ABSTRACT
We examined the relationship between the nuclear matrix and DNA in the
dihydrofolate reductase domain following irradiation of Chinese hamster cells
with UV light. The fraction of matrix-bound DNA increased in transcribed and non-transcribed regions during a 3 h period after irradiation. However,
no increase was observed with excision repair-deficient cells mutant for the ERCC1 gene. The major UV-induced lesion, the cyclobutane pyrimidine dimer, increased in
frequency in the matrix-bound DNA 1 h after irradiation, in both transcribed and non-transcribed regions, but decreased subsequently. This phenomenon was
also lacking in excision repair-deficient cells. These data demonstrate that recruitment of lesion-containing DNA to the nuclear matrix occurs following UV irradiation
and suggest that this recruitment is dependent upon nucleotide excision repair.
This is consistent with the concept of a `repair factory' residing on the
nuclear matrix at which excision repair occurs.
DNA within the interphase nucleus is complexed with histones and other proteins
in a highly compacted form known as chromatin. Also found within the nucleus is
an insoluble three-dimensional network of proteinaceous, non-histone fibers called the nucleoskeleton. The nucleoskeleton is
thought to play an architectural role in the nucleus by organizing higher order
chromatin structure and is often referred to as the nuclear matrix or scaffold.
It is now clear that many enzymatic functions affecting nucleic acids occur on
the nucleoskeleton, including DNA replication, transcription and RNA splicing.
Current models for the nucleus integrate the structural and functional
properties of chromatin and the nucleoskeleton in a coordinated fashion (
1
-
4
).
A common approach for studying specific interactions between chromatin and the
nucleoskeleton is to cleave the DNA with a nuclease, remove the bulk of DNA
fragments from the insoluble skeleton and examine the properties of the DNA
fraction that remains. One method involves extracting nuclei in a high salt
solution, which removes the majority of histones and other soluble proteins
from chromatin (
5
). The DNA may then be easily digested with a nuclease to generate a soluble
fraction and a small, insoluble portion that sediments with this salt-extracted nucleoskeleton, termed the `nuclear matrix' at this stage. An
alternate approach is to extract nuclei in a hypotonic solution containing the detergent lithium diiodosalicylate (LIS) to remove histones and other
proteins from the DNA (
6
). The DNA is then cleaved with a nuclease and a small fraction associated with
the detergent-extracted nucleoskeleton, called a `scaffold', is isolated. Both methods
use non-physiological ionic conditions, so there is a possibility that DNA-nucleoskeleton interactions may be artifactually created or
destroyed by the extraction procedure (
7
).
To reduce the potential for artifacts, a method has been developed to study
nuclei under physiological salt conditions (
8
). Nuclei are prepared in physiological buffer with agarose beads, which
prevents their aggregation and protects the chromatin from shearing. The DNA
may then be cleaved with nuclease, albeit inefficiently, and the fraction not
attached to the nucleoskeleton electroeluted in physiological buffer. Using
this `physiological salt' method, most of the conclusions drawn from studies of
high salt nuclear matrices have been confirmed. DNA replicated during S phase
is associated with the high salt nuclear matrix (
9
,
10
), the physiological salt nucleoskeleton (
11
) and replication `factories' observed
in situ
by fluorescence and electron microscopy (
12
,
13
). Genes being transcribed by RNA polymerases I and II are located on the high
salt nuclear matrix (
14
-
19
) and the physiological salt nucleoskeleton (
20
-
22
), as are nascent RNA transcripts and splicing intermediates (
20
,
22
-
25
), though the association of genes is not dependent upon the presence of the
RNA. Active RNA polymerase II elongation complexes are also found on the high
salt nuclear matrix (
16
,
26
) and physiological salt nucleoskeleton (
20
) and are observed
in situ
by fluorescence microscopy as discrete foci within the nucleus (
27
).
Cells exposed to chemical or physical agents that damage DNA are able to
mitigate the toxic and mutagenic effects of lesions in the DNA by various
mechanisms of lesion reversal, removal and tolerance (
28
). One ubiquitous mechanism, known as nucleotide excision repair (NER) (
29
,
30
), involves the recognition of a bulky lesion in DNA, incision of the strand of
DNA containing the lesion both 3' and 5' of the lesion, removal of the lesion-containing oligonucleotide and DNA repair replication and
ligation to close the resulting gap. This general mode of DNA repair is found
in many organisms from bacteria and yeast to rodent and human cells (
28
). In mammals NER is tremendously complex and involves ~30 known polypeptides (
31
). Nucleotide excision repair can act to remove lesions from the bulk of a cell
genome (global NER), but it can also act in a transcription-coupled fashion whereby lesions in a gene transcribed by RNA polymerase II
are repaired at a faster rate or to a greater extent than are lesions in the
overall genome. This sub-pathway, which has been observed for UV-induced lesions in bacteria (
32
), yeast (
33
) and mammalian cells (
34
,
35
), is due to enhanced repair of the transcribed DNA strand (
36
) and is dependent on active transcription (
32
,
37
-
40
).
