Fax: (011) 31 20 6977192
Efficient automated assessment of genetic abnormalities detected by fluorescence in situ hybridization on brush cytology in a Barrett esophagus surveillance population
Article first published online: 23 MAR 2007
Copyright © 2007 American Cancer Society
Volume 109, Issue 10, pages 1980–1988, 15 May 2007
How to Cite
Rygiel, A. M., van Baal, J. W. P. M., Milano, F., Wang, K. K., ten Kate, F. J., Fockens, P., Rosmolen, W. D., Bergman, J. J. G. H. M., Peppelenbosch, M. P. and Krishnadath, K. K. (2007), Efficient automated assessment of genetic abnormalities detected by fluorescence in situ hybridization on brush cytology in a Barrett esophagus surveillance population. Cancer, 109: 1980–1988. doi: 10.1002/cncr.22643
- Issue published online: 25 APR 2007
- Article first published online: 23 MAR 2007
- Manuscript Accepted: 18 JAN 2007
- Manuscript Revised: 10 JAN 2007
- Manuscript Received: 5 DEC 2006
- Dutch Cancer Society
- Barrett esophagus;
- brush cytology;
- automated fluorescence in situ hybridization analysis;
- cytogenetic abnormalities
Automated assessment of genetic abnormalities detected by fluorescence in situ hybridization (FISH) in brush cytology specimens from patients with Barrett esophagus (BE) may enhance the clinical applicability of this methodology. The objectives of this study were to validate a novel, automated, proprietary system (CytoVison SPOT AX) for the assessment of FISH abnormalities in BE brush cytology and, subsequently, to use this automated method for screening of a BE surveillance cohort.
FISH with DNA probes for chromosomes 9, 17, and Y, and for the 9p21 (p16), 17q11.2 (Her2/neu), and 17p13.1 (p53) loci was applied on brush cytology specimens from a surveillance cohort of 151 patients with BE. Validation of the automated system was performed by comparison of the automated FISH results with manual scores for the first 60 patients.
There was 98% concordance between manual and automated FISH analysis with κ values from 0.49 to 1 for the different probes. The loss of 17p13.1 (p53) was observed in only 5% of patients with no dysplasia (ND) and in 9% of patients with low-grade dysplasia (LGD) but increased to 46% in patients with high-grade dysplasia (HGD) (P < .005; Fisher exact test). Chromosomes 9 and 17 were observed in 6% of patients with ND, in 21% of patients with LGD, and in 62% of patients with HGD (P < .05). Ten percent of patients with ND had loss of the Y chromosome, which increased to 27% in patients with HGD (P< .05). The amplification of 17q11.2 (Her2/neu) was detected in 62% of patients with HGD (P < .001).
The current investigation indicated that the CytoVison SPOT AX is an objective, efficient system for the analysis of DNA-FISH on BE brush cytology and is applicable for analyzing large populations of BE patients. In the current study cohort, the loss of 17p13.1 (p53), Y chromosome loss, and polysomy of chromosomes 17 and 9 were correlated with increasing grade of dysplasia in patients with BE. Cancer 2007. © 2007 American Cancer Society.
