ABSTRACT
By first separating mutant from nonmutant DNA sequences on the basis of their
melting temperatures and then increasing the number of copies by high- fidelity DNA amplification, we have developed a method that allows
observation of point mutations in biological samples at fractions at or above
10
-6
. Using this method, we have observed the hotspot point mutations that lie in 100 base pairs of the mitochondrial genome in samples of cultured
cells and human tissues. To date, 19 mutants have been isolated, their
fractions ranging from 4*10-4
down to the limit of detection. We performed specific tests to determine if the
observed signals were artefacts arising from contamination, polymerase errors
during PCR or DNA adducts created during the procedure. We also tested the
possibilities that DNA replication mismatch intermediates, or endogenous DNA adducts that were originally present in the cells, were included
with true mutants in our separation steps and converted to mutants during PCR.
We show that while most of the mutants behave as double-stranded point mutants in the cells, some appear to arise at least in part
from mismatch intermediates or cellular DNA adducts. This technology is
therefore sufficient for the observation of the spectrum of point mutations in
human mitochondrial DNA and is a tool for discovering the primary causes of
these mutations.
From Mendel to the present era, the study of mutagenesis required recognition of
a phenotypic change in cells or organisms. Recently, approaches to detecting
rare (<10
-4
) mutations in DNA based on genotype rather than phenotype have been developed.
For example, allele specific PCR (
1
,
2
), ligase chain reaction (
3
-
6
) or high efficiency restriction digestion (
7
-
10
) permit detection of single base pair changes in known sequences with scanning ranges from 1 to 6
base pairs (bp).
In order to extend the scanning range of genotypic approaches, we previously coupled high-fidelity PCR with denaturing-gradient gel electrophoresis, DGGE, which was developed by Fischer and
Lerman (
11
) building on the studies of cooperative equilibria in macromolecules pioneered
in statistical mechanics (
12
-
14
). This allowed us to see point mutations in DNA isomelting domains of ~100 bp in cultured human cells (
15
,
16
).
In that work we were obliged to enrich for mutants in the
hprt
gene by phenotypic selection (6-thioguanine resistance). This selection increased the mutant fraction from
10
-7
to >= 10
-3
, a detectable level for our high-fidelity PCR/DGGE procedure. Since phenotypic selection based on selective
growth of mutant cells cannot be applied to human tissues, we continued to
develop technology which separated mutant sequences from the vast excess of
normal sequences by physical or chemical differences.
Recently, we have taken advantage of capillary gel electrophoresis (
17
) to create a separation method called `constant denaturant capillary
electrophoresis' (CDCE) (
18
). Separation by CDCE enabled us to measure mutant fractions as low as 10
-6
in model reconstruction experiments with mixtures of PCR fragments (
7
).
We calculated that a detection limit of ~10
-6
would allow us to see hotspots in human mitochondrial DNA if they occurred at a
mutation rate that is 20 times higher than nuclear point mutations, as has been
estimated based on studies of evolution (
19
-
22
). Also, mitochondrial DNA is present as a few hundred to thousands of copies
per cell, which reduces the size of the tissue sample and the amount of total
DNA that must be handled to obtain statistically reproducible observations (
23
).
As reported herein, the detection limit achieved was indeed sufficient to
observe the human mitochondrial point mutational spectrum for our 100 bp target
sequence in human cells and tissues. Key to this report are the series of tests
in which we explored the possibility that our observations were artefacts
generated by our procedures. While we find that the methodology described is
sufficient for identifying and measuring hotspot point mutations in
mitochondrial DNA, further technical improvements are needed to achieve our
goal of measuring mutational spectra in nuclear DNA where mutant fractions and
the number of target sequence copies per cell are both lower.
