Detection of stool DNA mutations before and after treatment of colorectal neoplasia
Whether stool DNA abnormalities arise solely from colorectal neoplastic lesions or are due to more pervasive field effects is not known. In the current study, the authors conducted a prospective multicenter study to evaluate the performance of stool-based DNA testing in a large cohort and to examine whether the findings before treatment persist after surgical resection and/or adjuvant therapy.
Patients with newly diagnosed colorectal carcinoma or advanced adenomas (AA) provided stool samples before therapy, 1–3 months after surgical resection, and 6–9 months postresection. Stool samples were analyzed using the multitarget DNA assay panel (MTAP) consisting of 23 markers: 21 mutations in the p53, K-ras, and APC genes, a microsatellite instability marker (BAT-26), and the DNA integrity assay (DIA), a marker of loss of apoptosis.
Overall, 49 of 91 individuals (54%) tested positive with the MTAP test. The sensitivity of the MTAP test was 63% for invasive tumors compared with 26% for AA. Individuals whose lesions had a more advanced TNM stage or were located distal to the splenic flexure were significantly more likely to have a positive MTAP test. Of the 79 samples collected at 1–3 months after surgical resection of the neoplasm, 14 (18%) had a positive MTAP result, 12 of which were positive for DIA only. Of those collected at 6–9 months of follow-up, 5 of 72 (7%) tested positive on the MTAP panel.
Although many samples collected 1–3 months after surgical resection of the colorectal neoplasm tested positive on the MTAP, most were negative by 6–9 months, indicating that stool DNA abnormalities disappear after treatment of the neoplastic lesions. Surgery and chemoradiation appear to induce transient DIA abnormalities that may be independent of the presence of neoplasia. Cancer 2006. © 2005 American Cancer Society.
Colorectal carcinoma is the third most common cancer in the U.S. and accounts for more than 56,000 deaths annually, second only to lung carcinoma.1 Innovations in techniques for DNA isolation, amplification, and analysis have made it possible to detect genetic mutations in colorectal tumors and in stool samples from patients with colorectal neoplasia. Mutations in tumor suppressor genes, such as APC and p53 can be found in 70% and 50% of colorectal tumors, respectively.2 K-ras, an oncogene, is mutated in approximately 50% of colorectal carcinomas.3, 4 Errors in DNA repair mechanisms resulting in basepair mismatches can manifest as instability in microsatellite markers such as BAT-26, which can be found in some tumors.5
There have been several studies examining the performance of techniques for stool DNA mutation analysis in detecting colorectal carcinoma. Ahlquist et al.6 tested archived stool specimens using a panel of 15 mutations in APC, K-ras, p53, and BAT-26, as well as highly amplifiable DNA (a marker for impaired apoptosis in colorectal tumors) and reported a sensitivity of 91% for colorectal carcinoma and 82% for adenomas measuring > 1 cm. Subsequent studies using various mutation marker panels have demonstrated sensitivities ranging from 37–71% for the detection of carcinoma and 27–57% for adenomas.7–10 Tagore et al.10 reported a sensitivity of 63.5% for colorectal carcinoma and 57.1% for advanced adenomas in distal lesions using a 23-marker panel. Similarly, Calistri et al.11 found alterations in K-ras, p53, APC, microsatellite markers, or DNA amplification in 62% of stool samples collected from subjects with colorectal carcinoma and reported that tumor location in the distal colon was associated with higher rates of detection of DNA alterations. Most recently, Imperiale et al.12 prospectively examined the performance of a multitarget assay panel among 4404 asymptomatic individuals undergoing screening colonoscopies, reporting sensitivities for detection of cancer and adenomas with high-grade dysplasia of 52% and 33%, respectively, with a specificity of 94%.
