in this region and in the non-transcribed strand. This indicates that the global genome repair pathway
is not intrinsically slow and at specific positions can be as efficient as the
transcription-coupled repair pathway.
When cells are subjected to UV light two major classes of lesions are introduced
into the DNA (
1
), the cyclobutane pyrimidine dimer (CPD) and the pyrimidine (6-4) pyrimidone photoproduct (6-4PP). Both lesions are substrates for the nucleotide excision
repair (NER) pathway. The CPD has been the most studied photoproduct, since its
detection can be achieved by the phage enzyme T4 endonuclease V (T4endoV),
which specifically recognizes CPDs and incises immediately 5' of the lesion (
2
). Repair of CPD lesions is heterogeneous throughout the genome. Gene-specific repair analysis showed that lesions in active genes are more
efficiently repaired than lesions in non-active DNA (
3
), primarily due to preferential repair of the transcribed strand over the non-transcribed strand. This phenomenon has been observed in mammalian cells (
4
),
Escherichia coli
(
5
) and
Saccharomyces cerevisiae
(
6
) and is dependent on transcription (
7
,
8
), indicating a role for the transcription process in efficient recognition of
DNA adducts. In yeast, repair of UV-induced CPDs requires the `core' NER enzymes Rad1, Rad2, Rad3, Rad4,
Rad10, Rad14, Rad25 and Ssl1 (reviewed in
9
). Besides these core enzymes, specific gene products are involved in the repair
of different DNA sequences. In
rad26
[Delta] mutants, efficient repair of the transcribed strand is severely impaired
(
10
), suggesting a specific function for Rad26p in the removal of CPDs from the
transcribed strand of active genes. Other gene products are specifically
involved in NER of non-transribed DNA. In
rad7
[Delta] and
rad16
[Delta] single and double mutants, repair of CPDs in non-transcribed strands of different active genes is completely
abolished (
11
). However, efficient repair of the transcribed strand is unaffected in these
mutants, indicating that the transcription-coupled repair pathway does not require these gene products. These
observations led to the postulation of two subpathways of NER, namely transcription-coupled repair (TCR), which is dependent on transcription and stimulated
by the
RAD26
gene product, and global genome repair (GGR), which requires the Rad7 and Rad16
proteins. Although it is clear that these proteins function in different
subpathways (
12
), it is still unknown how these proteins act at the molecular level.
Recently, it has been shown that variations in repair rate are not confined to
the gene-specific level.
In vivo
repair kinetics can vary even within a single DNA strand. CPDs are removed non-homogeneously from the
lacI
gene in
E.coli
(
13
) and from the
p53
and
PGK1
genes in human cells (
14
,
15
). Repair heterogeneity will have significant implications for mutagenesis,
since slow repair of specific DNA damages might underlie the hotspots for
mutation induction observed in various target genes in tissue culture (
16
) and in tumours (
17
). The objective of this study was to determine the kinetics of NER in
S.cerevisiae
at single nucleotide resolution. To obtain quantitative adduct detection in
yeast cells, a purification and end-labelling procedure was developed partly based on methodology previously
used to analyse repair in
E.coli
(
13
). This procedure allows the detection of
in vivo
DNA adduct incidence as well as the analysis of repair kinetics at the
nucleotide level. The
RPB2
locus was chosen as a target because this gene has been extensively used in
gene-specific repair analysis (
8
,
10
-
12
). In this report we show that fast repair of the
RPB2
locus starts near the transcription initiation site. The kinetics for this
efficient repair are identical for differently positioned CPDs in the
transcribed strand. In contrast, repair 5' of the transcription start site is very heterogeneous on both DNA
strands. These repair variations are not observed for CPDs in the non-transcribed strand of the transcription unit. However, both heterogeneous
repair found in the promoter and slow repair of the non-transcribed strand require the
RAD7
and
RAD16
gene products. Therefore, all CPDs positioned within these regions are
substrates for the global genome repair pathway.