TK6 cells (
24
) were maintained in exponential growth as spinner cultures by daily dilution
with RPMI-1640 media (Gibco BRL, Grand Island, NY) supplemented with 10% horse serum
(Gibco BRL) at densities not exceeding 10
6
cells/ml. Brush bronchoscopy samples of human lung epithelial cells were obtained through collaboration with
Dr M. Utell and Dr M. Frampton of the University of Rochester Medical Center.
Cells, tissue and DNA samples were stored at -70oC.
DNA was isolated from suspended cells or macerated tissue by digestion with
proteinase K (1 mg/ml) and RNAse A (0.1 mg/ml) in 0.5% SDS followed by ethanol
precipitation of DNA. DNA prepared in this way was easily restricted and
amplified. The method provided high DNA yields from 10
5
to 10
9
cultured cells or from milligrams to grams of tissue. The DNA was digested with
Rsa
I, which recognizes base pairs 10 009-10 012 of the mitochondrial genome, and
Dde
I, which recognizes base pairs 10 227-10 231 (New England Biolabs, Beverly, MA) at 2 U/[mu]g DNA, overnight. To prevent contamination with enriched mutants, the first stages of DNA isolation and mutant purification were
carried out in a separate clean room.
CDGE gels were 8% acrylamide, 1/40 bis-acrylamide, TAE (40 mM Tris-acetate, 1 mM EDTA), 0.4 mm thick. The gels were submerged in a tank
containing TAE heated to a precisely selected temperature around 70oC. As the temperature is increased, the electrophoretic mobility of a DNA
fragment with a biphasic melting profile decreases following a sigmoid curve.
We selected a temperature at the midpoint of the sigmoid curve of the wild-type fragment for our experiments in order to ensure the best separation
of both high and low T
m
mutants from the wild-type. Samples of 4-6 [mu]l of a restriction digestion were loaded and separated for 2 h
at 10 V/cm.
Fluorescein-labeled markers and samples were separated in alternating parallel lanes.
The markers were a mixture of PCR amplified
hprt
exon 3 wild-type and mutant homoduplexes and heteroduplexes with melting temperatures
similar to those of the mitochondrial DNA mutants to be detected. One lane of
each gel contained a second marker which was the wild-type target mitochondrial sequence with an altered high melting domain
primer (for sequence, see Materials and Methods: High-fidelity PCR). The melting behavior of this sequence was indistinguishable from that of the normal wild-type sequence, but it could not be amplified with the primers used
for subsequent PCR. This primer was designed as such to avoid contamination of
the tissue samples (the number of target sequence copies in a marker lane is
three orders of magnitude higher than in the sample lane). The marker bands
were visualized by illuminating the gel with the 488 nm argon laser used for
CDCE (see below) and marked while viewing the gel through a 520 nm low pass
glass filter (Oriel, Stratford, CT). The relative positions of the markers
provide information on the exact position to which the desired mitochondrial mutants migrated in the gel. The portions of the gel below and above the marker
wild-type band were cut so as to include the range of mobilities of most mutant
fragments.
The gel slices were finely ground between two glass slides and transferred into 1.5 ml test tubes containing 200 [mu]l of 200 mM NaCl. DNA was eluted by shaking for 20 min at 50oC, 1300 min
-1
. Gel particles were spun down, and the supernatant was collected. The DNA was
precipitated with 2 vol of ethanol and dissolved in ~15 [mu]l of water.
Our CDCE instrument has been described previously (
17
,
18
,
25
). Briefly, electrophoresis was performed in 30 cm long 75 [mu]m I.D. capillaries. A portion of the capillary was heated by a water jacket
connected to a constant temperature circulator. The jacket was positioned 5 cm
from the injection end of the capillary. The length of the jacket was 5 cm for
the enrichment of heteroduplexes and 15 cm for high resolution CDCE. The
temperature of the water jacket was ~65oC. The precise temperature to be used in a particular experiment and the time of fraction collection was determined in test
runs with appropriate standards.
