 |
INTRODUCTION |
The known contributions of the human DNA mismatch repair
(MMR)1 pathway to genomic
stability have expanded greatly over the past decade (1-3). A clear
relationship between the loss of a functional DNA MMR pathway and tumor
formation in both hereditary and sporadic settings has emerged (1, 2).
Loss of one of the protein activities essential for mismatch correction
results in a mutator phenotype, in which mononucleotide repeats within
coding sequences of genes that contribute to growth suppression or
apoptosis represent important mutational hot spots. In addition to its
role in repairing biosynthetic errors, the MMR pathway plays an
oversight role in DNA recombination events, limiting nonhomologous
strand exchange (4, 5). Lesion recognition and processing have also
been implicated in apoptotic events mediated by p53 and p73, although the specific sequence of events subsequent to lesion recognition is not
well defined (for review, see Ref. 6). Two commonly cited models are
the futile repair model, in which attempted repair events target the
strand without the lesion, and the direct signaling model, in which
assembly of the MMR machinery at the lesion initiates the apoptotic cascade.
The precise mechanism by which mismatch recognition results in strand
excision and resynthesis remains unresolved (1, 3, 7). This is
especially true for the events subsequent to mismatch recognition,
including the identification of the strand discrimination mechanism.
One model posits that mismatches are bound by the MSH2/MSH6 heterodimer
(MutS
), followed by an ATP-dependent, bidirectional translocation along the DNA to identify a strand discrimination signal
and orchestrate downstream events (8). A more recent analysis suggests
that MutS and MutL (MLH1/PMS2) together form a complex capable of
translocating along the DNA helix (9). A second model suggests that
MSH2/MSH6 is an activity analogous to a G protein, recognizing
mismatches when an ADP molecule is bound (10, 11). Exchange of ADP for
ATP in the ternary complex results in a hydrolysis-independent sliding
clamp that allows diffusion along the DNA, again presumably to initiate
later events in the repair pathway. In a very different formulation, it
has also been argued that mismatch-binding proteins, in the presence of
a heterodimer of MutL homologs and ATP, remain at the mismatch site
(3). This model was formulated using prokaryotic proteins and proposes
a role for ATP in mismatch verification (12), which appears to be
inconsistent with the two models cited above. Proliferating cell
nuclear antigen is also known to participate in the early stages
of mismatch correction (13), most recently as a modulator of mismatch
recognition (14) and by its ability to disrupt the complex of MutS and
MutL homologs from the mismatch site in Saccharomyces cerevisiae (15).
In contrast to the diverse efforts directed toward understanding how
the protein components interact to effect mismatch correction, the
known DNA template requirements for human MMR are few. A successful template for in vitro assays can be summarized as a circular
molecule containing a mismatch and nick positioned from 100 to 1,000 bp away in either DNA strand (16, 17), where repair efficiency is
diminished as the distance separating the two sites is increased (18).
The nick is essential for targeting repair to the strand where it
resides (16, 17, 19), although the ultimate strand signal in human
cells (or any organism outside a small number of related Gram-negative
prokaryotes) remains unknown. Recent biochemical analyses suggest that
mismatch processing in human cell extracts operates independently of
the CpG methylation status on the template, whether it is fully, hemi-,
or unmethylated (20; for an analysis using Xenopus
laevis extracts, see Ref. 21).
The possibility that a nick was sufficient to act as a strand signal
was drawn from the better characterized Escherichia coli model system (3, 22). In this organism, a site-specific DNA nick
represents an intermediate in mismatch correction (23, 24). The action
of MutH protein on a hemimethylated GATC sequence introduces a
strand-specific nick, and this reaction is largely dependent upon the
presence of a nearby mismatch (25). Furthermore, the presence of a nick
eliminates the need for MutH activity in a reconstituted assay system
(24). In parallel with the bidirectional capability of mismatch
correction in E. coli (26), extracts from human cells can
access a single strand nick placed in either strand and in either
orientation with respect to the mismatch (18). Mismatch processing
appears to function equally well when exonucleolytic removal of the
mismatch occurs 5' 
3' or 3' 5'.
Although a nick may act as a strand discrimination device in
vitro, its relationship with the authentic strand signal in
vivo remains unclear. Strand discontinuities, including nicks, are present during replication, but they are much more frequent in lagging
strand biosynthesis. Another potential means of identifying the
daughter strand is via the DNA end where strand elongation occurs, even
though it is sequestered by the replicative polymerase. This
possibility is supported by the fact that proliferating cell nuclear
antigen participates in an early step in mismatch correction, prior to
resynthesis, and it has been shown to interact directly with mismatch
recognition proteins (13, 27). Models for strand discrimination which
invoke the site of DNA synthesis are superficially inconsistent with
nick-directed repair because displacement of the replicative polymerase
to allow access to the MMR machinery is likely to reveal a DNA end
(3'-OH) within a gap rather than a nick. Restated, the ability to
identify and access a nick does not represent the same capability as
accessing the active site of replication.
In this work, we assessed the contribution of two major features of the
DNA template upon which the human MMR pathway acts. First, we asked
whether strand discontinuities beyond a single strand nick are
competent to target mismatch correction. Heteroduplexes containing a
single G·T mismatch and a site- and size-specific single strand gap,
positioned in either orientation (5' 3' or 3' 5' from the
discontinuity to the mismatch) were assayed using nuclear extracts
prepared from HeLa cells. The gap sizes chosen were intended to
represent those that might arise from specific DNA repair or synthesis
events and would plausibly represent targets for the human DNA
polymerases , 
, 
, or 
. Nicks and gaps of 4-210 nucleotides
were found to be effective strand signals, and intermediate sized gaps
were the most efficient in either orientation. Second, we asked whether
DNA ends were competent to direct MMR, and we found that neither blunt
ends nor small overhangs supported mismatch correction targeted to
either strand. MMR was supported efficiently on linear templates, and
we used this result to ask systematically whether DNA ends limited
mismatch correction as they were moved close to either the strand
signal or mismatch. In this way, we were able to define some of the
spatial and structural relationships required for efficient mismatch
correction in vitro.
| |
MATERIALS AND METHODS |
Construction of M13 Derivatives Competent for Assessing MMR Using
Restriction Analysis--
Manipulation of M13mp18 bacteriophage and
its derivatives was performed using standard methodology for culturing
filamentous phage (28) using the XL-1 Blue strain of E. coli
(Stratagene). Similarly, all DNA manipulation steps such as
precipitation, digestion, and ligation were carried out using standard
methodology (28). Before introducing a site competent for mismatch
construction, the unique HindIII site was eliminated from
M13mp18. A sample of the DNA was digested with HindIII, the
overhanging ends filled in with the Klenow fragment of DNA
polymerase I, and the molecule was recircularized using a blunt end
ligation. The original HindIII site was thus replaced with
an NheI site, giving a molecule we called M13
H.
