WO2008005559A2 - A strategy for detecting low abundance mutations - Google Patents

Abstract

The present invention relates, e.g., to a method for screening for mutations in a DNA sample, comprising (a) diluting and distributing the sample to obtain a plurality of aliquots which, on the average, contain between about 2-10 genome equivalents of the DNA; (b) PCR amplifying the DNA in a sufficient number of aliquots, with at least one set of PCR primers, so as to generate amplicons such that, if mutations are present at a frequency of about 1-10% in a given gene in the sample, the probability of there being a mutation in that gene in at least one of the amplicons is at least about 95%; and (c) screening the amplicons for the presence of mutations, using a method that can detect unspecified mutations at unspecified sites within an amplicon (e.g., temperature gradient capillary electrophoresis (TGCE) or cycle sequencing of the DNA).

Description

A STRATEGY FOR DETECTING LOW ABUNDANCE MUTATIONS

This research was supported by a U.S. government grant from NCI SPORE, number CA 62924. The government thus has certain rights in the invention. This application claims the benefit of the filing date of U.S. provisional application

60/819,440, filed July 7, 2006, which is incorporated by reference herein in its entirety.

FIELD OF THE INVENTION

This invention relates, e.g. , to a method for detecting the presence of one or more mutations, particularly low abundance mutations, in one or more genes of interest. The method comprises the use of limit dilution PCR (LD-PCR), coupled with a method that can detect a plurality of mutations, e.g. at undefined sites in the DNA, such as temperature gradient capillary electrophoresis (TGCE) or cycle sequencing.

BACKGROUND INFORMATION The detection of somatic mutations for cancer diagnostics and other applications (e.g. as a marker for ageing) has gained increasing importance in the last several decades. Thus far, the detection of a small number of mutant-containing cells among a large excess of normal (wild type) cells has proven difficult. Moreover, when DNA samples are obtained from secondary fluids (such as pancreatic duct juice, serum/plasma, etc.) rather than primary samples (such as tumor biopsies), mutant DNA from a particular gene may be present in only about 1-10% of the DNA assayed, thus increasing the challenge. ^•

Many of the current analytic procedures for detecting mutations are able to detect only a single, well-defined mutation and/or mutations at a well-defined site, and/or lack sufficient sensitivity to detect mutations that are present in low abundance (in a minor proportion of the cells, or at low abundance in a DNA preparation). There is a need for methods for accurately detecting low abundance genetic mutations, particularly undefined mutations at undefined sites in the DNA, in mixed populations of sequences.

Of particular interest is the detection/diagnosis of pancreatic cancer. Pancreatic ductal adenocarcinoma is the fourth leading cause of cancer death in the USA and has the lowest survival rate for any solid cancer (about 2%). This poor survival occurs in part because only about 15% of patients are diagnosed with pancreatic cancer while they have surgically resectable disease. Pancreatic cancer survival is better for patients with the smallest tumors. Such a poor survival is particularly of concern to patients with inherited susceptibility to the disease. Screening asymptomatic individuals with a significant risk of developing pancreatic cancer has demonstrated that pre-invasive pancreatic neoplasms can be detected by endoscopic ultrasound in some individuals. These results suggest that some form of clinical screening for high-risk individuals will eventually become an important part of their management.

When a pancreatic lesion is found, fine needle aspiration cytology may provide a diagnosis, but often it is difficult to determine if a lesion is neoplastic. Hence, many investigators have sought

• more accurate markers of pancreatic neoplasia, which could help differentiate pancreatic cancers from chronic pancreatitis and help identify pre-invasive pancreatic neoplasms such as intraductal papillary mucinous neoplasms (EPMNs) and pancreatic intraepithelial neoplasia (PanIN).

Pancreatic neoplasms evolve with many genetic and epigenetic alterations, but few such alterations are useful diagnostic markers. Genetic alterations of pancreatic neoplasia include oncogene (K-ras, BRAF) and suppressor gene mutations (p!6, p53, SMAD4, BRCA2, STKlJ, hMLHl, hCDC4, MKK4, and FancC). Mitochondrial mutations and microsatellite instability also occur in pancreatic cancers as do many gene expression changes. DNA methylation markers are currently undergoing evaluation as markers of pancreatic neoplasia.

The secondary fluid, pancreatic juice, is readily obtainable during endoscopic investigation. Of the noted genetic alterations, mutant K-ras is most readily detectable in such secondary fluids. However_^K-ra? mutations are not specific for invasive pancreatic cancer; they are also found in the pancreatic juice and in the stool of patients with chronic pancreatitis, individuals who smoke, and in

PanINs. Quantifying mutant K-ras levels in pancreatic juice helps to distinguish patients with pancreatic cancer from those with pancreatitis, but is not sufficient for pancreatic cancer diagnosis.

Mutations in the pi 6 and p53 genes are the most specific mutations for pancreatic cancer.

However, such mutations are difficult to identify in secondary fluids, using current methodology. The p53 gene is mutated in ~75% of pancreatic cancers. Mutations in the p53 gene are located throughout the gene (e.g. at any of several hundred nucleotides, within exons 5, 6, 7 and/or 8). Some current strategies can detect only specific mutations, or mutations at a specified nucleotide site in a DNA. Therefore, such methods are not suitable for identifying one of the myriad mutations that might be expected in patients with pancreatic cancer. Furthermore, among patients with pancreatic cancer, mutations seen in pancreatic juice are generally present at low concentration (about 1-10%). Although some methods currently exist which can detect a wider range of mutations, these methods can only detect mutant DNA when present at concentrations of about 5-10% or higher of the total DNA in a sample, so many low abundance mutations are missed using such methods of analysis.

Approximately 40% of pancreatic ductal adenocarcinomas harbor point mutations in exons 1 and 2 of pi 6. Another 40% of pancreatic cancers harbor homozygous deletions, and in most remaining cancers, pi 6 is inactivated by promoter methylation. Again, because pi 6 mutations are located through these two mentioned exons and are present in low abundance in secondary fluids, such mutations are not readily detected in secondary fluids, such as pancreatic duct juice, using currently available methods.

The detection in clinical samples of pi 6 andp53 mutations could facilitate pancreatic cancer detection, particularly early detection of this disease, if suitable assays were available to detect these mutations at the low concentrations often found in clinical samples.

DESCRIPTION OF THE DRAWINGS

Figure 1 shows a schematic of LD-PCR. PCR on diluted pancreatic DNA samples followed by heteroduplex analysis using TGCE discriminates between wild-type and mutant alleles.

Figures 2A-2D show examples of heteroduplex analysis of pl6 exon 2. Fig.2 A (top left) shows a heteroduplex pattern indicative of a mutation in exon 2 of pi 6 in the pancreatic juice DNA of a patient with pancreatic cancer (PJ-8 from patient No.8). Fig.2B (bottom left) shows a homoduplex pattern indicative of normal pl6 in pancreatic juice from a subject without a detectable pl6 exon 2 mutation in their cancer Fig. 2C (top right) shows a mutation in exon 2 of pi 6 in the patient's corresponding primary pancreatic cancer (patient No. 8). Fig.2D shows a normal pi 6 heteroduplex pattern from the pancreatic cancer DNA of a patient without a detectable pi 6 mutation.

DETAILED DESCRIPTION

This invention relates, e.g., to a method for detecting mutations that are at low abundance in a DNA sample. The method does not rely on the detection of specific mutations, or of mutations at specific nucleotide sites within a DNA molecule. The assay strategy involves PCR (polymerase chain reaction) amplification of DNA at limiting dilution (LD-PCR), followed by screening the PCR products for mutations, using a method that can identify an undefined mutation (rather than a specific mutation, or a mutation at a specific site). Suitable procedures for analyzing the amplified DNA (amplicons) for mutations include temperature gradient capillary electrophoresis (TGCE) or DNA sequence analysis (e.g. cycle sequencing).