In this study, the CDCE instrument used for identification of mutants has been
improved by adding a two wavelength detector. To detect DNA, the capillary was
illuminated by a 515 nm argon laser and emitted light was collected at a right
angle by a microscope objective. For two-channel detection of two fluorophores, collected light was split and directed into two detectors through
appropriate sets of filters. For fluorescein, a combination of a 540 nm
bandpass and a 530 nm long pass filter was used; for tetramethylrhodamine
(TMR), a single 580 nm bandpass filter was employed. The signals from the photomultipliers were recorded by a computerized data acquisition system.
The inner surface of the capillaries was coated with linear polyacrylamide
chains. The coating procedure was adapted from a methodology published by Hjerten (
26
). Capillaries were treated with 1 M NaOH for 2 h, washed sequentially with 1 M HCl and methanol, treated
overnight with [gamma]-methacryloxypropyltrimethoxysilane (Sigma, St Louis, MO) and washed
with methanol. The capillaries were then filled with a polymerizing solution of
6% acrylamide in TBE (89 mM Tris, 89 mM borate, 1 mM EDTA, pH 8.4), 0.1% TEMED and 0.025% ammonium persulfate, and left for several hours to polymerize.
Capillaries were filled with a fluid polyacrylamide matrix, which was replaced
before each run. The matrix was prepared as follows: 5% acrylamide solution in
TBE was deoxygenated by argon bubbling in an ice bath for 10 min and any
contact of the solution with air was avoided until polymerization was complete.
TEMED and ammonium persulfate were added to 0.03% and 0.003%, respectively. The polymerizing solution was immediately dispensed into 10 ml glass syringes and left at 2oC for several days. The matrix was dispensed from the 10 ml syringes into 100 [mu]l high pressure U6K syringes (Rainin, Emeryville, CA) which were used to
replace the matrix in capillaries through home-made teflon tube fittings.
About 10
8
total copies were electroinjected into a capillary and run at ~200 V/cm. Samples were electroeluted from the anode end of the capillary into 0.5 ml Eppendorf tubes with 5 [mu]l of 0.1* TBE, 0.1 mg/ml bovine serum albumin. High resolution CDCE runs were
performed at 100 V/cm using gel and running buffers containing 30 mM Na
+
(30 mM sodium borate, pH 8), in addition to TBE. Gels for high resolution CDCE
were prepared using a lower amount of ammonium persulfate (0.0015%).
PCR was performed in 10 [mu]l capillaries in an air thermocycler (Idaho Technology, Idaho Falls, ID)
with native
Pfu
thermostable polymerase (Stratagene, La Jolla, CA). The PCR mix included 10 mM KCl, 6 mM (NH
4
)
2
SO
4
, 20 mM Tris-HCl (pH 8.0), 2 mM MgCl
2
, 0.1% Triton X-100, 100 [mu]g/ml BSA, 0.2 [mu]M each primer, 0.1 mM dNTPs and 0.1 U/[mu]l of
Pfu
. Denaturation at 95oC, annealing at 57oC and extension at 72oC, each for ~10 s intervals, constituted the amplification cycle. After
the desired number of cycles, samples were incubated at 72oC for 2 min and then at 45oC for 30 min. The primers used were CW7 [Watson strand primer(ACC GTT
AAC TTC CAA TTA AC) base pairs 10 011- 10 031 of human mitochondrial genome] and J3 [Crick strand primer(GCG
GGC GCA GGG AAA GAG GT), complementary to base pairs 10 196-10 215]. The altered high melting domain primer used to generate CDGE
markers was [alternate Crick strand primer(GAA GAA TTT TAT GGA GAA AGG GTG CGC
CCG GGG GGA TAT AGG GTC GAA GC)]. The fluorescein moiety was linked to a
cyanoethyl phosphoramidite by a nine atom spacer arm. This species was coupled
to the 5' end of the oligonucleotide during commercial synthesis (Ransom Hill Bioscience, Inc., Ramona, CA). The tetramethylrhodamine (TMR) dye was coupled to the oligonucleotide via an amino C6 linker also during commercial synthesis (Biosynthesis, Louisville, TX).
