GASTROENTEROLOGY 2000;119:1219 –1227
`
`Colorectal Cancer Screening by Detection of Altered Human
`DNA in Stool: Feasibility of a Multitarget Assay Panel
`
`DAVID A. AHLQUIST,* JOEL E. SKOLETSKY,‡ KEVIN A. BOYNTON,‡ JONATHAN J. HARRINGTON,*
`DOUGLAS W. MAHONEY,* WILLIAM E. PIERCEALL,‡ STEPHEN N. THIBODEAU,* and
`ANTHONY P. SHUBER‡
`*Division of Gastroenterology and Hepatology, Department of Biostatistics, and Division of Molecular Genetics, Mayo Clinic, Rochester,
`Minnesota; and ‡Applied Research Group, EXACT Laboratories, Maynard, Massachusetts
`
`Background & Aims: Assay of altered DNA exfoliated
`into stool represents an intriguing approach to screen
`for colorectal neoplasia, but multiple markers must be
`targeted because of genetic heterogeneity. We explored
`the feasibility of a stool assay panel of selected DNA
`alterations in discriminating subjects with colorectal
`neoplasia from those without. Methods: Freezer-ar-
`chived stools were analyzed in blinded fashion from 22
`patients with colorectal cancer, 11 with adenomas >1
`cm, and 28 with endoscopically normal colons. After
`isolation of human DNA from stool by sequence-specific
`hybrid capture, assay targets included point mutations
`at any of 15 sites on K-ras, p53, and APC genes; Bat-26,
`a microsatellite instability marker; and highly amplifi-
`able DNA. Results: Analyzable human DNA was recov-
`ered from all stools. Sensitivity was 91% (95% confi-
`dence interval, 71%–99%) for cancer and 82% (48%–
`98%) for adenomas >1 cm with a specificity of 93%
`(76%–99%). Excluding K-ras from the panel, sensitivities
`for cancer were unchanged but decreased slightly for
`adenomas to 73% (39%–94%), while specificity in-
`creased to 100% (88%–100%). Conclusions: Assay of
`altered DNA holds promise as a stool screening ap-
`proach for colorectal neoplasia. Larger clinical investi-
`gations are indicated.
`
`Colorectal cancer remains the second leading cause of
`
`malignant mortality in industrialized nations, ac-
`counting for more than 10% of all cancer deaths.1 Be-
`cause of its orderly natural history and nonsurgical ac-
`cessibility, colorectal neoplasia appears ideally suited for
`preventive interventions. However, screening efforts have
`had relatively little impact partly because of performance
`limitations and low compliance rates with current
`screening tools. More optimally tailored screening tools
`are needed that would exhibit the combined features of
`high sensitivity and specificity for early-stage cancers and
`large premalignant adenomas, broad acceptability by the
`general population, affordability, and safety.
`
`Stool testing merits further consideration because its
`theoretical potential has not yet been achieved. Stool
`tests are noninvasive, require no cathartic preparation,
`can be performed on mailed-in specimens without a
`mandated health center visit, and may reflect the full
`length of the colorectum. Fecal occult blood testing
`regularly performed over a decade or more may lower
`colorectal cancer mortality by 15%–33%, primarily by
`detecting cancers at an earlier stage.2– 4 However, many
`cancers and most premalignant adenomas do not bleed
`and are missed.5,6 Furthermore, high false-positive rates
`as a result of frequent and trivial sources of occult bleed-
`ing lead to unnecessary colonoscopies, which drive up
`programmatic costs.7–9 More sensitive and specific mark-
`ers would improve the effectiveness and efficiency of stool
`screening.
`Neoplasm-specific DNA alterations have been well-
`characterized10,11 and represent
`intriguing candidate
`markers for stool screening. In contrast to blood, altered
`DNA arises from the neoplasm rather than from the
`circulation and is released into the lumen continuously
`via exfoliation rather than intermittently via bleeding.12
`Furthermore, DNA appears to be stable in stool13 and
`amplification techniques permit detection of minute
`amounts of analyte. Several investigators have recovered
`mutant DNA in stools from patients with colorectal
`cancer or adenomas.14 –20 Assays used have typically an-
`alyzed mutations on a single gene, especially K-ras be-
`cause of its small number of mutational sites. However,
`colorectal neoplasms are genetically heterogeneous10,11;
`no single mutation has been identified that is expressed
`across all colorectal neoplasms. Mutant K-ras, for exam-
`ple, is expressed by fewer than half.10,11,21 Thus, multiple
`
`Abbreviations used in this paper: L-DNA, “long” DNA; PCR, polymer-
`ase chain reaction.