A limiting dilution of DNA can be used to identify rare species of mutant PCR amplicons admixed with wild type DNA. The limiting dilution PCR strategy involves analyzing DNA samples for mutations by performing many PCR reactions on the sample, with each PCR amplifying only a few DNA templates. That is, DNA from a sample is diluted and distributed in a large number of aliquots, such that only a few DNA templates are present in each aliquot; and the DNA in each aliquot is amplified by PCR with primer sets that are specific for one or more segments of the DNA of interest (segments that are expected to contain mutants). In samples with low concentrations of mutant molecules, mutant DNA will be present in only a few of the amplicons, but in these amplicons, the mutant DNA will be at sufficient concentration to be detected by a detection method of the invention (e.g., TGCE or cycle sequencing). As a result, the majority wild-type DNA in the sample will no longer obscure the few mutant DNA molecules. For example, if one analyzes a sample of pancreatic juice DNA containing 3% mutant pi 6 DNA molecules, by diluting it and distributing the diluted DNA to 45 aliquots, each of which contains about 4 amplifiable DNA molecules; performing PCR amplifications on the DNA in the 45 aliquots; and then analyzing the resulting amplicons; one would screen 180 input DNA molecules (45 x 4) and expect to detect mutants in about 5 of the amplicons (3% of 180). The ratio of mutant to wild-type templates will be high in those aliquots containing mutants (on average, 1 mutant for every about 3 wild-type molecules). Currently, the most sensitive assays for detecting DNA alterations rely on PCR based strategies that can detect mutations present at 1/1000 or less of the concentration of wild-type DNA. However, the specificity of these assays is inherent to the specificity of PCR primers, and as a result DNA assays that have been created to identify mutations in clinical specimens rely on the detection of a specific mutation, or a mutation at a specific nucleotide site in a DNA molecule. These methods are not amenable to detecting mutations that can arise throughout a gene. For example, as noted above, the p53 gene can be mutated in a cancer at several hundred nucleotides. Current mutation detection strategies used to detect p53 mutations (such as single strand conformational polymorphism, p53 chip technologies and other strategies) can only detect mutant DNA present at a concentration of about 5-10% or higher of the total DNA concentration in a sample. Unfortunately, the concentration of mutant DNA in most clinical specimens is lower than can be reliably detected using these techniques. For example, mutations in pancreatic juice are generally present at low concentration (about 1-10%) among patients with pancreatic cancer. Methods of the present invention improve the lower limit of detection of low abundance mutations, such as p53 and/or pl6 mutations in pancreatic juice.

In one embodiment of the invention, temperature gradient capillary electrophoresis (TGCE) is used to analyze PCR products of limit dilution. This method is shown diagrammatically in Figure 1. By way of illustration, consider the screening of pancreatic juice from a subject having or suspected of having pancreatic cancer, for p53 and/oτ pi 6 mutations. In this illustration, one screens multiple PCR products (e.g., 24-48) amplified from, pancreatic juice DNA, using a limiting amount of DNA (e.g., about 6 molecules) for each PCR reaction. TGCE detects heteroduplexes created from mixes of mutant to wild-type DNA and can detect the majority of mutations in a PCR fragment as long as there is sufficient concentration of mutant DNA in a given sample (usually 10-50% of total DNA). To screen a DNA sample obtained from pancreatic juice by this method, one quantifies the DNA accurately, using a conventional procedure, then performs 48 PCR reactions with each PCR containing 6 molecules of DNA. Each PCR is then subjected to TGCE. As a result, 48 x 6 (288) molecules of DNA are screened for mutations. Since each PCR begins with 6 molecules of DNA, if a mutation is present in a well, the lowest ratio of mutant to wild-type DNA is 1/6 and this is within the lower limit of sensitivity of TGCE to detect heteroduplexes. Example IV illustrates the detection ofp53 and pi 6 mutations in the pancreatic juice of patients with pancreatic cancer, using limiting dilution TCGE.

In another embodiment of the invention, mutations in the limit dilution PCR products are detected by cycle sequencing rather than by heteroduplex detection. By way of illustration, such a procedure is accomplished by PCR amplifying DNA molecules at about 10 picogram amounts (comparable to about 3 molecules of human genomic DNA, measured using a conventional procedure, such as real time PCR). Most of the PCR wells (aliquots) contain 2-4 such DNA molecules, and when one of the molecules contains mutant DNA, the ratio of wild-type to mutant molecules enables the mutation to be detected by cycle sequencing in most cases. Similar to the limiting dilution strategy described above, performing 93 PCRs that contain about 3 molecules per well, about 279 DNA molecules are screened for mutations. Therefore, if one analyzes a clinical sample such as pancreatic juice for mutations ιnp53 (exons 5-8) andplό (exons 1 and 2), one can run four 96 well plates, and by sampling about 279 DNA molecules, there is about a 95% probability of sampling at least one mutant DNA molecule when the mutant to wild-type ratio of DNA is 1/100. This embodiment of the invention is described in Example VII. A variation of this embodiment, which might might increase the ability to detect mutations, would be to sequence 2 molecule instead of 3 molecule PCRs. More PCR products would need to be screened and more individual PCR wells would fail because of lack of input DNA.

Advantages of a method of the invention include that it is rapid, economical, simple, robust, highly sensitive, accurate, and does not require expertise in microdissection of cytologically suspicious cells. A method of the invention can detect mutations which are in low abundance (e.g., somatic mutations that are present in only about 1 -10% of the cells in a population, or that are only present in about 1-10% of the total DNA in a sample, such as in a clinical sample from a cancer patient); can detect unknown mutations (e.g., multiple mutations) at unspecified sites within a defined segment of DNA (or RNA that has been reverse transcribed) in a gene, e.g. at a hot spot for mutations; can simultaneously detect mutations in a variety of segments (e.g., exons) within a given gene of interest, or at sites within different genes; can be applied to small tissue samples (e.g. biopsies, fluids (such as pancreatic juice), or tumor samples); and/or can be readily miniaturized and/or adapted for use in a high throughput format. A method of the invention can be used for a wide variety of applications, including diagnostic methods (e.g. to detect cancer, such as pancreatic cancer); research purposes (e.g. to study factors involved in ageing, the pathogenesis of neoplasia or metastasis, etc.); or to identify mutations that are correlated with a disease or condition of interest, such as a cancer.

One aspect of the invention is a method of screening for mutations (variants) in a DNA sample, comprising a) diluting the sample and distributing the diluted sample to obtain a plurality of aliquots which, on the average, contain between about 2-10 genome equivalents of the DNA (e.g., of a mammalian DNA); b) PCR amplifying the DNA in a sufficient number of aliquots, with at least one set of PCR primers, so as to generate amplicons such that, if mutations are present at a frequency of about 1-

10% in a given gene in the sample, the probability of there being a mutation in that gene in at least one of the amplicons is at least about 95%, and c) screening the amplicons for the presence of mutations, using a method that can detect unspecified mutations at unspecified sites within an amplicon. A "genome equivalent," as used herein, is an amount of DNA that includes one copy of each allele in the genome.

In embodiments of the invention, each of the aliquots may contain between about 3 and 8 genome equivalents of DNA (e.g., between about 6 and 7 genome equivalents). All ranges used herein include the end points of the range. The term "about," as used herein, means plus or minus 10%. For example, about 8 genome equivalents includes 7, 8 or 9 genome equivalents. When analyzing amplicons by a method of the invention using a heteroduplex procedure, it is necessary to have at least two copies of a gene of interest (at least two genome equivalents) in an aliquot, in order to provide wild type DNA (e.g. from a second allele) for the formation of a heteroduplex.

The PCR step is carried out by selecting sets of PCR primers which flank segments of interest in a gene (e.g., exons which are thought or known to contain a plurality of mutations in a cancer of interest), hi one embodiment, in which it is desirable to identify previously unknown mutations present in a gene of interest, a series of adjacent regions of the gene of interest can be independently amplified by suitable PCR primers. In another embodiment, in which it is desirable to screen for the presence of low-abundance splice variants, mRNA is converted to DNA by RT-PCR (reverse transcriptase PCR), and PCR primers are selected to flank the splice site of an alternatively spliced variant. In this manner, different sized splice variants will be observed following heteroduplex analysis. A skilled worker will recognize suitable primer pairs to be used in a method of the invention.

