Objective
Chromosome 3q gain has been consistently observed in cervical intraepithelial neoplasia grades 2 and 3 (CIN 2,3) and squamous cell carcinomas of the cervix. There are a number of potential clinical uses of testing for 3q gain in liquid cytology specimens, including the identification of subsets of women with atypical squamous cells of undetermined significance or low-grade squamous intraepithelial lesion cytology who are at greatest risk of having CIN 2,3 and would thus benefit most from immediate colposcopy. The objective of this study was to establish the sensitivity and specificity of 3q gain for discriminating between CIN 2,3 and normal.
Study Design
Residual cytology specimens were collected from 199 women. Liquid-based cytology (LBC) was used for the selection of subjects, with women with high-grade squamous intraepithelial lesion or high-grade squamous intraepithelial lesion who had colposcopy and adjudicated biopsy-confirmed CIN 2,3 forming the disease-positive group (n = 28) and women doubly negative for both cytology and high-risk human papillomavirus (hrHPV) testing forming the disease-negative group (n = 171). A single slide was prepared from each residual LBC specimen and analyzed for 3q gain by fluorescent in situ hybridization, using a probe specific for the 3q26 region and a control probe for the chromosome 7 centromere. Two approaches were compared for the determination of 3q gain. The first was based on the analysis of an entire cervical cytology slide for the presence of rare cells with a high copy number (>4 copies) for the 3q locus. The second approach was based on the analysis of 400 cells to determine the percentage with 3 or more copies of the 3q locus.
Results
Using the approach based on the detection of rare cells with a high copy number (>4 copies) for the 3q locus, 26 of the specimens from women with CIN 2,3 and none of the 171 specimens from women who were both hrHPV and cytology negative was positive for 3q gain. This translates to a sensitivity of 92.9% (95% confidence interval [CI], 76.5–98.9%), a specificity of 100% (95% CI, 97.8–100%), a positive predictive value of 100% (95% CI, 86.7–100%), and a negative predictive value of 98.8% (95% CI, 95.9–99.8), for distinguishing CIN 2,3 from normal.
Conclusion
These data support the potential clinical use of 3q gain for the evaluation of women in a number of clinical situations, including women with atypical squamous cells of undetermined significance, low-grade squamous intraepithelial lesion, and those who are hrHPV positive.
Cervical cancer screening has been one of the most successful, albeit one of the most costly, public health screening programs. Although the burden of cervical cancer death remains high in countries without active screening programs, the main challenge for continued cervical cancer control in developed countries is likely to be the refinement of screening so that women who are at greatest risk for cervical cancer can be targeted for evaluation.
The sensitivity of a single cytology test for the detection of cervical intraepithelial neoplasia (CIN) grade 2 or above (CIN 2+) is only approximately 50-75%, which has led to the adoption of high-risk human papillomavirus (hrHPV) testing as an alternative or adjunct tool to cytology for cervical cancer screening. The specificity of screening with hrHPV testing, however, is lower than cervical cytology because many women become human papillomavirus (HPV) infected during their lifetime.
Although cervical infection with hrHPV is a necessary early event in cervical carcinogenesis, the majority of hrHPV infections is transient in nature, with only a small percentage persisting and progressing to CIN2+. Only a small number of hrHPV-infected women will eventually develop cancer. Thus, infection with hrHPV alone is not sufficient to cause cervical cancer, and other genetic events play an adjunctive role.
Chromosome 3q gain has been consistently observed in both CIN2,3 and invasive cervical squamous cell carcinomas. Several studies, using fluorescent in situ hybridization (FISH), have shown that gain in the chromosome 3q26 region is the most consistent genetic abnormality found in CIN2+. In addition, the frequency of 3q26 gain has been shown to increase with the severity of CIN. There are a number of potential clinical uses of testing for 3q gain, including the identification of subsets of women with atypical squamous cells of undetermined significance (ASC-US) or low-grade squamous intraepithelial lesion cytology who are at greatest risk of CIN2+ and would thus benefit most from an immediate colposcopy.
