Proc. Natl. Acad. Sci. USA
`Vol. 96, pp. 9236–9241, August 1999
`Genetics
`
`Digital PCR
`
`BERT VOGELSTEIN* AND KENNETH W. KINZLER
`The Howard Hughes Medical Institute and the Johns Hopkins Oncology Center, Baltimore, MD 21231
`
`Contributed by Bert Vogelstein, June 9, 1999
`
`The identification of predefined mutations
`ABSTRACT
`expected to be present in a minor fraction of a cell population
`is important for a variety of basic research and clinical
`applications. Here, we describe an approach for transforming
`the exponential, analog nature of the PCR into a linear, digital
`signal suitable for this purpose. Single molecules are isolated
`by dilution and individually amplified by PCR; each product
`is then analyzed separately for the presence of mutations by
`using fluorescent probes. The feasibility of the approach is
`demonstrated through the detection of a mutant ras oncogene
`in the stool of patients with colorectal cancer. The process
`provides a reliable and quantitative measure of the proportion
`of variant sequences within a DNA sample.
`
`In classical genetics, only mutations of the germ line were
`considered important for understanding disease. With the real-
`ization that somatic mutations are the primary cause of cancer (1)
`and may also play a role in aging (2, 3), new genetic principles
`have arisen. These discoveries have provided a wealth of oppor-
`tunities for patient management as well as for basic research into
`the pathogenesis of neoplasia. However, many of these oppor-
`tunities hinge on detection of a small number of mutant-
`containing cells among a large excess of normal cells. Examples
`include the detection of neoplastic cells in urine (4), stool (5, 6),
`and sputum (7, 8) of patients with cancers of the bladder,
`colorectum, and lung, respectively. Such detection has been
`shown in some cases to be possible at a stage when the primary
`tumors are still curable and the patients asymptomatic. Mutant
`sequences from the DNA of neoplastic cells have also been found
`in the blood of patients with cancer (9–11). The detection of
`residual disease in lymph nodes or surgical margins may be useful
`in predicting which patients might benefit most from further
`therapy (12–14). From a basic research standpoint, analysis of the
`early effects of carcinogenesis often depends on the ability to
`detect small populations of mutant cells (15–17).
`Because of the importance of this issue in so many settings,
`many useful techniques have been developed for the detection of
`mutations. DNA sequencing is the gold standard for the detection
`of germ-line mutations but is useful only when the fraction of
`mutated alleles is greater than ’20% (18, 19). Mutant-specific
`oligonucleotides sometimes can be used to detect mutations
`present in a minor proportion of the cells analyzed, but the
`signal-to-noise ratio distinguishing mutant and wild-type (WT)
`templates is variable (20–22). The use of mutant-specific primers
`and the digestion of PCR products with specific restriction
`endonucleases are extremely sensitive methods for detecting such
`mutations, but it is difficult to quantitate the fraction of mutant
`molecules in the starting population with these techniques (23–
`28). Other innovative approaches for the detection of somatic
`mutations have been reviewed (29–32). A general problem with
`these methods is that it is difficult or impossible to confirm
`independently the existence of any mutations that are identified.
`
`We therefore sought to develop an approach to the problem
`that would overcome some of the aforementioned difficulties.
`The strategy described in this paper involves separately amplify-
`ing individual template molecules so that the resultant PCR
`products are completely mutant or completely WT. The homo-
`geneity of these PCR products makes them easy to distinguish
`with existing techniques. Such separate amplifications are only
`useful in a practical sense, however, if a large number of them can
`be assessed simply and reliably. Techniques for such assessments
`were developed, with the output providing a digital readout of the
`fraction of mutant alleles in the analyzed population. A variety of
`applications for this technology are foreseeable.