In addition to the wide variation in sensitivity estimates using different samples and mutation analysis technologies, these earlier studies have provided little information regarding factors that may affect the sensitivity of stool DNA mutation analysis. Furthermore, to our knowledge, no studies to date have addressed the issue of whether the DNA mutations found in the stool of patients with colorectal neoplasia persist after treatment of the lesion through surgical resection and/or chemoradiation. The question of whether the stool DNA abnormalities represent ‘field effects’ or whether they arise directly from the neoplastic lesions has important implications for test specificity and utility for monitoring patients treated for colorectal neoplasms.
We conducted a prospective study to examine the performance of a 23-marker stool DNA assay panel in patients with endoscopically diagnosed colorectal carcinoma or advanced adenomas ≥ 1 cm. The primary objectives of the study were to 1) define the sensitivity of a refined multitarget assay panel (MTAP) for detecting DNA abnormalities in stool specimens collected from a large cohort of individuals with colorectal carcinoma or advanced adenomas; and 2) to prospectively examine whether the mutations detected in stool DNA before treatment persist after surgical resection and/or adjuvant therapy.
MATERIALS AND METHODS
Individuals with a newly diagnosed colorectal carcinoma or advanced adenoma measuring ≥ 1 cm were eligible to participate in the study. All potential subjects had been diagnosed with a colorectal neoplasm during a recent colonoscopy or sigmoidoscopy. Eligible individuals who had not yet undergone treatment of the neoplasm (surgical resection, chemotherapy, or radiation therapy) were enrolled in the study at six medical centers in the Boston area. To minimize any potential effect of the bowel-cleansing process for the original diagnostic procedure on detection of mutation abnormalities, subjects who underwent treatment of the neoplasm (surgical resection or neoadjuvant chemoradiation) less than 14 days after the endoscopic procedure were ineligible for the study. All subjects were enrolled under an Institutional Review Board (IRB)-approved protocol and provided written informed consent before enrollment.
Clinical characteristics of participants were collected at the time of subject enrollment and medical record reports (endoscopy, surgery, pathology, radiology reports) were reviewed for data corresponding to diagnosis and staging of the neoplastic lesion. The size of the neoplastic lesion was determined based on pathology reports from the surgically excised specimen. In cases in which no excision was performed, the size of the lesion was determined from the description provided in the endoscopy report. American Joint Commission on Cancer (AJCC) tumor classification was determined by review of clinical notes and pathologic and radiologic tests. Follow-up clinical information was obtained after collection of the final stool specimen. Available pathologic, radiologic, and endoscopic reports as well as physician notes were reviewed to assess evidence of disease recurrence or progression.
The baseline stool specimen was collected at least 14 days after the endoscopic procedure but before surgery or the commencement of chemoradiation treatment. The 14-day interval was chosen under the assumption that bowel function would return to normal within this interval after the preendoscopy bowel cleansing. Subsequent posttreatment stool samples were collected approximately 1–3 months and 6–9 months after surgery.
Research personnel contacted subjects and provided them with detailed oral and written instructions for stool collection. Subjects were instructed to collect one whole stool specimen for analysis using a specialized container that could be mounted on the toilet seat. Stool specimens were transported directly from the subject's home to EXACT Sciences' laboratory (Marlborough, MA) either at room temperature within 24 hours of collection or in chilled coolers within 72 hours of collection. At the laboratory, all specimens were weighed and then stored at −80°C until testing.
Multitarget DNA Assay Panel
The MTAP consisted of 23 markers: 21 discrete mutations in K-ras, APC, and p53, a microsatellite instability marker (BAT-26), and a marker of disordered apoptosis (DIA). The results of the MTAP were interpreted as positive if any component of the panel was positive.