PCR primers are selected which will amplify a segment of a nucleic acid of a desired size. Typical amplicons range in size from between about 100 and 1,000 bases pairs (bp), e.g. between about 200 and 800 bp. For amplicons that are shorter than about 200 bp, it can be difficult to detect a heteroduplex. And one cannot readily resolve differences in PCR products that are larger than about 800 bp in a capillary gel. If cycle sequencing is to be carried out, the optimal length of an amplicon to be analyzed is between about 10 and 800 bp. Improvements in gel chemistry in the future will likely allow larger PCR products to be screened for heteroduplexes. To a certain extent the length limitation is also determined by the ratio of deoxy to dideoxy nucleotides in the sequencing chemistry: too high a didoxy to deoxy nucleotide ratio results in predominantly short sequencing products, too low a ratio in mostly long sequencing products. Because of this sequencing chemistry, the accuracy of sequencing of the first about 25-50 base pairs of any PCR product is also typically reduced and the sequencing of a PCR product utilizes a sequencing primer at one end of the amplicon. For this reason, PCR products of at least 100 base pairs in length are generally sequenced.

The number of aliquots for PCR amplification can be determined empirically, depending on variety of factors, including the frequency of mutations, the number of genome equivalent templates in each aliquot, the number of sets of PCR primers, etc. In embodiments of the invention, the number of aliquots that are subjected to PCR analysis is determined by how low a concentration of templates containing mutant molecules one is trying to detect. A typical range would be about 40 to 200 aliquots for each PCR primer set, with each aliquot containing about 2 to 8 molecules. A skilled worker can readily determine the optimal limiting dilution PCR strategy for mutation detection. For example, when analyzing a pancreatic juice sample with 1 % mutant DNA, to have a 95% probability of sampling a mutant DNA molecule in the assay, 300 DNA molecules need to be amplified and screened for mutation. This can be achieved by screening 43 PCRs (from 43 aliquots) for heteroduplexes, with each PCR containing about 7 input DNA molecules. Screening more PCRs in this fashion increases the assay's limit of detection.

Increasing numbers of aliquots may be used as increasing numbers of PCR primer pairs are used in each PCR reaction. For example, one can amplify simultaneously several regions within a gene of interest (e.g., the four exons of p53 in which mutations are most often detected); regions of several different genes, such as p53 and pi 6; or other variations that will be evident to a skilled worker. For example, at least about 2, 4, 6, 8, 10, or more, sets (pairs) of primers can be used. Many aliquots (e.g., 1,000 or more) can be used in a method of the invention.

Methods for designing PCR primers and for carrying out PCR reactions, including reaction conditions, such as the presence of salts, buffers, ATP, dNTPs, etc. and the times and temperature of incubation, are conventional and can be optimized readily by one of skill in the art. See, e.g., Innis et αl., editors, PCR Protocols (Academic Press, New York, 1990); McPherson et αl., editors, PCR: A Practical Approach, Volumes 1 and 2 (IRL Press, Oxford, 1991, 1995); Barany( 1991) PCRMethods and Applications JL, 5-16; Diffenbach et al., editors, PCR Primers, A Laboratory Manual (Cold Spring Harbor Press); etc. To reduce or eliminate mutations that arise from errors which occur during PCR, a high fidelity DNA polymerase can be used. See, e.g., the discussion in Example VI. Any of a variety of reaction chambers (e.g., containers, wells of a plate, etc.) can be used, in any of a variety of formats. For example, one can employ 96 well PCR plates, 384 well PCR plates, 1536 well PCR plates, etc. Containers can be closed to form a leak-proof seal, in order to reduce or prevent cross-contamination of samples. Suitable formats for performing PCR reactions include computer-controlled thermal cyclers. In one embodiment of the invention, the detection of mutations in the amplified DNAs

(amplicons) is performed by temperature gradient capillary electrophoresis (TGCE). This method relies on the detection of heteroduplexes of DNA which form when a PCR product that contains templates of more than one sequence is re-annealed over a temperature gradient. Under optimized conditions, TGCE identifies most mutations when the ratio of mutant to wild-type DNA is 1 (typically PCR products are about 200-800 base pairs). TCGE can often detect mutant templates present within PCR products at concentrations as little as 10% of the wild-type template, but most mutations are not detectable at lower concentrations. Methods for performing TCGE are conventional and will be evident to a skilled worker. See, e.g., the discussion in Example I-D.

Variations of this method include the performance of multiple different sized PCRs during the same TGCE run, which can reduce assay cost. The sensitivity of TGCE for detecting mutations can also be improved by using GC clamps as part of the PCR products, and by using fluorescently labeled primers.

In one embodiment of this method, amplicons that have resulted in the formation of a heteroduplex are sequenced, using a conventional procedure, to characterize more precisely the nature of the mutation that gave rise to the heteroduplex.

In another embodiment of the invention, the detection of mutations in the amplified DNAs (amplicons) is performed by cycle sequencing. One sequencing method is dye-terminator sequencing, in which a sequencing chemistry reaction that includes DNA polymerases and nucleotides also includes four dideoxynucleotide chain terminators, one for each nucleotide, which are labeled with different fluorescent dyes. Once the dideoxy labeled nucleotide is incorporated into a template sequence, additional nucleotides can no longer be added and the sequence can be detected by automated high-throughput DNA sequence analyzers that detect the fluorescently labeled template. After a sequencing reaction, templates are resuspended in buffer after a clean up step, typically precipitation, and loaded onto a capillary sequencer. High throughput capillary sequencers now have as many as 384 capillaries and can run multiple reactions (4-8 or more depending on the length of the PCR products being sequenced) each day. Thus, thousands of PCR products can be sequenced in one day at current costs of less than $1 per sequence. Chromatograms of sequencing data are managed by software packages that can compare the sequencing data to identify sequence variants.

Another technology that can be used to detect mutations in limiting dilution PCRs is dHPLC (denaturing high performance liquid chromatography). In contrast to heteroduplex detection, single- strand conformation polymorphism (SSCP) detects base changes in single-strand DNA, by subjecting samples to denaturing electrophoresis. SSCP detection can also be applied to capillary electrophoresis and it is possible to combine SSCP with heteroduplex analysis to improve the detection of mutations (Kozlowski et al. (2005) Electrophoresis 26, 71-81; Kozlowski etal. (2001) Nucleic Acids Res 29, E71).

Samples for analysis can be obtained from any suitable source. Suitable subjects from which cells, tissues or fluids can be isolated include eukaryotes, such as plants or invertebrate or vertebrate animals, e.g. mammals (including pets, farm animals, research animals, and primates, including humans). The samples maybe, e.g., tumor samples, biopsy samples or other tissues. Alternatively, the samples may be secondary fluids, e.g. fluids which may contain DNA products of cancer cells, {e.g., fluids into which DNA from cancer cells has leaked). The analysis of such secondary fluids is advantageous because it offers a non-invasive sampling method;,and is especially useful for the detection of a cancer in an inaccessible tissue. Secondary fluids (bodily fluids) include, e. g, blood or blood fractions (serum/plasma), urine, seminal fluid, bronchoaveolar lavage, saliva, gastric or colonic washes, etc. For example, urine, stool or sputum samples can be tested for the presence of mutations associated with cancers of the bladder, colorectum, and lung, respectively. Mutant sequences from the DNA of neoplastic cells have been found in the blood of cancer patients, so the detection of residual disease in lymph nodes or surgical margins may also be useful in predicting which patients might benefit most from further therapy.

For the detection of pancreatic cancer, one can assay, e.g., tumor tissue obtaining during resection, pancreatic duct juice (sometimes referred to herein as "pancreatic juice" or "juice") obtained during endoscopy, fine needle aspirates of tumor masses, brushings of the pancreatic duct, bile duct or aspirates of cyst fluid.

In view of the sensitivity of methods of the invention, samples can be taken from small primary tumors, allowing a diagnosis at a stage when the primary tumors are still curable and the patients asymptomatic. Samples for analysis can also be obtained from cultured cells (e.g., primary cells or cell lines of interest). In one embodiment of the invention, the sample is a cell-free lysate. A method of the invention can be used to find a tumor mutation in a population of cells which is not purely tumor cells.