`
`MATERIALS AND METHODS
`Step 1: PCR Amplifications. The optimal conditions for PCR
`described in this section were determined by varying the param-
`eters described in Results. PCR was performed in 7-ml volumes in
`96-well polypropylene PCR plates (Marsh Biomedical Products,
`Rochester, NY). The composition of the reactions was 67 mM
`Tris (pH 8.8), 16.6 mM NH4SO4, 6.7 mM MgCl2, 10 mM
`b-mercaptoethanol, 1 mM dATP, 1 mM dCTP, 1 mM dGTP, 1
`mM dTTP, 6% (volyvol) DMSO, 1 mM primer F1, 1 mM primer
`R1, 0.05 unitsyml Platinum Taq polymerase (Life Technologies,
`Grand Island, NY), and one-half genome equivalent of DNA (see
`below for description of primers). To determine the amount of
`DNA corresponding to one-half genome equivalent, DNA sam-
`ples were serially diluted and tested via PCR. The amount that
`yielded amplification products in half of the wells, usually ’1.5 pg
`of total DNA, was defined as one-half genome equivalent and
`used in each well of subsequent digital PCR (Dig-PCR) experi-
`ments. Light mineral oil (50 ml; Sigma M-3516) was added to each
`well, and reactions were performed in a HybAid Thermal cycler
`(Middlesex, U.K.) at the following temperatures: denaturation at
`94° for 1 min; 60 cycles of 94° for 15 s, 55° for 15 s, 70° for 15 s;
`and 70° for 5 minutes. Reactions were analyzed immediately or
`stored at room temperature for up to 36 h before fluorescence
`analysis.
`Step 2: Fluorescence Analysis. The following solution (3.5 ml)
`was added to each well: 67 mM Tris (pH 8.8), 16.6 mM NH4SO4,
`6.7 mM MgCl2, 10 mM b-mercaptoethanol, 1 mM dATP, 1 mM
`dCTP, 1 mM dGTP, 1 mM dTTP, 6% (volyvol) DMSO, 5 mM
`primer INT, 1 mM molecular beacon (MB)-GREEN, 1 mM
`MB-RED, and 0.1 unitsyml Platinum Taq polymerase. The plates
`were centrifuged for 20 s at 6,000 3 g, and fluorescence was read
`at excitationyemission wavelengths of 485y530 nm for MB-
`GREEN and 530y590 nm for MB-RED. The fluorescence in
`wells without template was typically 10,000 to 20,000 specific
`fluorescence units (SFU), with about 75% emanating from the
`fluorometer background and the remainder from the MB probes.
`The plates were then placed in a thermal cycler for asymmetric
`amplification at the following temperatures: 94° for 1 min; 10–15
`cycles of 94° for 15 s, 55° for 15 s, 70° for 15 s; 94° for 1 min; and
`60° for 5 min. The plates were then incubated at room temper-
`ature for 10–60 min, and fluorescence was measured as described
`
`The publication costs of this article were defrayed in part by page charge
`payment. This article must therefore be hereby marked ‘‘advertisement’’ in
`accordance with 18 U.S.C. §1734 solely to indicate this fact.
`PNAS is available online at www.pnas.org.
`
`Abbreviations: Dig-PCR, digital PCR; MB, molecular beacon; SFU,
`specific fluorescence unit; WT, wild-type.
`*To whom reprint requests should be addressed. E-mail: vogelbe@
`welchlink.welch.jhu.edu.
`
`9236
`
`MYR1017
`Myriad Genetics, Inc. et al. (Petitioners) v. The Johns Hopkins University (Patent Owner)
`IPR For USPN 6,440,706
`
`Page 1 of 6
`
`The Johns Hopkins University Exhibit JHU2005 - 1 of 6
`
`

`

`Genetics: Vogelstein and Kinzler
`
`Proc. Natl. Acad. Sci. USA 96 (1999)
`
`9237
`
`above. Specific fluorescence was defined as the difference in
`fluorescence before and after the asymmetric amplification.
`REDyGREEN ratios were defined as the specific fluorescence of
`MB-RED divided by that of MB-GREEN. REDyGREEN ratios
`were normalized to the ratio of the positive controls (25 genome
`equivalents of DNA from normal cells, as defined above). We
`found that the ability of MB probes to discriminate between WT
`and mutant sequences under our conditions could not be deter-
`mined reliably from experiments in which they were tested by
`hybridization to relatively short complementary single-stranded
`oligonucleotides and that actual PCR products had to be used for
`validation.