Different options exist for the detection of 3q gain. To date, the most commonly used approach has been the utilization of liquid-based cytology (LBC) specimens and FISH to determine the number of cells with an increased copy number for the 3q26. LBC specimens offer the most attractive specimen type for evaluating increased copy number because they are widely used in the United States for both cervical cytology and hrHPV testing. However, applying FISH-based analysis for 3q gain using LBC specimens in a clinical setting is challenging. Most previous studies have chosen to select a small number of the cells and quantify the number with 3 or more copies of 3q26. However, significant numbers of cells with 3 or more copies of the 3q locus have been detected in normal cytological specimens, suggesting the possibility of a high false-positive rate from such an approach.
Screening all the cells present in a specimen for the detection of rare cells with high 3q copy numbers offers the potential for a test with higher sensitivity and specificity, but it is impractical using traditional manual microscopy. The power of automated microscopy analysis, however, offers the possibility of an approach to 3q gain analysis based on the screening of all the cells present for rare high 3q26 copy number cells.
The object of the work described herein was to compare alternative approaches to the detection of 3q gain. One approach, derived from the manual methods, analyzed a small proportion of the cells present on a LBC slide to determine the percentage with more than 2 3q FISH signals. The other approach utilized the power of automated fluorescence microscopy to analyze all the cells present on a LBC slide to detect rare cells with high 3q copy number present. Once we had identified the superior approach, we utilized it to determine the sensitivity and specificity of 3q gain to discriminate between CIN2,3 and specimens negative by both cytology and hrHPV.
Materials and Methods
Patient samples
Residual ThinPrep LBC specimens (Hologic, Bedford, MA) that were collected for routine cervical cancer screening and that otherwise would have been discarded were obtained from 3 independent clinical laboratories: West Coast Pathology Laboratories (Hercules, CA); Shiel Medical Laboratory (Brooklyn, NY); and Centro Diagnostico Italiano (Milan, Italy). In all instances, samples were deidentified before being forwarded for 3q analysis, and no patient information was provided. Because the samples were to be discarded, lacked patient identifiers, and no patient information was provided, patient informed consent and institutional review board review were not required.
Residual LBC samples that were diagnosed as negative for intraepithelial lesion or malignancy that were also hrHPV negative (by Hybrid Capture 2 [hc2]; QIAGEN, Germantown, MD) comprised the disease-negative group. A total of 187 disease-negative samples were obtained. Of these, 171 had sufficient residual material to allow us to make an additional slide for 3q gain analysis, using the ThinPrep T2000 processor, according to the manufacturer’s instructions (Hologic).
A total of 32 residual LBC specimens previously diagnosed as atypical squamous cells cannot exclude high-grade squamous intraepithelial lesion or high-grade squamous intraepithelial lesion were available from women who had colposcopy and biopsy. The cervical biopsy and/or conization specimens associated with these LBC specimens were reviewed by one of the authors (T.C.W.) in a blinded fashion with no knowledge of the cytology or FISH information attached to the specimens to confirm the presence of CIN2 or CIN3. Of the 32 residual LBC specimens, 28 had associated biopsy-confirmed CIN2 or CIN3 and sufficient cellularity for 3q gain analysis. These comprised the disease-positive group. Of the 28 specimens that comprised the disease-positive group, 7 were CIN2 and 21 were CIN3. p16 immunostaining had been previously done on tissue from 19 of the 28 CIN2 or CIN3, and all showed diffuse overexpression of p16.
HPV testing
The presence of HPV in the disease-positive samples was identified utilizing a polymerase chain reaction (PCR)-based method that has been described previously. Briefly, genomic DNA was extracted from an aliquot of the cytology specimen and amplified by PCR using consensus oligonucleotide primers specific for the L1 region of the HPV genome. Concurrently the integrity of the extracted DNA was evaluated by the amplification of beta-globin, a common housekeeping gene. HPV DNA-positive PCR products were subjected to digestion by the restriction endonucleases Hae III, Pst 1, and Rsa 1. Digested DNA fragments were separated on a 5% polyacrylamide gel and visualized by ethidium bromide staining. A digital image of the gel was captured, and the specific HPV genotype was determined by matching the restriction fragment patterns of the respective specimens to that of known HPV restriction fragment patterns.