Barrett esophagus (BE) is a metaplastic condition of the distal esophagus that is associated with an increased risk for developing esophageal adenocarcinoma (EAC). It is assumed that the malignant transformation of BE follows the sequence of no dysplasia (ND), low-grade dysplasia (LGD), and high-grade dysplasia (HGD) before invasive EAC occurs. For several decades, in regions of Western Europe and the United States, the incidence of EAC has been increasing rapidly.1–3 Because long-term survival of EAC patients is highly dependent on early diagnosis, the detection of BE in patients who are at high risk for developing HGD or EAC has become crucial.4 It has been demonstrated that the current surveillance strategies, based on biopsy and histopathologic staging for dysplasia, for identifying subpopulations with a greater risk of malignant progression are insufficient. The major difficulty is the interobserver and intraobserver variation when classifying biopsies for grade of dysplasia and the sampling error that may occur when random biopsies are used.5, 6 Despite the vigorous protocols that are applied for extensive and systematic sampling, malignant lesions still may be missed.7 Moreover, the transformation of BE into EAC may not necessarily follow the assumed sequence of histopathologic events.8 Thus, novel surveillance strategies should aim to improve the risk stratification of BE patients, for instance, by using objective, genetic markers to assess neoplastic progression. Therefore, during the past years, research on BE has focused on identifying biologic markers in the metaplasia-dysplasia-adenocarcinoma sequence, and numerous cytogenetic abnormalities have been described.9–12 Among the most important cytogenetic changes contributing to BE progression are chromosomal numeric aberrations, Y chromosome loss, loss of the p53 and p16 tumor suppressor loci, and amplification of the Her2/neu oncogene.10, 12–16 One of the methods for detecting these cytogenetic abnormalities is fluorescent in situ hybridization (FISH), which utilizes fluorescently labeled DNA probes to pericentromeric chromosomal regions or unique chromosomal loci to detect cells with numeric or structural chromosomal changes. Earlier studies with DNA-FISH for the detection of genetic abnormalities in BE frequently used biopsies or resection specimens.10, 12, 13, 15–17 An alternative way to obtain material for FISH analysis is by brush cytology. The advantage of brush cytology includes simplicity, lower cost, and the potential of this method to sample a larger area of the BE epithelium compared with random biopsies. Recent studies have demonstrated that DNA-FISH can detect cytogenetic changes successfully in interphase nuclei of brush cytology specimens from patients with BE.14, 18–20 Thus, DNA-FISH on brush cytology seems to be a promising method with which to screen patients with BE for cytogenetic abnormalities associated with dysplasia and/or malignancy. However, because of the laborious and time-consuming nature of manual DNA-FISH signal enumeration, the use of this method to screen large cohorts of patients with BE is not feasible. This disadvantage may be circumvented by using an automated DNA-FISH analysis system, which can provide hands-off and objective scores of genetic abnormalities and may provide an opportunity to apply DNA-FISH for the screening of large BE surveillance populations.
Therefore, the main objectives of this study were to validate the novel CytoVison SPOT AX counting system by determining the degree of concordance between manual and automated counting of FISH signals in BE brush cytology specimens and to evaluate further the FISH abnormalities in the premalignant stages of BE in brush cytology specimens from a surveillance population. For these purposes, we applied FISH on brush cytology specimens from 151 patients with BE in a surveillance population by using DNA probes for the pericentromeric regions of chromosomes 9, 17, and Y and for the locus-specific regions of 9p21 (p16), 17p13.1 (p53), and 17q11.2 (Her2/neu).
MATERIALS AND METHODS
The study included a surveillance cohort of 151 patients with BE who underwent routine surveillance by endoscopy and biopsy at the Academic Medical Center in Amsterdam between 2002 and 2006. The Ethics Committee of the Academic Medical Center approved the study. All patients signed informed consent for the use of their biopsy and brush cytology material. Only patients with proven, incomplete, intestinal-type metaplasia in biopsies that were taken during and prior to surveillance were included. All patients were on long-term proton pump inhibition of 40 mg to 80 mg daily to prevent reflux esophagitis. During endoscopy, the brush cytology specimens were taken prior to biopsy. The brushes of the normal squamous epithelium were taken from patients without dysplasia at ≥3 cm above the BE segment and were used for control purposes. Biopsies for routine histologic examination were taken at least every 2 cm in 4 quadrants and from all suspected visible lesions using the protocol described by Reid et al.5
Cytologic brush material was sampled using the Wilson-Cook (Winston-Salem, NC) brush type LCB-220-3-1.5-S. Directly before brushing, the mucosal surface was sprayed with acetylcysteine (50 mg/mL) for dissolving the mucus layer. After the procedure, the brushes were inserted into a vial with 20 mL of 5% acetylcysteine in 0.9% NaCl, mixed gently to obtain a homogeneous cell suspension, and then cytospin slides were processed. The cell suspension from the brush was poured into a 50-mL conical tube and centrifuged at 2100 revolutions per minute (rpm) for 10 minutes at 4°C. Most of the majority supernatant fluid was discarded, leaving the pellet in 5 mL of solution, and the cells were agitated to generate a cell suspension. A Cytospin (Shandon Cytospin 4 Cytocentrifuge; Thermo, Waltham, Mass) was used to generate a single layer of the cells on a glass slide. First, 50 μL of phosphate-buffered saline (PBS) were loaded into the cytospin chambers and centrifuged for 1 minute at 550 rpm at room temperature. Subsequently, up to 150 μL of cell suspension were loaded into the cytospin chambers and centrifuged for 2 minutes at 550 rpm at room temperature. The cytospin slides were dried overnight at room temperature and then stored at −80°C until FISH analysis was performed.