An oligonucleotide duplex containing a G·T mismatch was designed to
replace the sequence intervening between the M13H KasI and AvaII sites and generate two progeny phages for
substrate construction. The sequence, described below, included a
mismatch embedded within overlapping restriction sites to facilitate
strand-specific identification of MMR in the final substrates (29; see
Fig. 1). After digestion with AvaII and KasI,
removal of the putative intervening fragment, and ligation in the
presence of the replacement heteroduplex, transformation yielded single
plaques that were sequenced across the targeted region. It became
apparent that the KasI digestion failed, and a cryptic but
useful ligation reaction occurred. Plaques were isolated in which the
AvaII site was reconstituted correctly, but the
KasI overhang was trimmed from the oligonucleotide and ligated to the second AvaII overhang, which was apparently
also converted to a blunt end by replicative filling in
vivo. Two of the recovered plaques had sequences across the
modified region which read
5'-GGACCTAGGCAAGCT(T/C)TCGAGTCTGACC. The sequence in bold represents the parent M13H sequence after oligonucleotide incorporation (plain text) into the AvaII
site (5'-GGACC), and the bases in italics represent the filled-in
overhang. The double strand form of M13H clone 4 was sensitive to
XhoI (C), whereas M13H clone 7 was sensitive to
HindIII (T). Both included an additional StyI
site (CCTAGG) located adjacent to the mismatch site to facilitate
template remodeling.
To construct G·T heteroduplexes with 5'-strand discontinuities, large
scale preparations of both double (ds) and single strand (ss) DNA were
performed such that clone 4 was used as the double strand species and
clone 7 as the single strand species. Because the heteroduplex region
was introduced at the AvaII site of M13H, such constructs
place the mismatch 329 bp from the EcoRI site and a rich
source of other nearby unique digestion sites useful for the
construction of gapped regions. The DNA sequence reads 5' 3' from
nick or gap toward the mismatch. Repair directed by the nick or gaps is
therefore scored with HindIII, whereas mismatch correction
in the continuous strand is identified with XhoI (16,
29).
To construct comparable phages useful for preparing 3'-nicked or gapped
substrates, a 394-bp segment containing the mismatch region and
polycloning site of M13 was excised and inverted. The fragment was
first digested with NheI and StyI, which
recognize distinct sequences but yield complementary overhanging ends
upon digestion. The fragment was reinstalled in situ by
adding a 1/10 volume of 10× ligase buffer (New England Biolabs) and 80 units of T4 DNA ligase, then incubating for 24 h at 22 °C in
the presence of NheI and StyI. The presence of
these activities selects against regeneration of the parent phage. The
reaction was transformed into XL1-Blue cells, and the resulting plaques
were screened for loss of StyI and NheI
sensitivity. Two selected phages, designated 4F (XhoI site)
and 7F (HindIII site), were sequenced through the targeted
region to demonstrate sequence inversion between the StyI
and NheI sites.
For construction of 3'-substrates, clone 4F was used to prepare the
single strand species, and DNA derived from clone 7F was used to
amplify the double strand species. The substrate preparation protocol
described below results in the production of G·T heteroduplexes wherein the DNA sequence intervening between the mismatch and strand
discontinuity is preserved compared with the 5'-substrates. Mismatch
removal and resynthesis events that are nick- or gap-directed now yield
products that are sensitive to XhoI, whereas mismatch correction in the continuous strand is identified with
HindIII (16, 29).
Construction of a Nicked Circular DNA Containing a G·T
Mispair--
The preparation of heteroduplexes was performed using
minor modifications to published protocols (29). To construct a nicked substrate, 150 µg of the appropriate dsDNA was linearized by
digestion with EcoRI and precipitated by the addition of
1/10 volume of 3 M sodium acetate and 2 volumes of absolute
ethanol. The DNA was recovered by centrifugation for 20 min at
16,000 × g in a microcentrifuge at room temperature.
This precipitation protocol was followed for each ethanol precipitation
described subsequently, unless otherwise specified. The pelleted DNA
was then resuspended in 2.6 ml of buffer containing 50 mM
Tris-HCl, pH 8.0, 12.5 mM NaCl, 0.8 mM EDTA,
and 750 µg of ssDNA. The strands were denatured for 5 min at room
temperature by the addition of 60 µl of 10 N NaOH, and
the reaction was neutralized by the addition of a mixture of 200 µl
of 2.9 N acetic acid, 248 µl of 1 M potassium
phosphate buffer, pH 7.4, and 90 µl of 3 M KCl. The
reannealing reaction was incubated at 65 °C for 1 h and then
37 °C for a further 3 h. The solution was diluted with 1.5 volumes of distilled water and the DNA precipitated with ethanol. The
DNA was recovered by centrifugation in an SW41 swinging bucket rotor at
68,000 × g for 1 h at 20 °C. To remove the
excess ssDNA, the pellet was resuspended in 2.5 ml of 1 M
NaCl and passed three times through a column containing 1 ml of
benzoylated napthoylated DEAE-cellulose (Sigma) that had been
equilibrated in 1 M NaCl. The dsDNA was then precipitated directly from this solution by adding 2 volumes of absolute ethanol and
centrifugation as described above. To digest any residual linear DNA,
the pellet was resuspended in a 400-µl reaction containing 66 mM glycine at pH 9.4, 5 mM MgCl2,
0.5 mM ATP, and 35 units of exonuclease V enzyme (U. S.
Biochemical Corp.) and incubated at 37 °C for 1 h. The enzyme
was then inactivated by a 20-min incubation at 65 °C. The circular
substrate was fractionated from nucleotides and small oligonucleotides
by ultrafiltration; each reaction was concentrated to less than 100 µl in volume using a Microcon 100 centrifugal filter as described by
the manufacturer (Millipore). The filtrate containing small molecules
and buffer was discarded, and the process was repeated twice by
rediluting the sample to 400 µl with a buffer of 10 mM
Tris-HCl, pH 8.0, containing 1 mM EDTA.
Construction of a Circular DNA Containing a G·T Mispair and a
4-Nucleotide Gap--
To construct substrates containing 4-nucleotide
gaps, two additional phage templates were constructed which were
expanded by 4 bp. The double strand form of clones 7 (5'-template) and 4F (3'-template) were digested independently with EcoRI, and
the resulting overhangs were filled in with the Klenow fragment, as described above. After blunt end ligation and transformation, individual plaques were screened for lack of sensitivity to
EcoRI digestion and sequenced to confirm the addition of 4 nucleotides. These two constructs, called 7E and 4FE, were used
to produce the ssDNA partner in the annealing reaction described above.
The annealing reactions yield a gapped molecule containing the single strand sequence 5'-AATT in place of a nick generated by
EcoRI digestion. All further steps were carried out as per
the nicked circular DNA substrate.
Construction of a Circular DNA Containing a G·T Mispair and a
28-Nucleotide Gap--
To construct the templates containing gaps of
28 nucleotides, the protocol for preparing nicked substrates was
followed with one modification. Before digestion with EcoRI,
100 µg of the appropriate dsDNA was digested with 1 unit/µg
XbaI, and the product was precipitated with ethanol. The
first digestion was monitored for completeness by gel electrophoresis,
and DNA sequencing was used to verify the second digestion. One-fifth
of the EcoRI digestion was passed through each of five High
Pure PCR columns as described by the manufacturer to remove the 28-bp
fragment resulting from the double digest, and the eluates were
combined. All subsequent steps were performed as described for
templates containing a nick.