For asymmetric pre-amplification, the conditions were the same as described above except the
appropriate primer was omitted from the pre-amplification reaction mixture and exonuclease deficient
Pfu
polymerase (Stratagene, La Jolla, CA) was used instead of the exonuclease
proficient form.
Our approach to measuring mutational spectra (Fig.
1
) is based on sequential enrichment of mutants that lie within a target sequence
100 bp long. The target sequence comprises the low melting domain of a 200 bp
DNA fragment which has a biphasic melting profile suitable for separation of
point mutants by partially denaturing gel electrophoresis, i.e. a low melting
domain adjacent to a melting domain requiring a higher temperature for denaturation. Mutations in the low melting domain of such a sequence affect the rapid
equilibrium between the partially denatured and double stranded forms of the
molecule in a given temperature range, and therefore alter the mobility of the
DNA through a gel matrix. The enrichment of mutants is based on separation of
all of the mutant sequences present in a sample from the large excess of wild-type sequences.
It is critical for measurement of mutant fractions in our samples that we
accurately measure the number of copies of target sequence in each DNA sample.
Since our measurements of mutant fractions are based on comparisons with internal standards, knowledge of the
total
number of initial copies of the target sequence is essential for accurately
introducing internal standards at the desired initial fraction. Further, the
number of mutant copies in the sample must be known to ensure that it is
sufficient for the desired statistical power.
To measure the initial number of copies available for PCR in a DNA sample, the
sample was doped with a PCR amplified target sequence containing a A:T -> G:C mutation at bp 10 072 to serve as an internal standard. The number of
mutant copies in the stock dilutions of internal standard was inferred from the
amount of primer used in PCR given that the primer was completely exhausted.
The sample DNA and a known amount of internal standard was subjected to PCR
followed by CDCE. The areas of the wild-type and mutant peaks were determined. The ratio of these areas was used
to calculate the initial number of copies.
Critical to this calculation is the demonstration that wild-type and internal standard amplify with the same efficiency. The
amplification efficiency of the internal standard was equal to that of the wild-type (0.6 per cycle) within +-1%. The error resulting from this slight efficiency inequality is
smaller than the precision required in such an experiment. This method is
robust; the sample can be doped at any ratio of mutant to wild-type from 0.01 to 100 with good results.
Prior to enrichment for mutant sequences, samples are represented by ~2.5 [mu]g restriction digested cellular DNA containing ~4 * 10
8
copies of the unlabeled target sequence. The small fraction of target sequences
that bear mutations must now be separated from the vast excess of wild-type sequences and non-target DNA. This task could be simplified by PCR amplification of
the sample followed by CDCE separation. However, to achieve low backgrounds and, hence, low detection limits, it is important to significantly enrich for mutants before PCR is applied to the sample.
Slab gel CDGE separation (
28
) was used for the necessary initial enrichment in the examples reported here.
Under the partially denaturing conditions used for separation, `high T
m
mutants', i.e
.
mutants that increase the melting temperature of the target isomelting domain,
have higher electrophoretic mobility than the wild-type fragments. Conversely, the mobility of `low T
m
mutants' is lower than that of the wild-type. Therefore, by excising the portions of the gel below and above the
wild-type band, one enriches for high and low
Tm
mutants, respectively.
CDGE rather than CDCE was used at this stage because the large amount of non-target DNA and impurities in the samples create a high-resistance zone in the capillary which prevents separation. An
alternative approach has since been developed, in which CDGE is replaced by
CDCE in wide-bore capillaries (
29
).
The efficiency of CDGE enrichment of mutants was estimated using a sample doped
with a high T
m
and a low T
m
mutant as internal standards at fractions of 10
-2
each. Mutants at this fraction can be easily and reliably observed on CDCE
after PCR. (The samples used for mutational analysis are usually doped at 10
-5
-10
-4
, fractions that are too low to be observed without enrichment). The ratio of
mutant fraction after CDGE to the initial fraction represents the enrichment
efficiency. The enrichment was typically 100-200 for high T
m
mutants and 5-10 for low T
m
mutants.