We used directly labeled, fluorescent, chromosomal centromeric probes (CEP) for chromosomes 9, 17, Y and for the locus-specific probes (LSI) for regions of 9p21 (p16), 17p13.1 (p53), and 17q11.2-q12 (Her2/neu). The probes were obtained from Vysis (Downers Grove, IL). Dual-color probes were used combining CEP 9 SpectrumGreen/LSI, p16 (9p21) SpectrumOrange, and a single probe for LSI p53 (17p13.1) SpectrumOrange. The remaining probes were combined into 1 set, which contained CEP 17 SpectrumGreen, LSI Her2/neu (17q11.2-q12) SpectrumOrange, and CEP Y SpectrumAqua. DNA-FISH was performed according to the instructions provided by Vysis with slight modifications. Briefly, the cytospin slides were immersed in 2 × standard saline citrate (SSC) at 37°C for 10 minutes, then treated with 0.005% pepsin (Sigma, United Kingdom) in 0.01N HCl, pH 2. A 5-minute wash in PBS at room temperature arrested the enzymatic treatment, and slides were fixed in 5% formaldehyde at room temperature for 10 minutes. After 1 wash in PBS at room temperature for 5 minutes, slides were dehydrated in ethanol series and air dried. Then, 2 μL of probe from the mixture, which consisted of 7 μL hybridization buffer, 1 μL of each probe, and water to a 10-μL volume, were applied to each cytospin slide. The specimen and probes were codenatured at 80°C for 3 minutes, and the slides were incubated at 37°C for 48 hours. Thereafter, the slides were washed in 0.4 × SSC at 70°C for 5 minutes, in 2 × SSC/0.1% Nonidet P-40 at room temperature for 30 seconds, dehydrated in an ethanol series, and then air dried. Subsequently, 4′,6-diamidino-2-phenylindole (DAPI) diluted 1:1000 in mounting medium for fluorescence (Vectashield; Vector Laboratories, Inc., Burlingame, Calif) was applied to counterstain the nuclei.
FISH Analysis by the CytoVision SPOT AX System
The CytoVision Spot AX work station (Applied Imaging, Newcastle, United Kingdom) combines an automated Olympus BX61 fluorescent microscope with SPOT counting software. It is a high-throughput system for objective counting of fluorescent signals within interphase nuclei. The Spot software incorporates a trainable DAPI classifier that, based on the average size and circularity of the nuclei, excludes untypical shaped cells (damaged or smeared cells) from the analysis. The system automatically scans the slide area and captures the signals in 3 dimensions (Z-Stack utility), which help to distinguish overlapping spots. The system simultaneously captures up to 5 different probes and then counts the number of fluorescent signals per nucleus. The counting result is displayed as the number of cells with normal and abnormal spot counts for each of the probes used in the analysis. Within abnormal cells, those with signal loss (<2), gain, or amplification (>2) are distinguished. The output of the analysis is a gallery of images of each nucleus, which may be used for interactive review.
For each patient, between 100 to 200 interphase nuclei of BE cells were scored per slide. At the end of each analysis, the gallery of cell images was reviewed manually to correct for counting errors. Incorrectly counted cells were excluded from the analysis. The counting errors appeared to be caused by a high hybridization background, overlapping cells, autofluorescent signals, and cells that were out of focus. The percentage of incorrectly counted cells observed during the automated analysis varied from 5% to 35%. The automated analysis was performed blindly with respect to the results from manual FISH analysis and histology.
Manual Enumeration of FISH Signals for Validation of the Automated System
To validate the automated FISH analysis system, the first 60 patients with BE also were evaluated by manual analysis. These patients were evaluated without prior knowledge of histology findings or findings by the automated analysis. For each slide, 100 interphase nuclei of BE cells were evaluated using a fluorescent microscope (Olympus BX61, Germany). Damaged cells and cells with indistinct and blurry signals were not scored. In the first instance, the slides were analyzed by a single experienced scorer (A.M.R.). After a comparison of the manual and automated FISH results, all discordant FISH abnormalities were reviewed blindly by a second experienced scorer (K.K.K.).