Construction of a Circular DNA Containing a G·T Mispair and a
202- or 210-Nucleotide Gap--
To construct the 5'-gapped molecule
(202 nucleotides), 150 µg of dsDNA was sequentially digested with 1 unit/µg EcoRI followed by precipitation and digestion with
1 unit/µg BglI, each in recommended buffer for 4 h at
37 °C. For the 3'-gapped molecule (210 nucleotides) the dsDNA was
digested sequentially with DrdI and EcoRI. The
DNA was precipitated with ethanol, and the pellet was resuspended in 10 mM Tris, pH 7.6, 1 mM EDTA, and 300 mM NaCl (TNE). The excised 202- or 210-bp fragment was
removed from the digest by size exclusion chromatography using a
Sephacryl S500 column (Amersham Biosciences) which had also been
equilibrated in TNE. Fractions containing twice digested DNA were
pooled, precipitated by the addition of 2 volumes of absolute ethanol,
and centrifuged for 1 h at room temperature at 68,000 × g in an SW41 rotor. Subsequent steps were carried out as
described for the nicked circular DNA substrate.
Cell Lines and Nuclear Extract Preparation--
Nuclear extracts
were prepared from HeLa and LoVo cells according to Ref. 16. Both HeLa
and LoVo cells were provided by the National Cell Culture Center
(Minneapolis). Protein concentrations were determined with a Bradford
assay (Bio-Rad) using bovine serum albumin as the standard. MutS was
purified from HeLa cells as described (30).
Mismatch Repair Assays--
Repair assays were performed
essentially as described by Holmes et al. (16), using a
15-µl volume containing 20 mM Tris, pH 7.6, 50 µg/ml
bovine serum albumin, 1.5 mM ATP, 1 mM
glutathione, 100 mM KCl, and 0.1 mM dNTPs. 100 ng of DNA (~20 fmol) was used in each assay, and the reaction was
driven by the addition of 45 µg of HeLa nuclear extract protein.
Reaction components were combined on ice, with the HeLa extract added
last, then incubated for 20 min at 37 °C. The assays were terminated
by the addition of 45 µl of a termination solution (25 mM
EDTA, 0.67% SDS, 67 µg/ml proteinase K) and a further 20-min
incubation at 37 °C. The mixture was then extracted twice with 60 µl of phenol with the initial phenol phase being back extracted once
with 40 µl of TE. The aqueous phase that resulted (100 µl) was
diluted with an equal volume of distilled water followed by the
addition of 10 µl of 3 M sodium acetate and 2 volumes of
absolute ethanol. Precipitated DNA was pelleted by a 20-min
centrifugation in a microcentrifuge at 16,000 × g at
room temperature.
In each case where a circular template was used in a MMR assay, all DNA
molecules recovered from the extract were linearized by the addition of
1 unit of AlwNI in buffer 2 (New England Biolabs) supplemented with 20 µg/ml ribonuclease A and 100 µg/ml bovine serum albumin. The extent of mismatch correction was scored
concurrently by including either 0.5 unit of HindIII or 1 unit of XhoI/assay, depending upon the strand assessed (29).
The resulting DNA fragments were separated by electrophoresis through a
15-cm 1% agarose gel at 300 V/h. Gels were stained with ethidium
bromide, and a digital image was captured under UV illumination using a
ChemiImager 4000 CCD cooled camera (Alpha Innotech). The fraction of
DNA in each band was determined by a densitometric analysis using NIH
Image software, version 1.62.
Effect of Salt Concentration and Nuclear Extract Amount on
MMR--
Assays were set up so that the final salt concentration of
components entering the assay was 100 mM with respect to
added cations (primarily potassium). When complete assays were tested, the ionic strength was found to be equivalent to a solution of 125 mM KCl, based on a standard titration curve. In our hands, maximal repair was observed over a range from 125 to 150 mM. Repair assays failed when the salt concentration
exceeded 200 mM KCl. Although the assays reported in this
work were done at the lower end of the range, no improvement in repair
activity for any substrate was observed at higher salt concentrations.
The amount of nuclear extract used in each assay was determined by
titrating the repair efficiency using 100 ng of heteroduplex DNA. The
maximal level of repair was reached at ~20-30 µg of extract, and
50 µg was chosen as a standard assay amount to ensure that extract
components did not limit repair efficiency. Time courses were performed
to verify that no significant amount of mismatch correction occurred
beyond the 20-min assay time course.
The MMR assay employed here (16) has several limitations that must be
recognized. Foremost, it is a nonlinear assay (18) that reports the
maximal amount of repair that may be obtained under a specific set of
reaction conditions. For a given amount of template, no increase in the
extent of MMR can be obtained by increasing the amount of nuclear
extract that provides the requisite activities (16, 18). The substrates
rely on a strand break to direct repair, and sealing the required break
by a DNA ligase can preclude further mismatch processing (18).
Similarly, in the reconstituted repair pathway from E. coli
(24), high concentrations of ligase can suppress mismatch correction.
HeLa extracts prepared as described are relatively unstable at
37 °C,2 which complicates
the critical assessment of parameters beyond the magnitude of mismatch
correction. Under the conditions used here, patterned on earlier
characterizations of this assay method (16, 31), typical values
reported for nick-directed repair of G·T mismatches ranged from 38 to
50% (9.4-12 fmol).
Mismatch Repair Assays with LoVo and Purified
MutSa--
Mismatch repair assays employing LoVo nuclear extracts were
carried out essentially as with HeLa repair assays with the exception that 50 µg of LoVo nuclear extract was used in place of the HeLa extract. In the assays that were supplemented with purified MutS, 300 ng of purified protein was added to each reaction. This represents a saturating amount of MutS; adding more than ~100 ng to the assay
did not enhance repair efficiency.
Analysis of Mismatch Repair on Linear
Templates--
Repair assays using linear substrates were performed
exactly as they were for circular substrates, except that the DNA was digested in a buffer compatible with both enzyme and the assay (New
England Biolabs buffer 2 or 4) and introduced directly into the assay.
For example, 450 ng of substrate was digested for 1 h at 37 °C
with the indicated enzyme in a 25-µl volume as specified by the
supplier. A sample containing 50 ng was removed and evaluated for
completeness of the reaction by agarose gel analysis. Aliquots containing 100 ng (20 fmol) were added directly to each assay and the
concentration of monovalent cations adjusted to 125 mM. Similar assay results were obtained when the substrates were
precipitated from the restriction digest or concentrated by
ultrafiltration and introduced into the assay (not shown). When the
relative efficiency of the MMR pathway on linear and circular templates
was compared (Fig. 4), the substrate containing a 28-nucleotide gap
entering each assay was treated under restriction digest conditions.
AlwNI restriction enzyme (1 unit/assay) was added where
indicated to generate the linear substrate. When nicked templates were
mock digested the repair efficiency dropped modestly, which is
reflected in Figs. 5 and 7.
In experiments in which the substrates were linearized near the
mismatch, a modification of the scheme for scoring repair was required.
The 5'-nicked template was digested with BspHI in addition
to HindIII, generating fragments of 2,651 (repaired) and
2,816 bp (unrepaired) on template linearized prior to the assay with
DrdI. The 3'-nicked substrate was digested with
PacI in addition to XhoI, generating repaired and
unrepaired fragments of 2,165 and 2,299 bp, respectively, on template
linearized prior to the assay with PvuI. The repair value in
each assay was calculated by comparing band intensities of the unique
repaired and unrepaired fragments, with consideration given to the
small repair-specific fragments not resolved in the assay.