The
S.cerevisiae
wild-type strain used for this study was W303-1B. The isogenic
rad
mutant strains used were MGSC104
rad7
[Delta]::
LEU2
and MGSC126
rad16
[Delta]::
LEU2
(
11
). All strains were kept on selective YNB medium (0.67% yeast nitrogen base, 2%
glucose, 2% bacto-agar) supplemented with the appropriate markers. Cells were grown in
complete medium (YEPD; 1% yeast extract, 2% bacto-peptone, 2% glucose) at 28oC under vigorous shaking.
Yeast cells diluted in chilled phosphate-buffered saline were irradiated with 254 nm UV light (Philips TUV 30W) at
a rate of 3.5 J/m
2
/s. Cells were collected by centrifugation, resuspended in complete medium and
incubated for various times in the dark at 28oC prior to DNA isolation (
18
). DNA samples were purified on CsCl gradients (
19
).
Table 1
Oligonucleotide primers
Samples of 20 [mu]g DNA, containing ~1 * 10
9
copies of the yeast genome, were digested with an appropriate restriction
endonuclease and precipitated according to standard procedures (
22
). Dynal M-280
®
streptavidin beads were used to enrich the desired chromosomal DNA target.
After 3 min incubation at 93oC, 1 pmol biotinylated oligonucleotide (Table
1
) complementary to the fragment of interest was annealed in 100 [mu]l Beads-Binding buffer (BB buffer; 10 mM Tris-HCl, pH 9.0, 1.5 mM MgCl
2
, 50 mM KCl and 0.1% Triton X-100) for 30 min. To increase specificity, the annealing temperature was
chosen 2oC beneath the predicted
T
m
of the primers used. Subsequently, 10 [mu]l (1 mg/[mu]l) streptavidin-coated magnetic beads were added (pre-washed with BB buffer) and incubated for 15 min with
occasional gentle agitation to avoid bead sedimentation. Using the Dynal
magnetic partical concentrator, the immobilized templates were washed once with
BW solution (2.0 M NaCl, 5.0 mM Tris-HCl, pH 7.5, 0.5 mM EDTA), three times with BB buffer and once with TE
(10.0 mM Tris-HCl, pH 8.0, 1 mM EDTA). The captured DNA fragments were eluted from the
beads by incubating for 3 min at room temperature in 10 [mu]l 0.1 M NaOH.
End-labelling conditions were used as described (
13
) with some modifications. An oligomer was designed to be complementary to the 3'-end of the desired DNA fragment with a six nucleotide non-complementary dGTP or dTTP stretch (Table
1
). This nucleotide stretch is used as a template to extend the free 3'-OH end of the restriction fragment of interest with [[alpha]-
32
P]dNTP (either dATP or dCTP depending on the primer used) and Taq DNA
polymerase. The reaction mixture was generated by sequential addition of 10 [mu]l 0.1 M NaOH containing the purified DNA fragment (see above), 37 [mu]l BB buffer, 1.0 [mu]l 1 M HCl, 1.0 pmol oligonucleotide, 0.2 [mu]l [[alpha]-
32
P]dA/CTP (3000 Ci/mmol) and 0.2 U SuperTaq polymerase (HT Biotechnology Ltd). Samples were denatured for
3 min at 93oC and subjected to four consecutive rounds of denaturing (30 s at 93oC), annealing (30 s at
T
m
) and extension (90 s at 72oC) to optimize end-labelling. To assure complete extension, 1 [mu]l 10 mM dA/CTP was added followed by two additional cycles.
Phenol/chloroform extraction was performed to exclude Taq activity in the later
steps.
CPDs were identified using T4endoV. Since incision is most efficient on a double-stranded (ds) DNA substrate, the end-labelled fragments were subjected to a hybridization protocol. A 200-fold molar excess of complementary strand, synthesized by
linear amplification, was added, followed by 3 min incubation at 93oC and gradual cooling to room temperature. Native gel electrophoresis
showed that all labelled DNA fragments were in the dsDNA configuration. The DNA
was precipitated, redissolved and divided into two equal parts. One was
incubated with T4endoV, while the other was mock treated. Samples were
subjected to spin column chromatography and lyophilized to small volumes.