Elution from the CDGE gel slices and the first PCR represents the statistical
`bottleneck' in the procedure since the number of mutant copies is at a
minimum. Any mutant to be measured with adequate precision (that is, not worse
than +- 20%, 95% confidence limits) should be represented by at least 100
copies at the bottleneck in the procedure. Thus, if one began with 4 * 10
5
TK6 cells or 4 * 10
8
mitochondrial DNA copies, a mutant with a mutant fraction of 10
-6
would be present as 400 copies. Losses in DNA isolation and enrichment steps
would reduce this number to ~200, and an estimated efficiency of the first cycle of PCR of 0.5 would
mean the bottleneck number would be 100 mutant copies, the minimum necessary to
give the desired lower limit of statistical uncertainty.
After enrichment by CDGE, the samples were subjected to PCR, which generated 10
12
copies of the target sequence, reduced the relative amount of non-target DNA to an insignificant level and provided for fluorescent labeling
of the target. In essence, PCR made the samples suitable for convenient CDCE
separations.
During PCR, the mutants were converted into heteroduplexes with the excess of
wild-type strands by continuing to subject the sample to temperature cycles
even after the point where the primers were exhausted. All such heteroduplexes have significantly lower melting temperatures than the parent wild-type homoduplex, a fact which allows collection of all of the heterodouplexes as a single
fraction greatly reduced in the number of wild-type strands by DGGE, CDGE or CDCE (
30
).
PCR introduces `noise' into the analysis. It is important to amplify with a DNA
polymerase of sufficiently high fidelity so that errors created by copying
excess wild-type strands do not interfere with observation of mutants in the samples (
31
).
Pfu
polymerase was used in this work because it is thermostable and has high
fidelity, with an error rate as low as 2 * 10
-6
errors per base per doubling (
32
). In our target sequence the rate appears to be even lower, 7 * 10
-7
(André
et al
., in preparation). This is significantly lower than all other DNA polymerases
we have screened for fidelity. On the other hand,
Pfu
shares with other DNA polymerases the ability to convert wild-type sequences into byproducts that have lower melting temperatures than
the wild-type homoduplex. These byproducts can interfere with the mutant enrichment
step on CDCE. Among these byproducts are incomplete or exonucleolytically
processed products missing from one to several nucleotides. We have experienced
fractions of these byproducts as high as 50% of the total in some PCR
preparations. The problem has been reduced but not eliminated by use of an
increased
Pfu
concentration (4-fold higher than the manufacturer's suggestion) and a 30 min post-PCR incubation at 45oC. Under these conditions the CDCE enrichment is ~30-fold, limited by collection of 3% of the wild-type as tail on the main peak and low melting
temperature wild-type byproducts.
A post-PCR CDCE separation is shown in Figure
2
A. A sample of TK6 cellular DNA containing two prominent low T
m
mutants with initial fractions of ~4 * 10
-4
and 3 * 10
-4
was chosen for demonstration purposes. The four heteroduplexes of the two mutants (marked by Xs) are ~0.2% of the wild-type peak (which implies 5-fold CDGE enrichment) and are hardly distinguishable among the
wild-type variants (noise) which are present at a similar fraction (marked by
asterisks).
Once the mutants have been enriched, our goal is to separate, measure, isolate
and sequence the individual mutants. We convert mutant heteroduplexes into
homoduplexes by stopping PCR when the molar amount of unused primers still
exceeds that of the products.
Homoduplexes are then separated via CDCE under `high resolution separation
conditions' that include increasing the length of the zone, decreasing the
running voltage, and increasing the salt concentration in the electrophoretic
buffer. As we show in a recent paper (
33
), the improvement in resolution with these modifications results from
increasing the average number of partial meltings and reannealings a molecule
undergoes while in the heated zone of the capillary.