Analysis of Denatured DNA by Alkaline Gel
Electrophoresis--
Mismatch-containing templates were assayed for
mismatch correction as described above and digested with
AlwNI. Precipitated DNA was resuspended in 15 µl of
aqueous solution containing 50 mM NaOH and 1 mM
EDTA, and 15-cm alkaline 0.9% agarose gels were poured and run at
4 °C (60 V for 12.5 h) as described (28). After gel
neutralization, the DNA was transferred onto a nylon membrane
(Osmonics, Inc.) and cross-linked using a 45-s exposure to short wave
UV illumination using an Alpha Innotech transilluminator. The membranes
were pretreated at 65 °C for 1.5 h in 6× SSC buffer (0.15 M NaCl and 0.15 M sodium citrate at pH 7.0 is
1× SSC) containing 0.25% nonfat milk powder, whereupon the
radiolabeled probe (32P-GGTTCTGGTGGCGGCTCTGA-3') was
introduced and the hybridization reaction allowed to proceed at
52 °C (28). The membranes were washed with three changes of 2× SSC
buffer containing 0.5% SDS at room temperature and exposed to Kodak
X-Omat AR film. The probe sequence was complementary to the
discontinuous DNA strand near the AlwNI site (see Fig.
6).
| |
RESULTS |
A set of eight dsDNA molecules that all contain a single G·T
mismatch was constructed to address three mechanistic questions about
the human MMR pathway. First, because DNA nicks represent the only
characterized strand discrimination signal available for targeting MMR,
we wanted to determine whether other types of DNA discontinuities or
ends were also competent. This allows us to investigate whether
specific DNA templates that might arise during DNA replication or
recombination are capable of directing mismatch excision and
resynthesis. Second, we wanted to ask whether the MMR pathway could
compete effectively with different classes of gap-filling replication
events. In models that seek to explain strand discrimination based on
the DNA terminus where elongation occurs (see Refs. 27 and 32),
mismatch correction must be able to compete effectively with DNA
synthesis using polymerases that function during either leading or
lagging strand biosynthesis. Third, we sought to characterize a minimal
DNA template for the human MMR pathway, using DNA constructs that
systematically limited the DNA template close to either the mismatch or
the strand discrimination signal. This deletion-based mapping strategy
allows an approximation of the physical template size required for
specific stages in mismatch correction, such as the initiation of an
excision tract.
Comparison of 3'- and 5'-Substrates--
The eight G·T templates
can be broken down into two groups based on strand polarity. Four
templates represent nicks or gaps where the DNA sequence reading from
the nick or gap to the mismatch reads 5' 3' (5'-substrates; Fig.
1, top), whereas four
represent the opposite polarity (3'-substrates; Fig. 1,
bottom). Inversion of the 394-bp region within the
replicative phage form of M13 does confer substantial differences
between the 3'- and 5'-substrates which must be considered. The
heteroduplexes are constructed from a single strand and double strand
phage partner, and the discontinuous strand is always derived from the
duplex. This means that one substrate is technically a G·T and the
other a T·G mismatch (see Fig. 1) and that the distance separating
the nick and mismatch is offset by 4 bp when comparing the two groups.
Because MSH2/MSH6 can be inferred to bind the mismatch asymmetrically
(33), they do logically represent two different structures for repair.
However, both substrates present an identical sequence for mismatch
recognition, centered within 10 flanking bp of palindromic DNA. They
are presumably bound by MSH2/MSH6 in precisely the same way. An
identical duplex sequence separates the two key sites (mismatch and
strand signal). They become distinct when one considers the
relationship between the mismatch and the strand signal. Most
significantly, during the repair excision step, the G-containing strand
would be removed in one case and the T-containing strand in the
other.
View larger version (27K):
[in this window]
[in a new window]
|
Fig. 1.
Relationship between DNA mismatches and
strand discontinuities. In all cases, a G·T mismatch was
constructed within overlapping HindIII/XhoI
restriction sites (29) to facilitate assessment of strand-specific
repair. The upper portion of the figure shows templates with
5'-discontinuities, where the strand polarity runs 5' 3' from
strand break to mismatch. Gaps of 202, 28, or 4 nucleotides were
constructed in addition to a nicked substrate. Each gap terminates at
the site generated by digestion with EcoRI. The lower
portion of the figure shows the corresponding 3'-substrates, where
all gaps initiate with the EcoRI terminus and extend 4, 28, or 210 nucleotides further away from the mismatch. Competition between
excision to remove the mismatch and the later stages of gap-filling
replication on the 5'-substrates shown in the upper portion,
and with the early stages of gap-filling synthesis for the
3'-substrates in the lower portion.
|
|
Gapped Regions of DNA Are Effective Strand Discrimination Signals
for Mismatch Correction--
The group of four substrates containing a
nick or gap where the strand polarity runs 5' 3' from the
discontinuity toward the mismatch was assayed first (Fig.
2, top panel). In the case of
the gapped molecules, the substrates allow for a gap-filling replication event that proceeds toward the mismatch (refer to Fig. 1).
Mismatch correction events must therefore compete with the late stages
of replication and ligation for the available DNA end. Each substrate
was assayed for MMR using nuclear extracts prepared from HeLa cells as
the source of the repair apparatus, and the recovered molecules were
linearized with AlwNI and scored for mismatch correction by
restriction analysis (see Ref. 29 and "Materials and Methods").
Unrepaired heteroduplexes were insensitive to digestion with
HindIII or XhoI and migrated as linear DNA
species, whereas sensitivity to these enzymes reflects site-specific
mismatch correction and yields two diagnostic bands of 3,537 and 3,739 bp.
View larger version (48K):
[in this window]
[in a new window]
|
Fig. 2.
Strand-specific repair of the 5'- or
3'-nicked or gapped substrates. The top panel shows the
repair efficiency of the 5'-nicked or gapped molecules. DNA fragments
resulting from digestion by HindIII at the corrected
mismatch site are visible as a doublet below (see "Materials and
Methods"). No mismatch correction targeted to the continuous (closed)
strands was detectable with XhoI (odd numbered
lanes), whereas mismatch removal and resynthesis in the open
strands varied from 49 to 80% (even numbered lanes).
nd indicates not detectable; na means not
attempted. In the bottom panel, the repair efficiency of the
analogous 3'-substrates is compared after passage through HeLa extract.
Strand-specific repair directed to the discontinuous strand resulted in
reconstitution of the XhoI restriction site, and no repair
was detectable in the continuous strands. In all cases, the repair
values are given ± 1 S.D. from the mean value determined, based
on 3-11 independent assays.
|
|
The effectiveness of gaps that ranged from 4 to 202 nucleotides as a
strand discrimination signal was compared with that of a nick, where
each signal terminates 329 bp from the mismatch. For each size gap, the
magnitude of repair was greater than that found for nick-directed
repair (Fig. 2). A gap of either 4 or 28 nucleotides represents the
most effective signal (~80% of the molecules were repaired),
although the largest gaps of 202 bp were still more effective than a
nick (67% compared with 49% of the molecules were repaired,
respectively). In each instance, repair was directed to the nicked or
gapped strand and reconstituted the HindIII restriction
enzyme sequence (Fig. 2, upper panel, even numbered
lanes). No digestion of these substrates with XhoI restriction enzyme was detected (Fig. 2, upper panel,
odd numbered lanes). Under the assay conditions used, the
mean assay value for each gap size represents a significant difference
in MMR efficiency compared with nick-directed repair using a Student's
independent samples t test (p 
0.0005).