Nucleotide excision repair may also occur in association with the nucleoskeleton
(
41
,
42
). Mullenders
et al
. (
43
) found that in UV-irradiated human cells, DNA containing repair patches was enriched on the
nuclear matrix. Furthermore, in cells deficient in the ability to perform
global NER this effect was enhanced and in cells deficient in the ability to
perform transcription-coupled NER the effect was lost. This suggested that repair patches
synthesized in expressed genes are associated with the nuclear matrix as a
consequence of the association of active genes with the matrix (
43
). In normal cells the phenomenon disappears at higher UV doses (
42
,
44
,
45
), where transcription-coupled repair may decrease relative to global DNA repair due to the
inhibition of transcription (
46
). Repair patches in non-transcribed DNA must then either occur away from the matrix or occur on
the matrix and be released soon after the repair synthesis step. There is
evidence to suggest that repair patches, unlike newly replicated DNA resulting
from S phase synthesis, are easily released from the nucleoskeleton under
physiological conditions (
47
).
We sought to investigate the relationship between damaged DNA, in both
transcribed and non-transcribed regions, and the high salt nuclear matrix in both repair-proficient and NER-deficient hamster cells. We chose to examine DNA-matrix associations, which may reflect the earlier NER
stages of recognition and incision. Using cells with an amplified dihydrofolate
reductase (
DHFR
) domain we examined the association of restriction fragments in transcribed and
non-transcribed regions of this domain with the nuclear matrix after UV
irradiation. The fraction of DNA associated with the nuclear matrix increased,
to a varying degree, in several regions of the domain after irradiation and the
frequency of the predominant UV-induced lesion, the cyclobutane pyrimidine dimer (CPD), increased in
matrix-bound DNA 1 h after irradiation, but then decreased at later times. These
phenomena were not seen in cells mutant for ERCC1 (excision repair cross-complementing group 1), which is absolutely required for NER (
48
,
49
) and is thought to participate in the incision of DNA 5' of the lesion (
50
,
51
). This suggests that the results we observed derive from a direct involvement
of excision repair, rather than a generalized stress response. These data lend
support to a model for global NER in which damage in the genomic DNA is
recognized and brought into association with the nuclear matrix. Repair then
occurs at a matrix-bound `repair factory' consisting of an incision complex (including
ERCC1), the transcription initiation factor TFIIH and enzymes for DNA synthesis
and ligation. After repair synthesis is completed the DNA then loses its
functional association with the matrix-bound repair factory.
B11 and UVL-10-PT cells are both Chinese hamster ovary (CHO) fibroblasts containing
an ~50-fold chromosomal amplification of the
DHFR
gene region. B11 cells are repair proficient, whereas UVL-10-PT cells were derived from UVL-10 cells, which belong to excision repair cross-complementing group 1 (ERCC1) (
48
), previously called ERCC2. The UVL-10-PT cells were generously provided by M. S.Tang (M. D. Anderson
Cancer Center, University of Texas). The cells were grown in minimal essential
medium (Gibco) supplemented with 10% dialyzed fetal bovine serum, glutamine,
non-essential amino acids and 0.5 [mu]M methotrexate in a 37oC humidified atmosphere containing 5% CO
2
. Two days before experimentation cells were split 1:6 into medium containing 0.1 mM thymidine and 1 [mu]Ci/ml [
3
H]thymidine added to label the DNA. Ten hours before irradiation the medium was replaced
with fresh medium of the same composition. At the time of irradiation the cells
had not yet grown to confluence.
For irradiation, the medium was removed from the cells and reserved, then the
cells were rinsed with phosphate-buffered saline (PBS) at 37oC and irradiated with a dose of 10 J/m
2
UV from a germicidal lamp (Westinghouse IL782-30) at an incident rate of 0.39 J/m
2
/s at 254 nm, as determined by an International Light IL254 photometer. After
irradiation the medium was replaced and cells were incubated for 0.5, 1, 2 or 3
h at 37oC. Cells incubated for 0 h were washed with ice-cold PBS immediately after irradiation and harvested. Irradiations
were staggered for the various time points so that all cells were harvested at
the same time.