Methods for isolating samples for analysis are conventional and well-known in the art. Although much of the discussion herein is directed to methods for detecting mutations in DNA, such as genomic DNA, a method of the invention can also be applied to the analysis of RNAs. RNA can be converted to cDNA and amplified, using conventional procedures, such as RT-PCR (reverse transcriptase PCR), and the resulting DNA assayed as described herein. The method can detect mutations is RNAs such as, e.g., mRNA (transcribed mutations in coding sequences, splice variants, etc.), tRNA, rRNA, microRNAs, etc. Methods of RT-PCR are conventional and well- known in the art.

A variety of types of mutations can be identified by a method of the invention. Although much of the discussion herein is directed to "mutations," it is to be understood that any type of variant DNA, including naturally occurring variants, e.g. allelic differences, SNPs, etc. can be detected by a method of the invention. A DNA mutation may differ from the wild type by a single base (point mutations, including transversions, transitions, base substitutions, etc.), two or more non-contiguous bases (including mutations that result in frame shifts), small or large deletions or insertions (e.g. deletions or insertions of between about 1-50, 1-25,1-10 bp, etc.), inversions, truncations, combinations thereof, etc. Chromosomal translocations (e.g., which are characteristic of leukemias or lymphomas) and gene amplifications can also be detected. The mutations can be in any type of nucleic acid, including, e.g., genomic cellular DNA, mitochondrial DNA (mtDNA), messenger RNA (mRNA), viral DNA or RNA genomes, etc.

A method of the invention may be used for a variety of applications. One aspect of the invention is a method for testing a subject for (diagnosing) pancreatic cancer, comprising screening a sample from the subject for mutations \np53 and/or pi 6 genes, using a method of the invention. An increase in the number of mutations in thsp53 and/or the pi 6 genes, compared to a baseline value, indicates that the subject has or is likely to have pancreatic cancer. The baseline value may be, e.g, a reference standard, or the number of such mutations in a subject known not to have pancreatic cancer (such as a "normal" control or a subject having chronic pancreatitis). Alternatively, an empirical cut-off value may be used, as described, e.g., in Example rv.

Such a method can be used in conjunction with other methods for diagnosing pancreatic cancer. For example, since not every pancreatic neoplasm has mutations in p53 or pi 6, the detection of these mutations can be combined with quantification of mutant K-rαs, aberrantly methylated DNA, telomerase activity, or the detection of LOH (loss of heterozygosity) in microdissected samples to facilitate pancreatic cancer diagnosis. Such other methods can be carried out before atest of the invention, as part of a preliminary screen.

Another aspect of the invention is a method for diagnosing cancers other than pancreatic cancer for which somatic mutations are diagnostic. For example, one can use a method of the invention to detect pi 6 and p53 mutations in lung or biliary tract cancers, or mutations in other genes, particularly when mutations are not limited to a few single nucleotide hotspots, such as mutations of the EGFR gene found in non-small cell lung carcinomas, or mutations in the cluster region pfAPC (codons 1254-1631) found in colorectal cancer.

Another aspect of the invention is a method to identify previously unknown somatic mutations that are associated with a cancer of interest. DNA-containing samples from the cancer can be analyzed by a method of the invention, and amplicons in which heteroduplexes are observed can be further sequenced to characterize the mutation. Once such mutations are identified, and shown to be correlated with the presence of the cancer or, in some cases, to be causal, the newly identified mutations can serve as the basis for diagnosis of the cancer.

Another aspect of the invention is a method to determine the future risk of cancer in a patient having one of certain chronic inflammatory diseases (e.g., ulcerative colitis, primary sclerosing cholangitis or chronic pancreatitis) that are associated with an increased risk of cancer, typically 10- 20 or more years after the onset of the chronic inflammatory disease. Such chronic inflammatory conditions are characterized by the accumulation of somatic mutations (e.g., in exons 1 and 2 of pi 6 in chronic pancreatitis and primary sclerosing cholangitis, and in the mutation cluster region of APC (codons 1254—1631) in colon neoplasms, polyps and colorectal cancer). Somatic mutations of the p53 gene have been described in these conditions and are thought to be associated with neoplasia and cancer risk. For all these conditions, there is widespread chronic inflammation that affects much of the organ or gland, and often if there is a focus of neoplasia, it is microscopic and not visible with imaging modalities. Therefore, one aspect of the invention is to periodically (e.g., about once every year or two years) screen subjects having one of these conditions, starting, e.g., about 10 years after the incidence of the inflammatory condition, in order to detect cancer or advanced precancerous lesions (e.g., to determine if a subject is likely to develop cancer). Suitable secondary fluids for testing will be evident to a skilled worker and include, e.g., plasma, colonic washings (ulcerative colitis), bile (primary sclerosing cholangitis) or pancreatic fluids (chronic pancreatitis). Markers for neoplasia, such as, e.g.,p53 mutations, can be detected even if present at low concentrations (such as about 1%). Preferable genes for use in this method are those in which mutations are concentrated in particular regions of the gene. For example, for subjects with long-standing ulcerative colitis, screening for mutations in p53 is preferred; for subjects with long-standing primary sclerosing cholangitis, one can screen for mutations in p53, pi 6, or both. Another aspect of the invention is a method to screen a sample from the circulation of a subject (e.g. a non-symptomatic subject) for the presence of cancer in the subject, using a method of the invention. The detection of, e.g., p53, pl6 and/or APC mutations in the circulation would indicate a high probability of cancer and necessitate a search for cancer.

Another aspect of the invention is a method for predicting the response of a subject to a chemotherapy procedure, comprising testing a secondary fluid (e.g. from the circulation) from the subject for the presence of somatic mutations by a method of the invention. For example, somatic mutations of the EGFR gene in lung cancers predict response to EGFR inhibitors. It is often not possible to access cancer tissue without thoracotomy if the lung cancer is beyond the reach of a needle biopsy. Detecting mutations in the plasma or serum or bronchoalveolar lavage require a strategy that can detect such mutations when they are at low concentrations. Many other targets of novel therapies are being identified, such as c-kit and other tyrosine kinases, and could be detected in a similar fashion.

Methods for isolating DNA or RNA and other molecular biology methods used in the invention can be carried out using conventional procedures. See, e.g., discussions in Sambrook, et al. (1989), Molecular Cloning, a Laboratory Manual, Cold Harbor Laboratory Press, Cold Spring Harbor, N.Y.; Ausubel et al. (1995). Current Protocols in Molecular Biology, N.Y., John Wiley & Sons; Davis et al. (1986), Basic Methods in Molecular Biology, Elseveir Sciences Publishing,, Inc., New York; Hames et al. (1985), Nucleic Acid Hybridization, IL Press; Dracopoli et al. (current edition) Current Protocols in Human Genetics, John Wiley & Sons, Inc . ; and Coligan et al. (current edition) Current Protocols in Protein Science, John Wiley & Sons, Inc.

Methods of the invention can be readily adapted to a high throughput format, using automated (e.g. robotic) systems, which allow many measurements to be carried out simultaneously. Furthermore, the methods can be miniaturized (e.g. , carried out in reaction buffers of about 25 μl, 1 μl, 0.1 μl, or less).

The order and numbering of the steps in the methods described herein are not meant to imply that the steps of any method herein must be performed in the order in which the steps are listed or in the order in which the steps are numbered. The steps of any method disclosed herein can be performed in any order which results in a functional method. Furthermore, the method may be performed with fewer than all of the steps, e.g., with just one step.

The phrase "a method for testing a subject ..." is not meant to exclude tests in which no mutations are found. In a general sense, this invention involves assays to determine whether a subject has the recited mutations, irrespective of whether or not such mutations are detected.

Any combination of the materials useful in the disclosed methods can be packaged together as a kit for performing any of the disclosed methods. For example, reagents for performing PCR and for heteroduplexing could be packaged along with suitable PCR primers. Components for performing cycle sequencing may also be included. If desired, the reagents are packaged in single use form, suitable for carrying one set of analyses.