Parallel assays were performed for the substrates containing a 3'-nick
or gapped region, and the results are shown in the bottom
panel of Fig. 2. In these substrates, the overlapping restriction sites used to score repair (HindIII and XhoI)
were maintained, but the strand signals now reside in the T-containing
strand (refer to Fig. 1). Note that a replication event that fills the
gap must initiate at the DNA end that serves as a strand discrimination signal. Mismatch correction events therefore compete with replication initiation and elongation for the available 3'-terminus. As in the
5'-substrates above, repair of mismatches present in molecules with
gaps of either 4 or 28 nucleotides (~80%) was significantly more
efficiently than nick-directed repair. The template containing a gap of
210 nucleotides was repaired with an efficiency similar to the
nick-containing molecules (44% compared with 41% of the molecules,
respectively). In each instance, only repair directed to the nicked or
gapped strand that reconstituted the XhoI restriction enzyme
was detected.
Gap-directed Repair Events Depend upon the MSH2/MSH6 Heterodimer
(MutS)--
Previous work has established the fact that the long
patch and nick-directed correction of base-base mismatches depends upon the presence of the MutS heterodimer (30, 34). We wanted to
determine whether the analogous gap-directed mismatch correction described above was also dependent upon MutS. A simple explanation for the increased efficiency of gap-directed repair might be that gaps
lead to mismatch-independent excision of the mismatched region. We
therefore carried out repair assays using nuclear extracts prepared
from LoVo cells, which are defective in MMR because of a partial
deletion of the MSH2 gene (35). In these extracts, neither
MSH2 nor MSH6 protein is detectable by Western analysis, apparently
because the lack of MSH2 leaves the partner MSH6 susceptible to
proteolysis (30, 34). The absence of this protein complex renders LoVo
defective in the ability to correct mismatches unless exogenous MutS
is supplied.
Shown in Fig. 3 are assays using LoVo
extract in the presence and absence of added MutS. The upper
panel shows substrates with 5'-strand discontinuities, whereas the
lower panel shows the molecules with 3'-discontinuities. In
each case, no repair activity was detected in assays using LoVo extract
alone (lanes 1, 3, 5, and
7). Treatment with LoVo did result in gap filling and
ligation to yield covalently closed molecules (data not shown; for
methodology, see to Fig. 6). This result shows that gaps are subject to
repair-independent synthesis and ligation events over the time course
of the MMR assay. When MutS was supplied to the assay, mismatch
excision and resynthesis were clearly present. The magnitude of the
repair events was uniformly lower compared with the MMR-competent HeLa
extract, as found previously for nick-directed repair events (30), but
the relative efficiency of repair among each was maintained. Again,
gaps of 4 or 28 nucleotides were repaired the most efficiently,
independent of strand polarity.
View larger version (75K):
[in this window]
[in a new window]
|
Fig. 3.
MMR assays using LoVo extracts in the absence
and presence of added MutS. Mismatch correction was not
detectable (nd) in LoVo extracts alone (odd numbered
lanes), which lack the MSH2 protein (35) and functional MutS
heterodimers (30, 34). The addition of 300 ng of purified MutS
(even numbered lanes) restored mismatch correction to a
detectable level that mirrored the pattern found using HeLa
extracts.
|
|
Mismatch Correction Is Efficient on Linearized Substrate
Molecules--
Our next objective was to truncate the template
systematically and delineate the minimal spatial requirements around
the required nick and mismatch sites. This was achieved by linearizing
the template using the available unique restriction sites around the target site. To test the validity of this approach, we first needed to
determine whether linear substrate molecules were competent to assess
MMR using human extracts and whether exonucleolytic degradation
substantially affected the free DNA ends. Normally, nicked circular
molecules have been used in such assays (16, 17), although linear
molecules with exonuclease-resistant ends have been used as a template
for assessing repair using E. coli extracts (26).
The substrate chosen for the initial experiments was the G·T template
containing a 28-nucleotide gap (5' 3' from gap to mismatch). Repair
efficiencies were compared using circular and linear M13 templates that
were predigested with AlwNI, which cuts at least 3,000 bp
away from either the gap or mismatch site. This digestion is normally
used to linearize the molecules after extract treatment at a site
remote from the mismatch, so that restriction enzyme sensitivity at the
mismatch site yields two diagnostic DNA fragments. In these
experiments, both the control circular and digested molecules were
incubated in restriction enzyme buffer at 37 °C and then added
directly to a MMR assay adjusted so that the overall concentration of
monovalent cations was 125 mM.
When a circular template was recovered from HeLa nuclear extract and
not digested with restriction enzymes (Fig.
4, lane 1), the DNA was
converted primarily to a high molecular weight species that includes
catenated molecules (36, 37). In comparison, when the linearized
template was treated under identical conditions, it was recovered from
the assay without forming either catenates or multimers by ligation
(lane 2). When these linear molecules were treated with
HindIII to assess mismatch correction (lane 3),
more than 80% of the molecules were sensitive, indicating a magnitude
of repair comparable with that found for the circular derivative. A
similar experiment was repeated for all eight substrate molecules, and
in each case the circular and linearized templates were repaired with
similar efficiency. Finally, sequencing to the termini of the linear
substrates after passage through HeLa revealed no appreciable
nucleolytic degradation of the ends of the molecule (data not
shown).
View larger version (37K):
[in this window]
[in a new window]
|
Fig. 4.
Linear DNA molecules are repaired efficiently
by HeLa nuclear extract. Linear or circular substrates with a
5'-gap of 28 nucleotides were assayed under standard conditions.
Lane 1 shows that DNA treated under mock restriction digest
conditions and assayed was recovered as a high molecular weight
catenated species (36, 37). Lanes 2 and 3 show an
identical assay, except that the DNA was digested with the single
cutter AlwNI before entering the assay. The DNA is recovered
as a linear monomer in the absence of further treatment (lane
2), and digestion with HindIII to identify molecules
repaired at the mismatch site is shown in lane 3.
|
|
dsDNA Breaks Do Not Support Mismatch Correction--
Given that
linear molecules with internal strand discontinuities support MMR, we
wanted to establish whether a dsDNA break, instead of a single strand
discontinuity, could serve to direct repair when positioned close to
the mismatch. In this case, mismatch correction might be directed
toward either strand, eliminating strand bias. When substrates already
nicked at the EcoRI site were fully digested to generate a
DNA end presenting a 5'-overhang of 4 nucleotides in place of the nick,
no mismatch correction was observed in either strand (data not shown).
In experiments described subsequently, the nicked templates were
linearized with restriction enzymes that produced blunt ends,
3'-overhanging ends and 5'-overhanging sequences at sites other than
the EcoRI site. In none of these experiments did
linearization detectably affect which strand was repaired or the amount
of repair that occurred; all of the observed MMR events could be
accounted for as dependent upon the presence of a nick or gap. We also
tested these linear molecules for strand excision events using
denaturing alkaline gels (see "Materials and Methods"), assuming
that the early stages of mismatch correction might be successful but
that strand resynthesis might fail. No evidence for any
mismatch-dependent processing events was found (data not shown).