The cells were washed with ice-cold PBS and harvested in PBS by scraping with a rubber policeman. Cells
were pelleted and nuclei were prepared by vortexing in an ice-cold solution of 10 mM NaCl, 10 mM Tris-HCl, pH 7.4, 3 mM MgCl
2
, 0.5% NP-40 and 0.2 mM phenylmethylsulfonyl fluoride (PMSF), as described (
39
). PMSF was added to all buffers at 0.2 mM in the following steps of the
procedure, until DNA purification.
Nuclear matrix was prepared according to the protocol of Dijkwel and Hamlin (
52
). The nuclei were washed in cold 50 mM KCl, 5 mM Tris-HCl, pH 7.4, and 10 mM MgCl
2
and then extracted in 2 M NaCl. The pellet was then washed three times in cold
50 mM KCl, 20 mM Tris-HCl (pH 7.4), 5 mM MgCl
2
for
Kpn
I digestion. The pellet was digested in 5 ml buffer with 50 U/ml
Kpn
I for 1 h at 37oC and matrix-bound DNA was separated from unbound DNA by centrifugation. Digestion
with
Kpn
I was then repeated with 70 U/ml for 1 h and again with 100 U/ml for 1 h. The
matrix-bound DNA and pooled supernatant DNA were then purified.
Purified DNA was quantified by fluorometry (
53
) and
3
H was measured by liquid scintillation counting. To probe for the presence of
different DNA fragments, 1 [mu]g each (by
3
H radioactivity) of matrix-associated DNA and supernatant DNA were re-cut with
Kpn
I, loaded on neutral agarose gels and electrophoresed. The DNA was transferred
to Hybond N
+
membrane (Amersham) and probed for the DNA fragments indicated in Figure
1
using nick-translated
32
P-labeled probes. The radioactive membranes were exposed to Kodak X-omat AR film without intensifying screens, for times such that the
band intensities remained within the linear range of the film. Autoradiographic
bands on the films were quantified by densitometry using a Hewlett Packard
ScanJet IIp flat bed scanner and NIH Image software. The scanner was calibrated
with NIH Image to provide a linear response using a Kodak standard gray scale.
The `association factor' of DNA with the nuclear matrix was calculated as the
autoradiographic signal intensity of the matrix-bound DNA divided by the signal in total DNA. The signal in total DNA was
derived as the weighted average of the matrix-bound and supernatant DNA signal intensities.
B11 and UVL-10-PT CHO fibroblasts were grown and irradiated with UV light as
described in Materials and Methods. Cells were harvested and nuclei were
prepared and extracted using a standard high salt nuclear matrix procedure (
52
). During the time between harvesting and high salt extraction (<2 h), when cells and nuclei were manipulated at 4oC, we found that there was no overall removal of either pyrimidine(6-4)pyrimidone photoproducts or CPDs (the two major UV-induced lesions), with an assay (
55
-
57
) using monoclonal antibodies (
58
) against each of the two kinds of photoproducts (data not shown). The matrices
were washed and incubated with a restriction endonuclease to remove DNA not
closely associated with the nuclear matrix. We used the enzyme
Kpn
I, which generates large restriction fragments, and our preparations contained 2-10% of the total DNA in the matrix fraction. Equal amounts of DNA from
matrix and supernatant fractions were electrophoresed on agarose gels,
transferred to nylon membranes and probed to measure the relative amounts of
specific DNA fragments. A
Kpn
I restriction map of the
DHFR
region in these cells, including the probes used, is shown in Figure
1
.
In studies with unirradiated B11 cells and a second CHO cell line (CHOC400) (D.
R. Koehler and P. C. Hanawalt, manuscript in preparation) we found no
conclusive evidence for strong high salt matrix attachment sites (association
factor > 1) within the
REP3
gene or sequences downstream of
DHFR
or within the
DHFR
gene, as reported in conflicting studies (
52
,
59
). We obtained similar results using either
Kpn
I or
Bam
HI endonucleases and found that the association factor of the resulting DNA
fragments was positively correlated with fragment size (most strongly in the
transcribed region). A strong matrix attachment site residing downstream of
DHFR
was found in an
Eco
RI fragment previously described (
52
), yet we did not find this attachment site in
Kpn
I or
Bam
HI fragments overlapping the same region. We confirmed these results in B11
cells with the physiological salt method of Jackson
et al
. (
8
) using
Kpn
I,
Bam
HI and
Eco
RI restriction enzymes. In all cases ribosomal DNA was enriched on the high salt
nuclear matrix (Fig.
2
) and physiological nucleoskeleton, as reported by others using these procedures
(
19
,
21
). Results with the excision repair-deficient UVL-10 PT cells were virtually identical. This is also evident in the
association factors for DNA fragments observed immediately following UV
irradiation (
t
= 0; Fig.
3
B and C), in which the UVL-10-PT cells have a similar pattern of association as the B11 cells: no
strong attachment sites, though slightly greater association factors for the
fragments detected by probes pZH-34 and pZH-17. In preliminary experiments we found no significant difference in
the association of
Kpn
I or
Bam
HI fragments with the nuclear matrix in unirradiated cells compared with cells
harvested immediately after irradiation with 10 J/m
2
(
t
= 0 in the experiments below).
We have shown that changes occur in the association of DNA with the nuclear
matrix in the
DHFR
domain of CHO fibroblasts after UV irradiation. First, the matrix association
factor for DNA in the
DHFR
region was increased, up to 2.5-fold in some areas, over a period of 3 h following UV, in both transcribed
and non-transcribed regions of the domain (Fig.
3
A and B). Second, the frequency of CPDs increased 50% in the matrix-bound DNA 1 h after irradiation and then declined, in both the transcribed
and non-transcribed regions (Fig.
4
A and B). Neither of these effects were evident in a CHO fibroblast deficient in
excision repair (Figs
3
C and
4
C), suggesting that the phenomena observed are not a generalized response to
cellular stress, but are specifically due to the action of NER. For a
comparison of the significance of the magnitude of the changes we observed,
consider the finding that the fraction of nuclear matrix-bound vitellogenin II DNA is increased only 3-fold in chick liver after a primary or secondary stimulation of
vitellogenin transcription with estradiol (
15
). This estradiol treatment resulted in a 120-fold (primary) or 1400-fold (secondary) increase in vitellogenin mRNA levels (
15
). We do not know if the small (<35%) drop in the matrix association factor for some DNA fragments in the repair-deficient UVL-10-PT cells after irradiation is significant (Fig.
3
C). There was no change in matrix associations of DNA from repair-proficient cells immediately after irradiation (data not shown), so it
seems unlikely that alterations in chromatin structure due to the presence of
photodimers leads to the loss of matrix association.
Since the observed increase in matrix-bound DNA and CPD frequency occurs in both transcribed and non-transcribed regions of DNA, we consider these effects related to
global genomic DNA repair, but not necessarily transcription-coupled repair (
36
). We sought to investigate whether this phenomenon occurs in the genome as a
whole using a monoclonal antibody-based method that we and others have successfully applied to detect CPDs
and pyrimidine(6-4) pyrimidone photoproducts in the DNA of bacteria (
57
,
60
), yeast (
56
), maize (
55
) and mammalian cells (
61
; D. R. Koehler, unpublished data). Our matrix-associated and supernatant DNA prepared by high salt fractionation,
however, was refractory to repeated attempts at analysis using this technique.
In formulating a hypothesis to explain these changes in nuclear matrix-DNA association we must first consider the differential processing of the
two major kinds of UV-induced lesions in the cell. Hamster cells remove CPDs from the
transcribed strands of active genes, but are generally deficient in removal of
these lesions from the overall genome (
34
,
36
,
62
,
63
). In contrast, the second most frequent UV-induced lesion, the pyrimidine(6-4)pyrimidone photoproduct, is removed very rapidly from the overall
genome, reaching completion in ~3 h (
61
; D. R. Koehler, unpublished data) and this removal may be enhanced in active
genes (
64
). Therefore, regardless of whether CPDs or pyrimidine(6-4)pyrimidone photoproducts or both lesions are responsible for the NER-dependent increase in matrix-bound DNA observed in Figure
3
A and B, we must be aware that CPDs are not subject to complete repair in non-transcribed regions of the genome. Additionally, the increase in CPD
frequency we observed in matrix-bound DNA 1 h after UV (Fig.
4
A and B) may involve NER, but probably does not result in successful repair of
these lesions, especially in the non-transcribed regions. The difference in time course between the increase in
matrix-bound DNA, which remains elevated for at least 3 h, and the increase in
CPD frequency only at 1 h is mysterious, but could be related to differential
processing of the two major UV lesions in CHO cells. In human cells,
pyrimidine(6-4)pyrimidine photoproducts are rapidly repaired and CPDs are also removed
efficiently (though more slowly) from the total genome and both lesions are
removed more rapidly from genes in a transcription-coupled manner (
35
,
36
,
46
,
65
).