`© 2000 by the American Gastroenterological Association
`0016-5085/00/$10.00
`doi:10.1053/gast.2000.19580
`
`Geneoscopy Exhibit 1065, Page 1
`
`

`

`1220 AHLQUIST ET AL.
`
`GASTROENTEROLOGY Vol. 119, No. 5
`
`DNA alterations must be targeted to achieve high neo-
`plasm detection rates.
`An assay system was developed which targets a spec-
`trum of DNA alterations that occur with colorectal
`neoplasia. This multicomponent assay panel
`targets
`point-mutations at any of 15 mutational hot spots on
`K-ras, APC, and p53 genes; Bat-26, a marker of mi-
`crosatellite instability22,23; and highly-amplifiable or
`“long” DNA (L-DNA). Observations by other investiga-
`tors17 of higher fecal DNA yields from patients with
`colorectal cancer than from controls provided the basis to
`include this latter marker in the assay panel.
`The aim of this blinded clinical pilot investigation was
`to assess the potential of a multitarget fecal DNA assay
`panel to discriminate selected patients with colorectal
`neoplasia from those without neoplasia using colonos-
`copy as the criterion standard.
`
`Materials and Methods
`Design and Subjects
`The investigation was approved by the Mayo Clinic
`Institutional Review Board and comprised 2 clinical pilot
`studies. Stools for each were selected from a freezer archive to
`yield subject groups with verified colorectal adenocarcinoma,
`colorectal adenomas $1.0 cm, and colonoscopically normal
`colons. Subjects were chosen to provide a balanced age and
`gender representation across groups and a mixed distribution
`of neoplasms from both proximal and distal colorectal sites
`(Table 1). Most patients with cancer had been referred with a
`known diagnosis or with a radiographically suspicious mass.
`All subjects with adenomas and normal colons were asymp-
`tomatic and undergoing surveillance because of either a family
`history of colorectal neoplasia or a previous personal history of
`neoplasia.
`Pilot study 1 was conducted to explore the diagnostic
`discrimination of the fecal DNA assay panel by confirming the
`appropriateness of preset positivity levels for point mutations
`and BAT-26, by establishing a cutoff level of positivity for
`L-DNA, and by correlating specific mutations found in stool
`with those in matched tumors. In pilot study 2, the assay panel
`was applied using assay parameters established from pilot
`study 1, and the major focus of study 2 was to examine
`specificity in a separate group. As indicated in Table 1, 7 of the
`10 patients with neoplasms in pilot study 2 had also been
`tested in pilot study 1. Because separate unthawed fecal ali-
`quots were assayed on these 7 patients, within-stool reproduc-
`ibility of assay results could be determined. All assays were
`performed by technicians blinded to the clinical data. On
`pathology review, 1 of the adenomas originally in pilot study
`1 was found to have high-grade dysplasia with a focus of
`invasive cancer and was reassigned from the adenoma to the
`cancer group. In pilot study 2, 2 of the original 20 normal
`controls were excluded, one because of an inadequate colono-
`scopic examination due to a poor preparation and the other
`
`Table 1. Demographic and Colorectal Neoplasm
`Characteristics of Subject Groups
`
`Cancers
`n
`Sex (M/F)
`Age (yr)a
`Tumor site, prox/
`dist
`Tumor size (cm)a
`Tumor stage,
`Dukes AB/CD
`Adenomas
`n
`Sex (M/F)
`Age (yr)a
`Polyp site, prox/
`dist
`Polyp size (cm)a
`Normal colons
`n
`Sex (M/F)
`Age (yr)a
`
`Pilot 1
`
`Pilot 2
`
`Combined
`
`21
`11/10
`69 (38–88)
`
`8
`4/4
`73 (54–83)
`
`22b
`11/11
`70 (38–88)
`
`10/11
`4.9 (2.6–11)
`
`4/4
`3.9 (2.5–11)
`
`14/15
`4.1 (2.5–11)
`
`13/8
`
`5/3
`
`13/9
`
`9
`4/5
`69 (61–76)
`
`2
`1/1
`74 (72–76)
`
`11
`5/6
`73 (61–76)
`
`5/4
`1.5 (1–5)
`
`1/1
`4 (1–7)
`
`6/5
`2 (1–7)
`
`10
`5/5
`69 (53–77)
`
`18
`9/9
`67 (50–74)
`
`28
`14/14
`68 (50–77)
`
`prox, proximal to splenic flexure; dist, splenic flexure or distal.