`Oligonucleotides and DNA Sequencing. Primer F1 was 59-
`CATGTTCTAATATAGTCACATTTTCA-39; primer R1 was
`59-TCTGAATTAGCTGTATCGTCAAGG-39; primer INT was
`59-TAGCTGTATCGTCAAGGCAC-39; MB-RED was 59-Cy3-
`CACGGGCCTGCTGAAAATGACTGCGTG-Dabcyl-39; MB-
`GREEN was 59-fluorescein-CACGGGAGCTGGTGGCG-
`TAGCGTG-Dabcyl-39. MBs (33, 34) were synthesized by Mid-
`land Scientific, and other oligonucleotides were synthesized by
`Gene Link (Thornwood, NY). All were dissolved at 50 mM in TE
`buffer (10 mM Tris, pH 8.0y1 mM EDTA) and kept frozen and
`in the dark until use. PCR products were purified with QIAquick
`PCR purification kits (Qiagen, Chatsworth, CA). In the relevant
`experiments described in the text, 20% of the product from single
`wells was used for gel electrophoresis, and 40% was used for each
`sequencing reaction. The primer used for sequencing was 59-
`CATTATTTTTATTATAAGGCCTGC-39. Sequencing was
`performed by using fluorescently labeled Applied Biosystems Big
`Dye terminators and an Applied Biosystems 377 automated
`sequencer.
`
`RESULTS
`Principles Underlying Dig-PCR. The two steps comprising
`Dig-PCR are outlined in Fig. 1A. First, the DNA is diluted into
`multiwell plates so that there is, on average, one template
`molecule per two wells, and PCR is performed. Second, the
`individual wells are analyzed for the presence of PCR products
`of mutant and WT sequence by using fluorescent probes.
`As the PCR products resulting from the amplification of single
`template molecules should be homogeneous in sequence, a
`variety of standard techniques could be used to assess their
`presence (see Introduction). Fluorescent probe-based technolo-
`gies, which can be performed on the PCR products ‘‘in situ’’ (i.e.,
`in the same wells), are particularly well suited for this application
`(31, 33–40). We chose to explore the utility of one such technol-
`ogy, involving MBs, for this purpose (33, 34). MB probes are
`oligonucleotides with stem–loop structures that contain a fluo-
`rescent dye at the 59 end and a quenching agent (Dabcyl) at the
`39 end (Fig. 1B). The degree of quenching via fluorescence energy
`resonance transfer is inversely proportional to the sixth power of
`the distance between the Dabcyl group and the fluorescent dye
`(41). After heating and cooling, MB probes reform a stem–loop
`structure that quenches the fluorescent signal from the dye. If a
`PCR product whose sequence is complementary to the loop
`sequence is present during the heatingycooling cycle, hybridiza-
`tion of the MB to one strand of the PCR product will increase the
`distance between the Dabcyl and the dye, resulting in increased
`fluorescence.
`A schematic of the oligonucleotides used for Dig-PCR is shown
`in Fig. 1C. Two unmodified oligonucleotides are used as primers
`for the PCR reaction. Two MB probes, each labeled with a
`different fluorophore, are used to detect the PCR products.
`MB-GREEN has a loop region that is complementary to the
`portion of the WT PCR product that is tested for mutations.
`Mutations within the corresponding sequence of the PCR prod-
`uct should impede its hybridization to the MB probe significantly
`(33, 34). MB-RED has a loop region that is complementary to a
`different portion of the PCR product, one not expected to be
`mutant. It thus should produce a signal whenever a well contains
`
`FIG. 1. Schematic of Dig-PCR. (A) The basic two steps involved:
`PCR on diluted DNA samples is followed by addition of fluorescent
`probes that discriminate between WT and mutant alleles and subse-
`quent fluorometry. (B) Principle of MB analysis. In the stem–loop
`configuration, fluorescence from a dye at the 59 end of the oligonu-
`cleotide probe is quenched by a Dabcyl group at the 39 end. On
`hybridization to a template, the dye is separated from the quencher,
`resulting in increased fluorescence (modified from Marras et al.; ref.
`56). (C) Oligonucleotide design. Primers F1 and R1 are used to
`amplify the genomic region of interest. Primer INT is used to produce
`single-stranded DNA from the original PCR products during a sub-
`sequent asymmetric PCR step (see Materials and Methods). MB-RED
`is an MB that detects any appropriate PCR product, whether it is WT
`or mutant at the queried codons. MB-GREEN is an MB that
`preferentially detects the WT PCR product.
`
`a PCR product, whether that product is WT or mutant in the
`region tested by MB-GREEN. Both MB probes are used together
`to detect the presence of a PCR product and its mutational status
`simultaneously.
`Practical Considerations. Numerous conditions were opti-
`mized to define conditions that could be reproducibly and
`
`Page 2 of 6
`
`The Johns Hopkins University Exhibit JHU2005 - 2 of 6
`
`

`

`9238
`
`Genetics: Vogelstein and Kinzler
`
`Proc. Natl. Acad. Sci. USA 96 (1999)
`
`generally applied to Dig-PCR-based projects. As outlined in Fig.
`1A, the first step of Dig-PCR involves PCR-amplification from
`single template molecules. Most protocols for amplification from
`small numbers of template molecules use a nesting procedure,
`wherein a PCR product resulting from one set of primers is used
`as template in a second PCR employing internal primers (42, 43).
`Because many applications of Dig-PCR are expected to require
`hundreds or thousands of separate amplifications, such nesting
`would be inconvenient and could lead to contamination prob-
`lems. Hence, conditions were sought that would achieve robust
`amplification without nesting. The most important of these
`conditions involved the use of a polymerase that was activated
`only after heating (44, 45) and of optimized concentrations of
`dNTPs, primers, buffer components, and temperature. The con-
`ditions specified in Materials and Methods were defined after
`individually optimizing each of these components and proved
`suitable for amplification of several different human genomic
`DNA sequences. Though the time required for PCR was not
`particularly long (’2.5 h), the number of cycles used was high and
`excessive compared with the number of cycles required to amplify
`the ‘‘average’’ single template molecule. The large cycle number
`was necessary because the template in some wells might not begin
`to be amplified until several PCR cycles had been completed. The
`large number of cycles ensured that every well (not simply the
`average well) would generate a substantial and roughly equal
`amount of PCR product if a template molecule were present
`within it.
`The second step in Dig-PCR involves the detection of these
`PCR products. It was necessary to modify the standard MB probe
`approach in order for it to function efficiently in Dig-PCR
`applications. Theoretically, one separate MB probe could be used
`to detect each specific mutation that might occur within the tested
`sequence. By inclusion of one MB corresponding to WT se-
`quence and another corresponding to mutant sequence, the
`nature of the PCR product would be identified. Though this
`strategy could obviously be used effectively in some situations, it
`becomes complex when several different mutations are expected
`to occur within the same tested sequence. For example, in the
`c-Ki-Ras gene example explored here, 12 different base substi-
`tutions resulting in missense mutations could theoretically occur
`within codons 12 and 13, and at least 7 of these are observed in
`naturally occurring human cancers. To detect all 12 mutations as
`well as the WT sequence with individual MBs would require 13
`different probes. Inclusion of such a large number of MB probes
`would raise the background fluorescence and the cost of the
`assay. We therefore attempted to develop a single probe that
`would react with WT sequences better than any mutant sequence
`
`within the tested sequence. We found that the length of the loop
`sequence, its melting temperature, and the length and sequence
`of the stem were each important in determining the efficacy of
`such probes. Loops ranging from 14 to 26 bases and stems ranging
`from 4 to 6 bases, as well as numerous sequence variations of both
`stems and loops, were tested during the optimization procedure.
`For discrimination between WT and mutant sequences (MB-
`GREEN probe), we found that a 16-bp loop with a melting
`temperature of 50–51° and a 4-bp stem of sequence 59-CACG-39
`were optimal. For MB-RED probes, the same stem with a 19- to
`20-bp loop with a melting temperature of 54–56° proved optimal.
`The differences in the loop sizes and melting temperatures
`between MB-GREEN and MB-RED probes reflected the fact
`that only the GREEN probe is designed to discriminate between
`closely related sequences, with a shorter region of homology
`facilitating such discrimination.
`Examples of the ratios obtained in replicate wells containing
`DNA templates from colorectal tumor cells with mutations of
`c-Ki-Ras are shown in Fig. 2. In this experiment, 50 genome
`equivalents of DNA were added into each well before amplifica-
`tion. Each of six tested mutants yielded ratios of REDyGREEN
`fluorescence that were significantly in excess of the ratio obtained
`with DNA from normal cells (1.5 to 3.4 in the mutants, compared
`with 1.0 in normal DNA; P , 0.0001 in each case, Student’s t test).
`The reproducibility of the ratios can be observed in Fig. 2. Direct
`DNA sequencing of the PCR products used for fluorescence
`analysis showed that the REDyGREEN ratios depended on the
`relative fraction of mutant genes within the template population
`(Fig. 2). Thus, the DNA from cells containing one mutant c-Ki-Ras
`allele per every two WT c-Ki-Ras alleles yielded a REDyGREEN
`ratio of 1.5 (Gly12Arg mutation), whereas the cells containing
`three mutant c-Ki-Ras alleles per WT allele had a ratio of 3.4
`(Gly12Asp). These data suggested that wells containing only
`mutant alleles (no WT) would yield ratios in excess of 3.0, with the
`exact value dependent on the specific mutation.
`Fluorescent probes such as those of the MB type are generally
`included in the PCR mix and followed in real time. Though this
`mode is the most convenient for many applications, we found it
`useful to add the MB probes after the PCR amplification was
`complete (Fig. 1). This procedure allowed us to use a standard
`multiwell plate fluorometer to analyze sequentially a large num-
`ber of multiwell plates containing preformed PCR products and
`bypassed the requirement for multiple real-time PCR instru-
`ments. Additionally, we found that the fluorescent signals ob-
`tained could be considerably enhanced if several cycles of asym-
`metric, linear amplification were performed in the presence of the
`MB probes. Asymmetric amplification was achieved by including
`
`FIG. 2. Discrimination between WT and mu-
`tant PCR products by MBs. Separate PCR prod-
`ucts (n 5 10), each generated from ’50 genome
`equivalents of DNA of cells containing the indi-
`cated mutations of c-Ki-Ras, were analyzed with
`the MB probes described above. Representative
`examples of the PCR products used for MB anal-
`ysis were purified and sequenced directly. In the
`cases with Gly12Cys and Gly12Arg mutations,
`contaminating nonneoplastic cells within the tu-
`mor presumably accounted for the relatively low
`ratios. In the cases with Gly12Ser and Gly12Asp,
`there were apparently two or more alleles of
`mutant c-Ki-Ras for every WT allele; both these
`tumors were aneuploid.
`
`Page 3 of 6
`
`The Johns Hopkins University Exhibit JHU2005 - 3 of 6
`
`

`

`Genetics: Vogelstein and Kinzler
`
`Proc. Natl. Acad. Sci. USA 96 (1999)
`
`9239
`
`an excess of a single internal primer (primer INT in Fig. 1C) at
`the time of addition of the MB probes.
`Dig-PCR on DNA from Tumor Cells. The principles and
`practical considerations described above were illustrated with
`DNA from two colorectal cancer cell lines, one with a mutation
`in c-Ki-Ras codon 12 and the other in codon 13. Representative
`examples of the MB-RED fluorescence values obtained are
`shown in Fig. 3. There was a clear biphasic distribution, with
`‘‘positive’’ wells yielding values in excess of 10,000 SFU and
`‘‘negative’’ wells yielding values less than 3,500 SFU. Gel elec-
`trophoreses of 127 such wells indicated that all positive wells but
`no negative wells contained PCR products of the expected size
`(Fig. 3). The REDyGREEN fluorescence ratios of the positive
`wells are shown in Fig. 4. Again, a biphasic distribution was
`observed. In the experiment with the tumor containing a
`Gly12Asp mutation, 64% of the positive wells had REDy
`GREEN ratios in excess of 3.0, whereas the other 36% of the
`positive wells had ratios ranging from 0.8 to 1.1. In the case of the
`tumor with the Gly13Asp mutation, 54% of the positive wells had
`REDyGREEN ratios .3.0, whereas the other positive wells
`yielded ratios ranging from 0.9 to 1.1. The PCR products from 16
`positive wells were used as sequencing templates (Fig. 4). All the
`wells yielding a ratio in excess of 3.0 were found to contain mutant
`c-Ki-Ras fragments of the expected sequence, whereas WT
`sequence was found in the other PCR products. The presence of
`homogeneous WT or mutant sequence confirmed that the am-
`plification products usually were derived from single template
`molecules. The ratios of WT to mutant PCR products deter-
`mined from the Dig-PCR assay were also consistent with the
`fraction of mutant alleles inferred from direct sequence analysis
`of genomic DNA from the two tumor lines (Fig. 2).
`Dig-PCR on DNA from Stool. As a more practical example of
`the intended use of Dig-PCR, we analyzed the DNA from stool
`specimens of patients with colorectal cancer. A representative
`result of such an experiment is illustrated in Fig. 5. From previous
`analyses of stool specimens from patients whose tumors con-
`tained c-Ki-Ras gene mutations, we expected that 1–10% of the
`c-Ki-Ras genes purified from stool would be mutant. We there-
`fore set up a 384-well Dig-PCR experiment. As positive controls,
`48 of the wells contained 25 genome equivalents of DNA (defined
`in Materials and Methods) from normal cells. Another 48 wells
`
`FIG. 3. Detecting Dig-PCR products with MB-RED. SFU of
`representative wells from an experiment employing colorectal cancer
`cells with Gly12Asp or Gly13Asp mutations of the c-Ki-Ras gene.
`Wells with values .10,000 SFU are shaded yellow. PAGE analyses of
`the PCR products from selected wells are shown. Wells with fluores-
`cence values ,3,500 SFU had no PCR product of the correct size,
`whereas wells with fluorescence values .10,000 SFU always contained
`PCR products of 129 bp. Nonspecific products generated during the
`large number of cycles required for Dig-PCR did not affect the
`fluorescence analysis. M1 and M2 are molecular length markers used
`to determine the size of fragments (indicated on the left in base pairs).
`
`served as negative controls (no DNA template added). The other
`288 wells contained an appropriate dilution of stool DNA.
`MB-RED fluorescence indicated that 102 of these 288 experi-
`mental wells contained PCR products (mean 6 SD of 47,000 6
`18,000 SFU), whereas the other 186 wells did not (2,600 6 1,500
`SFU). The REDyGREEN ratios of the 102 positive wells sug-
`gested that five contained mutant c-Ki-Ras genes, with ratios
`ranging from 2.1 to 5.1. The other 97 wells had ratios ranging from
`0.7 to 1.2, identical to those observed in the positive-control wells.
`To determine the nature of the mutant c-Ki-Ras genes from stool
`in the five positive wells, the PCR products were sequenced
`directly. The four wells with REDyGREEN ratios in excess of 3.0
`were completely composed of mutant ras sequence (Fig. 5). The
`sequence of three of these PCR products indicated Gly12Ala
`mutations (GGT to GCT at codon 12), whereas the sequence of
`the fourth indicated a silent C-to-T transition at the third position
`of codon 13. This transition presumably resulted from a PCR
`error during the first productive cycle of amplification from a WT
`template. The well with a ratio of 2.1 contained an ’1:1 mix of
`WT and Gly12Ala mutant sequences. Thus 3.9% (4 of 102) of the
`c-Ki-Ras alleles present in this stool sample contained a Gly12Ala
`mutation. The mutant alleles in the stool presumably arose from
`the colorectal cancer of the patient, as direct sequencing of PCR
`products generated from DNA of the cancer identified the
`identical Gly12Ala mutation (not shown).
`
`DISCUSSION
`Dig-PCR represents another example of the power of PCR, in
`combination with more recently developed detection technol-
`ogies, to provide opportunities for genetic analysis. There are
`several precedents for the approach described here. For ex-
`ample, PCR-amplification from single cells isolated by physical
`separation or dilution has been used to address a variety of
`interesting biologic questions (46–49). Gel electrophoretic and
`sequence analysis of single alleles, produced by amplification
`of diluted DNA or from cloning of PCR products, has also
`proven useful in several areas of investigation (43, 48, 50–53).
`In situ amplification of single alleles by using rolling-circle
`amplification represents another exciting strategy for extract-
`ing genetic data that would be impossible to obtain from more
`standard analyses of bulk DNA populations (54).
`Dig-PCR can be used to detect mutations present at rela-
`tively low levels in the samples to be analyzed. The limit of
`detection is defined by the number of wells that can be
`analyzed and the intrinsic mutation rate of the polymerase
`used for amplification. The 384-well PCR plates are commer-
`cially available, and 1,536-well plates are on the horizon,
`theoretically allowing sensitivities for mutation detection at
`the ’0.1% level. It is also possible that Dig-PCR can be
`performed in microarray format, potentially increasing the
`sensitivity by another order of magnitude. This sensitivity may
`ultimately be limited by polymerase errors. The effective error
`rate in PCR as performed under our conditions was ,0.3%,
`i.e., in control experiments with DNA from normal cells, none
`of 340 wells containing PCR products had REDyGREEN
`ratios .3.0. Any individual mutation (such as a G-to-C
`transversion at the second position of codon 12 of c-Ki-Ras) is
`expected to occur in ,1 in 50 polymerase-generated mutants
`(there are at least 50 base substitutions within or surrounding
`codons 12 and 13 that should yield high REDyGREEN ratios).
`Determining the sequence of the putative mutants in the
`positive wells, by direct sequencing as performed here or by
`any of the other techniques described in the Introduction,
`provides unequivocal validation of a prospective mutation; a
`significant fraction of the mutations found in individual wells
`should be identical if the mutation occurred in vivo. Signifi-
`cance can be established through rigorous statistical analysis,
`as positive signals should be distributed according to Poisson
`probabilities. Moreover, the error rate in particular Dig-PCR
`
`Page 4 of 6
`
`The Johns Hopkins University Exhibit JHU2005 - 4 of 6
`
`

`

`9240
`
`Genetics: Vogelstein and Kinzler
`
`Proc. Natl. Acad. Sci. USA 96 (1999)
`
`FIG. 4. Discriminating WT from mutant PCR
`products obtained in Dig-PCR. REDyGREEN
`ratios were determined from the fluorescence of
`MB-RED and MB-GREEN as described in Ma-
`terials and Methods. The wells shown are the same
`as those illustrated in Fig. 3. The sequences of
`PCR products from the indicated wells were de-
`termined as described in Materials and Methods.
`The wells with REDyGREEN ratios .3.0 each
`contained mutant sequences, whereas those with
`REDyGREEN ratios of ’1.0 contained WT se-
`quences.
`
`experiments can be determined precisely through perfor-
`mance of Dig-PCR on DNA templates from normal cells.
`Dig-PCR is as easily applied to RT-PCR products generated
`from RNA templates as it is to genomic DNA. For example, the
`fraction of alternatively spliced or mutant transcripts from a gene
`could be determined easily by using fluorescent probes specific
`for each of the PCR products generated. Similarly, Dig-PCR
`could be used to quantitate relative levels of gene expression
`within an RNA population. For this amplification, each well
`
`would contain primers that are used to amplify a reference
`transcript expressed constitutively as well as primers specific for
`the experimental transcript. One fluorescent probe would then be
`used to detect PCR products from the reference transcript, and
`a second fluorescent probe would be used for the test transcript.
`The number of wells in which the test transcript is amplified
`divided by the number of wells in which the reference transcript
`is amplified provides a quantitative measure of gene expression.
`Another group of examples involves the investigations of allelic
`
`FIG. 5. Dig-PCR of DNA from a stool sample. The 384
`wells used in the experiment are displayed. Those colored
`blue contained 25 genome equivalents of DNA from nor-
`mal cells. Each of these registered positive with MB-RED,
`and the REDyGREEN ratios were 1.0 6 0.1 (mean 61
`SD). The wells colored yellow contained no template DNA,
`and each was negative with MB-RED (i.e., fluorescence
`,3,500 SFU.). The other 288 wells contained diluted DNA
`from the stool sample, prepared by alkaline extraction (57).
`Those registering positive with MB-RED were colored
`either red or green, depending on their REDyGREEN
`ratios. Those registering negative with MB-RED were
`colored white. PCR products from the indicated wells were
`used for automated sequence analysis.
`
`Page 5 of 6
`
`The Johns Hopkins University Exhibit JHU2005 - 5 of 6
`
`

`

`Genetics: Vogelstein and Kinzler
`
`Proc. Natl. Acad. Sci. USA 96 (1999)
`
`9241
`
`Table 1. Potential applications of Dig-PCR
`Application
`Base substitution mutations
`
`Determine presence or extent of amplification
`
`Probe 2 detects
`Probe 1 detects
`Mutant or WT alleles WT PCR products
`
`Translocated allele
`
`Sequence from another part
`of same chromosome arm
`Common exons
`
`First transcript
`
`Reference transcript
`
`First mutation
`
`Second mutation
`
`Marker from test
`chromosome
`
`Marker from reference
`chromosome
`
`Example
`Cancer gene mutations in stool, blood, and lymph
`nodes
`Residual leukemia cells after therapy (DNA or RNA) Normal or translocated
`alleles
`Sequence within
`amplicon
`Minor exons
`
`Chromosomal translocations
`
`Gene amplifications
`
`Changes in gene expression
`
`Allelic discrimination
`
`Allelic imbalance
`
`Alternatively spliced products Determine fraction of alternatively spliced transcripts
`from same gene (RNA)
`Determine relative levels of expression of two genes
`(RNA)
`Two different alleles mutated vs. one mutation in each
`of two alleles
`Quantitative analysis with nonpolymorphic markers
`
`status when two mutations are observed in the sequence analysis
`of a standard DNA sample. To distinguish whether one variant is
`present in each allele (vs. both occurring in one allele), cloning of
`PCR products generally is performed. The approach described
`here would simplify the analysis by eliminating the need for
`cloning. Other potential applications of Dig-PCR are listed in
`Table 1. When the goal is the quantitation of the proportion of
`two relatively common alleles or transcripts rather than the
`detection of rare alleles, techniques such as those employing
`TaqMan and real time PCR (31, 33–38, 40) provide an excellent
`alternative to Dig-PCR. Advantages of real time PCR methods
`include their simplicity and the ability to analyze multiple samples
`simultaneously. However, Dig-PCR may prove useful for these
`applications when the expected differences are small (e.g., only
`’2-fold, as with allelic imbalances; ref. 55).
`The ultimate utility of Dig-PCR lies in its ability to convert
`the intrinsically exponential nature of PCR to a linear one. It
`should thereby prove useful for experiments requiring the
`investigation of individual alleles, rare variantsymutations, or
`quantitative analysis of PCR products.
`We thank the members of our laboratory for advice and support; J. Jen
`for stool and tumor samples; D. Sidransky, J. Jen, and other colleagues
`for critical reading of the manuscript; and staff of the DNA Analysis
`Facility of the Johns Hopkins University Genetic Resources Core Facility
`for DNA sequencing. This work was supported by National Institutes of
`Health Grants CA 43460, CA 57345, and CA 62924.
`
`1. Vogelstein, B. V. & Kinzler, K. W. (1998) The Genetic Basis of Human Cancer
`(McGraw–Hill, Toronto).
`2. Lee, C. M., Weindruch, R. & Aiken, J. M. (1997) Free Radical Biol. Med. 22,
`1259–1269.
`3. Ozawa, T. (1997) Physiol. Rev. 77, 425–464.
`4. Sidransky, D., Von Eschenbach, A., Tsai, Y. C., Jones, P., Summerhayes, I.,
`Marshall, F., Paul, M., Green, P., Hamilton, S. R., Frost, P., et al. (1991)
`Science 252, 706–709.
`5. Sidransky, D., Tokino, T., Hamilton, S. R., Kinzler, K. W., Levin, B., Frost,
`P. & Vogelstein, B. (1992) Science 256, 102–105.
`6. Smith-Ravin, J., England, J., Talbot, I. C. & Bodmer, W. (1995) Gut 36, 81–86.
`7. Mills, N. E., Fishman, C. L., Scholes, J., Anderson, S. E., Rom, W. N. &
`Jacobson, D. R. (1995) J. Natl. Cancer Inst. 87, 1056–1060.
`8. Mao, L., Hruban, R