Criteria for Determining FISH Abnormalities
To establish the frequencies of artifacts that resulted from background hybridization variation, the probes used in the study were applied to normal squamous epithelium obtained from 20 patients who had BE without dysplasia. Signals from 100 and 200 interphase nuclei of squamous cells were evaluated by manual and automated FISH analysis, respectively. From these counts, cut-off values were calculated for each probe and for each method separately as the mean percentage of cells with signal gain or loss plus 3 × standard deviation. BE material was considered abnormal when the number of cells with abnormal counts, for any probe, was equal to or greater than the cut-off value. The cut-off values of each probe for both FISH analysis methods are presented in Table 1.
|FISH analysis method||CEP9||CEP17||CEP Y||LSI p16||LSI p53||LSI Her2/neu|
For validation of the automated system, an overall percentage of concordance between manual analyses and automated FISH analyses was assessed for all 6 probes that were hybridized for the 60 test cases. Sensitivity of the automated FISH analysis to detect an abnormality (at least 1 abnormality present) was determined for 26 patients who were classified as abnormal by the manual FISH analysis, which we considered the gold standard for the purposes of the current study. Likewise, the specificity of the automated analysis was calculated by using the 34 patients who had no abnormalities according to the manual FISH analysis. The κ value was determined for each probe separately to adjust for the chance of agreement: κ = 1 implies perfect agreement, and κ = 0 suggests that the agreement is not better than that obtained by chance. The κ value was judged as providing agreement that was good if κ > 0.80, substantial if 0.61 < κ < 0.80, moderate if 0.41 < κ < 0.60, fair if 0.21 < κ < 0.41, and poor if κ < 0.20. Differences in the frequency of abnormalities were compared using the Fisher exact test, and P values <.05 were considered statistically significant.
Patients and Histopathology
Of 151 patients with BE, 123 were men, and 28 were women; the median age was 60 years (range, 26–84 years); and the median BE segment length was 3 cm (range, 1–13 cm). The studied cases included 114 patients with ND, 11 patients with LGD, 13 patients who were indefinite for dysplasia (IND), and 13 patients with focal or diffuse HGD. For validation of the automated FISH analysis system, 60 BE patients with ND (n = 42 patients), LGD/IND (n = 10 patients), and HGD (n = 8 patients) were analyzed by both manual and automated FISH analysis.
Validation of the Automated FISH Analysis System
FISH abnormalities that were assessed by the manual and automated analyses are shown in Table 2. Figure 1 depicts examples of BE cell images with FISH abnormalities that were captured by the automated system. Twenty-six of 60 (43%) and 25 of 60 (41%) patients with BE had abnormal FISH findings in the manual and automated FISH analyses, respectively. We observed a high overall concordance (calculated for all 6 probes) of 98% between the 2 methods. The sensitivity and specificity of the automated FISH analysis to detect an abnormal case (patients who presented with at least 1 abnormality) were 96% and 100%, respectively. Table 3 shows a comparison of manual and automated FISH analyses as well as the κ values, which indicate the degree of agreement for each probe. We observed perfect agreement between manual and automated FISH analyses for the assessment of trisomy and/or tetrasomy of chromosome 9 and amplification of 17q11.2 (Her2/neu; κ = 1). Furthermore, we observed good agreement for detecting the losses of 9p21 (p16), 17p13.1 (p53) loci, and the Y chromosome (κ > 0.80) with a low level of discrepancies (1–2%). Moderate agreement was observed for assessment of the trisomy and/or tetrasomy of chromosome 17 with a 3% discrepancy between the 2 methods (κ = 0.49).