Figure
3
(upper curve) shows a high resolution separation of the same sample as the one
shown in Figure
2
B after it was converted to homoduplex form. The reason that the two separations
look dissimilar is that each pair of heteroduplexes present in Figure
2
B yielded a single homoduplex peak in Figure
3
A.
In our first studies of mitochondrial hotspot mutants, each mutant peak was
isolated, amplified and sequenced. It became clear soon thereafter that the
same mutants appeared in multiple samples and that a more efficient mode of
mutant identification was possible. All sequenced mutants were labeled with
tetramethylrhodamine (TMR) and combined together to form a `standard set' of markers. This TMR-labeled `standard set' could then be coinjected with an aliquot of fluorescein-labeled sample under study and both signals could be observed on a
single CDCE separation by means of a two-wavelength detector. In this way, mutants in the fluorescein-labeled sample could be identified based on comigration with the rhodamine-labeled previously isolated hotspot mutants. Further, any
novel mutants could be identified for subsequent isolation and sequencing. This
time saving approach is illustrated in Figure
3
.
Comigration of an unknown peak with a known standard is not proof of identity
(non-comigration is proof of nonidentity). Two different mutants could
potentially co-migrate leading to false identification. In order to provide a more
rigorous test of peak identity, we developed an identification procedure based
on on-column hybridization of sample peaks with the mutants of the standard set
(in preparation). The CDCE instrument used for identification by hybridization
included two consecutive heating jackets. The sample and the standards were co-separated in the first jacket as for identification by co-migration and the resulting set of peaks was stopped inside the second jacket. The temperature in the second jacket was increased to denature the DNA, then
decreased to permit reannealing of single strands, and finally the sample was
electrophoresed through the rest of the second jacket to the detector. If a
peak of the sample was not identical to the standard mutant with which it co-migrated during the pre-hybridization part of CDCE separation, such hybridization converted
the sample mutant into two heteroduplexes with the standard mutant present in
excess. Such heteroduplexes are always less stable than the homoduplexes from
which they were derived, and they moved to a different position during the post-hybridization portion of CDCE separation. This resulted in the
`disappearance' of the sample peak from the spectrum. This approach is as
precise as sequencing but does not require isolation of individual mutants
prior to identification. The identities of all the mutant peaks reported in
this paper were verified either by direct sequencing or on-column hybridization.
The area under each peak in Figure
3
is proportional to the initial mutant fraction in this sample. The initial
mutant fraction of mutant peaks p3, p1.5 and p1 can be determined by comparison
of the areas under these peaks to the peak representing the 10
-4
internal standard. In this sample, the original fractions of mutants represented by p3, p1.5 and p1 are estimated to be 5 * 10
-5
, 4 * 10
-4
and 3 * 10
-4
, respectively.
Given that our procedure involves ~10
9
-fold amplification, the relative sizes of the peaks we observe may depend
not only on the original mutant fractions but also on the relative
amplification advantage or disadvantage of a particular mutant (allelic preference). To discover if the mutant fractions we report deviated significantly from the original fractions due to allelic preference, we checked the relative
amplification efficiencies of most of the mutants in the standard set. Mutants
were mixed together and amplified 1000-fold in the same PCR reaction and the relative sizes of the peaks before
and after the amplification were compared, as suggested by Keohavong and Thilly
(
31
). The amplification efficiency of the mutants was within +-6% of the efficiency of amplification of the wild-type (0.6 per cycle). Given that the overall amplification of
mutants during our procedure is ~10
9
-fold, these differences in amplification efficiencies will lead to errors
in the measurement of mutants within a factor of two. These errors may be
corrected by the appropriate weighting of the mutant frequencies as determined
by comparison with the internal standards.
Since it is not possible to check the amplification efficiencies of all theoretically possible mutants, we cannot exclude the possibility that
some mutants with a severe amplification disadvantage were originally present in the samples but were missed during our isolation
procedure.