Kits may supply reagents in pre-measured amounts so as to simplify the performance of the subject methods. Optionally, kits of the invention comprise instructions for performing the method. Other optional elements of a kit of the invention include suitable buffers, packaging materials, etc. The kits of the invention may further comprise additional reagents that are necessary for performing the subject methods. The reagents of the kit may be in containers in which they are stable, e.g., in lyophilized form or as stabilized liquids. In the foregoing and in the following example, all temperatures are set forth in uncorrected degrees Celsius; and, unless otherwise indicated, all parts and percentages are by weight.

EXAMPLES Example I - Materials and Methods A. Patients and Samples

Lymphocytes were obtained from 7 healthy controls. The pancreatic juice and pancreatic •-cancer tissues of 20 patients with Stage 1 or 2 disease undergoing Whipple resection between 2000 and 2003 were included. Pancreatic duct juice was also obtained during endoscopy from 8 patients with pancreatitis and 8 patients with a normal pancreas by clinical evaluation undergoing ERCP (endoscopic retrograde cholangiopancreatography) as part of their diagnostic workup. These patients were studied as part of a protocol approved by the Johns Hopkins Joint Committee for Clinical Investigation. The pancreatic cancers of 7 patients were grown in athymic nude mice to enable human pancreatic cancer cells to grow along with mouse stroma. Genetic characterization of the pancreatic cancer DNA (such as identifying homozygous deletions) is facilitated because the pancreatic cancer is not admixed with human stroma. Mutation analysis of the remaining 13 patients was performed in microdissected fresh frozen pancreatic cancer tissues. Pancreatic juice was collected from these 20 individuals during their pancreatic surgery by aspirating the pancreatic juice present just after the pancreatic duct is cut. B. Cell Lines The cancer cell lines used in this study include 9 pancreatic cell lines (AsPcI, BxPc3,

CAPANl, CAPAN2, CFPACl, COLO357, HS766T, MiaPaCa2 and Panel) and 16 breast cancer cell lines (MDA-468, T47D, BT-20, BT-474, BT-483, BT-549, CAMA-I, HCC- 1937, MDA-MB- 134, MDA-MB-157, MDA-MB-231, MDA-MB-361 , MDA-MB-436, HS-578T, MCF-7, and SK- BR3). DNA was extracted from tissue samples using the QIAGEN DNeasy Tissue Kit (QIAGEN Inc, Valencia, CA) C. PCR Amplification

The primers used are described in Table 1. Table 1. Summary of pi 6 and p53 primers

Set Exon/gene PCR product Annealing Name Primers (5 '-3') Size (bp) temp ("C)

1 1/CDKH2A 204 60 pl6 IF GGG AGC AGC ATG GAG CCG (SEQ IDNO: 1) pl6 IR AGT CGC CCG CCA TCC CCT (SEQ ID NO:2)

2 2/CDKN2A 440 59 pl6 2F CGG TGA GGG GGC TCT ACA CAA G (SEQ ID

NO:3) p 16 2R GCT GAA CTT TCT GTG CTG GAA AAT G (SEQ ID NO:4)

3 5-6/p53 455 59 p53 GTG CCC TGA CTT TCA ACT CTG T (SEQ ID

5A NO:5) p53 ACC CCT CCC AGA GAC (SEQ ID NO:6)

6B

4 7-8/p53 650 56 p53 CCT CAT CTT GGG CCT GTG TT (SEQ ID NO:7)

7A p53 TCC TCC ACC GCT TCT TGT (SEQ ID NO: 8)

8B

10μl PCRs were performed in 96-well PCR plates (2OmM Tris-HCl (pH 8.4), 3.OmM MgCl₂, 0.2mM of each dNTPs, 0.5μM of each primer F (forward) and R (reverse), 0.2 units/μl Platinum Tag polymerase, Invitrogen, Carisbad, CA). 10% DMSO (vol/vol) was added to amplify pl6. Input DNA is quantified using a real-time Q-PCR assay (Quantifier, Applied Biosystems, Foster City, CA). Since about 3 pg is the molecular weight of one DNA molecule (genome equivalent), the equivalent of about 7 molecules of amplifiable DNA (2 lpg) was used in the limiting dilution heteroduplex analysis. PCR was carried out in a ThermoHybaid Thermal cycler (Middlesex, U.K.) at 94°C for 2 min; 60 cycles of 94°C, 30s; 60⁰C (pl6 exon I)₅59°C (plό exon 2 and p53 exon 5-6), or 56°C (p53 exon 7-8), 45s; 72°C for 1 min; and 72°C for 10 minutes. PCR products were visualized on 2% agarose gels prior to heteroduplex analysis. C. Heteroduplex Analysis PCR products were diluted in a 1 : 1 ratio with IX AmpliTaq PCR buffer containing 1.5mM

MgCl₂) in 96-well PCR plates and overlaid with mineral oil. To form heteroduplexes, samples were thermocycled: typically 95°C for 3 min, 95°C-80°C for 3°C/min, 80°C-50°C for l°C/min, 50⁰C for 20min, 50°C-45°C for l°C/min, 45°C-25°C for 2°C/min, 4°C hold. Samples were then subjected to temperature gradient capillary electrophoresis (TGCE) using a SCE9610 automated sequencer (SpectruMedix Corporation, State College, PA). Heteroduplex analysis of PCR products were performed twice, subjecting PCR products to two different temperature protocols during capillary electrophoresis. The temperature protocol used varied with the size of the PCR product (e.g.400bp PCR products underwent a ramping between 50-60⁰C over 30 minutes). In a typical analysis, of 96 PCR products screened for heteroduplexes, only 2 PCRs had discordant results between the first and second temperature protocol runs. Heteroduplexes were identified using Revelation Mutational Discovery Software, version 2.4 (SpectruMedix Corporation, State College. PA). The software is programmed to score each chromatogram and to automatically call abnormal chromatograms suggestive of heteroduplexes that are above a set mutation score of 200. All chromatograms scored as mutant were reviewed and chromatograms with significant changes in peak characteristics (a second peak after the homoduplex peak with a height ~50% or higher relative to the main peak or when a major shift took place in the position of a peak) were scored as heteroduplexes (Figure 2).

E. Sequencing analysis

PCR products were purified using QIAquick PCR Purification Kit (QIAGEN Inc, Valencia, CA). Sequencing was performed using BigDye terminator mix vl .1 (Applied Biosystems, Foster City, CA). Sequence analysis was performed with ABI PRISM 377XL DNA sequencer (The Perkin- Elmer Corporation, Wellesley, MA).

F. Statistics Chi-square tests were used to determine differences in the proportion of PCR products that formed heteroduplexes. A two-tailed P value of less than 0.05 was used to assess statistical significance. Statistical analysis was performed using STATA version 8.2 software.

Example II - Sensitivity and Specificity of Heteroduplex assay by limiting dilution PCR We performed multiplex reactions on samples with known mutations using DNAs from pancreatic and breast cancer cell lines to determine the best limiting dilution PCR strategy for mutation detection. Initially, DNA from six cancer cell lines with different mutations in either pi 6 or p53 were analyzed including 4 base substitutions and 2 insertion/deletions. Characteristics of all these sequence variants are shown in Table 2. Table 2. Characteristics of sequence variants of cancer cell lines used in the study

No Name Gene/Fragments sequence change localization in the fragments

1 CAPAN2 pl6/exon 1 6bp insertion codon 11

2 AsPcI pl6/exon 2 2bp deletion codon 69

3 HS-578T p53/exon 5 G>T codon 157

4 BxPc3 p53/exon 6 A>G codon 220

5 CFPACl p53/exon 7 A>G codon 242

6 BT-474 p53/exon 8 G>A codon 285

The number of DNA molecules per PCR and the ratio of mutant and wild type DNA was adjusted from 1 : 1 to 1 :20 and screened for heteroduplex changes using TGCE. TGCE could detect heteroduplexes as long as the mutant to wild-type ratio of input DNA was 1 :6 ratio or more (1:4 etc), but at higher ratios such as 1:8 or 1:10 some mutations were not detected, and no mutations were detected at lower ratios using our protocol (1 :20). To further confirm that PCRs in which 1 of 7 DNA molecules (21 pg of DNA) would reliably identify mutations, we examined 8 additional DNA samples from breast cancer cell lines with known mutation in p53. Mutant and wild type DNA was mixed at a ratio of 1:6. Heteroduplex changes were found in all PCR products from these mutant/wild type DNA mixtures. We therefore chose to amplify DNA samples and screen the PCR products for heteroduplexes by including about 7 DNA molecules for each PCR. This way PCRs that amplified a mutant DNA molecule would have a ratio of mutant to wild-type DNA of about 1 :6, a concentration sufficient to allow the mutation to be detected by TGCE.