Approximately equal amounts (measured as c.p.m.) were loaded on 6% denaturing
acrylamide gels alongside Maxam-Gilbert sequencing reactions. After drying, autoradiograms were prepared
from the gels.
Autoradiograms were scanned using an LKB Ultrascan XL densitometer (Pharmacia)
and analysed using ImageMastertm software (Pharmacia). Background levels were subtracted and gel band
intensities were corrected for loading variations. Serial dilutions of Maxam-Gilbert sequencing reactions were used to determine the linear range of
the autoradiograms. Quantification data were obtained from experiments carried
out in triplicate. Repair plots were established for each CPD that gave a
sufficient signal to background ratio and were within the linear range of Kodak
X-OMATtm-AR scientific imaging films. The time at which 50% of the
initial damage (signal at
t
= 0) was removed was calculated from these plots.
Oligonucleotides specific for the
S.cerevisiae RPB2
locus were used to map CPDs along the
RPB2
promoter (oligonucleotides RR1TB and RR1NB), the transcription initiation site
(oligonucleotides RD8TB, RH9TB and RP7NB) and within the transcription unit
(oligonucleotides RE2TB and RR2NB). All oligonucleotide primers listed in Table
1 contain 5'-biotinylated ends and therefore could also be used in the oligo-directed purification protocol. All primers indicated were
also synthesized without the six base non-complementary extension to generate PCR fragments for Maxam-Gilbert sequencing.
Maxam-Gilbert sequencing reactions
Maxam-Gilbert sequencing ladders were obtained according to standard procedures
(
21
) using PCR fragments identical to the chromosomal DNA fragment under analysis. After the sequencing reactions the fragments were
32
P-labelled using the tailed oligonucleotides (as described earlier). In this way, a 3'-end-labelled product identical to the chromosomal DNA fragment used
in the repair analysis was obtained.
Saccharomyces cerevisiae
cells were irradiated with UV light. DNA was isolated directly after
irradiation and at several post-incubation time points. The DNA was digested with an appropriate
restriction endonuclease. The experimental system to detect CPDs at nucleotide
resolution as described in Materials and Methods is outlined in Figure
1
. In brief, purification of the fragment of interest was obtained by annealing
to a complementary biotinylated oligonucleotide. DNA hybrids were captured with
paramagnetic streptavidin-coated beads. After extensive washing the target strand was eluted from
the immobilized oligonucleotide. The DNA was labelled using an oligo-directed end-labelling procedure (
13
). Since incision of the DNA strand 5' of the CPD by T4endoV is most efficient on dsDNA (
2
), the complementary strand was added in excess and annealed to the target.
Samples were treated or mock treated with T4endoV, concentrated and subjected
to denaturing PAGE alongside Maxam-Gilbert reactions of the corresponding sequence labelled in an identical
manner. The positions of the individual CPDs are indicated by the length of the
fragments, while the intensity of each signal is a measure of the frequency of
the photoproduct at that particular position. The obtained distribution pattern
reflects the
in vivo
CPD levels at single nucleotide resolution. This method obviates the need for
PCR amplification, thereby circumventing disproportionate adduct distribution
patterns due to site-to-site variations in amplification (
22
) and ligation efficiencies (
23
). After post-incubation, repair of CPDs at specific nucleotides will result in a
decrease in the T4endoV- specific signal compared with the intensity detected directly after
irradiation.
The initial distribution pattern of CPDs after UV irradiation can be seen in
Figure
2
and in the
t
= 0 lanes of Figures
3 ,
4 ,
5 and
7
. As expected, lesions are exclusively found at adjacent pyrimidines.