`aMedian (range).
`bBecause stools from 7 cancer patients from pilot 1 were repeated in
`pilot 2, total number of unique cancer patients was 22.
`
`because of the subsequent finding of a malignant ileal carci-
`noid tumor.
`
`Stool Collection, Processing, and Storage
`All stools had been collected within days before ca-
`thartic preparation for a scheduled colonoscopy, which served
`as the criterion standard. Any previous instrumentation had
`occurred $2 weeks before stool collections from colorectal
`cancer patients and $1 year for patients with adenomas and
`normal controls. To prevent toilet water artifact,24 a plastic
`bucket device was used to collect whole stools. Stools in sealed
`buckets were received within 12 hours of defecation at the
`on-site processing laboratory where they were tested by
`Hemoccult (see below) and promptly frozen at 280°C in
`multiple aliquots. Frozen single fecal aliquots of at least 6 g
`per subject were sent in batches on dry ice for blinded DNA
`analyses at EXACT Laboratories (Maynard, MA).
`
`Multitarget DNA Assay Panel
`Total nucleic acid preparation. All stool samples
`were thawed at room temperature and homogenized in an
`excess volume (.1:10, wt, vol) of EXACT buffer A (EXACT
`Laboratories) using an EXACTOR stool shaker (EXACT Lab-
`oratories). After homogenization, a 4-g stool equivalent of each
`sample was centrifuged to remove all particulate matter, and
`the supernatants were incubated at 37°C after addition of
`proteinase K (0.5 mg/mL) and sodium dodecyl sulfate (0.5%).
`The supernatants were subsequently extracted with Tris-satu-
`rated phenol (GIBCO BRL, Grand Island, NY), phenol/chlo-
`roform/isoamyl alcohol (25:24:1), and chloroform. Total nu-
`
`Geneoscopy Exhibit 1065, Page 2
`
`

`

`November 2000
`
`STOOL SCREENING FOR ALTERED DNA 1221
`
`cleic acid was then precipitated (1/10 volume 3 mol/L NaAc
`and an equal-volume isopropanol), removed from solution by
`centrifugation, and resuspended in TE (0.01 mol/L Tris [pH
`7.4] and 0.001 mol/L EDTA) buffer containing RNase A (2.5
`mg/mL). For each group of samples prepared, process positive
`control samples as well as component negative controls were
`included.
`Sequence-specific purification and amplification.
`Sequence-specific DNA fragments were purified from the total
`nucleic acid preparations by performing oligonucleotide-based
`hybrid captures. For each sample, 7 unique hybrid capture
`reactions were performed in duplicate. Each capture reaction
`was carried out by adding 300 mL of sample preparation to an
`equal volume of 6 mol/L guanidine isothiocyanate solution
`(GIBCO BRL) containing biotinylated sequence-specific oli-
`gonucleotides (20 pmol; Midland Certified Reagent Co., Mid-
`land, TX). After a 2 -hour incubation at 25°C, streptavidin-
`coated magnetic beads were added to the solution, and the
`tubes were incubated for an additional hour at room temper-
`ature. The bead/hybrid capture complexes were then washed 4
`times with 13 B1W buffer (1 mol/L NaCl, 0.01 mol/L
`Tris-HCl [pH 7.2], 0.001 mol/L EDTA, and 0.1% Tween 20),
`and the sequence-specific captured DNA was eluted into 35
`mL L-TE (1 mmol/L Tris [pH 7.4] and 0.1 mol/L EDTA) by
`heat denaturation.
`Polymerase chain reaction (PCR) amplifications (50 mL)
`were performed on MJ Research Tetrad Cyclers (Watertown,
`MA) using 10 mL of captured DNA, 13 GeneAmp PCR
`buffer (PE Biosystems, Foster City, CA), 0.2 mmol/L dNTPs
`(Promega, Madison, WI), 0.5 mmol/L sequence-specific pri-
`mers (Midland Certified Reagent Co., Midland, TX), and 5 U
`Amplitaq DNA polymerase (PE Applied Biosystems, Nor-
`walk, CT). All sequence-specific amplification reactions were
`performed in identical thermocycler conditions. After an ini-
`tial denaturation of 94°C for 5 minutes, PCR amplification
`was performed for 40 cycles consisting of 1 minute at 94°C, 1
`minute at 60°C, and 1 minute at 72°C, with a final extension
`of 5 minutes at 72°C. For PCR product analysis, 8 mL of each
`amplification reaction was loaded and electrophoresed on a 4%
`ethidium bromide–stained NuSieve 3:1 agarose gel (FMC,
`Rockland, ME) and visualized with a Stratagene EagleEye II
`(Stratagene, La Jolla, CA) still image system (Figure 1).