To determine the specificity of heteroduplex changes identified in limiting dilution PCR products by TGCE, seven lymphocyte DNA samples from healthy controls were analyzed. For each sample, multiple PCRs were performed with 12-24 pg of input DNA (equivalent to 4-8 DNA molecules per PCR) and screened for heteroduplexes. Three heteroduplex changes (one inplό exon 1 and two inp53 exon 5-6) were found in 967 PCRs analyzed yielding an assay specificity of 99.7%. The 3 PCRs with false positive heteroduplexes likely arose either from sequence alterations generated during PCR₃ from incorrect heteroduplex formation, or calls during the TGCE step. DNA samples generating PCRs with heterσduplexes above healthy control levels are likely to contain sequence alterations.

Example IH - Detection of p53 and pi 6 mutations in pancreatic juice using TCGE

Before applying our limiting dilution heteroduplex detection strategy to pancreatic juice, we determined the utility of TGCE to detect heteroduplexes in undiluted pancreatic juice DNA. For each patient's pancreatic juice, two PCR products were amplified using 20ng of undiluted pancreatic juice DNA and screened for heteroduplexes. Among individuals with pancreatic cancer undergoing pancreaticoduodenectomy, heteroduplex changes were detected in the pi 6 gene in 3 of 20 (15%) pancreatic juice samples, and for p53 in 2 of 20 (10%). In contrast, using the same approach no heteroduplex changes were detected in 8 pancreatic juice DNA samples obtained from individuals with chronic pancreatitis but was found in 1 of 8 (12%) individuals undergoing pancreatic evaluation who were not found to have clinical evidence of pancreatic disease (Table 3).

Table 3

Undiluted Pancreatic Juice DNA Pancreatic Juice DNA Analyzed at

Limiting dilution

p16 exon p16 p53 p53 p16 p16 p53 p53 exon

1 exon 2 exon 5/6 exon 7/8 exon 1 exon 2 exon 5/6 7/8

PJ- 1 Wt mt Wt Wt 2:43 4:38 0:45 0:45

PJ-2 wt Wt Wt Wt 5:40 3:42 0:44 3:27

PJ-3 Wt Wt Wt Wt 2:44 3:42 0:43 4:41

PJ-4 Wt Wt Wt Wt 0:45 6:39 0:45 0:42

PJ-5 Wt Wt Wt Wt 0:45 0:43 0:43 1:37

PJ-6 mt Wt Wt Wt 0:45 0:45 0:43 1:44

PJ-7 Wt Wt Wt Wt 0:44 0:45 0:44 0:41

PJ-8 Wt Wt mt Wt 2:42 6:39 0:44 3:41

PJ-9 Wt Wt Wt mt 1:44 0:45 0:45 1:43

PJ-10 Wt Wt Wt Wt 0:45 0:45 0:44 1:33

PJ-11 Wt mt Wt Wt HD HD 0:45 1:42

PJ-12 Wt mt Wt Wt 0:45 3:42 0:43 0:43

PJ-13 Wt Wt Wt Wt 0:45 4:41 2:44 2:42

PJ-14 Wt Wt Wt Wt HD HD 0:37 2:21

PJ-15 Wt Wt Wt Wt 0:45 1:42 0:32 1:33

PJ-16 Wt Wt Wt Wt HD HD 1:45 0:43

PJ-17 Wt Wt Wt Wt HD HD 3:42 0:39

PJ-18 Wt Wt Wt Wt HD HD 1:21 3:42

PJ-19 Wt Wt Wt Wt 0:45 0:44 0:31 1:43

CPJ-1 Wt Wt Wt Fail 2:45 0:45 0:43 0:42

CPJ-2 Wt Wt Wt wt 1:41 0:45 0:45 2:39

CPJ-3 Wt Wt Wt Wt 0:45 0:45 0:45 1:37

CPJ-4 Wt Wt Wt Wt 1:44 0:44 0:45 0:45

CPJ-5 Wt Wt Wt wt 1:44 0:45 0:37 1:41

CPJ-6 Wt Wt Wt Wt 2:43 0:45 0:40 0:40

CPJ-7 Wt Wt Wt Wt 0:45 2:41 0:44 0:35 CPJ-8 Wt Wt Wt Wt 1:44 0:44 0:43 4:41

CPJ-9 wt Wt wt Wt 0:45 0:45 1:44 0:35

CPJ-10 Wt mt Wt Wt 0:44 1:43 0:41 0:34

CPJ-11 Wt Wt Wt Wt 1:44 0:43 0:45 0:45

Normal

Pt 1 Wt Wt Wt Wt 0:27 0:37 0:44 0:35

2 Wt Wt Wt Wt 1:44 0:45 0:42 0:45

3 nd nd nd nd 0:45 0:44 0:45 0:45

4 Wt Wt Wt Wt 0:45 2:43 0:35 1:31

5 Wt Wt Wt Wt 2:43 0:45 0:45 0:45

Spousal

DNA

1 0:22 0:20 0:19 0:37

2 0:22 NA 1:45 0:38

3 0:23 0:38 0:47 0:36

4 0:23 0:21 0:46 0:45

5 NA 0:40 0:42 0:45

6 1:44 0:45 0:45 0:45

7 0:45 0:45 1:44 0:45

Example IV - Heteroduplex assay by LD-PCR on DNA from pancreatic juice

We next screened our pancreatic juice samples for p53 andplό mutations using the LD-PCR heteroduplex assay. For each of the exons of p53 andplό being evaluated, pancreatic juice DNAs were subjected to 45 PCRs with 7 DNA molecules per PCR (21 pg of DNA per PCR) for a total of ~315 DNA molecules screened. Three additional control wells were analyzed including a positive ^•control (a DNA sample with a 1 :6 mixture of mutant cancer cell line DNA and wild type DNA)₃ one PCR of wild type DNA₅ and a negative control PCR (water instead of DNA).

After LD-PCR and TGCE, 11 of 20 (55%) juice samples from patients with pancreatic cancer had at least one PCR that formed heteroduplexes in the pi 6 gene and 16 of 20 (80%) had one or more PCRs with heteroduplexes in p53. Heteroduplexes were detected less frequently in pancreatic juice from those with pancreatitis and pancreas controls (Table 3). Five (83%) of the six heteroduplexes detected using undiluted DNA had detectable heteroduplexes in the corresponding exon using LD-PCR. These results indicate that LD-PCR and TGCE identifies mutations in pancreatic juice that would otherwise be missed by using conventional TGCE (p<0.01).

Using limiting dilution, we identified 99 heteroduplexes inp53 and pi 6 among 3214 PCRs of pancreatic juice DNA samples analyzed from patients with pancreatic cancer. This proportion was significantly higher than the proportion found in pancreatic juice DNA from patients with chronic pancreatitis (17 of 1369, p<0.001)_} a normal pancreas (10 of 1336, pO.OOl), or in healthy control lymphocyte DNA (3 of 967, p<0.001). The number of heteroduplexes detected in the pancreatic juice of patients with chronic pancreatitis was also higher than that found in lymphocyte DNA from healthy controls (p=0.02, Chi-Square) indicating that our assay is finding low levels of mutant p53 and pl6 in the pancreatic juice of some patients with chronic pancreatitis.