Fluorescent in situ hybridization
A single slide was prepared from each of the residual liquid cytology specimens using the Thinprep method, according to manufacturer’s instructions (Hologic). The slides were hybridized with a probe for the chromosome 3q26 region, labeled with spectrum gold (Abbott Molecular, Des Plaines, IL) and a probe for the centromeric alpha-repeat sequence of chromosome 7, labeled with spectrum Aqua (Abbott Molecular). The probe for the 3q26 region comprised a series of BAC clones spanning a 500 kb region containing the TERC and the PIK3CA loci.
Hybridization was performed using standard FISH methods. Briefly, slides were pepsin treated, further fixed with 2% formaldehyde, washed, and dehydrated in ethanol series. Probe mix in hybridization buffer was added, and the sample was denatured and hybridized overnight. Following FISH, the slides were counterstained with 4[prime],6[prime]-diamino-2-phenylindole (DAPI) and coverslipped with antifade-containing mounting medium in preparation for scanning. Control slides were processed in parallel to monitor the efficiency of the FISH process.
Slide analysis
Slides were analyzed using an automated system, based on the Ikoniscope/Ikonisoft robotic microscopy instrument (Ikonisys Inc, New Haven, CT). This instrument offers fully automated scanning of fluorescence microscopy slides and allows approaches based on both impartial analysis of randomly selected small numbers of cells from a slide and comprehensive analysis of large numbers of cells to be considered. By offering fully robotic, unattended slide scanning, the instrument allows more than 100 slides to be loaded at a time, if desired, with the results available for review as the scanning of each slide is completed.
Two approaches were utilized for detection of 3q gain. The first utilized the onco FISH cervical scanning algorithm, specifically developed for the detection and analysis of rare cells with high copy numbers (>4 copies) of the 3q26 locus in a background of a large cell number. The second approach, derived from the approach taken in several previous studies, determined the percentage of cells with 3 or more copies of the 3q locus in a subset of 400 cells analyzed from each slide.
Statistical analysis
Standard statistical analysis was used to assess the ability of 3q gain to distinguish CIN2+ from normal.
Results
Determination of 3q gain
Initially a subset of 36 samples, 25 from the disease-negative group and 11 from the disease-positive group, was analyzed using the 2 alternate approaches for the automated detection of 3q gain to compare the 2 approaches. The 2 scanning algorithms, method 1 and method 2, are summarized in Figure 1 .
The slides were first scanned utilizing method 1, the onco FISH cervical scanning algorithm. This approach initially scans the entire slide using a ×4 objective and the DAPI fluorescence channel to identify the cell deposition area on the slide to assure that all cells are evaluated. This leads to the generation of a focus/exposure slide map, used in subsequent low- and high-magnification scanning, to avoid the over- or underexposure of the optical fields visited.
The total number of nuclei present and the number of nuclei with more than 2 signals for the 3q26 locus are then determined using the low-magnification ×20 objective. This process involves scanning the entire cell deposition area in 2 fluorescence channels, 1 for the nuclear counterstain (DAPI) and 1 for the 3q26 FISH signal. After delineation of the perimeter of each nucleus, the FISH 3q26 signals in the nucleus are counted.
The nuclei are then ordered based on a combination of the number of 3q26 FISH signals they contain and cellular morphology to prioritize those abnormal nuclei most likely to have high levels of 3q gain. The slide locations containing the 400 nuclei with the highest likelihood of having high levels of 3q gain are subsequently imaged in high magnification using a ×40 dry objective. Each optical field is imaged in 3 fluorescence channels, for DAPI, 3q26, and the chromosome 7 centromere. After nucleus delineation, 3q26 and chromosome 7 FISH probe signals are counted for the determination of 3q gain.