`
`Figure 1. Agarose gel electrophoretic analysis of K-rasPCR products.
`Amplification results representing 6 unique stool DNA samples (1– 6)
`amplified in duplicate (lanesa1 b) with appropriate negative (lane7)
`and positive (lane 8) control amplifications. Similar results were ob-
`tained for other amplicons including p53, APC, and Bat-26.
`
`Figure 2. Point mutational results for APC codon 1378 position 1.
`Point mutation results representing 6 unique stool DNA samples
`(5–10 and 15–20) analyzed in duplicate (lanes a 1 b). Wild-type
`reactions (lanes 1–10) and corresponding mutant reactions (lanes
`11–20) were analyzed by polyacrylamide gel electrophoresis. Positive
`wild-type results (lanes 1–10) served as internal sample specific
`controls. Each set of reactions (lanes 11–20) included mutation-
`specific positive controls representing 1% (lane13) and 5% (lane14)
`mutant DNA populations and negative control samples (lanes 1, 2,
`11, and 12) containing wild-type DNA only. Within this analysis, a
`single-stool DNA sample was positive for both wild-type (lane8, a 1
`b) and mutation-specific reactions (lane 18, a 1 b). All other stool
`DNA samples were positive for the wild-type reaction (lanes 5–7, 9,
`and 10, a1 b) and negative for the APC 1378 mutation (lanes15–17,
`19, and 20, a 1 b).
`
`Point mutation and Bat-26 analysis. The presence
`or absence of point mutations or Bat-26 –associated mutations
`was determined by using a modified solid-phase minisequenc-
`ing method.25 Point mutation targets included codons K12p1,
`K12p2, and K13p2 on the K-ras gene; codons 1309 delta 5,
`1367p1, 1378p1, and 1450p1 on the APC gene; and codons
`175p2, 245p1, 245p2, 248p1, 248p2, 273p1, 273p2, and
`282p1 on the p53 gene. These targets were selected for assay
`because they correspond to the highest frequency mutational
`sites observed in available tissue databases. From these data-
`bases, theoretical diagnostic yields for cancer were estimated to
`be 44% for p53,26 41% for K-ras,27 and 19% for APC28
`markers. For all gene targets, both wild-type and mutant-
`specific reactions were performed. Within the wild-type reac-
`tions, radionucleotide bases complementary to the wild-type
`base were added (Figure 2). For each point mutation–
`specific reaction, radionucleotide bases complementary to
`the expected mutant bases were added in addition to unla-
`beled dideoxy nucleotides complementary to the wild-type
`base (Figure 2). Bat-26 mutations associated with a deletion
`of 4 –15 base pairs (bp) were identified by size discrimina-
`tion of reaction products (Figure 3). We estimated that the
`theoretical yield by this microsatellite instability marker for
`cancer detection would be at least 15% based on reported
`observations in tissue.29
`L-DNA analysis. L-DNA was performed by analyz-
`ing the relative intensity of each sample-specific PCR product.
`For each stool sample analyzed, 7 unique PCR amplification
`
`Geneoscopy Exhibit 1065, Page 3
`
`

`

`1222 AHLQUIST ET AL.
`
`GASTROENTEROLOGY Vol. 119, No. 5
`
`blocks using a previously described extraction technique.30
`This DNA was sent to EXACT Laboratories where point
`mutation assays by the single-base extension method (see
`above) were performed in a blinded fashion.
`
`Statistical Analysis
`Sensitivity and specificity were estimated relative to
`the results of colonoscopy in the usual manner; 95% confidence
`intervals (CIs) for these estimated parameters were based on
`the exact binomial distribution. Comparisons of proportions
`between various subgroups were based on the Fisher exact test,
`and McNemar’s matched-pairs test for proportions was used to
`compare sensitivities between the fecal DNA assay panel (with
`and without K-ras) and Hemoccult. Interobserver variability
`for L-DNA was assessed in pilot study 1 using a weighted k
`statistic because the individual scores were ordinal in nature.