To evaluate the sensitivity and specificity of our LD-PCR marker panel for pancreatic cancer, we empirically chose a cut-off of 3 or more PCR products with heteroduplex changes per gene region assayed to distinguish pancreatic juice samples from patients with pancreatic cancer from those without cancer. By using this criterion, 10 of 20 (50%) pancreatic juice samples from patients with pancreatic cancer had increased amounts of mutant jσi<5 DNA and 6 of 20 (30%) had increased mutant p53. In contrast only 1 of 8 (12%) juice samples from patients with chronic pancreatitis had increased mutant p53 and none had increased mutant pi 6. Overall, elevated levels of mutant/* 16 or p53 were found in 12 of 20 (60%) pancreatic juice samples from patients with pancreatic cancer and only in 1 of 16 (6.6%, p=0.001) control pancreatic juice samples (1 of 8 with chronic pancreatitis, p=0.023; and 0 of 8 patients with a normal pancreas, p=0.004) (Table 3). These results indicate that using LD-PCR and TGCE to screen pancreatic juice for pi 6 and p53 mutations can help distinguish individuals with pancreatic ductal adenocarcinoma from those without pancreatic neoplasia.

Example V - p53 and pl6 alterations in primary pancreatic cancer

We analyzed the pancreatic cancer DNA from 20 patients that underwent pancreaticoduodenectomy for mutations in p53 and pi 6. Alterations of pi 6 were found in 55% (11/20) of the pancreatic cancers including five point mutations, 5 homozygous deletions (confirmed by multiplex PCR using larger primer sets), and one 12-base pair deletion (Table 4).

Table 4. Sequencing results of 20 primary pancreatic adenocarcinoma cases

Patient Age, gender, Cancer pl6 p53 number race stage exon 1-2 exon 5-8

1 66,f,c PT31bMX GCG>ACG, codon 148, Arg>Trp Wt

2 77,m,c T3NlbMX Del/33-36 CG1>TGT, Codon 273, Arg>Cys

3 64,f₃c T3N0MX AGA>ACA, Codon 138, Arg>Trr Wt

T3N1BM

4 77Ac X Wt Wt

5 70,f,aa T3N0MX Wt CGT>CAT, Codon 273, Arg>His

6 73,m,c T2N1MX Wt CGOCAC, Codon 175, Arg>His

7 67,f_;c T3N0Mx Wt CGOCAC, Codon 175, Arg>His

T3N1BM

8 54,m,c X CGG>CAG, Codon 24, Arg>Gln Wt

9 6S,m,c T2N0MX Wt CAOCGC, Codon 297, His>Arg

10 61, m, c T3N0MX GTG>GCG, Codon 28, Val>Ala CGOTGG, Codon 248, Arg>Trp

11 81,f,c T3N0MX Homodel TGT>CGT, Codon 277, Cys>Arg

12 64,m,c T3N0MX Wt CGOCAC, Codon 175, Arg>His

13 65Ac T3N0MX Wt TGT>CGT, Codon 275, Cys>Arg

14 71,f,other T3N0MX Homodel CGG>TGG, Codon 248, Arg>Trp 15 82,m,c T3N1MX Wt GAG>AAG, Codon 271, Glu>Lys

16 45,m,c T3N1MX Homodel CCOCGC, Codon 177, Pro>Arg

T3N1BM

17 69,f,c X Homodel ACOATC, Codon 140, Thr>IIe

18 77,m,c T3N1MX Homodel CGG>TGG, Codon 248, Arg>Trp GGT>GGC, Codon 262, Silent

19 59,f,c T2N0MX CGOCCG, Codon 29, Arg>Pro Mutation

T3N1BM

20 73,f,c X Wt GTG>ATG, Codon 272, Val>Met

Four of the five homozygous deletions were in pancreatic cancer xenografts. p53 mutations were detected in 80% (16/20) of the pancreatic cancers.

We compared the results of mutational analysis of the 20 patients with pancreatic cancers to their pancreatic juice results. Because healthy control DNA could yield a rare heteroduplex alteration in our assay (<l/300 PCRs)₃ we set a threshold of heteroduplexes in more than one of 45 PCRs screened as indicating a mutation in a pancreatic juice sample. By this criterion, mutations in either p53 or pl6 were detected in 12 of the 20 (60%) pancreatic juice samples from patients with pancreatic cancer (Table 4). The concordance rate between heteroduplexes identified in pancreatic juice samples and the site of mutations detected by sequencing of primary tumors was 48 of 70 (68%). Usually the discordance arose from detecting heteroduplexes in pancreatic juice DNA samples that were not detected in the corresponding pancreatic cancer. These findings suggest that the LD-PCR assay also detected mutations arising elsewhere in the pancreas (such as from PanlN) or from other regions of the cancer reflecting intra-tumoral heterogeneity. It is also possible that despite careful microdissection and molecular analysis, one or more of pancreatic cancer mutations were not detected. Previous studies comparing pancreatic cancer and matching pancreatic juice samples have found that mutations in pancreatic juice that are not found in the patient' s primary pancreatic cancer.

Example VI - Discussion In this study, we demonstrated that limiting dilution PCR followed by TGCE is more sensitive at detecting low-abundance mutations in pancreatic juice samples than conventional TGCE. We chose the markers mutant p53 and pi 6 as these alterations are considered highly specific to the development of neoplasia. Their specificity for neoplasia is reflected in our results. Using our assay 60% of patients with pancreatic cancer had mutant p 53 and ox pi 6 DNA in their pancreatic juice compared to only 1 of 16 controls (with either chronic pancreatitis or a normal pancreas). These results indicate that the detection of p53 andpl 6 mutations in pancreatic juice using our assay can aid in the diagnosis of pancreatic cancer. Our results indicate that the detection of pi 6 mutations in pancreatic juice is also useful as an indicator of pancreatic cancer.

The high specificity of our limiting dilution PCR assay for detecting mutations is indicated by finding only 3 of 967 PCRs positive for heteroduplexes inp53 or pi 6 when amplifying healthy control DNA. The most likely explanation -for the low-frequency heteroduplexes in control DNA samples is Taq polymerase-induced amplification errors. To minimize DNA polymerase errors during PCR, we used a high-fidelity DNA polymerase (Pfu, Platinum Taq) (Girald-Rosa et al. (2005) Clin Chem 5_L 305-11), which has a low error-rate (1.3 errors/10^"6 nucleotides/cycle). With this high specificity, one can detect mutations at lower concentrations by screening more PCR products for mutations. This can be important for detecting advanced PanINs. PanINs are small neoplasms, 5mm or less, and are thought to be the commonest precursor to invasive pancreatic ductal adenocarcinoma. PanINs have been shown to harbor mutations in p53 and pi 6. PanINs are too small to be detected directly by imaging. Using molecular assays to predict whether or not a pancreas harbors advanced PanIN can help predict future pancreatic cancer risk among high-risk individuals undergoing pancreas screening. Since not all pancreatic cancers harbor mutations in p53 or pi 6, it is not surprising that our assay did not identify evidence of mutations in the pancreatic juice of all patients with pancreatic (cancer. It is likely that the sensitivity of our panel could be improved without significantly affecting disease specificity by adding other markers of pancreatic cancer such as quantifying levels of aberrantly methylated DNA or K-rαs mutations in pancreatic juice. The specificity of these markers for cancer increases by quantifying their levels in pancreatic juice, as it is uncommon for individuals without cancer to have mutant K-rαs or aberrantly methylated DNA above a certain threshold concentration (-1%).

We report herein a sensitive assay for detecting mutations (e.g., inp53 anά/oτplό) and show that detecting a threshold concentration of these mutations in pancreatic juice helps predict the presence of pancreatic cancer. The strategy of limiting dilution PCR and heteroduplex detection using TGCE may also be useful for the detection of mutations in other clinical settings.