We determined the limit of detection of this method by doping purified wild-type sequences with purified mutants of known quantity. Shown in Figure
4
A are the results when wild-type PCR fragments purified from possible mutants by CDGE were doped with
a low T
m
mutant at 10
-4
and with a high T
m
mutant at 10
-5
, mixed with non-amplifiable herring carrier DNA, and subjected to all the procedures
described above except for the initial DNA isolation. The data sets are
separated as `high T
m
mutants' and `low T
m
mutants' because they are processed separately at stages succeeding CDGE.
When such a sample is enriched in mutants by CDGE, the peaks representing both
of the reference mutants are about 10 times larger than the largest peaks of
the background (Fig.
4
A). Thus, the limit of detection, defined as the frequency of a mutant that
yields a peak comparable to the largest background peaks, should be considered 10
-5
for low T
m
mutants and 10
-6
for high T
m
mutants. When CDGE enrichment is not used, as may be seen in Figure
4
B, the background noise is 5-10-fold higher than when it is used. Comparison of Figure
4
A and B therefore demonstrates that the CDGE step improves the limit of
detection of the method ~10-fold, from 10
-5
to 10
-6
for high T
m
mutants and from 10
-4
to 10
-5
for low T
m
mutants, thus illustrating the need for a CDGE enrichment step prior to PCR.
Since the limit of detection depends on the size of background peaks, it is
important to know the sources of background mutants. Most of the background
peaks do not co-migrate with mutants originating from
Pfu
replication errors (André
et al.
, manuscript in preparation). We therefore assume that the background mutations
most probably arose from mutagenic DNA lesions that were generated during our
procedures. The difference in detection limits between high and low T
m
mutants could reflect that more of these lesions co-migrate with low T
m
mutants than with high T
m
mutants.
The approach described herein was tested by measuring mutational spectra of
cultured cell populations and human tissue samples. Shown in Figure
5
B are high resolution CDCE runs of a sample derived from a culture of TK6 cells,
while Figure
5
C shows the same data for a sample of human bronchial epithelial cells. Figure
5
A shows the outcome of subjecting CDGE-purified mitochondrial wild-type homoduplexes to the procedure as a negative control for the
entire analytical cascade.
There are two classes of errors which could potentially `create' some of the
mutants we observe. The first class involves artefacts generated by our
procedures themselves. Contamination of samples and PCR-created mutants are included in this first class. PCR errors could arise
as misincorporation events during primer extension over a wild-type substrate or as a result of replicative bypass of DNA adducts formed
in our procedures. The errors of this class are readily detected by comparison of a spectrum to the
negative control, such as the one shown in Figure
5
A.
Errors arising during mutant isolation procedure
The mutants that we detect with this method are represented in our initial
samples by as few as 400 molecules. Consequently, an obvious problem is the
potential for contamination from purified PCR preparations of these mutant
sequences. Although there are a number of ways to reduce the probability of
such accidents, we have chosen the alternative of setting up a clean laboratory
equipped with high-throughput air filters. No PCR products of purified mutant sequences are
allowed into this laboratory except for very dilute stocks used to introduce
internal standards. DNA isolation, CDGE, and preparing aliquots for PCR are
performed in this laboratory. Our quality control procedures involve routine negative controls which have shown no evidence of contamination in the past few years, but which did detect intermittent contamination in
earlier studies.
In order to determine whether any of the mutant peaks were caused either by
Pfu
polymerization errors or by replicative bypass of DNA adducts created during
our procedures, we subjected CDGE-purified wild-type homoduplex containing appropriate internal standards to all of
the steps of our procedure. The background peaks observed (Fig.
5
A) were much smaller than the mutant peaks isolated from cellular DNA (Fig.