`The weighted k statistic was estimated for each of the 7 PCR
`amplification products separately and for the pooled observa-
`tions across all PCR amplifications. Among the 560 PCR
`amplifications scored, there were 7 instances in which 1 of the
`duplicate amplifications was noninformative; such instances
`were considered discordant so as to obtain a conservative
`estimate of interobserver variability.
`
`Results
`Analyzable human DNA was recovered in all
`subjects. When detected, mutant DNA accounted for
`1%–24% of total human DNA recovered in stools from
`cancer patients and for 1%–7% from those with large
`adenomas.
`
`Pilot Study 1
`Cancers. At least 1 point mutation among the
`15 targeted sites on K-ras, APC, and p53 genes was
`present in stools from 11 (52%) of the 21 cancers. Bat-26
`was positive in 4 cases, 1 of which also had a point
`mutation, yielding a detection rate of 14/21 (67%) when
`combined with point mutation components. Samples
`that were assigned an “A” score on .8 amplifications
`were considered L-DNA positive, because all of the
`colonoscopically normal controls fell below this cutoff.
`Using this definition, L-DNA alone was positive in
`stools from 14 (67%) of the 21 cancers. With all com-
`ponent markers together, the fecal DNA assay panel
`detected 19 (90%) of the 21 cancers (Table 2).
`Tissue was available for DNA extraction and point
`mutation analyses on 19 of the 21 cancers. Point muta-
`tion results on tissue and stool were concordant in 12
`cases (63 %): at least one pair of identical mutations was
`found in stools and matched tumors in 7 cases, and all
`targeted mutations were absent in both stool and tumors
`in 5 cases. In the 7 discordant cases, a mutation was
`
`Figure 3. Polyacrylamide gel electrophoresis analysis for deletions
`within Bat-26. Results representing 10 unique stool DNA samples
`(lanes 1–10) analyzed for deletions within Bat-26. The upper (U)
`region of the gel contains reaction products representing the wild-type
`full-length product. Presence of reaction products within the lower (L)
`region of the gel is indicative of deletions within the Bat-26 polyA tract
`sequence. In addition to the stool samples analyzed (lanes1–10), a
`no-DNA negative control (lane 11) and a wild-type-only DNA positive
`control (lane12) were analyzed. Deletion-positive controls containing
`1% (lane13) and 5% (lane14) mutant DNA (15 base deletions) were
`also analyzed within each assay to assure that resolution between
`wild-type and deleted sequences was achieved. Note that 2 of the
`samples (lanes 2 and 9) contain deletions within Bat-26, while all
`remaining samples contain the wild-type number of bases.
`
`products were generated in duplicate (or 14 amplifications per
`subject) and independently scored by 2 technicians. PCR
`product intensities were scored as high (A), medium (B), or
`low (C) by visual examination of the gel image. Figure 1
`illustrates examples of samples scored as A amplifications
`(lanes 1ab 1 4ab), B amplifications (lanes 2ab, 5ab, and 6ab),
`and C amplifications (lane 3ab). The cutoff score to indicate a
`positive result was determined in pilot study 1.
`Sequence information for all capture probes and primers will
`be available on request to the corresponding author.
`
`Fecal Occult Blood Testing
`All stools collected from subjects with adenomas and
`normal colonoscopy had been tested in blinded fashion by
`Hemoccult II (SmithKline Diagnostics, Sunnyvale, CA) im-
`mediately on receipt and before freezing. From a single stool,
`2 aliquots sampled from opposite ends of the specimen were
`each smeared onto 2 windows of a Hemoccult II test card for
`a total of 4 windows. A single drop of peroxide catalyst was
`promptly added to each window, and a blue color reaction
`within 60 seconds on at least 1 of the 4 test windows was
`called a positive result for that stool. Positive and negative
`controls were tested with each run. Stools were selected from
`the archive for study without knowledge of Hemoccult status.