Example VII - Detection of mutations by LD-PCR, followed by cycle sequencing

In this study, DNA molecules are PCR amplified at about 10 picogram concentrations (about 3 molecules of DNA, measured using real time PCR). By adding 10 picograms of template DNA per

PCR, most PCR aliquots will contain 2-4 DNA molecules. When one of the molecules in a PCR contains mutant DNA, the ratio of mutant to wild-type molecules (1 :1 to 1 :3) will enable the mutation to be detected by cycle sequencing of the resulting PCR product. In a typical scenario, 93 PCRs are performed where each aliquot contains about 3 molecules per well, and about 279 DNA molecules are screened for mutations. Therefore, if one analyzes a clinical sample such as pancreatic juice for mutations \np53 (exons 5-8) saxάplό (exons 1 and 2), one uses 4 sets of PCR primers (p53 exon 5-6, p53 exon 7-8, pl6 exon 1, plό exon 2). One 96 well plate is used for each primer, 93 wells containing about lOpg of pancreatic juice DNA, 1 well containing a water lane, one well containing a wild type DNA control and one well containing a mutant DNA control. By sampling about 279 DNA molecules, there is about a 95% probability of sampling at least one mutant DNA molecule when the mutant to wild-type ratio of DNA is 1/100. Each PCR is sequenced and the ability to detect the mutation by sequencing approaches the accuracy of sequencing to detect such a mutation. There is a small chance that a mutation could by chance be mixed in an aliquot that contains 5 or more templates and the other 4 templates are wild type. If so, the cycle sequencing reaction may not be able to detect the mutation. Of note, this ratio of wild-type to mutant molecules is still generally detectable by heteroduplex analysis. The number of wells containing 5 or more templates will depend on the accuracy of pipetting. Also, some wells will not contain any DNA but the average number of DNA templates screened in a 96 well plate is still the same, hi many clinical samples the concentration of mutant molecules is in the 2-10% range, and so this limiting dilution strategy for sequencing is useful. The availability of low-cost high-throughput sequencing makes this assay approach feasible to use in the clinical setting. Direct comparison of limiting dilution sequencing and limiting dilution heteroduplex detection can be used to determine which method has better sensitivity and specificity for mutation detection. Since this assay strategy is designed to detect mutations at low concentrations, the false positive rate could be reduced by combining sequencing and heteroduplex analysis to each sample.

From the foregoing description, one skilled in the art can easily ascertain the essential characteristics of this invention, and without departing from the spirit and scope thereof, can make changes and modifications of the invention to adapt it to various usage and conditions and to utilize the present invention to its fullest extent. The preceding preferred specific embodiments are to be construed as merely illustrative, and not limiting of the scope of the invention in any way whatsoever. The entire disclosure of all applications, patents, and publications cited above (including U.S. provisional application 60/819,440, filed July 7, 2006) and in the figures, are hereby incorporated in their entirety by reference.

Claims

We claim:

1. A method for screening for mutations in a DNA sample, comprising a) diluting and distributing the sample to obtain a plurality of aliquots which, on the average, contain between about 2-10 genome equivalents of the DNA, b) PCR amplifying the DNA in a sufficient number of aliquots, with at least one set of PCR primers, so as to generate amplicons such that, if mutations are present at a frequency of about 1- 10% in a given gene in the sample, the probability of there being a mutation in that gene in at least one of the amplicons is at least about 95%, and c) screening the amplicons for the presence of mutations, using a method that can detect unspecified mutations at unspecified sites within an amplicon.

2. The method of claim 1, wherein the screening method is temperature gradient capillary electrophoresis (TGCE).

3. The method of claim 1, wherein the screening method is cycle sequencing of the DNA.

4. The method of any of claims 1-3, wherein each aliquot contains between about 3-8 genome equivalents of DNA.

5. The method of any of claims 1-3, wherein each aliquot contains between about 6-7 genome equivalents of DNA.

6. The method of any of claims 1-5, wherein the number of aliquots in which DNA is amplified is at least about 40 for each PCR primer set screened.

7. The method of any of claims 1-5, wherein the number of aliquots that are amplified is at least about 150 for each PCR primer set screened.

8. The method of any of claims 1-5, wherein the number of aliquots that are amplified is at least about 300 for each PCR primer set screened.

9. The method of any of claims 1-8, wherein the frequency of mutations in the gene is about 10%

10. The method of any of claims 1-8, wherein the frequency of mutation in the gene is about 9%.

11. The method of any of claims 1-8, wherein the frequency of mutation in the gene is about 8%.

12. The method of any of claims 1-8, wherein the frequency of mutation in the gene is about 7%.

13. The method of any of claims 1-8, wherein the frequency of mutation in the gene is about 6%.

14. The method of any of claims 1-8, wherein the frequency of mutation in the gene is about 5%.

15. The method of any of claims 1-8, wherein the frequency of mutation in the gene is about 4%.

16. The method of any of claims 1-8, wherein the frequency of mutation in the gene is about 3%.

17. The method of any of claims 1-8, wherein the frequency of mutation in the gene is about 2%.

18. The method of any of claims 1-8, wherein the frequency of mutation in the gene is about 1%.

19. The method of any of claims 1-5, wherein there are about 6-7 mammalian genomic DNAs in each of about 45 aliquots; and mutations that are present at a frequency of about 3% are detected.

20. The method of any of claims 1-19, wherein at least one set of PCR primers is used.

21. The method of any of claims 1-19, wherein at least two sets of PCR primers are used.

22. The method of any of claims 1-19, wherein at least about ten sets of PCR primers are used.

23. The method of any of claims 1-22, wherein the sample is from a subject.

24. The method of any of claims 1-23, wherein the sample is a clinical sample.

25. The method of claim 23, wherein the subject is a plant, an invertebrate animal or a vertebrate animal.

26. The method of claim 23 or 24, wherein the subject is a mammal.

27. The method of claim 26, wherein the mammal is a human.

28. The method of claim 26 or 27, wherein the subject is suspected of having a cancer.

29. The method of claim 28, wherein the cancer is pancreatic cancer, lung cancer, biliary tract cancer, colorectal cancer, or non-small cell lung carcinoma.

30. The method of claim 28, wherein the cancer is pancreatic cancer.

31. The method of any of claims 1-30, wherein the mutations are somatic mutations.

32. The method of any of claims 1-31, wherein the mutations are point mutations, small insertions and/or small deletions.

33. The method of any of claims 1-32, wherein the sample is from a tumor tissue.

34. The method of any of claims 1-32, wherein the sample is a secondary fluid.

35. The method of claim 34, wherein the secondary fluid is pancreatic duct juice, serum/plasma, urine, seminal fluid, bronchoaveolar lavage, saliva, bile, gastric or colonic washes, or stool.

36. The method of claim 34, wherein the secondary fluid is from the circulation (serum/plasma).

37. The method of claim 30, wherein the sample is pancreatic duct juice.

38. The method of any of claims 28-37, wherein the mutations are in the p53, pi 6, APC and/or EGFR genes.

39. The method of claim 30, wherein the mutations are in the p53 and/or pi 6 genes.

40. A method of testing a subject for pancreatic cancer, comprising screening for mutations in a sample from the subject by the method of any of claims 1-40, wherein the presence of an increased number of mutations in the p53 and/or theplό genes, compared to a baseline value, indicates that the subject has or is likely to have pancreatic cancer.

41. The method of claim 40, further comprising sequencing the DNA of the amplicons which resulted in the formation of heteroduplexes.

42. The method of any of claims 1-41, wherein in the method is high throughput.

43. The method of any of clams 40-42, wherein the subject suffers from ulcerative colitis, the sample is from plasma or colonic washings, and the mutations screened for are in p53; the subject suffers from primary sclerosing cholangitis, the sample is from plasma or bile, and the mutations screened for are \np53,pl6 or both; or the subject suffers from chronic pancreatitis, the sample is from plasma or a pancreatic fluid, and the mutations screened for are in p53.

44. The method of any of claims 40-42, further comprising testing for aberrantly methylated DNA, telomerase activity, or the detection of LOH (loss of heterozygosity) in microdissected samples.

45. A kit, comprising containers) suitable for holding a dilution sample, and/or reagents suitable for obtaining dilution samples, and/or regents for performing PCR, and/or reagents for performing TCGE, and/or reagents for performing cycle sequencing.