The clinical laboratory processed samples in an automated manner without knowledge of the clinical diagnosis. Stools samples were thawed at room temperature and then homogenized. Aliquots of stool were centrifuged to remove particulate matter. DNA was precipitated and resuspended in TE buffer (0.01 mol/L of Tris [pH 7.4] and 0.001 mol/L of ethylenediamine tetraacetic acid [EDTA]). Sequence-specific DNA fragments were purified from the total nucleic acid preparations by performing oligonucleotide-based hybrid captures. For each sample, seven unique hybrid capture reactions were performed. The detection of 21 separate mutations (3 in K-ras, 10 in APC, and 8 in p53) was performed on the polymerase chain reaction (PCR) products amplified from captured DNA using a modified solid-phase minisequencing method as previously reported.10 Minisequencing products were analyzed on an ABI PRISM 3100 Capillary Electrophoresis unit (Applied Biosystems, Foster City, CA). Point mutations were considered positive if capillary electrophoresis values exceeded site-specific quantified detection thresholds. Size discrimination of reaction products was used to detect BAT-26 mutations associated with a deletion of 4–15 base pairs.
Ahlquist et al.6 previously reported the sample preparation methodology used for the DIA assay. Captures of four separate loci (APC, p53, BRCA1, and BRCA2) were performed according to the methods described.6 PCR was performed using primers that amplified 1.3-kilobase (kb), 1.8-kb, and 2.4-kb fragments. PCR products were run on agarose gels and were imaged on a Stratagene (La Jolla, CA) EagleEye II still image system. Image files were imported into National Institutes of Health (NIH; Bethesda, MD) Image V. 1.62 software for Macintosh (Apple Computer, Inc., Cupertino, CA). Densitometry values were derived from the integration of the area under the peak heights for each band of interest. Individual samples were considered positive if 3 or more of the 12 target fragments had densitometry values higher than the predetermined cutoff.
Subjects were categorized based on the pathology and staging identified before preoperative neoadjuvant therapy or at subsequent surgery. Neoplastic lesions were classified as invasive carcinoma or advanced adenomas. Advanced adenomas were defined as any adenoma measuring ≥ 1 cm containing low-grade dysplasia (LGD) or high-grade dysplasia (HGD). If no residual tumor was identified at surgery or if the tumor was limited to lymph nodes, the subject was excluded because it was assumed that all endoluminal neoplastic tissue had been removed on the initial endoscopy, thus preventing neoplastic cells from being shed in baseline samples collected after the endoscopic resection. The sensitivity of the multitarget DNA assay panel was assessed for all neoplastic lesions and for invasive carcinomas and adenomas separately, using 95% confidence intervals (95% CI) based on the exact binomial distribution.
A total of 135 subjects with colorectal carcinoma or adenoma were enrolled in the study between September 2001 and April 2003 (Table 1). Samples from 44 subjects were not available for analysis: 10 subjects provided a baseline stool sample that was inadequate (< 30 g) for processing, 24 subjects did not provide a baseline sample before surgery or the beginning of their adjuvant treatment, and 2 subjects provided samples but were later found not to meet the study inclusion criteria. Eight additional individuals who submitted baseline stool samples were excluded from the analysis of test sensitivity because there was no intraluminal residual lesion found at surgery or follow-up endoscopy. Therefore, 91 subjects (67%) were eligible and provided a baseline stool sample that was adequate for analysis.
Table 1. Evaluable Stool Samples
Baseline Stool DNA Abnormalities
Results of the baseline stool analysis from the 91 evaluable individuals (68 with invasive carcinoma and 23 with advanced adenomas) are presented in Table 2. The age range for the study participants was 37–83 years with a median age of 67 years. Forty-five percent of the participants were female. Thirty individuals (44%) had invasive carcinoma of TNM Stage III or IV and 38 (56%) had TNM Stage I or II disease. Of the 23 patients with advanced adenomas, 12 (52%) had HGD and 11 (48%) had LGD. Forty-three percent of neoplastic lesions (39 of 91 lesions) were located proximal to the splenic flexure. The median size of the neoplastic lesions at the time of surgical resection was 3.0 cm (range, 0.1–11 cm).