In vivo
adduct levels are heterogeneous throughout the gene. On average, the order of preference for CPD induction is TT > TC [approx] CT > CC, and increased levels of induction are observed when one or more pyrimidines
are positioned 5' of the dinucleotide. These observations are consistent with experiments
using cloned end-labelled DNA for irradiation (
24
,
25
). Not all potential dimer sites result in significant photoproduct formation
when cells are irradiated
in vivo
. Previously, others have detected photofootprints in yeast (
26
,
27
) in which the absence of
in vivo
photoproduct induction was attributed to local protein-DNA interaction. When we compared the
in vivo
distribution pattern to the pattern obtained from DNA irradiated
in vitro
, clear differences were observed, e.g. formation of photoproducts at nt -127 to -117 is significantly lower
in vivo
. (Fig.
2
A). However, other potential dimer sites, e.g. nt -388 (5'-TT-3') and -400 (5'-TC-3') do not show
detectable photoproduct levels either after
in vivo
or after
in vitro
irradiation at 70 J/m
2
(data not shown). At 400 J/m
2
low levels of CPDs are induced at these sites both
in vitro
and
in vivo
(Fig.
2
B), demonstrating that the lack of CPD induction at 70 J/m
2
is not the result of DNA protection
in vivo
. Therefore, cold spots for DNA damage induction are not only influenced by DNA-interacting proteins but also by the sequence context.
To determine the repair rates of UV-induced CPDs
in vivo
at single nucleotide resolution, cells were irradiated at a UV dose of 70 J/m
2
and incubated to allow DNA repair. Figure
3
shows repair of CPDs along the transcribed strand of the
RPB2
locus in an ORF fragment at position +2214 to +2689. DNA adduct levels were
determined directly after UV irradiation and following 20, 40 and 120 min
incubation. After 20 min incubation, over 80% of the signal present at
t
= 0 is removed for all CPDs (Fig.
3
A). To study in more detail whether repair rates vary between different CPDs
shorter intervals were used. Figure
3
B shows CPD levels at 5, 10 and 15 min after UV irradiation. Repair plots were
produced for each individual CPD to determine the
t
1/2
value as the time at which 50% of the signal present at
t
= 0 has disappeared.
t
1/2
values of 8 +- 1 min were found for dinucleotides in the transcribed strand. No
significant variations in repair rate were observed for differently positioned
CPDs, nor for different dipyrimidine combinations (e.g. TT, CT, TC and CC).
The removal of CPDs from the non-transcribed strand of the
RPB2
locus at this region is distinct.
t
1/2
values are 120 +- 3 min for each CPD examined in this fragment, but for this strand also
no significant variation in repair rate between individual dinucleotides could
be observed (data not shown). Thus, individual dinucleotides in the transcribed
strand are repaired 15 times more efficiently compared with lesions positioned
in the non-transcribed strand, but no significant positional repair variations were
observed between different dinucleotides in both DNA strands.
Fast repair of CPDs was observed in the transcribed strand of the
RPB2
locus compared with repair in the non-transcribed strand. To determine whether the start of this fast repair
coincides with the start of transcription, repair analysis was performed on the
transcribed strand around the transcription initiation site. The initiation
site was previously designated at nt 269 (+- 25 bp) 5' of the ATG using S1 nuclease digestion (
28
). To allow more accurate correlation of CPD repair with transcription initiation, we used primer
extension to refine the mapping of the major transcription initiation site at
278 bp 5' of the ATG (data not shown). All sequence positions mentioned are
calculated according to this position (nt +1). Figure
4
shows induction and repair of CPDs along the template for transcription of
RPB2
from position -60 to +135. CPDs induced immediately 5' of the transcription initiation site show moderate repair rates.
t
1/2
values calculated for lesions at nt -3 (5'-TC-3'), -4 (5'-CC-3'), -6 (5'-TT-3') and -18 (5'-TT-3') were 26, 24, 24 and 27 min respectively. However, fast repair of the template strand is observed for CPDs at dinucleotide +23 and for all CPDs which are 3'
of this position (
t
1/2
= 8 min). Nucleotide +23 is the first position 3' of the transcription initiation site with detectable adduct formation.
Although potential dimer sites are present at DNA positions +1 (5'-TC-3') and +17 (5'-TT-3'), these did not result in
detectable CPD incidence
in vivo
or
in vitro
. Thus, fast repair of the transcribed strand starts within 23 bases from the
transcription initiation site and continues downstream into the transcribed
strand. The CPDs analysed downstream of this dimer site are repaired with equal
efficiency and exhibit identical repair rates to the CPDs within the analysed
ORF fragment further downstream.
Removal of CPDs from the promoter region of the
RPB2
locus was analysed in both DNA strands. For both strands a considerable repair heterogeneity is observed. For example, in the template strand, repair
at positions -132 (5'-TT-3') and -135 (5'-TTTT-3') is not
detectable after 120 min of repair, whereas
t
1/2
for nt -183 (5-CTTTT-3'), separated from the latter by <50 nt, is 14 min (Fig.
5
). Hence, at least a 10-fold variation in repair between specific nucleotides can be observed in
the upstream region of the
RPB2
locus. Figure
6
shows a schematic representation of dimer removal along the promoter and the
transcription initiation site for both strands of the
RPB2
gene. An interesting observation is that specific dimer sites positioned
outside the transcribed regions of the
RPB2
gene are repaired with comparable rates to CPDs in the transcribed strand. For
example, CPDs between nt -70 and +15 in the non-template strand are removed with
t
1/2
values of the order of 10 min (Fig.
6
).
Repair rates within the promoter region were shown to be heterogeneous, in
contrast to the slow and uniform repair of the non-transcribed strand of the active gene. Neither DNA sequence was
transcribed and therefore repair of CPDs within these sequences should be
dependent on the global genome repair pathway (
12
). This suggests that at specific positions, the global genome repair pathway
can be very fast. To investigate whether quickly repaired CPDs positioned
within the promoter are indeed substrates for this pathway, repair analysis was
performed in a
rad7
[Delta] disruption mutant, which is disturbed in global genome repair (
11
). In this mutant, no repair could be observed for sequences upstream of the
transcription initiation site in the template strand (Figs
7
and
8
). Also, CPDs positioned in the non-template strand were not repaired, neither near the transcription initiation site (Fig.
8
) nor in the ORF fragment (data not shown). This indicates that repair of each
CPD within these sequences is completely dependent on the
RAD7
gene product. Furthermore, a repair gradient could be observed near the
transcription initiation site (Fig.
7
). CPDs positioned 5' of nt -18 (5'-TT-3') were not repaired at all in
rad7
[Delta], whereas moderate repair was observed at positions -3 (5'-TC-3'), -4 (5'-CC-3') and -6 (5'-TT-3'), which were completely repaired after 120 min. Subsequent fast transcription-coupled repair was observed from nt +23 onwards. This repair gradient at the transcription initiation site is only seen for the template strand. A schematic representation
of the repair
t
1/2
values is depicted in Figure
8
. Identical results were obtained for a
rad16
[Delta] mutant (data not shown), which is also deficient in the global genome
repair pathway.
Intragenic repair variation is of considerable interest with regard to the
mutagenic potential of carcinogens. In both human tumours (
17
) and tissue culture (
16
), hotspots for mutation induction have been found in different target genes.
Recent data suggest that slow repair of DNA damage at specific sites underlies
the observed mutation hotspots (
13
,
14
), although mutation spectra are also biased by phenotypic selection. Another
parameter that is influenced by the sequence context is the initial
distribution of DNA lesions (
24
,
25
; this study). Since mutations are produced when DNA lesions are by-passed by DNA polymerase during DNA replication, the mutation frequency at
any nucleotide position depends on, besides the mutagenic potential of the
lesion itself, the product of both parameters.
We have developed an assay to detect DNA lesions in the yeast
S.cerevisiae
. With this method, we have analysed repair of UV-induced CPDs along the
RPB2
locus. Since different laboratories have used this locus to study repair at the
gene-specific level,
data obtained in both assays could be compared. CPDs within the transcribed strand are repaired with a
t
1/2
of ~8 min, whereas CPDs positioned in the non-transcribed strand are repaired with a
t
1/2
of ~120 min. Repair rates of CPDs positioned in the transcribed strand appear
higher than repair rates observed in gene-specific repair analysis using this locus (
8
,
11
,
12
). Since in those data CPD frequencies were averaged over kilobase-length DNA fragments of which only part was transcribed, repair kinetics
were influenced by slowly repaired CPDs in non-transcribed regions of the DNA fragment.
Fast repair of the transcribed strand is observed downstream of the
transcription initiation site from dinucleotide position +23 onwards. However,
since no significant number of lesions were detected between this position and
the transcription initiation site, transcription-coupled repair might start farther upstream. An indication of fast repair
prior to dinucleotide 23 is the observation that CPDs at positions -3 and -4 are repaired in a
rad7
[Delta] mutant, albeit with reduced efficiency. Positions for the onset of
transcription-coupled repair have been determined in different organisms and different
genes and do not seem to coincide exactly with the transcription initiation
site. In
E.coli
, fast repair of CPDs in the
lacI
and
lacZ
genes starts 10 and 32 bp respectively downstream of the transcription
initiation site (
29
). In the human
PGK1
gene, fast repair starts in a region 140 bases downstream of the transcription
initiation site (
15
). Recently, however, repair analysis along the UV-inducible human
JUN
promoter showed that for this gene fast repair of the transcribed strand starts
upstream of the transcription initiation site (
30
). These authors suggested that the presence of the general transcription factor
TFIIH, which has a dual role in transcription and NER (
31
), results in locally increased repair efficiency. Although this explanation
seems plausible, it might only be true for promoters with very high
transcriptional activity, since this phenomenon is absent in the
PGK1
gene in human cells and the
RPB2
gene in
S.cerevisiae
(
15
; this work)
.
Fast repair of the transcribed strand of the
RPB2
gene continues from position +23 downstream into the gene, with uniform repair
rates for differently positioned CPDs. This observation strengthens the
hypothesis that the elongating RNA polymerase has a role in efficient repair of
DNA lesions from transcribed regions (
32
). It has been shown that elongating RNA polymerase is blocked by CPDs in the
transcribed strand
in vitro
(
33
,
34
). The uniform rates at which individual sites are repaired in the transcribed
strand can be explained assuming that recognition of the damage by the RNA
polymerase determines the repair rate of individual lesions. Once transcription
is initiated, lesions are recognized by the RNA polymerase with equal
probability under conditions where little variation in transcription rate
exists throughout the gene. Provided that each CPD blocks transcription to the
same extent, this results in identical recognition rates and therefore uniform
repair rates. However, uniform repair within the transcribed strand of the
RPB2
gene is in contrast to the repair heterogeneity observed in the transcribed
strand of the human
p53
gene (
14
). Slow repair was observed at dinucleotide positions frequently mutated in skin
cancer, suggesting that repair variation strongly influences mutation induction
in the
p53
gene. This profound repair heterogeneity in the transcribed strand of the
p53
gene is not a general rule in human genes, since only moderate repair
variations are observed in the human housekeeping gene
PGK1
and the UV-inducible
JUN
locus. Also, in the
E.coli
lacI
gene, repair variation in the transcribed strand is confined to one slowly
repaired dinucleotide (
13
). Furthermore, it has been shown for this gene (
29
) that modest repair heterogeneity observed immediately downstream of the
transcription initiation site converts to fast and uniform repair upon
induction of the gene with isopropylthiogalactosepyranoside, suggesting a more
uniform repair pattern when transcription activity is increased. These latter
observations suggest that the transcribed strands of active genes are repaired
in general with little or no repair heterogeneity. Repair variations in the
transcribed strand of the
p53
gene might be explained by a low transcription rate. Reduced transcription-coupled repair probably leads to a more prominent role of global genome
repair, with possible heterogeneity, in repair of CPDs from the transcribed
strand of the
p53
gene, especially since in human cells repair of the transcribed strand is only
2-fold more efficient compared with the non-transcribed strand. In support of this hypothesis, repair of CPDs
from the transcribed strand is more efficient in
JUN
and
PGK1
compared with
p53
when repair rates are averaged. Although repair efficiency clearly influences
mutation induction in both strands of a target gene (
35
,
36
), the observed uniform repair rates of CPDs within the transcribed strand imply
a more prominent role for CPD induction levels in the distribution of UV-induced mutations, since CPD induction is heterogeneous and dependent on
the sequence context. Further support for this suggestion awaits analysis of
DNA damage incidence, repair and mutation spectra in a yeast locus (in
progress).