`
`Tissue Processing and Assay
`DNA from colorectal cancers in pilot study 1 was
`obtained at the Mayo Clinic after microdissection of sections
`from the original paraffin-embedded, formalin-fixed tumor
`
`Geneoscopy Exhibit 1065, Page 4
`
`

`

`November 2000
`
`STOOL SCREENING FOR ALTERED DNA 1223
`
`Table 2. Positivity Rates of the Fecal DNA Assay Panel by Subject Group From Each Pilot Study and From Combined Studies:
`Component Markers and All Markers Together
`
`K-ras
`
`APC
`
`p53
`
`Bat-26
`
`L-DNAa
`
`n (%)
`
`95% CI
`
`All markers
`
`Pilot 1
`Cancers (21)
`Adenomas (9)
`Normals (10)
`Pilot 2
`Cancers (8)
`Adenomas (2)
`Normals (18)
`Combined studies
`Cancers (22)b
`Adenomas (11)
`Normals (28)
`Combined studies (excluding K-ras)
`Cancers (22)b
`Adenomas (11)
`Normals (28)
`
`4
`0
`0
`
`2
`1
`2
`
`4
`1
`2
`
`—
`—
`—
`
`5
`3
`0
`
`2
`0
`0
`
`5
`3
`0
`
`5
`3
`0
`
`3
`0
`0
`
`1
`0
`0
`
`3
`0
`0
`
`3
`0
`0
`
`4
`0
`0
`
`2
`0
`0
`
`5
`0
`0
`
`5
`0
`0
`
`14
`5
`0
`
`4
`1
`0
`
`14
`6
`0
`
`14
`6
`0
`
`19 (90)
`7 (78)
`0 (0)
`
`8 (100)
`2 (100)
`2 (11)
`
`20 (91)
`9 (82)
`2 (7)
`
`20 (91)
`8 (73)
`0 (0)
`
`70%–99%
`40%–97%
`0%–31%
`
`63%–100%
`16%–100%
`1%–35%
`
`71%–99%
`48%–98%
`1%–24%
`
`71%–99%
`39%–94%
`0%–12%
`
`aL-DNA refers to “long” or nonapoptotic DNA.
`bBecause stools from 7 cancer patients from pilot 1 were repeated in pilot 2, total number of unique cancer patients was 22.
`
`detected in the stool only in 4 cases and the tissue only
`in 3 cases.
`Adenomas and normal controls. The fecal DNA
`panel detected 7 (78%) of the 9 adenomas $1 cm. All
`positives were the result of APC mutations or elevated
`L-DNA (Table 2). The assay panel was negative in all 10
`colonoscopically normal patients.
`Perfect
`for L-DNA.
`Interobserver
`variability
`agreement between the 2 technicians scoring L-DNA
`ranged from 88% to 96% for duplicate testing on the 7
`PCR amplification products; perfect agreement was
`.93% for 6 of the 7 products. Weighted k values for
`individual PCR products ranged from 0.36 6 0.09 to
`0.74 6 0.07; the pooled weighted k value across all 7
`PCR products was 0.58 6 0.12.
`
`Pilot Study 2
`
`All 8 of the cancers and both of the adenomas
`were detected by the fecal DNA assay panel using the
`same assay parameters as in pilot study 1 (Table 2). Of
`the 7 cancer patients in this series who had also been
`evaluated by separate fecal aliquots in pilot study 1, at
`least 1 identical DNA alteration was reproduced in all 7
`instances; all component markers of the assay panel were
`concordant in 5 instances and at least 1 marker was
`discordant in the other 2 instances. The assay panel was
`positive in 2 (11%) of the 18 colonoscopically normal
`controls. Both of these false-positive cases were a result of
`K-ras mutations.
`
`Both Studies Combined
`Sensitivity and specificity. Using the full panel
`of component markers, the sensitivity of the fecal DNA
`assay panel for the 22 cancers was 91% (95% CI, 71%–
`99%) and for the 11 adenomas $1 cm was 82% (95%
`CI, 48%–98%) with a specificity of 93% (95% CI,
`76%–99%). If K-ras markers were excluded from the
`panel, then sensitivity for cancer was unaffected at 91%
`(95% CI, 71%–99%) but decreased slightly for adeno-
`mas to 73% (95% CI, 39%–94%) while specificity in-
`creased to 100% (95% CI, 88%–100%).
`For all neoplasms (cancers and adenomas), L-DNA
`proved to be the most informative marker and alone
`detected 20 (61%) of the 33 unique lesions. Bat-26 and
`p53 markers were positive with cancer but not with
`adenomas in this initial series. In stools from the 20
`cancer patients with a positive DNA assay panel, 3
`component markers were positive in 3 cases, 2 markers
`positive in 5 cases, and a single marker positive in 12
`cases. In stools from the 9 test-positive adenoma patients,
`2 component markers were positive in 2 cases and a
`single marker was positive in the other 7.
`In this highly selected subject group, the positive
`predictive value for colorectal neoplasia by the fecal
`DNA panel (excluding K-ras markers) was 100% (28/28)
`and the negative predictive value was 85% (28/33).