Dependence of Mismatch Correction Activity on Template Length
Adjacent to the Nick--
The 5'- and 3'-nicked substrate molecules
were singly truncated with a series of restriction enzymes that create
a double strand break close to the nick. These experiments were
directed toward establishing the minimal spatial requirements of the
MMR apparatus for initiating an excision event from the nick site and
for the polymerization event that restores the excision tract. Samples
of the 5'-nicked substrate were digested independently with
SmaI, AccI, and NheI (Fig.
5), whereas samples of the 3'-nicked substrate were digested with only the former two enzymes. The NheI site in this template was disrupted during construction
of the 3'-substrates and was therefore unavailable. Digestion produces a linear molecule with ends located 14 (blunt), 33 (5'-overhang), and
53 bp (5'-overhang) away from the 5'-nick, respectively. The linear
templates were assayed directly for effectiveness as a repair substrate
under standard assay conditions (see "Materials and Methods"),
using the same template digested at a remote site (AlwNI) as
a control. Note that in this and subsequent experiments, where any of
the DNA templates was digested or treated under mock digest conditions
prior to assay, the overall magnitude of nick-directed repair
efficiency was reduced from 50% to ~40%.
View larger version (63K):
[in this window]
[in a new window]
|
Fig. 5.
MMR assays conducted with templates
linearized close to the nick site. In the box below the
figure is a diagram approximately to scale representing the spatial
relationship among the mismatch, nick, and the panel of digestion
sites. The numbers in parentheses adjacent to the
enzyme names indicate the distance from the linear end to the nick.
Lanes 1-4 show the results of assays using the 5'-nicked
substrate; lanes 5-7 represents the 3'-nicked substrate.
For both 5'- and 3'-nicked substrates, a control assay was performed in
which the substrate was linearized with AlwNI at a remote
site and assayed. The diagnostic fragments are labeled R
(repaired) or U (unrepaired).
|
|
Shown in Fig. 5 (left panel) are assays using 5'-nicked
substrate linearized close to the nick site. A diagram drawn to scale showing the spatial relationship between these sites is shown below it.
No mismatch correction was detectable on templates linearized 14 or 33 bp away from the nick (SmaI or AccI), whereas
38% repair was measurable with the NheI linearized
substrate. For comparison, similar repair efficiencies were obtained in
control assays using 5'-nicked substrate molecules digested with
AlwNI so that the closest linear end was located 3,540 bp
from the nick. This result indicates that, in substrates containing a
5'-nick, restoration of the correct base pair at the mismatch site is
inhibited by a double strand break placed within 33 bp of the nick.
With the slightly greater flanking site size of 53 bp, the mismatch
site was repaired at the control level.
An analogous experiment was performed to assess the viability of
templates truncated near the 3'-nick for competence in mismatch correction. Digestion with SmaI or AccI was used
to produce linear ends located 18 or 33 bases from the nick (refer to
the diagram in Fig. 5, right panel). As a benchmark, 35% of
the 3'-nicked substrate linearized at a remote site (AlwNI)
was repaired in this assay. Equivalent repair efficiency was found when
only 33 bp extended beyond the 3'-nick (39%), and nearly half-of the
control level of repair was detectable when only 18 bases extend beyond the nick (16%). This suggests that the MMR machinery can access a nick
placed within 18 bp of a blunt DNA end, and sufficient template is
present to allow loading and function of the machinery required for
mismatch-dependent, nick-directed strand excision.
The failure to observe mismatch correction using a restriction-based
assay, as in the case of the molecules linearized proximal to the
5'-nick site described above, does not imply that either mismatch
recognition or subsequent processing steps failed. This assay simply
quantifies the extent to which replication has restored the sequence
and returned enzymatic sensitivity to the scored site. When truncation
is made close to a critical site, it is plausible that strand excision
associated with mismatch removal could also eliminate the DNA primer
required for resynthesis or that the short primer remaining was not
sufficient to initiate resynthesis. We therefore wanted to establish
whether substrate truncation prevented excision of the tract that
removes the mismatch, or if mismatch removal was effected but repair
synthesis was precluded. This was achieved by analyzing the products of
MMR assays on truncated substrates using denaturing alkaline agarose
gels (Fig. 6). After transfer of the
denatured DNA fragments to a nitrocellulose membrane, the DNA fragment
appropriate for analysis of the excision tract was probed using a
radiolabeled oligonucleotide (see Fig. 6 and "Materials and
Methods").
View larger version (27K):
[in this window]
[in a new window]
|
Fig. 6.
Excision tract formation in G·T substrates
linearized near the nick or gap site. ssDNA molecules were
resolved using alkaline gel electrophoresis (see "Materials and
Methods") and probed with a 32P-labeled oligonucleotide
complementary to the G-containing strand (refer to the
diagram below the figure). Probing the denatured substrate
in the absence of HeLa treatment reveals a fragment that runs from the
AlwNI-generated DNA end to the gap or nick region and
includes the mismatch site (4,069 nucleotides; see lane 1).
The mismatch site lies 330 nucleotides from the discontinuity in these
5'-heteroduplexes. Upon HeLa treatment, ligation of the nick
(lane 2) or a gap-filling replication followed by ligation
(lane 3) results in a full-length M13 template of 7276 nucleotides which represents both repaired and unrepaired molecules.
Substrates linearized 14 bp beyond the discontinuity with
SmaI before HeLa treatment are unable to support resynthesis
and yield a family of excision products (the smeared band in
lanes 4 and 5) which extends beyond the mismatch
site (indicated by a dashed line). The broad excision tract
is centered ~300 nucleotides past the mismatch site, as depicted in
the diagram on the right.
|
|
As shown in Fig. 6, full-length G·T substrates containing either a
5'-nick or a 5'-gap of 4 nucleotides were assayed and probed after
resolution of MMR intermediates and products on a denaturing agarose
gel. Prior to HeLa treatment, the probe identified a ssDNA fragment of
4,069 nucleotides (lane 1); this represents the sequence from the AlwNI site to the discontinuity, including the
mismatch site. After HeLa treatment, a full-length product appeared
where the nick or gap is sealed (lanes 2 and 3,
respectively). Based on restriction analysis performed in parallel,
this larger band comprises both repaired and unrepaired molecules where
the strand break has been sealed (18). This experimental approach was
used to verify that the substrate nicks or gaps were sealed after
treatment with either HeLa or LoVo nuclear extracts.
In contrast, when the same substrates were linearized 14 bp away from
the discontinuity with SmaI prior to the assay, evidence for
an excision tract large enough to remove the mismatch is apparent for
both the nicked and gapped substrates (Fig. 6, lanes 4 and 5). Such intermediates have been documented previously when
repair synthesis has been inhibited with aphidicolin (18) and can
extend several hundred nucleotides beyond the mismatch. These products appear as a smeared family of fragments degraded from the nick site
(lane 4) or gap site (lane 5) to yield smaller
products. When the substrate contains a nick, a substantial fraction of the molecules is apparently unprocessed, consistent with the proportion of molecules that is not repaired when the molecules are assayed by
restriction analysis (refer to Fig. 2). When the 4-nucleotide gap was
positioned close to the SmaI-generated end (lane
7), all of the molecules appear within a smeared band centered at
a site consistent with removal of roughly 500 nucleotides. No evidence for unprocessed molecules was found in this assay. These results show
that the early steps in the MMR pathway, through strand excision, occur
on templates where only 14 bp extends beyond the nick or gap site but
that resynthesis to correct the mismatch fails. The explanation for
this failure may be as simple as the exonucleolytic loss of the DNA
primer required for strand resynthesis.
Dependence of Mismatch Correction Activity on Template Length
Adjacent to the Mismatch--
The 5'- and 3'-nicked substrates were
also singly truncated relatively close to the G·T mismatch to address
the template requirements close to this site. As in the previous set of
experiments, the sequence intervening between the nick and mismatch
site was preserved between the 5'- and 3'-substrates. This strategy,
however, was limited by the need to place the linear end sufficiently
far from the mismatch to allow for detection of the small DNA fragment resulting from digestion at the reconstituted HindIII or
XhoI restriction sites in the gel-based assay. Because no
additional unique restriction sites were incorporated into the template
molecules, the sites present in the modified M13 template were used for linearization.
Samples of the 5'-nicked substrate were digested with DrdI
(3'-overhang) and NgoMIV (5'-overhang) that linearized the
substrate 165 and 311 bases from the mismatch, respectively (Fig.
7, left panel). Identification
of repaired and unrepaired DNA products after HeLa treatment was
accomplished by digestion with enzymes BspHI and
HindIII (see "Materials and Methods"). In the case of the NgoMIV-linearized substrate, note that the unrepaired
molecules are 150 bp larger. We found that linearization close to the
mismatch site had little effect on the rate of repair. Substrate
digested at the remote, control site (AlwNI) or the
NgoMIV site was repaired with equal efficiency. Substrate
digested at the closest site (DrdI) was repaired with an
efficiency similar to the undigested (circular) substrate. These
results indicate that for the 5'-substrates, a linear end positioned as
close as 165 bases away from the G·T mismatch is sufficient to allow
the MMR pathway to act upon these substrates, including resynthesis
through the mismatch site.
View larger version (67K):
[in this window]
[in a new window]
|
Fig. 7.
Mismatch repair competence of substrates
linearized close to the mismatch. Lanes 1-3 show the
results of assays using linearized 5'-nicked substrate; assays of
3'-nicked substrates are shown in lanes 4-6. Below each
panel is a diagram approximately to scale
representing the spatial relationship among the mismatch, nick, and the
panel of digestion sites. The numbers in
parentheses adjacent to the enzyme names indicate the
distance from the linear end to the mismatch. The unrepaired band
(U) specific to the NgoMIV-digested substrate is
150 bp larger than its DrdI counterpart. Lanes 1 and 2, respectively, show the results of the MMR assays
carried out using DrdI- or NgoMIV-linearized
5'-nicked substrates. Substrate linearized at a remote site
(AlwNI) is presented for comparison (lane 3).
Lanes 4-6 show a parallel experiment performed on the
3'-nicked substrate linearized with either PvuI or
BglI prior to the MMR assay (see illustration
below). Successful mismatch correction results in a diagnostic band of
1,217 bp (R), either 136 or 165 bp smaller than the
unrepaired counterpart resulting from PvuI or
BglI digestion, respectively. MMR was not scored reliably in
lane 4, so no quantitation is reported
(na).
|
|
We next determined the efficiency with which the 3'-nicked substrate
was repaired when a linear end was placed either 136 bp
(PvuI, 3'-overhang) or 165 bp (BglI, 3'-overhang)
away from the mismatch (see Fig. 7, right panel). Shown in
Fig. 7 are typical results of assays using PvuI- (lane
4) and BglI (lane 5)-linearized substrates
possessing a 3'-nick after scoring digests with PacI and
XhoI (see "Materials and Methods"). Control experiments
in which the substrate was digested with AlwNI more than
3,000 bp away were repaired efficiently (37%). Substrate digested 165 bp away from the mismatch yielded detectable repair (25%), whereas the
closest truncation did not yield a repair band that could be reliably
scored. These data suggest that mismatch correction can be supported on
a 3'-nicked template truncated as close as 165 bp away from the
mismatch. Based on the results described in Fig. 6, we suggest that the
early stages of mismatch correction, up to strand resynthesis, are
likely to be accommodated on this substrate but that resynthesis is
limited because of the excision of a required priming template.
DNA Ends Close to a Mismatch Do Not Deter Repair Directed from a
Remote Nick--
We designed an experiment to determine whether the
MMR pathway was competent to access a remote nick as a strand
discrimination signal in the presence of a proximal DNA end. Fig.
8 shows the efficiency of mismatch
correction when a 5'-nick was positioned 884 bp away from a G·T
mismatch and linearized 165, 311, or 3,537 bp beyond the mismatch.
Because a relatively small region of DNA separates the mismatch and one
end in two of the assays, all substrates were linearized with
BspHI after HeLa treatment to facilitate discrimination of
repaired and unrepaired molecules by HindIII sensitivity
(see the diagram in Fig. 8). Fragments that correspond to
repaired and unrepaired substrate are designated R and
U, respectively, in Fig. 8. Fragments not designated
R or U were common to digests of both repaired
and unrepaired molecules. When assayed, no substantial difference in
competence among these three templates was observed. These data
indicate that a linear end did not deter communication between the
mismatch and nick site, even when the mismatch was 5-fold closer to the
nearest free end than it was to the nick. Quantitatively similar
results were obtained in assays conducted with a 5'-nicked substrate
where the nick and mismatch were separated by 1,037 bp, although the
overall repair efficiency was lower (data not shown).
View larger version (37K):
[in this window]
[in a new window]
|
Fig. 8.
MMR assays on asymmetric DNA templates.
A G·T heteroduplex containing a 5'-nick placed 884 bp away was
linearized with DrdI, NgoMIV, or AlwNI
(refer to the diagram below the figure). In the case of
DrdI digestion, this creates a DNA end 165 bp beyond the
mismatch (with respect to the nick). After treatment with HeLa nuclear
extract, the DNA molecules were digested with BspHI to
linearize them and HindIII to reveal DNA fragments
diagnostic for mismatch correction. Recovered molecules are marked as
repaired (R) or unrepaired (U). All three
substrates yield a characteristic fragment of 2,649 bp which spans the
BspHI site and the mismatch site. Compensation for the small
repair-specific fragments that could not be resolved by the gel system
was made for the DrdI- and NgoMIV-linearized
substrates. The three substrates were repaired independently of the
proximity of the G·T mismatch to the DNA end.
|
|
| |
DISCUSSION |
At the outset of this work, a DNA nick was the only structure
known to be competent to allow strand discrimination for MMR assays
using human nuclear extracts in vitro. We wanted to
determine whether other types of DNA discontinuities or ends were also
competent. We therefore compared the efficiency with which a nick can
direct the repair of a G·T mismatch placed ~300 bp away with the
efficiency of single strand gaps of 4, 28, and 202 (or 210) bp. In each
case, the efficiency of mismatch correction was higher for the gapped species than for the nick. In contrast, neither blunt ends nor single
strand overhangs of 4 nucleotides were competent to direct MMR to
either strand.
Although we did not discriminate among models that might explain these
repair differences, we offer two simple working models to account for
them. One possibility is that nicks can be sealed by the action of a
single enzymatic activity, a DNA ligase, whereas the gaps all require
polymerase activity prior to strand sealing by ligase. Alternatively,
the DNA ends available within the gaps represent sites where activities
involved in DNA repair or replication may be loaded, which might
involve prior loading of proliferating cell nuclear antigen at the
3'-end (38). In this model, gaps might facilitate recruiting repair
machinery to a site near the mismatch. Such a model offers the testable
prediction that the rate at which the strand signal, either nick or
gap, is accessed might be fundamentally different between the nicked
and gapped molecules. A critical kinetic analysis of key steps in the
MMR pathway is essential to make such a distinction.
Our second objective was to determine whether the MMR pathway competes
effectively with different classes of gap-filling replication events.
In models that seek to explain strand discrimination based on the DNA
terminus where elongation occurs (27, 32), mismatch correction must be
able to compete with DNA synthesis performed by polymerases that
function during either leading or lagging strand biosynthesis. Although
we are unable to generate DNA mismatches by replication and ask whether
the polymerase active site may be used as a strand discrimination
signal, our assays test this hypothesis indirectly. By providing DNA
gaps of different sizes, which presumably serve as templates for DNA
polymerases with different tract length specificities, the MMR pathway
was allowed to compete with replication for an available DNA end. In
the case of 5'-discontinuities, the polymerase must copy the template
toward the mismatch, and mismatch processing presumably competes with
the late stages of replication and ligation for the DNA terminus. In
the case of 3'-discontinuities, polymerase loading and synthesis must
compete with the MMR pathway for access to the site of elongation.
The choice of the gap sizes was intended as a device to recruit
different eukaryotic polymerases to compete with mismatch correction.
For example, we expected that human polymerase would be most likely
to fill in the smallest gap of 4 nucleotides, whereas replicative
polymerases such as or might fill in the largest gaps (42). The
latter two polymerases have been associated with specific gap-filling
repair events in nucleotide excision repair (39), so we also tested an
intermediate gap size of 28 nucleotides. Polymerase has not been
associated with any specific repair events (39), but it also represents
a polymerase capable of synthesis within the intermediate size gaps.
Although it is beyond the scope of this work to characterize the extent
to which each DNA polymerase contributed to the gap-filling events
tested, this represents the rationale for the experimental design.
It is clear that gap sizes of 4 or 28 nucleotides, presented in either
strand polarity, were the most effective in strand discrimination using
the HeLa nuclear extract assay system. The largest gaps (
202
nucleotides) were the least efficient strand discrimination signals.
These data suggest that not all DNA ends are equally effective signals
or that the largest gaps are accessed more efficiently by a replication
apparatus than they are by MMR. It should be noted, however, that the
substrates all presented an arbitrary distance of ~300 bp separating
discontinuity and mismatch. This situation may not reflect the optimal
physical layout for communication between the sites of mismatch
recognition and ongoing replication.
One hypothesis for the mechanism by which strands may be discriminated
posits that MMR is directly coupled with genomic replication (3, 40).
The intimate involvement of proliferating cell nuclear antigen in both
early and late stages of the MMR process supports this notion (13, 14).
It deserves mention that if the polymerase active site represents the
primary strand discrimination information for mismatch correction,
tract excision to remove the mismatch is expected to proceed primarily
from 3' to 5'. This raises the question of why the human pathway, as
judged using assays from nuclear extracts in vitro, is
competent to access strand discontinuities efficiently when the
excision proceeds from 5' to 3'. This stands in contrast to the better
characterized E. coli model system, in which strand
discrimination may be separated from replication, accessing transiently
hemimethylated sites in either orientation to define the daughter
strand. Both pathways are nonetheless bidirectional (18, 26), able to
operate with similar efficiencies given a strand discontinuity in
either orientation.
One often cited mechanistic possibility that preserves the requirement
for a bidirectional competence of the human pathway is based on the
fact that lagging strand biosynthesis is discontinuous and that nicks
might be used as strand markers in this strand (4, 40). This
possibility suggests that mismatch correction is targeted in a
fundamentally different manner between the leading and lagging strands.
The bidirectional capability of the human MMR pathway as well as the
efficiency of gaps presented in either strand polarity as strand
signals might be unified based on the observation that, in the lagging
strand, the succeeding Okazaki fragment is synthesized toward the
mismatch-containing fragment. Both a DNA gap and an approaching
replication complex are available close to the mismatch, with a 5' 3' polarity reading toward the mismatch. Such a model might require
that the 5' 3' exonuclease recruited to remove the mismatch,
e.g. human exonuclease I (41), be able to excise both an RNA
primer and the intervening DNA. At a minimum, it represents a situation
in which a DNA gap could be accessed by a MMR pathway with a
bidirectional capability.
One striking finding resulting from the deletion-based mapping strategy
was that heteroduplex molecules containing a nick or 4-nucleotide gap,
digested to leave a blunt end only 14 bp beyond the break
(5'-orientation), were processed efficiently. Communication between the
DNA mismatch site and the discontinuity must result in the loading of
activities with a mandate to excise one DNA strand in the direction of
the mismatch. Our data suggest that little more than one helical turn
of DNA beyond the strand break is sufficient to direct mismatch
excision in a strand-specific manner. This requirement held true for
either strand polarity with respect to the mismatch. However,
resynthesis to restore the mismatch site in these assays depended upon
the presence of a DNA primer to initiate resynthesis, which was
presumably removed in the case of the 5'-nicked species.
Finally, we infer from our data that ssDNA ends within a gap are at
least as effective as a DNA nick in strand discrimination. By contrast,
linear DNA ends with short overhangs, such as those generated by
EcoRI digestion, do not contain enough information to allow
targeting of repair to either strand. These observations might be
important in contexts such as recombination, where the MMR pathway is
known to participate (4, 5). Based on our results, a DNA end generated
by strand invasion into a homologous duplex is predicted to be an
effective means of targeting mismatch correction to the invading
strand. This assumes that the mismatch and DNA end are close enough for
effective communication and that mismatch excision leaves a priming
sequence for replication. The ability of small gaps to direct MMR might
also suggest that mismatch removal can be coupled to other DNA repair
events, such as base excision repair or nucleotide excision repair
processes that remove from 1 up to ~35 nucleotides to form a
transient single strand gap (42). It is beyond the scope of this work
to consider all of the possible implications, but the fact that MMR can
be activated by small gaps may also be important within the context of
lesion-induced apoptosis mediated by the MMR pathway (43). Such a
scenario might arise when two lesions that can be removed by multiple
repair pathways occur close to each other because processing of the
first lesion may generate a strand signal appropriate to trigger a
MMR-dependent processing event of the second lesion.
TOP
ABSTRACT
INTRODUCTION
To whom correspondence should be addressed: Dept. of Biology,
Indiana University, 1001 E. Third St., Bloomington, IN 47405. Tel.: 812-856-4184; Fax: 812-855-6705; E-mail:
jdrummon@bio.indiana.edu.