Table 2. Results of the Baseline Samples
|Total subjects||91||49||54% (43–64%)|
|Pathologic classification|| || || |
| Invasive carcinoma||68||43||63% (50–75%)|
| TNM Stage I||18||7||39%|
| TNM Stage II||20||14||70%|
| TNM Stage III||29||21||72%|
| TNM Stage IV||1||1||100%|
| Adenoma||23||6||26% (10–48%)|
|Location|| || || |
| Proximal||39||15||38% (23–55%)|
| Distal||52||34||65% (51–78%)|
|Tumors by size|| || || |
| < 2.0 cm||20||10||50%|
| 2.0–4.9 cm||32||21||66%|
| > 5.0 cm||13||10||77%|
|Adenomas by size|| || || |
| < 2.0 cm||7||1||14%|
| 2.0–4.9 cm||13||3||23%|
| > 5.0 cm||3||2||67%|
The overall sensitivity of the MTAP for detecting neoplastic lesions was 54% (Table 2). Among the 68 invasive carcinoma cases, 43 had a positive test resulting in a sensitivity of 63% (95% CI, 50–75%). When examined by pathologic stage, the assay detected 72% of Stage II, III, and IV tumors compared with 39% of Stage I carcinoma (P < 0.02). The MTAP detected 6 of the 23 adenomas (26%; 95% CI, 10–48%) and appeared to be more sensitive for HGD compared with LGD (33% vs. 18%, difference not significant), although the analysis was limited by small numbers. When examined based on location, 65% of lesions distal to the splenic flexure were detectable using the MTAP, compared with 38% of proximal lesions (P = 0.02).
The breakdown of results based on panel markers, disease stage, and anatomic location is shown in Tables 3 and 4. Of the 49 samples that tested positive on the MTAP, 23 (47%) had 2 or more positive markers. K-ras was the most commonly identified gene mutation, found to be present in 20 of 49 positive tests (41%), including distal (n = 13) and proximal (n = 7) lesions. The DIA marker also was found to be present in 20 of 49 positive tests (41%); however, DIA was found to be positive exclusively in subjects with tumors in the distal colon and was not found in proximal tumors or advanced adenomas. APC and p53 mutations were found in 17 of 49 (35%) and 13 of 49 (27%) of the cases with detectable DNA abnormalities, respectively. Although BAT-26 alterations were found in only 6 specimens (12%), all occurred in subjects with lesions located proximal to the splenic flexure.
Table 3. Distribution of Mutations in Relation to Stage
|TNM Stage III & IV||22/30 (73%)||9||5||8||1||14|
|TNM Stage I & II||21/38 (55%)||8||8||5||5||6|
|Advanced adenoma||6/23 (26%)||3||4||0||0||0|
Table 4. Distribution of Mutations in Relation to Location
Treatment Effects on Stool DNA Abnormalities
Of the 91 evaluable subjects who submitted baseline stool samples before undergoing treatment of their neoplasm, 79 (87%) submitted a follow-up stool sample at approximately 1–3 months and 72 (79%) at approximately 6–9 months after surgical resection. Clinical follow-up information for the samples received 6–12 months after surgery was available for 52 of the 79 subjects who provided follow-up specimens (66%). Results of the follow-up specimens are shown in Tables 5 and 6.
Table 5. Comparison of Follow-Up and Baseline Stool DNA Results in Positive Cases after Surgical Resection (1–3 Months Postoperative)a
|MG-044||II||p53 (175p.2)||p53 (245p.2)||NA||Negative|
Table 6. Comparison of Follow-Up and Baseline Stool DNA Results in Positive Cases Posttreatment (6–9-Month Postsurgical Resection)a
|MG-033||II||K-ras (k12p.2)||Negative||Chemo/rad||K-ras (k12p.2)|
Among samples collected at 1–3 months after surgical resection of the neoplasm, 14 of 79 (18%) had a positive MTAP result; of the 14 positive results, 12 were positive for DIA and 2 were positive for a DNA mutation (Table 5). Ten of the 12 patients with positive DIA assay after surgery had had a negative DIA assay at baseline. Both subjects with DNA mutations were found to have a p53 mutation; one had the same p53 mutation that had been found in the baseline stool sample and the other had concomitant diagnoses of colorectal carcinoma and leukemia with a negative baseline MTAP.
At 6–9 months postresection, only 5 of 72 samples (7%) had a positive MTAP result; 2 were positive for DIA alone and 3 had detectable DNA mutations (Table 6). Neither of the individuals who were positive for DIA alone were found to have evidence of recurrent neoplasia on follow-up tests. However, of the three subjects with detectable DNA mutations, one subject had known liver metastases, one had indeterminate hepatic lesions, and one had an elevated carcinoembryonic antigen level but no other objective evidence of metastatic disease.
We designed our study with two objectives in mind. The first goal was to provide a more refined estimate of the sensitivity of the currently available technology for the detection of stool DNA abnormalities in patients with colorectal neoplasia compared with prior studies, which had limitations due to the use of archival stool specimens, highly selected patient populations, small sample sizes, or the inclusion of more readily accessible distal colonic tumors. The second objective was to address the biologic question of the origin of the detected DNA abnormalities. Although the most likely answer was that mutational abnormalities that were being detected in stool arose from the tumor itself, and would resolve with the removal of the primary lesion, the specific question of whether some of these changes might be more pervasive, or ‘field effects,’ in the colon had not been addressed. The results from the current study address both of these questions.
In the current study, which to our knowledge represents the largest cohort of subjects with colorectal neoplasia analyzed to date, the stool MTAP detected 63% of invasive carcinomas and 26% of advanced adenomas using a single prospectively collected stool sample. Among the 49 specimens that tested positive on the assay, the DNA abnormalities detected most often were K-ras mutations and DIA alterations (each found in 41% of positive tests). The sensitivity of the MTAP was significantly higher for larger tumors at a more advanced pathologic stage and for tumors located in the distal as compared with the proximal colon.
Previously published studies using a variety of fecal DNA markers or panels have reported sensitivities of fecal DNA tests ranging from 37–91%, depending on the composition of the assay panel.6–12 Although Ahlquist et al.6 reported a sensitivity of 91% for colorectal carcinoma using a 15-mutation panel examining fewer mutations than the one used in the current study, their analysis tested archived stool specimens from 22 colorectal patients, the majority of whom had relatively large tumors (mean, 4.1 cm; range, 2.5–11 cm) compared with those assayed in the current study. In a recent study using prospectively collected samples of colon carcinomas distal to the splenic flexure, Tagore et al.10 analyzed 80 samples (52 colorectal carcinoma samples and 28 advanced adenoma samples) using the same MTAP panel as used in the current study and reported a sensitivity of 63.5% for detecting colorectal carcinoma. Our study, which included both proximal and distal carcinomas, found a nearly identical sensitivity estimate of 63%, confirming that this is what one can reasonably expect for test sensitivity in a patient population with malignancy with the currently available stool DNA tests.
We did not compare the performance of the MTAP test with other noninvasive tests for colorectal carcinoma screening. In the recent study of stool DNA testing in asymptomatic, average-risk subjects, Imperiale et al.12 reported that MTAP detected 16 of 31 carcinomas (51.6%) compared with 4 of 31 carcinomas (12.9%) detected by fecal occult blood testing, with nearly identical specificity for the two tests. Our study of the performance of the MTAP test on samples from 91 patients with colorectal neoplasia sheds some light on the parameters that may affect MTAP test performance and can hopefully help target areas for further refinement in stool DNA detection technology. We found increasing disease stage of the neoplasm and distal location to be significant predictors of a positive MTAP test result. It is a reasonable hypothesis that abnormal DNA may be shed from larger lesions in larger amounts than from smaller lesions. Furthermore, abnormal DNA shed from distal lesions may have less time to undergo degradation and therefore may be more readily detectable. Improvements in DNA isolation technology will likely improve test sensitivity and may lead to improved detection of proximal neoplasms as well as smaller tumors.
Sensitivity for the detection of large adenomas in the current series was 26%, which is higher than the 18% sensitivity for advanced neoplasia reported by Imperiale et al.12 but lower than the 57.1% reported by Tagore et al.10 In terms of broad implementation of the stool MTAP as a screening test, the inability to detect adenomas is an important limitation when compared with invasive tests such as colonoscopy13 or computed tomography colography.14 The addition of other DNA mutation markers as well as analyses that detect other mechanistic events in colorectal pathogenesis, such as DNA methylation,15 may lead to further improvements in test performance, particularly for advanced adenomas.
The second component of the current study evaluated the effect of treatment on stool DNA abnormalities. We were surprised to find that treatment effects on detection of stool DNA abnormalities depended on the type of abnormality being assessed, with clear differences noted between mutational changes and the DIA assay, a marker of loss of apoptosis. DIA abnormalities were found in 15% of specimens collected at 1–3 months after surgical resection. Many of the patients with posttreatment abnormalities did not have detectable DIA abnormalities when their neoplasm was first detected. DIA abnormalities disappeared in all but 2 subjects (3%) by 6–9 months after surgical resection of the lesion. These results suggest that both chemoradiation and surgery induce transient increases in long DNA that were detectable by the DIA component of the MTAP. The mechanisms underlying this alteration in stool DNA over time are unclear, but may be related to the effect of inflammation (and the detection of fecal leukocyte DNA) in the immediate postoperative period.
In contrast, stool DNA panel results, exclusive of DIA, were found to be well correlated with the presence or absence of tumor. Of 42 subjects who had positive stool DNA mutation tests at baseline, DNA mutations were detected in 2 subjects at 1–3 months, and only 3 subjects had detectable stool DNA mutation abnormalities at 6–9 months after surgical resection. Because of the limited availability of clinical follow-up information on these subjects, it is unclear whether these positive results were the result of residual or recurrent disease, represent an uncommon field effect, or represent a false-positive assay. Additional studies are needed to further examine this finding.
The follow-up results raise some interesting issues regarding test specificity in implementing the stool MTAP test in the population at large. Similar to the current study, most prior studies were performed on patients with colorectal neoplasia, and a positive DIA assay is one of the most common reasons for test positivity. Because our study clearly demonstrates the potential effects of other parameters that may affect DIA results, it will be important to evaluate similar potential effects of this marker in other patient populations. If the presence of inflammation is the primary mechanism behind this DIA positivity, then test interpretation in patients with inflammatory bowel disease, for example, may not be able to include DIA as a reliable marker for the prediction of neoplasia. Further studies are needed to clarify the etiology of DIA positivity.
The results of the current study demonstrate that stool DNA mutation abnormalities disappear after removal/treatment of the colorectal neoplasm. Additional studies with longer clinical follow-up are needed to assess the utility of the stool MTAP test for the posttreatment monitoring of patients with colorectal neoplasia. Although the sensitivity of stool DNA testing for detecting colorectal neoplasia appears to be lower than that of conventional colonoscopy, noninvasive screening alternatives may have a role in the care of some patients.
The sensitivity of a multitarget assay panel for stool DNA abnormalities was 63% for colorectal carcinomas and 26% for advanced adenomas. Mutational DNA abnormalities resolved for most patients with treatment of the primary lesion, whereas markers of abnormal apoptosis were transiently affected by treatment.
The authors thank the following physicians for patient contribution: Stanley Ashley, David Berger, Ronald Bleday, Elizabeth Breen, David Brooks, Robert Cima, James Cusack, Charles Ferguson, Carlos Fernandez, John Jao, Fredrick Makrauer, Harvey Mamon, Daniel Matloff, Norman Miller, Albert Namias, David Rattner, Paul Shellito, Ken Tanabe and Michael Zinner. They also thank study staff Lisa Abel, Jameel Ali, Marie Bonneau, Anne Burgess, Kerry Collier, Leslie Cheryl Frank, Heidi Judge, Rebecca Liberman, and Sheila Wilson for assistance in patient recruitment and sample procurement, and the clinical laboratory staff at Exact Sciences for processing the samples.