`Clinical correlates. Positive results by the com-
`ponent marker Bat-26 were significantly associated with
`proximal colorectal cancer site, and an association of
`positive L-DNA results with distal tumor site was sug-
`
`Geneoscopy Exhibit 1065, Page 5
`
`

`

`1224 AHLQUIST ET AL.
`
`GASTROENTEROLOGY Vol. 119, No. 5
`
`gested (Table 3). All 5 Bat-26 –positive tumors were
`located proximal to the splenic flexure (4 in the cecum or
`ascending colon). Conversely, 10 of the 14 L-DNA–
`positive cancers were located distal to the splenic flexure.
`Outcomes by the fecal DNA assay panel were not
`associated with subject age or gender or with lesion size
`or stage. Because all adenomas were classified as low-
`grade dysplasia and all but one as tubular, neither his-
`tologic grades nor architectural type could be evaluated
`in this small series.
`Comparison of the fecal DNA assay panel and
`Hemoccult tests. Hemoccult testing had been per-
`formed only on the asymptomatic adenoma and control
`subjects. Whether or not K-ras was included, the DNA
`assay panel detected substantially and significantly more
`of the adenomas $1 cm than did the Hemoccult test
`based on a single stool per subject (Table 4). Hemoccult
`was negative in stools from all 28 normal controls. Thus,
`at matched specificities of 100% in this small series, the
`DNA assay panel (minus the K-ras marker) detected 8
`(73%) of the 11 adenomas compared with none (0%) by
`the Hemoccult test (P , 0.008).
`
`Discussion
`The rationale and appeal of stool screening for
`colorectal cancer would be strengthened if markers sub-
`stantially more accurate than occult blood were used. On
`the basis of this pilot investigation, the aggregate assay
`of multiple genetic markers exfoliated into stool repre-
`sents a promising alternative. The panel of neoplasm-
`associated DNA alterations targeted in this study highly
`discriminated patients with colorectal cancer or large
`adenomas from those with endoscopically normal colons.
`The fecal DNA assay panel detected 20 (91%) of the
`22 colorectal cancers in this blinded pilot. A likely factor
`contributing to high neoplasm detection rates by the
`
`Table 3. Components of the Fecal DNA Assay Panel
`Associated With Colorectal Cancer Site
`
`Positive markers by colorectal site
`
`Proximal to
`splenic flexure
`
`Distal to
`splenic flexure
`
`4/9
`2/4
`5/10 (50%)
`
`4/9
`0/4
`4/10 (40%)
`
`0/12
`0/4
`0/12 (0%)
`
`10/12
`4/4
`10/12 (83%)
`
`Marker
`
`Bat-26
`Pilot 1
`Pilot 2
`Combined studiesa
`L-DNA
`Pilot 1
`Pilot 2
`Combined studiesb
`
`aP 5 0.01.
`bP 5 0.07.
`
`Table 4. Comparison in Adenoma Detection Rates by the
`Fecal DNA Assay Panel and Hemoccult Test
`
`DNA assay panel (all markers)
`DNA assay panel minus K-ras
`Hemoccult test
`
`Adenoma detection rate
`
`9/11 (82%)a
`8/11 (73%)b
`0/11 (0%)
`
`NOTE. Both tests were performed on same single stool per patient.
`aP 5 0.004.
`bP 5 0.008.
`
`DNA assay panel, in addition to the multiple markers
`targeted, was the efficient isolation of human DNA from
`stool. Amplifiable human DNA was recovered from all
`stools in this pilot compared with recovery rates of
`45%–90% by others.14 –20 This difference may, in part,
`be accounted for by the more effective removal of PCR
`inhibitors, known to be present in stool,31 by the se-
`quence-specific hybrid capture method used. However, it
`must be emphasized that the cancers included in this
`pilot study were largely symptomatic and may behave
`differently from asymptomatic cancers with respect to
`marker shedding. Another variable that could influence
`sensitivity is the representativeness of the fecal aliquot
`tested, and the degree of tissue-stool discordance ob-
`served would suggest that specimen sampling was sub-
`optimal in this pilot study. The effect of different sam-
`pling techniques on test performance will require further
`evaluation.
`A major limitation of stool screening has related to
`the very low sensitivity of occult blood as a marker for
`premalignant adenomas.5,6 The functional inadequacy
`of fecal occult blood tests to detect adenomas is evi-
`denced practically by their failure in controlled trials
`to meaningfully reduce the cumulative incidence of
`colorectal cancer.2– 4 In the present small initial series,
`the fecal DNA assay panel detected a remarkably high
`proportion (73%– 82%) of adenomas $1 cm, whereas
`single-stool Hemoccult testing in the same asymp-
`tomatic patients detected none. The results by this
`multitarget DNA assay approach must be corrobo-
`rated in larger clinical studies, but the clear implica-
`tions are for improved cancer prevention and, perhaps,
`less frequent screening.
`The presence of high-integrity DNA, or L-DNA,
`proved to be the most informative component marker
`of the fecal DNA assay panel and alone detected 61%
`of neoplasms. We speculate that longer template
`DNA is an epigenetic phenomenon consistent with
`the known abrogation of apoptosis that occurs with
`colorectal cancer.32,33 There appears to be abundant
`exfoliation of nonapoptotic cells from neoplasms34; in
`contrast, colonocyte shedding from normal mucosa is
`
`Geneoscopy Exhibit 1065, Page 6
`
`

`

`November 2000
`
`STOOL SCREENING FOR ALTERED DNA 1225
`
`relatively sparse, and sloughed cells appear to be
`largely apoptotic.34,35 Furthermore, normal cells rap-
`idly undergo apoptosis after detachment from their
`basement membrane.36 Because a hallmark of apopto-
`sis is the cleavage of DNA by endonucleases into
`fragments of 180 –200 bp,37,38 it follows that human
`DNA in normal stools would exist primarily in frag-
`mented or “short” forms. However, stools from pa-
`tients with colorectal neoplasia should contain subsets
`of both nonapoptotic or “long” DNA arising from
`dysplastic cells and “short” DNA from normal mu-
`cosa. Because the primers used in the L-DNA assay
`were set to amplify human DNA sequences longer
`than 200 bp, an increase in the number of amplifiable
`template molecules serves as a logical marker for
`exfoliated dysplastic cells. In preliminary experiments,
`we have found that amplification of even longer hu-
`man DNA sequences (e.g., 1800 –2400 bp) from stool
`may allow for a more dramatic and objective separa-
`tion of cancer patients from controls.39 Further opti-
`mization and validation of this marker are needed.
`High specificity is a requisite for cost-effective screen-
`ing. Two false-positives were identified in this small
`series, both a result of K-ras mutations. Mutant K-ras
`can potentially arise from nonneoplastic sources, such as
`pancreatic hyperplasia40 or hyperplastic aberrant crypt
`foci throughout the colorectum,41 or even normal-ap-
`pearing colonic mucosa.42 Exclusion of mutant K-ras
`improved assay specificity to 100% with no loss in the
`cancer detection rate and only minimal loss in adenoma
`detection. The finding of a fecal DNA alteration in a
`patient with normal colonoscopic findings could indicate
`a supracolonic neoplasm. Indeed, the control patient
`disqualified from the present study because a malignant
`ileal carcinoid tumor was subsequently found to have a
`positive fecal DNA assay panel due to the L-DNA com-
`ponent. Further observations are needed to determine the
`positive predictive value of each component marker of
`the fecal DNA assay panel for supracolonic neoplasms in
`colonoscopy-negative subjects and the implications on
`screening algorithms. Finally, 2 recent studies have
`shown that Bat-26 polymorphisms occur in some African
`Americans,43,44 which could potentially yield false-posi-
`tive fecal Bat-26 results. As such, alternative markers of
`microsatellite instability that are less affected by these or
`other polymorphisms might be considered.
`Stool findings in the present study were consistent with
`certain biological features of colorectal neoplasia previously
`observed in tissue. Multiple DNA alterations were more
`frequently recovered in stools from patients with cancer
`than from those with adenomas, in keeping with the known
`
`accumulation of gene mutations that occurs during the
`adenoma-to-carcinoma progression.10,11 The microsatellite
`instability marker, Bat-26, was positive only in stools of
`patients with proximally located colon cancer, reflecting a
`site-association well established from tissue studies.45 Fi-
`nally, neither Bat-26 nor p53 markers were positive with
`adenomas in this stool study, which would be in accord
`with previous tissue studies reporting that these 2 markers
`are rarely expressed before adenomatous transformation to
`high-grade dysplasia or frank cancer has occurred.46 – 48 Re-
`sults from these 2 component markers of the fecal DNA
`assay panel may have been different had adenomas with
`high-grade dysplasia been included.
`Detection of altered human DNA in stool using a
`multitarget assay panel appears to be a feasible method to
`screen both cancers and premalignant adenomas of the
`colorectum. Larger clinical studies are clearly indicated
`using optimal sample preparation and marker selec