All signals are detected by analyzing 3-dimensional images collected in each fluorescence channel. The 3-dimensional structure of the potential FISH signals is analyzed for a number of quantitative features before the signal is accepted as true. Following scanning, a gallery of any cells determined to have a high 3q26 copy number is available for expert review, which for this study was done blinded to whether the sample was from the disease-negative group or the disease-positive group. Specimens were determined to be positive for 3q gain if 2 or more cells with 5 or more copies of the 3q locus were detected ( Figure 2 ). Using this approach, in the 11 cases from the disease-positive group, all were positive for 3q gain. In the 25 cases from the disease-negative group, all were negative for 3q gain ( Table 1 ).
| Variable | 3q Gain | CIN2,3 (n = 28) | Cytology(–)/hc2(–) (n = 25) |
|---|---|---|---|
| Method 1: entire slide | |||
| ≥2 cells with >5 3q signals | 3q Gain (+) | 11 | 0 |
| 3q Gain (–) | 0 | 25 | |
| Method 2: evaluation of 400 cells | |||
| >1% of 400 cells with >2 3q signals | 3q Gain (+) | 10 | 1 |
| 3q Gain (–) | 19 | 6 | |
| >3% of 400 cells with >2 3q signals | 3q Gain (+) | 5 | 6 |
| 3q Gain (–) | 7 | 18 | |
The slides were then scanned using method 2 to collect 400 scorable nuclei for 3q gain determination ( Figure 1 ). Slides were initially scanned using a ×4 objective and the DAPI fluorescence channel to identify the cell deposition area on the slide. High-magnification (×40) fields were then imaged in 3 fluorescence channels, for DAPI, 3q26, and the chromosome 7 centromere. All nuclei contained within those fields were imaged, and, after nucleus delineation, 3q26 and chromosome 7 FISH probe signals for each individual nucleus were counted.
As with method 1, following scanning, a gallery of all cells imaged at high magnification was available for blinded expert review. Given the importance of blinding in this study, the FISH analysis was performed completely independently of any knowledge of the cytology and histology data. To ensure blinding, the FISH slides were barcoded independently and the data compiled only at the conclusion of the study. The percentage of the first 400 scorable nuclei collected that contained more than 2 3q26 FISH signals was recorded. To be scorable, a nucleus was required to have at least 1 signal in each of the 2 FISH channels.
Based on previous work that has demonstrated the potential for detection of 3q amplification in cervical cytology specimens as a genetic test for cervical dysplasia, specimens were determined to be positive for 3q gain if 1% or greater or 3% or greater of the 400 cells had more than 2 copies of the 3q locus. Using this approach, 10 of 11 cases from the disease-positive group were positive for 3q gain based on 1% of 400 cells with more than 2 3q FISH signals and 5 of 11 were positive based on a threshold of 3%. In the 25 cases from the disease-negative group, 6 of 25 were negative for 3q gain based on the presence of 1% of 400 cells with more than 2 3q FISH signals and 18 of 25 were negative based on a threshold of 3% ( Table 1 ).
Sensitivity and specificity of 3q gain for discriminating between disease-positive and disease-negative specimens
After the initial comparison of the performance of the 2 methods, all 199 cytological specimens were analyzed for the presence of 3q gain, based on the detection of 2 or more cells with 5 or more copies of the 3q locus, utilizing the onco FISH cervical scanning algorithm (method 1, described in the previous text). The prevalence of 3q gain in these 199 specimens, 28 from the disease-positive group and 171 from the disease-negative group, is summarized in Table 2 . All women in the group of 171 that were disease negative (ie, negative by both cytology and hrHPV) were also negative for 3q gain. Of the 28 disease-positive specimens (ie, women with CIN2+), 26 were positive for 3q gain using this methodology. This translates to a sensitivity if 92.9% (95% confidence interval [CI], 76.5–98.9%), a specificity of 100% (95% CI, 97.8–100%), a positive predictive value (PPV) of 100% (95% CI, 86.7–100%), and a negative predictive value of 98.8% (95% CI, 95.9–99.8).