Knowing the characteristics of DNA repair in human and hamster cells, our
results can be integrated with those of Mullenders
et al
. (
43
) to form a more complete model of the role of the nuclear matrix in NER. In
repair-proficient human cells, after a UV dose of 5 J/m
2
, Mullenders
et al
. (
43
) found that DNA repair patches were preferentially located close to the nuclear
matrix for up to 2 h after irradiation. Additionally, patches created shortly
after irradiation could not be chased from the matrix over the course of 1 h.
In xeroderma pigmentosum group C (XP-C) cells, which are deficient in all NER that is not transcription-coupled (
46
,
66
,
67
), repair patches were even more frequently found near the nuclear matrix. In
Cockayne's syndrome (CS) cells, which lack transcription-coupled repair but have no demonstrated defect in global NER (
68
,
69
), there was a slight depletion of repair patches near the nuclear matrix. At a
higher UV dose of 30 J/m
2
the enrichment of repair patches at the matrix is lost in normal cells, but
still apparent in XP-C cells (
42
,
45
). This may be due to an overall inhibition of transcription and thus
transcription-coupled repair, relative to global DNA repair, at the higher dose (
46
).
The data of Mullenders
et al
. (
43
) likely indicate that repair patches created in genes, both by the rapid
transcription-coupled mechanism and by the total genomic repair system, remain
associated with the nuclear matrix by virtue of their location in actively
transcribed sequences (
43
). Repair patches created in transcriptionally silent DNA then either occur
distantly from the matrix or also occur on the matrix and are released after
completion. Our data support the latter hypothesis. Our observation that the
matrix association factor for DNA in the
DHFR
domain increased after UV in an NER-dependent fashion and in transcriptionally active and silent regions is
consistent with damaged DNA being recruited to the matrix for repair. The high
salt treatment used to prepare nuclear matrix may precipitate functional DNA
associations with NER proteins on the matrix during the excision repair event,
but before repair synthesis has been completed. After repair synthesis, the
lesion-free DNA would then be released from the matrix. In fact, under physiological conditions repair patches are easily removed from the nucleoskeleton, in
marked contrast to replicative S phase DNA synthesis (
47
). The increase in CPD frequency observed on the matrix 1 h after UV suggests
that the `defect' in hamster cells that prevents significant global CPD repair
occurs in a step subsequent to recruiting the DNA to the matrix, such as later
recognition and/or incision stages of NER.
It is of interest to speculate on the role of ERCC1 in the association of UV-damaged DNA with the nuclear matrix. The initial recognition of DNA damage
probably involves the XPA protein, which is defective in persons with xeroderma
pigmentosum group A (
70
,
71
). ERCC1 has been shown to interact with XPA protein both
in vitro
and
in vivo
(
72
,
73
). ERCC1 may also enhance the DNA damage binding ability of XPA
in vitro
Our data lend support to the concept of a `repair factory' (
82
,
83
) localized on the nucleoskeleton, comprised of many or most of the enzymes
involved in the excision and repair synthesis stages of NER. Excision repair,
like transcription and replicative DNA synthesis, occurs very inefficiently
in vitro
and may require proper spatial orientation of the ~30 polypeptides (
31
) involved. Human cell-free systems for studying NER, in which UV-damaged plasmids are incubated with cell extracts, typically result in the
removal of <5% of the lesions.
In situ
DNA repair occurs at many discrete foci in the nucleus that are unrelated to
DNA and UV lesion density, co-localize with some but not all transcription foci and contain PCNA (
82
). It would be informative to localize proteins involved in repair and
transcription, such as TFIIH, in relation to the nucleoskeleton under
physiological conditions and to examine the distribution of these proteins
after high salt extraction to prepare nuclear matrix.
We are indebted to M. S. Tang (University of Texas, M. D. Anderson Cancer
Center) for allowing us to use the UVL-10-PT cells developed in his laboratory prior to publication. We wish
to thank M. Ljungman, A. K. Ganesan and C. A. Smith for helpful discussions and
A. K. Ganesan, K. K. Bowman and C. A. Smith for critical review of the
manuscript. This work was supported by a Natural Sciences and Engineering
Research Council of Canada (NSERC) post-graduate scholarship to D.R.K. and a Program Project Grant from the
National Institute on Aging and an Outstanding Investigator Grant from the
National Cancer Institute to P.C.H.
REFERENCES
Return