Repair rates for the non-transcribed strand also do not exhibit significant positional variations,
with
t
1/2
values of the order of 120 min. This is in contrast to the profound
heterogeneity in repair of CPDs located upstream of the transcription
initiation site, where at least a 10-fold variation can be observed between individual lesions depending on the
dinucleotide position. We suggest two possible explanations for the observed
differences in repair of these distinct non-transcribed DNA regions. One possibility is that repair of non-transcribed DNA exhibits uniform slow repair rates throughout the
genome and behaves like the non-transcribed strand of the
RPB2
gene, except at positions in the genome with a more open or disturbed chromatin
structure. Chromatin perturbations at promoter sequences might render the DNA
more accessible to repair proteins, as they do for the transcription initiation
machinery.
The other possibility is that repair heterogeneity is an intrinsic feature of
global genome repair. CPDs are repaired with profound variations depending on
the chromatin organization and accessibility to DNA repair proteins.
Heterogeneous repair of non-transcribed DNA turns to uniform repair only when transcription on the
opposite strand disturbs or randomizes the local chromatin organization (
37
). In this hypothesis, transcription leads not only to uniformity of repair
rates in the transcribed strand, but also in the non-transcribed strand, although with reduced efficiency, since recognition of
the damage still depends on different factors. Also, a combination of both
possibilities could underlie the repair characteristics observed.
Fast repair patches within the promoter are not the consequence of abberrant
transcription, since this repair heterogeneity is totally dependent on the Rad7
and Rad16 proteins. This indicates that global genome repair at specific
positions can be very efficient and even comparable with repair observed for
CPDs in the transcribed strand. Thus, global genome repair is not necessarily
inefficient. This observation suggests that slow repair of specific CPDs is due
to inhibition rather than to the previously assumed intrinsic slow repair rate
of the global genome repair pathway for CPDs. Bulky chemical adducts and 6-4PPs are repaired more efficiently by the global genome repair pathway
than are CPDs (
38
). It has been suggested that the more profound disturbance of the DNA
conformation at the site of damage underlies this difference. CPDs are minor
distorting lesions compared with bulky adducts (
39
,
40
) and probably therefore less well recognized. One can envisage that dimers
positioned at dinucleotides which are arranged in a nucleosomal structure are
not accessible to DNA repair proteins unless specific gene products rearrange
the DNA structure. However, all CPDs examined in the non-transcribed DNA of the
RPB2
locus, i.e. the fast and slowly repaired lesions, require the RAD7 and RAD16
proteins, indicating that differences in repair rates for individual CPDs do
not result from the action of these proteins at specific positions.
In summary, we have analysed repair at the nucleotide level in the yeast
S.cerevisiae
. This report presents the methodology to study nucleotide excision repair
in vivo
at single nucleotide resolution in yeast. Since no amplification steps are used,
the
in vivo
damage distribution levels are measured quantitatively. We have shown that
heterogeneity in repair of CPDs is observed within the
RPB2
locus. Fast repair of the transcribed strand starts at or directly downstream
of the transcription initiation site and exhibits uniform kinetics. Also, no
significant variations in the repair rate are observed for differently
positioned CPDs in the non-transcribed strand. However, profound variations are observed in the
promoter region of this gene. Both heterogeneous repair within both strands of
the promoter and slow repair of CPDs in the non-transcribed strand are totally dependent on the
RAD7
and
RAD16
gene products, which indicates that repair of CPDs by the global genome repair
pathway can be efficient for non-transcribed DNA.
We thank Drs R.A.Verhage for discussion and critical reading of the manuscript, H.den Dulk for advice regarding the methodology and M.van Nierop and E.E.A.Verhoeven for technical assistance.
REFERENCES
Return