|2||LGD||M||−9p, −17p, +9 (tetrasomy)||−9p, −17p, +9 (tetrasomy)|
|5||HGD||M||−Y, −9p, −17p,† +17 (trisomy)||−Y, −9p, +17 (trisomy)|
|8||HGD||M||−Y, −17p, +17 (tetrasomy),‡ +++17q||−Y, −17p, +++17q|
|15||HGD||M||−Y, +9 (trisomy), +17 (trisomy/tetrasomy),‡ +++17q||−Y, +9 (trisomy), +++17q|
|19||ND||M||+9 (trisomy)||+9 (trisomy)|
|22||HGD||M||−9p, +++17q||−9p, +++17q|
|23||HGD||M||−9p, −17p, +++17q||−9p, −17p, +++17q|
|FISH abnormality||No. of patients (%)||κ*|
|9p21 (p16) loss||17 (28)||16 (27)||0.96|
|17p13.1(p53) loss||6 (10)||5 (8)||0.90|
|CEP Y loss||4 (8)†||5 (10)†||0.88†|
|CEP 9 trisomy/tetrasomy||2 (3)||2 (3)||1|
|CEP 17 trisomy/tetrasomy||3 (5)||1 (2)||0.49|
|17q11.2 (Her2/neu) amplification||7 (12)||7 (12)||1|
Overall, there were 6 discordant FISH abnormalities (Table 2), which subsequently were reviewed blindly by a second experienced scorer (K.K.K.). Two of 6 discordant FISH results that were obtained by manual FISH analysis turned out to be false positive. These included the loss of 17p13.1 (p53) and loss of the Y chromosome (Table 2; Patients 5 and 17, respectively). Three other abnormalities—a trisomy and/or tetrasomy of chromosome 17 (Table 2; Patients 8 and 15) and the loss of 9p21 (p16) (Table 2; Patient 9)—were missed by the automated system. In 1 patient, the loss of the Y chromosome was detected by automated FISH analysis but was missed by the manual method (Table 2, Patient 4).
Cytogenetic Abnormalities in the Cohort of BE Patients
Using the same DNA probe set and the automated FISH analysis system, the frequency of FISH abnormalities in brush cytology material from the rest of surveillance population was determined. FISH outcomes of the total BE surveillance cohort (n = 151 patients) were compared with the grade of dysplasia as assessed by routine histopathology of biopsies that were taken at the same endoscopic procedure as the brush cytology material (Table 4). Figure 2 depicts the most commonly observed genetic abnormalities in our BE cohort compared between patients with ND, patients with IND/LGD, and patients with HGD. The most frequently observed cytogenetic abnormality in patients with ND was the loss of 9p21 (p16), which was observed in 30% of patients. Loss of 9p21 (p16) also was detected in 33% of patients with IND/LGD and in 46% of patients with HGD (P > .05). Five percent of patients with ND displayed loss of 17p13.1 (p53). In addition, no patients with IND and only 9% of patients with LGD had this abnormality, which increased significantly to 46% in patients with HGD (P < .005). Trisomy and/or tetrasomy of chromosomes 17 and 9 were present in 4% and 3% of patients with ND, respectively. Trisomy and/or tetrasomy of chromosome 17 was observed in 17% of patients with IND/LGD and in 46% of patients with HGD (P > .05). Chromosome 9 trisomy and/or tetrasomy was present in 9% of patients with IND/LGD and in and 15% of patients with HGD (P > .05). The combined frequencies of the trisomy and/or tetrasomy of chromosomes 17 and 9, indicating aneusomy and/or aneuploidy, were 6% in patients with ND. These frequencies increased significantly to 21% in patients with IND/LGD and to 62% in patients with HGD (P < .05). Ten percent of patients with ND displayed loss of the Y chromosome. This abnormality was not detected in any patient with IND or LGD, but it was present in 27% of patients with HGD (P < .05). Amplification of the 17q11.2 loci (Her2/neu) was observed in 62% of patients with HGD (P < .001). Loss of chromosomes 9 and 17 as well as gain of chromosome Y were very rare findings in both patients with and without dysplasia (>3%).