5
B and C). This demonstrates that the background generated by our procedure is
small enough not to interfere with real cellular mutants. In a variation on
this control, to test whether detectable levels of DNA adducts leading to noise
`mutants' were generated during cell lysis, we mixed the purified target human
mitochondrial fragment with rat cells and carried out the entire procedure. No
interfering signals were obtained in this control experiment either.
Two additional sources of false mutants are potentially mutagenic lesions of
cellular (as opposed to instrumental) origin which yield mutants when bypassed
by a polymerase during PCR, and replication mismatches not yet repaired at the
time of DNA isolation. These two types of errors cannot be recognized by any of
the tests described above. In order to discriminate cellular lesions and
mismatches from true mutants, we developed a method that exploits a
distinguishing property of these species: adducts or sequence changes are
present in only one of the two DNA strands, while the other strand is an
unadducted wild-type sequence. True mutants, in contrast, contain a changed sequence at
the same base pair in both DNA strands of the original sample.
To preserve this distinguishing information, we developed a modified PCR procedure which includes asymmetric pre-amplification. Two aliquots of a CDGE-enriched sample are subjected to repeated primer elongation cycles
with only one or the other of the two PCR primers. During this `linear
amplification' stage, only one strand of the sample DNA is used as a template.
Thus after preamplification one aliquot will contain a vast excess of `Watson' copies while the other will contain excess `Crick' copies as single strands. Then the other primer is added and exponential amplification proceeds.
Asymmetric pre-amplification of a real double stranded mutant results in equal fractions
of a particular mutant when the opposite DNA strands are compared. However, in
the case of a mismatch or a pre-mutagenic lesion, the mutant will arise only from copying the strand which
contains the mismatch or premutagenic lesion. This results in an asymmetric
distribution of the mutant between the spectra obtained from the two aliquots
which begin by copying only the Watson or the Crick strand.
Figure
6
shows the results of asymmetric pre-amplification as applied to low T
m
mutants isolated from a lung epithelium sample from a different person than
that shown in Figure
4
C. The sample contains the peaks for p1 (G:C -> A:T at bp 10 068) and p3 (G:C -> A:T at bp 10 038). The results show that p1 yields the same mutant
fraction regardless of the strand used in linear amplification. We regard this
as strong evidence that this signal arises from a true mutant in this sample.
However, the signal designated p3 shows a significantly larger mutant fraction
when linear amplification copies the Watson as opposed to the Crick strand.
These results suggest that a significant portion of the signal designated p3
arose either as a mismatch intermediate or as an adduct located in the Watson
strand that was converted into a mutant by
Pfu
polymerase. Another mutant, yet unsequenced peak designated X in Figure
8
, demonstrates the opposite distribution, which suggests that it originates in
significant part from a mismatch or a lesion located on the Crick strand. As a
rule, high T
m
mutants behaved symmetrically in the assay. The signal represented by peak p3,
however, has behaved asymmetrically in all samples analyzed to date, while the
signal represented as peak p1 showed different degrees of asymmetry in
different samples.
The authors would like to acknowledge Dr J. S. Hanekamp (MIT) and Dr G. Hu
(Shanghai Cell Biology Institute) for sharing experimental results and valuable
discussions, and Dr S. Singer (Brigham and Women's Hospital) and Drs M. Utell,
M. Frampton and A. Torres (University of Rochester Medical Center) for help in
obtaining tissue samples. This work was supported by grants from the National
Institute for Environmental Health Sciences: PO1-ESO7168, Mutagenic Effects of Air-Borne Toxicants, P42-ESO4675, Superfund Basic Research and PO1-ESO3926, Genetics and Toxicology, and a grant from the
U.S. Department of Energy, Office of Health and Environmental Research DE-FGO2-86ER60448, Comparative Mutagenesis of Human Cells
In Vitro
and
In Vivo
.
*To whom correspondence should be addressed. Tel: +1 617 253 6226; Fax: +1 617
258 5424; Email: khrapko@wccf.mit.edu
REFERENCES
Return