ABSTRACT
Bulky lesions in the template strand block the progression of RNA polymerase II
(RNAP II) and are repaired more rapidly than lesions in the non-transcribed strand, which do not block transcription. In order to better
understand the basis of this transcription-coupled repair we developed an
in vitro
system with purified transcription and nucleotide excision repair proteins and
a plasmid containing the adenovirus major late promoter and a thymine dimer in
the template strand downstream of the transcription start site. The footprint
of RNAP II stalled at the thymine dimer, obtained using DNase I,
[lambda]
exonuclease and T4 polymerase 3
' ->
5
'
exonuclease, covers
~
40 nt and is nearly symmetrical around the dimer. The ternary complex formed at
the lesion site is rather stable, with a half-life of
~20 h. Surprisingly, addition of human repair proteins results in repair of
transcription-blocking dimers in the ternary complex. The blocked polymerase neither
inhibits nor stimulates repair and repair is observed in the absence of CSB
protein, the putative human transcription-repair coupling factor.
DNA is template/substrate for enzyme systems which perform replication,
transcription, repair and recombination. While the interactions of the
individual enzyme systems with DNA are relatively well understood, the
processing of DNA by more than one enzyme system has not been investigated in
detail. It appears that when transcription and replication proceed in the same
direction one can bypass the other with minimal interference; in contrast, when
the replication fork opposes a transcribing RNA polymerase, replication pauses
momentarily and then proceeds without disrupting the transcription complex (
1
-
3
) in both pro- and eukaryotic systems.
The joint actions of transcription and repair systems on damaged DNA lead to a
phenomenon called transcription-coupled repair (
4
,
5
). Phenomenologically, transcription-repair coupling manifests itself by a faster rate of repair of
transcribed DNA, in particular the template strand of transcribed DNA (
5
,
6
), compared with non-transcribed DNA. In
Escherichia coli
, the mechanistic basis of transcription-repair coupling is known: upon encountering a lesion in the template
strand RNA polymerase (RNAP) stalls and forms a stable ternary complex; the DNA lesion within the complex is not readily accessible to the excision nuclease and as a
consequence repair is inhibited (
7
). A transcription-repair coupling factor (TRCF) encoded by the
mfd
gene (
8
) releases the stalled RNA polymerase and recruits the damage recognition
subunit of excision nuclease to the lesion site and thus accelerates the rate
of damage recognition and hence repair (
9
).
In humans as well, transcription by RNAP II increases the rate of repair of the
transcribed strand (
6
). Furthermore, humans appear to have a functional homolog of TRCF which is
encoded by the
CSB
/
ERCC6
gene (
10
,
11
). However, an
in vitro
system for transcription-repair coupling in humans is not available at present. Nevertheless, progress
has been made towards developing such a system. In particular, it has been
found that a cyclobutane thymine dimer (T<>T), which is subject to transcription-coupled repair (
4
), constitutes an absolute block for RNAP II (
12
) when present in the template strand but not when in the coding strand. In contrast, an
acetylaminofluorene adduct, which does not undergo transcription-coupled repair (
13
), only causes brief pausing of RNAP II when located in the template strand (
14
). In the present study we have investigated the properties of the ternary
complex of RNAP II at a T<>T in the template strand and the effect of the ternary complex on human
excision nuclease. Our data show that RNAP II makes a stable ternary complex at
the T<>T site (
t
1/2
~20 h) in which RNAP II covers ~40 nt around the dimer in a nearly symmetrical manner. In contrast to
the prokaryotic system, the stalled RNA polymerase, even in the absence of the
presumptive coupling factor CSB, does not interfere with removal of the damage
by the excision nuclease system.
pPU192, illustrated in Figure
1
, posesses the adenovirus major late promoter (MLP) and a single thymine-thymine dimer located in the template strand at positions 149-150 downstream of the transcription start site. pPU192 also
posesses a T7 RNAP promoter which is co-directional with the MLP. The dimer is located at nt 252-253 downstream of the T7 RNAP transcription start site. The DNA
sequence in the region of the dimer is given in Figure
4
A. pPU192 was labeled with 32
P at either one of two locations indicated in Figure
4
A and was constructed by previously described methods (
15
). pMLU112 contains the MLP and a downstream sequence (`U-less cassette') such that the first 112 nt of the transcript contains no U
(
16
).
Transcription was reconstituted with purified recombinant (TBP, IIB, IIE and
IIF) and native (IIH and RNAPII) human proteins (except yTBP) as described
previously (
17
,
18
), with some modifications. Reactions were in transcription buffer (60 mM HEPES,
pH 7.9, 6 mM Tris, pH 7.9, 108 mM KCl, 6.4 mM MgCl2
, 2.1 mM EDTA, 4 mM dithiothreitol, 2.8 mM [beta]-mercaptoethanol, 5.5% glycerol and 3% polyethylene glycol). Template
(~20 ng) was mixed with general transcription factors (GTFs) for 30 min at 28oC and then rNTPs were added to 625 [mu]M each. To label transcripts, CTP was added to only 1.5 [mu]M and several [mu]Ci of [[alpha]-32
P]CTP were added. [alpha]-Amanitin when used was at 10 [mu]g/ml. Incubation continued for 45 min. To analyze transcripts,
reactions were extracted with phenol/chloroform, precipitated and RNA was
resolved on sequencing gels. To footprint or repair transcribed
template/substrate, transcription reactions were diluted with 2 vol. of a
solution so as to arrive at repair buffer conditions (8.7 mM Tris, pH 7.9, 30
mM HEPES, pH 7.9, 61 mM KCl, 13 mM NaCl, 5.4 mM MgCl2
, 0.9 mM EDTA, 2 mM dithiothreitol, 0.9 mM [beta]-mercaptoethanol, 5% glycerol, 1% polyethylene glycol, 1.9 mM ATP,
0.21 mM each GTP, CTP and UTP, 20 [mu]M each dNTP, 133 [mu]g/ml BSA and 17 [mu]g/ml pMC1). pMC1 was added as competitor DNA for non-specific DNA binding proteins. At this point reagents for
footprinting and/or repair were included and further processing of reactions
was as described below.
Transcription by T7 RNAP was with 12 U enzyme (Promega) directly in repair
buffer.
DNase I (Gibco BRL) digestion was with several hundred units of enzyme for 5 min
in the presence of 2 [mu]M CaCl2
. Products were extracted with phenol, precipitated and resolved on 8 or 10%
sequencing gels. Digestion with [lambda] exonuclease (Pharmacia) was with 27 U enzyme in repair buffer after
cleaving the plasmid with
Pvu
II and
Hae
III. Digestion with 2 U T4 polymerase (BMB) was after digesting the plasmid with
Hae
III and then diluting the 30 [mu]l reaction to 100 [mu]l with 1* commercial T4 polymerase reaction buffer.
Preparation of cell-free extract (CFE) was as described previously (
19
). Excision assay with CFE was under the conditions described above for 25 min
at 29oC. After repair, extraction with phenol and precipitation in ethanol, the
DNA was resolved on an 8 or 10% sequencing gel to identify the radiolabeled 24-29 nt products of the nucleotide excision reaction.
Excision assay with a reconstituted system of partially purified native human
general repair factors (GRFs) was also used. GRFs were generated based upon a
reproducible purification scheme previously described (
20
), with the exception that we used native human XPA rather than recombinant
protein from
E.coli
. All of the repair proteins passed through the first column, DE-52. We found by complementation analysis that XPA passed through the
second column, affigel blue, and highly active material was obtained following
two additional passes through the affigel blue column. The ERCC-1-XPF complex, which is eluted from the affigel blue column with low
salt, was further purified on a heparin column and a second preparation of
ERCC1-XPF was additionally purified on an MBP-XPA affinity column. XPC and TFIIH, which partially co-elute from the affigel blue column at intermediate salt, were
pooled and separated from one another on an SP-Sepharose column. RPA, XPG and CSB, which co-elute when the affigel blue column is washed with high salt, were
also separated from one another on an SP-Sepharose column. In some experiments we used recombinant XPG purified
from insect cells (
21
). In our system, omission of any repair protein individually, except XPG,
abolished repair. XPG is known to partially co-purify with TFIIH by this procedure (
20
). The optimal amount of each GRF used for repair was determined empirically.
This reconstituted repair system was found to be several times more active than
CFE. An activity (described in Results) which removes RNAP II stalled at a T<>T was largely removed from the reconstituted system. The RNAP releasing
activity co-purified with TFIIH and XPC through the DE-52 and affigel blue columns but it did not bind to the third column,
SP-Sepharose, which retained both XPC and TFIIH.
To assay repair of lesions where RNAP II was stalled, transcription reactions
were first photoreactivated with 5 nM
E.coli
DNA photolyase. This procedure was used to remove T<>T located on templates where RNAP II was not stalled. In controls conducted in
the absence of photoreactivating light, this addition of photolyase was found
to inhibit nucleotide excision repair by <20%. After photoreactivation, reactions were diluted into repair buffer with
GRFs and incubated at 29oC for 25 min in the absence of photoreactivating light. In parallel
reactions, DNase I footprinting was performed during an additional 5 min
incubation (after the 25 min reaction) to determine the percentage of templates
having a stalled polymerase. In additional parallel reactions, after
photoreactivation, a low level of pPU192 (3-12% of the original 20 ng added) was added to samples that had undergone
both mock transcription in the presence of [alpha]-amanitin and then photoreactivation. After excision repair,
reactions were then digested with proteinase K, extracted with phenol,
precipitated with ethanol and resolved on an 8% sequencing gel. In processing
DNase I digestion products, the proteinase K digestion was omitted. Products
were quantified with a phosphorimager.
With the reconstituted transcription system we first confirmed the observation
that a T<>T in the template strand blocks progression of RNAP II (
12
). Radiolabeled transcripts made from pPU192 were compared with transcripts made
from a control undamaged plasmid. The presence of the dimer resulted in the
formation of a truncated transcript (data not shown). The truncated RNA was ~145-150 nt in length, indicating that RNAP II transcribes very close to
the lesion before it stops (Fig.
1
), as previously reported (
12
). A low level of transcription past the dimer site (~1%) was observed, a result of either transcriptional bypass of the dimer or
trace contamination of the damaged pPU192 with undamaged plasmid.
To examine the status of RNAP II upon encountering a T<>T we conducted footprinting experiments. To obtain the footprint the
radiolabel was incorporated at the 13th phosphodiester bond downstream of the T<>T in pPU192 and following incubation with GTFs and RNAP II, the DNA was
digested exhaustively with DNase I and then analyzed on sequencing gels. Using
this procedure 30-48 nt fragments of the transcribed strand were protected from DNase I
(Fig.
2
, lane 2). No protection was observed when transcription was inhibited with [alpha]-amanitin (lane 1), by the absence of rNTPs or by omission of RNAP
II (data not shown). With another template in which the dimer was present in a
different sequence context the stalled polymerase protected a 36-54 nt region from DNase I (data not shown). Using DNase I protection we
measured the stability of the ternary complex formed at a T<>T site. Figure
2
shows the results of an experiment conducted at room temperature over an 8 day
period. The ternary complex is quite stable, with a half-life of ~20 h.
To understand the interactions of the transcription apparatus with the
nucleotide excision repair system it is useful to know the positioning of the
stalled RNAP II around the dimer. Therefore, the boundaries of the DNase I
footprint of stalled RNAP II were determined. Footprints such as in Figure
2
were gel purified and then digested with either [lambda] 5' -> 3' exonuclease or the T4 DNA polymerase 3' -> 5' exonuclease. Digestion by these
nucleases is impeded when they encounter the dimer (
22
). The results (not shown) indicate that RNAP II covers 13-16 nt 3' of the dimer and 8-18 nt 5' of the dimer. The DNase I footprinting was
complemented by direct exonucleolytic footprinting with [lambda] and T4 DNA polymerase exonucleases and using linearized DNA with a
ternary complex. A stalled RNAP II blocks [lambda] exonuclease 26 nt 5' of the dimer and T4 3' -> 5' exonuclease 18, 19 and 23 nt 3' of the dimer (Fig.
3
).
To examine repair of transcription-blocking lesions, repair-competent extracts were added to reactions in which RNAP II was
stalled at a dimer. Results could not be interpreted because, unexpectedly, the
extract, whether prepared from wild-type HeLa or
CSA
-
or
CSB
-
mutant cells, released RNAP II from the template, as determined by footprinting
experiments (data not shown). An RNAP II `release' factor, or `factor R', was
partially purified and its activity is shown in Figure
5
. In this assay of factor R, RNAP II was stalled by nucleotide starvation at the
end of a 112 nt long U-less cassette and then incubated with and without factor R. UTP was then
added and transcription proceeded to the
Pvu
II site where the template had been cleaved. Factor R prevented elongation upon
addition of UTP. Thus, both physical and functional assays show that CFEs
contain an activity which disrupts the ternary complex making it impossible to
study transcription-repair coupling in extracts.
To examine repair where RNAP II was stalled we performed the experiments
outlined in Figure
6
. First, a reconstituted transcription system was used to form stalled ternary
complexes at the dimer in pPU192. This system transcribes 3-10% of input plasmids. The remaining plasmids were photoreactivated by
treatment with
E.coli
DNA photolyase plus near-UV light. This treatment does not repair dimers where RNAP II is stalled (
12
) and thus results in two major populations of DNA: (i) plasmid with no dimer
and no RNAP II; (ii) plasmid with RNAP II stalled at a dimer. At this point,
instead of performing repair with CFE, which removes polymerase, or highly
purified GRFs, which do not efficiently repair circular substrates (data not
shown), we used partially purified human GRFs.
For comparison with the human transcription system, we tested the effect of a T7
RNAP blocked at the dimer on the human repair enzyme. Transcription of pPU192
from the T7 RNAP promoter was controlled by adding or withholding polymerase
and rNTPs. In contrast to the human polymerase, T7 RNAP is highly efficient and
transcribes nearly 100% of templates. Therefore, we simply incubated pPU192
briefly with T7 RNAP and rNTPs to form stalled elongation complexes, then
repaired the DNA with human CFE. The results in Figure
8
show that, in contrast to human RNAP II, T7 RNAP II stalled at the dimer
prevents repair of the dimer by human excision nuclease.
The progress of polymerases, including human RNAP II, has been shown to be
impeded by bulky adducts located in the template strand but is not blocked by
adducts in the complementary strand (
12
). This blockage of RNAP II is probably an early event in the human
transcription-stimulated repair reaction (
14
), as it is in
E.coli
. Footprints of human RNAP II stalled by nucleotide starvation have been
described. Ternary complexes stalled at different sites demonstrated DNase I
footprint sizes that ranged from 42 to 58 nt. Interestingly, more of the coding
strand was protected than the template strand and complexes stalled at
different sites on the template exhibited slightly different footprint sizes (
25
-
27
). Exonuclease III footprints of ternary complexes generated by nucleotide
starvation were 33-39 bp in size (
28
). Regions of protection obtained with the different footprinting procedures
were fairly symetrical around the 3'-end of the RNA and stalled elongation complexes were reported to be
stable. Thus, the footprint and stability of RNAP II blocked at a dimer are
generally similar to those of RNAP II stalled by nucleotide starvation and the
structural feature that elicits preferential repair may be no more than the
combination of a stationary polymerase in a ternary complex and the bulky DNA
adduct.
By using purified transcription and repair proteins we were able to examine
repair in the absence of removal of RNAP II. With this system and using
enzymatically purified ternary complexes we observed repair of the dimer within
the complex. The presence of the polymerase had no detectable inhibitory or
stimulatory effect on repair. This result was unexpected for several reasons.
First, in a comparable system with purified
E.coli
transcription and repair proteins, the stalled polymerase did inhibit repair (
7
). Also, stalled human RNAP II was found to partially or completely `cover' the
incision sites of human excision nuclease. Furthermore, in the ternary complex,
the 3' incision site and lesion may be in a region or adjacent to a region of
single-stranded DNA or a DNA-RNA hybrid, as shown in Figure
4
. Finally, RNAP II blocked at a T<>T is known to prevent repair of the T<>T by
E.coli
DNA photolyase (
12
). Our results do not have the resolution to answer whether or not the ternary
complex is removed or remains bound after the dual incision.
Our finding that repair occurs in a ternary complex containing a transcription
bubble terminating at a photodimer suggests that a dimer adjacent to a
synthetic transcription bubble should be an efficient substrate for human
excision nuclease. This prediction has recently been confirmed in a model
system in which the T<>T was preceded on the 3'-side by a bubble of 10 mismatched nucleotides (Mu and Sancar,
unpublished results). The effect of an RNA-DNA hybrid on repair has not yet been tested.
In
E.coli
, RNAP stalled at a dimer inhibits nucleotide excision repair of the dimer
in vitro
(
7
). The
E.coli
Mfd protein couples repair to transcription by removing the polymerase stalled
at lesions in the template strand and delivering the repair enzymes to the
lesion. In UV-irradiated
E.coli
cells which are
mfd
-
and lack the transcription-repair coupling factor, specific inhibition of repair by stalled RNAP
results in an elevated frequency of mutation, specifically in the template
strand (
29
). Since human RNAP II stalled at a dimer does not inhibit repair, it is
predicted that the strand bias for mutation in
mfd
-
cells is absent from
CSA
-
and
CSB
-
human cells, which lack the presumed human coupling factors.
A relevant observation has been made with Rad 26 disruption mutants of
Saccharomyces cerevisiae
. Rad 26, which is homologous to human CSB, is the yeast transcription-repair coupling factor (
30
). In certain genetic backgrounds, transcription-stimulated repair was observed even in the absence of Rad 26 protein (
31
). This finding is consistent with transcription-repair coupling resulting from a combination of coupling factor-independent and -dependent pathways. It is possible that the coupling factor-independent pathway involves a release from inhibition.
Chromatin structure inhibits repair (
32
,
33
). Consequently, when a lesion in the template strand blocks RNAP II it may
become more accessible to repair enzymes than when constrained within a
nucleosome.
Recent investigations of coupling factor-dependent repair have shown that purified CSB protein does not remove a
polymerase stalled at a dimer (
18
) as does its
E.coli
counterpart (Mfd) (
9
). However, CSB does bind to human RNAP II (Selby and Sancar, unpublished
result) and to human GRFs (
18
,
34
) and thus may enhance the repair rate by recruiting repair enzymes to the
transcription-blocking lesion, as is the case in
E.coli
(
9
).
We thank Drs D.Svoboda and J.-S.Taylor for providing the dimer-containing 20mer used for constructing pPU192. This work was
supported by grants from the NIH (A.S. and D.R.), the Howard Hughes Medical
Institute (D.R.), grant CTR 3852 from the Council for Tobacco Research USA Inc.
(C.P.S. and A.S.) and by a NIH predoctoral grant to R.D.
REFERENCES
Return