|FISH abnormality||No. with abnormality/Total no. of patients (%)|
|9p21 (p16) loss||34/114 (30)||4/13 (31)||4/11 (36)||6/13 (46)|
|17p13.1 (p53) loss||5/114 (5)||0||1/11 (9)||6/13 (46)|
|CEP Y loss||9/89* (10)||0||0||3/11* (27)|
|CEP 9 loss||2/114 (2)||0||0||1/13 (8)|
|CEP 17 loss||1/114 (1)||0||0||0|
|CEP 9 trisomy/tetrasomy||3/114 (3)||0||1/11 (9)||2/13 (15)|
|CEP 17 trisomy/tetrasomy||4/114 (4)||2/13 (15)||2/11 (18)||6/13 (46)|
|CEP Y gain||1/89* (1)||0||0||0|
|17q11.2 (Her2/neu) amplification||0||0||0||8/13 (62)|
We validated the automated CytoVison SPOT AX system for the analysis of FISH results in BE brush cytology specimens using DNA probes for chromosomes 9, 17, Y and for the locus-specific regions of 9p (p16), 17p (p53), and 17q (Her2/neu). Upon validation, we observed a high concordance of 98% between the automated FISH analysis and the manual method. We also demonstrated that the automated FISH analysis had excellent sensitivity (98%) and specificity (100%) to detect an abnormal case when we considered manual analysis as the gold standard. Moreover, we demonstrated good to excellent agreement between both FISH-analysis methods separately for each of the probes that were used, as indicated by κ values. Excellent and good agreement was achieved for the assessment of trisomy and/or tetrasomy of chromosome 9 (κ = 1), 17q11.2 (Her2/neu) amplification (κ = 1), loss of 9p21 (p16; κ = 0.96), loss of 17p13.1 (p53; κ = 0.90), and loss of Y chromosome (κ = 0.88). Moderate agreement was observed only for the assessment of trisomy and/or tetrasomy of chromosome 17 (κ = 0.49). In total, there were 6 discordant cases between the manual and automated FISH analyses. Of the 6 discordant cases, automated FISH analysis truly failed to assess only 3 of them, including the loss of 9p21 (p16) in 1 case and trisomy and/or tetrasomy of chromosome 17 in 2 cases. Discrepancies in the assessment of these abnormalities were caused by the clumping of cells, which means that an inhomogeneous cell distribution may have a negative influence on the performance of the automated system. Overlapping nuclei from the cells clumps were excluded from analysis by using the DAPI classifier in the SPOT software; however, as long as cells in the clumps can be distinguished from each other, automated assessment of these clumps remains possible. For the automated analysis, thus, it is important to use good-quality FISH samples with sufficient, evenly spread cells. Most important is the revision of the gallery of images provided by the system for each case, which is required to exclude incorrectly counted cells.
From this part of the study, we conclude that the CytoVision SPOT AX counting system may be applied successfully to assess genetic abnormalities that are detected by DNA-FISH in brush cytology specimens from patients with BE. This automated system is highly applicable to analyze large quantities of slides and omits manual scoring, which is tedious and time-consuming. Therefore, the application of this system provides an opportunity to extend the use of FISH on brush cytology to larger quantities of cells, which may be important, especially when cytology samples are taken from a long BE segment.
In addition, the imaging system provides an objective, consistent way to score FISH signals and enables the revision of previously analyzed samples, because all images are stored digitally. If necessary, the user easily may repeat the entire review in a rapid manner through the use of image galleries, whereas similar manual verification is labor-intensive and is not always possible (eg, because of the loss of FISH signal intensity over time). The automatic collection of cell images also has the advantage that the investigator can compare the morphology of cells within 1 patient or among different patients with BE, a capacity that may be useful for the standardization of diagnostic decision and quality control in the future. In the past, several software programs have been developed to score fluorescent spots in interphase nuclei; however, their feasibility was never tested on large sample sizes; and, in contrast to the CytoVision SPOT AX system, none of those systems could analyze more than 1 fluorochrome simultaneously in a 3-dimensional mode (Z-stack utility).21, 22 To our knowledge, this is the first report demonstrating the feasibility of an automated, high-throughput system for the accurate assessment of FISH signals in brush cytology specimens from patients with BE.
In the second part of the study, we proceeded with screening a cohort of patients with BE by using the automated system and the same panel of FISH probes. We evaluated the frequencies of cytogenetic abnormalities in brush cytology specimens from a surveillance cohort of 151 BE patients and compared the results with the histopathologic grade of dysplasia. We observed the loss of 9p21 (p16) in a substantial portion of patients with ND (30%), which confirmed that this alteration occurs early in BE.14, 23, 24 In our cohort, the loss of 9p21 (p16) did not correlate significantly with increasing stage of dysplasia (P > .05), possibly because of the small sample size of the population with dysplasia. Because the loss of 9p21 (p16) had a relatively high frequency in patients who had ND compared with patients who had LGD and HGD, this abnormality itself may not be indicative as a marker for dysplasia in BE. The loss of 17p13.1 (p53), which was a relatively rare event in the ND (5%) and LGD (9%) groups but was observed in 46% of patients in the HGD group, was correlated significantly with increasing grade of dysplasia (P < .005). Thus, the loss of 17p13.1 (p53) may be indicative of the presence of dysplasia in BE; in fact, it was demonstrated previously that the loss of p53 precedes dysplasia and is associated with an increased risk of progression to HGD and to EAC.14, 19, 25, 26 Ten percent of the patients with ND had loss of the Y chromosome, a proportion comparable to what has been reported in other in situ hybridization studies.12, 13, 19 We did not observe the loss of this chromosome in patients with IND or LGD, as demonstrated previously by Krishnadath et al.12 and Doak et al.19 This abnormality was observed in 27% of our patients with HGD, a proportion that also is lower than that documented in earlier studies, in which loss of the Y chromosome was documented either in all or in the majority of patients with HGD. In those studies, however, loss of the Y chromosome appeared to be mostly focal and was confined to a relatively low percentage of cells.12, 19 Previously Walch et al.16 demonstrated that, in the metaplasia-dysplasia-adenocarcinoma sequence, polysomy of chromosome 17 already was present in 50% of patients with LGD but was not observed in patients with ND. Our results indicated that polysomy (trisomy and/or tetrasomy) of chromosomes 17 and 9 was detectable by DNA-FISH with low frequency in patients with ND (6%). The combined frequency of these abnormalities increased further through IND/LGD (21%) to HGD (62%; P < .05) and may have predictive value for malignant transformation in patients with BE. Polysomy of these chromosomes may reflect aneuploidy, which is indicative of BE progression, as demonstrated in DNA flow cytometry studies, which demonstrated that increased tetraploidy/anueploidy was correlated with BE malignancy.27–31 We also observed the amplification of 17q11.2 (Her2/neu) in >60% of patients with HGD, a proportion that is concordant with previous FISH studies that demonstrated a high prevalence of Her2/neu amplification only in patients with HGD or EAC.15, 16, 18 Our data indicate that the amplification of 17q11.2 (Her2/neu) is a late event in BE progression and, thus, may be a useful, specific marker for detecting HGD in BE.
In conclusion, for the second part of this study, we evaluated cytogenetic abnormalities in the premalignant stages of BE in 151 surveillance patients. In general, we observed that the genetic abnormalities detected by FISH on cytology specimens were in agreement with FISH genetic abnormalities found in biopsy specimens, as reported in literature. Because brush cytology has the potential to sample the entire surface of BE epithelium, theoretically, we may assume that, compared with random biopsies, this method diminishes sampling errors. This would be reflected in a higher number of genetic abnormalities assessed in brush cytology samples than in randomly taken biopsies. In the current study, however, we did not compare the cytology FISH results with FISH results that were obtained from biopsy specimens and, thus, cannot draw this conclusion. We hope that a future study comparing these 2 methodologies will enlighten us on this matter.
In summary, the current study demonstrated that the automated assessment of cytogenetic abnormalities detectable by DNA-FISH in brush cytology from a BE surveillance population is feasible and reliable. Using this method, we detected several important cytogenetic abnormalities in patients with BE with ND that correlated with an increasing grade of dysplasia. However, future follow-up of the surveillance cohort will be required to prove the true predictive value of these abnormalities. We believe that the potential of automated FISH analysis for the accurate assessment of important genetic changes can improve the efficacy of future surveillance programs.
Supported by the Dutch Cancer Society
- 7Optimizing endoscopic biopsy detection of early cancers in Barrett's high-grade dysplasia. Am J Gastroenterol. 2000; 95: 3089–3096., , , .Direct Link:
- 8Long-term follow-up of Barrett's high-grade dysplasia. Am J Gastroenterol. 2000; 95: 1888–1893., , , , , .Direct Link:
- 9DNA abnormalities as marker of risk for progression of Barrett's esophagus to adenocarcinoma: image cytometric DNA analysis in formalin-fixed tissues. Am J Gastroenterol. 2004; 99: 1887–1894., , , , , .Direct Link:
- 25Predictors of progression in Barrett's esophagus II: baseline 17p (p53) loss of heterozygosity identifies a patient subset at increased risk for neoplastic progression. Am J Gastroenterol. 2001; 96: 2839–2848., , , et al.Direct Link: