Residual cervical specimens in PreservCyt solution were obtained from the University of Washington, Seattle, Washington, and the Mayo Clinic, Rochester, Minnesota.
Testing for human papillomavirus (HPV) is used in the triage of women with a cervical cytology of atypical squamous cells of undetermined significance (ASCUS). A fluorescent in situ hybridization assay was developed for the detection of HPV using the catalyzed receptor deposition technique (HPV-CARD). In this study, the utility of this assay was tested for the detection of HPV in liquid-based cervical cytology specimens.
A total of 195 liquid-based cytology specimens were analyzed using the HPV-CARD assay. The results from the assay were compared with HPV polymerase chain reaction (PCR) and typing results. The number of HPV-infected cells and the staining pattern was correlated with the cytology classification.
A 91% concordance between HPV-CARD and PCR was observed for the detection of high-risk HPV viruses. A 78% concordance was observed for specimens that were negative for HPV. In ASCUS, low-grade squamous intraepithelial lesion (LSIL), and high-grade squamous intraepithelial lesion (HSIL) categories, the average number of HPV-positive cells per slide was 19 cells, 127 cells, and 450 cells, respectively. The number of cells with a punctate staining, suggestive of HPV integration, was 21% in ASCUS, 34% in LSIL, and 46% in HSIL specimens.
Cervical cancer is the second most common form of cancer in women worldwide. It is well established that persistent human papillomavirus (HPV) infection is the main cause of cervical cancer.1–3 More than 40 HPV types have been shown to infect the genital tract, and of these, 15 to 20 types may be classifiable as high-risk types that are associated with the development of cervical carcinoma.4 The majority of HPV infections will clear with time. However, a certain proportion of women with high-risk HPV infection eventually develop high-grade dysplasia or cervical cancer,5–7 suggesting that HPV infection is necessary but not sufficient for the development of cervical cancer.
Several studies have investigated the relations between HPV viral load and the development of high-grade dysplasia. Viral load has been measured for different categories of cytology specimens and biopsy preparations, primarily using the HC2 (Hybrid Capture 2) assay (Digene, Gaithersburg, Md) and polymerase chain reaction (PCR) approaches. One of the advantages of an in situ hybridization (ISH) approach is that HPV can be evaluated in the morphologic context of cervical specimens, thereby allowing distinction between HPV-infected epithelial cells and uninfected cells. Another advantage of ISH for HPV detection is the ability to distinguish among different states of HPV infection (ie, integrated vs episomal). Initial studies for the in situ detection of HPV demonstrated that the signal patterns match the physical status of HPV as determined by PCR and restriction digest patterns of HPV DNA.8–10 A diffuse signal throughout the nucleus was found to be correlated with episomal HPV, whereas a punctate signal within the cell nucleus was indicative of integrated HPV. Several studies suggest that integration of HPV into the cellular genome might be an important event for the progression to high-grade dysplasia or cervical cancer. In the early stages of cervical dysplasia, HPV is predominantly present in the episomal form, whereas in the majority of cervical carcinomas HPV is integrated into the genome.11–13
A historical issue concerning the use of ISH for viral detection has been its low sensitivity. Conventional ISH methods have a detection limit of approximately 50 viral copies per cell.14, 15 The introduction of the catalyzed receptor deposition (CARD) system in the late 1980s resulted in the development of more sensitive ISH systems. CARD is based on the deposition of labeled tyramide molecules by peroxidase activity.16–19
In this study, we tested the performance of a fluorescent-based HPV-CARD assay. The objectives of the current study were: 1) to determine the effectiveness of the HPV-CARD assay for detecting high-risk HPV in thin-layer cervical specimens; 2) to evaluate the staining pattern of HPV in infected epithelial cells (diffuse, punctate, or mixed); and 3) to estimate the number of HPV-infected epithelial cells for different categories of cervical specimens.
MATERIALS AND METHODS
A total of 195 cytology specimens comprised of normal, atypical squamous cells of undetermined significance (ASCUS), low-grade squamous intraepithelial lesion (LSIL), and high-grade squamous intraepithelial lesion (HSIL) diagnostic categories were included in the study. The distribution of the specimens was as follows: 26 cases with normal cytology, 15 cases with ASCUS cytology, 106 cases with LSIL cytology, and 48 cases with HSIL cytology.
ThinPrep slides were prepared using the ThinPrep 2000 Processor (Cytyc Corporation, Marlborough, Mass) according to the manufacturer's instructions. Slides were dried overnight at room temperature and then stored at –20°C until hybridized. Slides were prepared from residual cell suspensions that were 6 to 24 months old and had been stored in PreservCyt fixative at 4°C.
Cervical epithelial cell lines CaSki, HeLa, SiHa, and C-33 A were purchased from American TypeCulture Collection (ATCC; Manassas, Va). These cells were propagated in RPMI-1640 medium, supplemented with 2 mM of L-glutamine, 10 mM HEPES, and 10% fetal bovine serum at 37°C in an atmosphere of 5% carbon dioxide. Cells were grown to a 60% to 70% confluence level and harvested using a standard trypsinization procedure. After washing collected cells 3 times in phosphate-buffered saline (PBS), cells were resuspended in a cold PreservCyt fixative solution (Cytyc Corporation). Cell suspensions were stored at 4°C for up to 3 months. Slides were prepared from these suspensions by depositing cells on ThinPrep slides using the ThinPrep 2000 Processor according to the manufacturer's instructions (Cytyc Corporation). Slides were air-dried overnight at room temperature before hybridization.
HPV Probe Source, Composition, and Biotin Labeling
HPV-16 and HPV-18 were obtained from ATCC. HPV-30 and HPV-45 were obtained from Dr. Ethel-Michele de Villiers from DKFZ (Heidelberg, Germany). HPV-51 and HPV-58 were obtained from Klara Abravaya (Abbott Molecular, Des Plaines, Ill).
The selection of HPV types included in the probe mix was based on the following criteria: the HPV type is in the public domain, the HPV type had been classified as high risk, and the HPV type has high homology to other high-risk HPV types currently not available in the public domain. Because high-risk HPV-56 is not in the public domain, HPV-30 was used as a substitute due to its high homology with HPV-56. In addition, HPV-30 has high homology with HPV-53 and HPV-66, which have been classified as probable high-risk types.
Based on a BLAST2 sequence analysis, presented in Table 1, it was proposed that this HPV probe cocktail should hybridize to all HPV types sharing >50% sequence homology with any of the 6 HPV types in the probe mix at lower stringency hybridization conditions. Therefore, the stringency of the probe mix was modified by addition of salt to enable the 6 HPV probes to detect the majority of known high-risk types based on DNA homology.
Table 1. Homology Analysis of the 6 HPV Types Included in the Probe Cocktail and Known High-risk HPV Types*
HPV types included in HPV-CARD assay probe mix
HPV indicates human papillomavirus; CARD, catalyzed receptor deposition technique.
Homology was determined using the BLAST 2 SEQUENCES program and reflects the total percentage of homology between indicated HPV types.
HPV-68 has not been completely sequenced. The E2, E4, and E5 genes have not been reported in GenBank.
Probable high-risk HPV types. Classification is based on 0 control and 1 to 3 positive cases.
For convenience, entire HPV DNA sequences were cloned into the pBluescript SKII (−) plasmid (Stratagene, La Jolla, Calif) and were included into our probe cocktail (for HPV types 16, 18, 30, 45, 51, and 58). HPV constructs that included both insert and vector were individually labeled with biotin by means of nick-translation using the BioNick DNA Labeling System (Invitrogen, Carlsbad, Calif). The median size of the final biotinylated probe was between 100 and 400 basepairs (bp). Biotin-labeled pBluescript was used as a negative control in hybridization experiments with various cell lines and clinical specimens to ensure specificity of the HPV probe signal. In addition, each probe was tested individually to confirm that each probe was detecting the expected HPV type.
The percentage of biotin incorporation was determined for each biotinylated probe using the FluoReporter Biotin quantitation assay kit (Molecular Probes, Eugene, Ore) and the percent incorporation for biotin was approximately 1% to 3%. Each probe was labeled individually and then all 6 labeled HPV types were combined into a single HPV probe mix. The final composition of the hybridization cocktail was as follows: LSI hybridization buffer (Vysis/Abbott Molecular), saline-sodium citrate (SSC), Cot-1 DNA, human placental DNA, and biotinylated HPV DNAs (HPV types 16, 18, 30, 45, 51, and 58). Each reaction contains 2 μg of human placental DNA, 1 μg of Cot-1 DNA, and 25 ng of each biotinylated HPV DNA (150 ng total HPV DNA) in LSI hybridization buffer containing 4× SSC.
Sample Pretreatment and Hybridization
ThinPrep slides were soaked in 2× SSC at 73°C for 2minutes, followed by incubation in pepsin (0.5 mg/mL in 10 mM HCl) at 37°C for 10 minutes. The slides were then soaked in 1× PBS at room temperature for 5 minutes, fixed in 1% neutral-buffered formalin at room temperature for 5 minutes, and soaked in 1× PBS for 5 minutes. Slides were then dehydrated in an ethanol series of 70%, 85%, and 100% ethanol for 1 minute in each solution and air dried. The probe hybridization mixture (10 μL) was applied to each slide, a coverslip was placed over the solution, and the coverslip was sealed to the slide with rubber cement. The slides with probe mix were codenatured at 72°C for 2 minutes and then hybridized at 37°C for 16 to 18 hours on a hybridization platform (HyBrite; Vysis/Abbott Molecular). After hybridization, slides were washed in 2× SSC/0.3% NP-40 for 2 minutes at 48°C and then 2× SSC/0.1% NP-40 for 1 minute at room temperature.
HPV Probe Detection
Detection of the biotin-labeled HPV probes was performed using the Alexa Fluor 488 TSA (tyramide signal amplification) kit (Invitrogen) using the manufacturer's directions. Additional modifications included blocking of endogenous peroxidase activity by incubating hybridized slides in 3% hydrogen peroxide for 30 minutes at room temperature before application of streptavidin-HRP conjugate (SA-HRP). SA-HRP was diluted 1:100 in blocking reagent and incubated at 37°C for 25 minutes (both incubations were performed in a humidified chamber). The biotin-labeled HPV probe-SA–HRP complex was visualized by incubation with Alexa Fluor 488 labeled tyramide (1:100 dilution) for 10 minutes at room temperature. Nuclear counterstain DAPI was applied and a coverslip was placed over the counterstain solution.
Slides were analyzed under a fluorescence microscope using ×40 magnification. HPV-positive cells were visualized using a single bandpass green filter (Vysis/Abbott Molecular). HPV staining was classified according to the criteria previously described.8, 9, 20 The entire surface area of each slide was analyzed in the majority of cases. In cases with a large number of HPV-positive cells, the first 100 HPV-positive cells were analyzed.
HPV Detection by PCR
In 164 of 195 samples, HPV infection and typing was obtained. The remaining samples were not tested for HPV presence due to insufficient material for PCR testing. An aliquot of cell suspension (1–6 mL, depending on density) was sent to Esoterix (Eden Prairie, Minn, or Calabasas Hills, Calif) for HPV detection and typing. The presence of HPV was determined by PCR using MY09/11 primers and HPV type analysis was performed by restriction fragment length polymorphism (RFLP).
Enumeration results were analyzed using JMP statistical software (version 5; SAS Institute Inc, Cary, NC).
Comparison of HPV-CARD and HPV PCR
ThinPrep slides were tested with the HPV-CARD assay and the results of the HPV PCR assay were blinded until HPV-CARD analysis was completed. The results demonstrated that the HPV-CARD assay was able to detect 12 of the 13 high-risk HPV types at the following rates (for single infections): HPV-16 in 29 of 31 cases, HPV-18 in 4 of 4 cases, HPV-31 in 6 of 6 cases, HPV-33 in 1 of 1 case, HPV-35 in 2 of 2 cases, HPV-39 in 1 of 2 cases, HPV-45 in 1 of 1 case, HPV-52 in 3 of 3 cases, HPV-56 in 3 of 3 cases, HPV-58 in 5 of 5 cases, HPV-59 in 3 of 3 cases, and HPV-68 in 1 of 2 cases. In addition, HPV-66 and HPV-53, which more recently have been classified as high-risk types,4 were also detected by the HPV-CARD assay in 9 of 9 and 5 of 5 cases, respectively. None of the samples tested were infected with HPV-51.
Infections with a single type of HPV were more common among all the cytologic and histologic categories tested. In all, 71% of cases had a single HPV infection (87 of 126 cases), whereas 31% (39 of 126 cases) had multiple HPV infections as determined by PCR. The prevalence of HPV types in specimens with a single infection was determined for each cytologic category. HPV-16 was most frequent in the HSIL group (58% [19 of 33] cases). In the LSIL group, HPV-16 was present in 17% (8 of 45 cases), and in the ASCUS category, HPV-16 was present in 4 cases (50%). The most prevalent type in the LSIL group was HPV-66 (20% [9 of 45] cases).
Table 2 summarizes the results comparing the HPV-CARD assay with PCR.21 The concordance between the testing modalities was 91% for HPV-positive cases (115 of 126 cases) and 79% for HPV-negative cases (30 of 38 cases). Eleven PCR-positive cases were missed by HPV-CARD. Of those, 2 cases were infected with HPV-16, 1 with HPV-39, and 1 with combined HPV-39 and HPV-72 infection. These samples should have been detected by HPV-CARD assay. There were 8 cases that were PCR-negative but positive by the HPV-CARD assay. The number of HPV-positive cells in these cases ranged from 1 to 114 HPV-positive cells. Three reviewers examined these cases and all were in agreement that the samples were truly HPV-positive.
Table 2. Concordance Between HPV PCR and HPV-CARD Assays for the Detection of HPV in Cervical Specimens
Quantitation of HPV-Infected Epithelial Cells and Evaluation of HPV Physical State in Different Clinical Categories
The HPV-CARD assay was optimized using cervical carcinoma cell lines containing a known number of HPV copies and known integration status: CaSki (up to 600 copies of HPV-16), HeLa (up to 50 copies of HPV-18), and SiHa (1–2 copies of HPV-16).22 Cervical carcinoma cell line C-33 A, which is not infected with HPV, was used as negative control. As shown in Figure 1, the HPV assay generated a punctate signal in all 3 HPV-positive cell lines. The size of the HPV signal increased correspondingly with the increase of HPV DNA copy numbers. The HPV signal was localized to the nucleus as shown by nuclear staining with DAPI. In the HPV-negative cell line C-33 A, no HPV signals were detected (data not shown). In clinical specimens, 3 categories of HPV staining patterns were observed: a homogenous (diffuse) staining of the nucleus suggestive of episomal HPV (Fig. 2A), a punctate staining suggestive of integrated HPV (Fig. 2B), and a mixed HPV staining suggestive of the simultaneous presence of episomal and integrated forms of HPV within the same cell nucleus (Fig. 2C).
The total number of epithelial cells on each specimen slide was estimated after the HPV-CARD assay. The number of cells varied from a low of 1000 cells per slide to a maximum of approximately 100,000 cells per slide. On average, a slide contained approximately 23,000 epithelial cells. The absolute number of HPV-positive cells was evaluated for each specimen. The mean and median numbers of HPV-positive cells per slide for different cytologic categories are depicted in Figure 3. In the group of cytologically normal specimens, only 1 case was found to be positive with a total of 6 HPV-positive cells. On average, this represents 0.2 HPV-positive cells per slide for this group of normal cases. A wide variation in the quantity of HPV-infected cells among different specimens was found for all dysplastic groups. In the ASCUS group, the quantity of HPV-infected cells ranged from 0 to 104 cells per slide with a mean of 19 cells per slide and a median of 3 cells per slide. In the LSIL group, the average quantity of HPV-positively stained cells was 127 cells per slide, ranging from 0 to 1000 cells, with a median value of 46. The variability in the quantity of infected cells was found to be greatest in the HSIL group, extending from 0 to 6250 cells with an average of 450 cells per slide and a median of 74 cells per slide. The difference in average value for HPV-infected cells between groups was statistically significant (P < .05).
Evaluation of the physical state of HPV revealed that cells displaying a diffuse, punctate, and mixed HPV pattern were present in all dysplastic categories tested, excluding cytologically normal samples. Because the mixed type of pattern suggests integration has already occurred, we combined quantitative values for integrated and mixed HPV into 1 category for presenting quantitative results. In the normal cytology group, only a diffuse staining was present in HPV-infected cells found in this study. In the ASCUS category, 21% of infected cells (4 of 19 cells) demonstrated punctate/mixed staining suggestive of HPV integration, and 79% (15 of 19 cells) of HPV-positive cells demonstrated a diffuse HPV staining pattern suggestive of an episomal state of HPV. In LSIL, the proportion of infected cells with the punctate HPV staining increased to 34% (43 of 127 cells), and diffuse staining was observed in 66% of all HPV-positive cells in the LSIL group. The difference between ASCUS and LSIL groups was statistically significant (P < .0001). In more advanced lesions (HSIL), a punctate/mixed staining was found in 46% of all HPV-positive cells (207 of 450 cells). However, 54% of HPV-positive cells (243 of 450 cells) still showed diffuse staining. The difference between the LSIL and HSIL categories with respect to HPV integration was also found to be statistically significant (P = .0489).
The goal of development efforts was to produce an in situ assay that will 1) reliably detect high-risk HPV types, 2) be sensitive enough to detect a single HPV-infected cell among thousands of cells present in a thin-layer cervical preparation, and 3) be sensitive enough to detect HPV-infected cells containing only a few copies of virus.
ISH is a powerful technique that allows unique molecular information to be obtained within intact cellular structures. The most commonly used approaches for detecting HPV DNA sequences in the host cells are based on either a target amplification technique with subsequent hybridization (HC2 assay) or with the PCR-based approaches for common HPV genes. These approaches demonstrate high analytical sensitivity but are unable to address important questions such as the physical status of the virus (episomal vs integrated) and the percent of infected cells in a sample.
ISH approaches have the ability to overcome these difficulties but were reported to suffer from low sensitivity. Hesselink et al.23 compared a chromogenic ISH assay for the detection of HPV with PCR-based and hybrid-capture-based techniques and reported inferior results using the ISH assay. Similar results were reported by other groups.24 To increase the sensitivity of the assay, 2 approaches were used: first, we used a fluorogenic substrate instead of a chromogenic substrate and, second, we employed the tyramide signal amplification system. Using a fluorogenic substrate with DAPI staining allows confirmation that the HPV stain colocalizes within the cell nucleus, reducing the chance that staining is an artifact. By using a fluorogenic substrate, a higher signal-to-noise ratio was observed when compared with a chromogenic substrate. Use of the tyramide signal amplification system allowed for the detection of only few copies of HPV DNA per cell, as demonstrated by detection of integrated HPV in the cell lines HeLa and SiHa. A homogenous staining (diffuse) of the nucleus is suggestive of episomal HPV, a punctate staining is suggestive of integrated HPV, and a mixed HPV staining is suggestive of the simultaneous presence of episomal and integrated forms of HPV within the same cell nucleus. Similar findings have been reported for histologic sections.25 Recently, Hopman et al.25 demonstrated that integrated HPV could be unmasked in the specimens with a mixed staining pattern and determine whether true integration is present through the use of a harsh pretreatment protocol.
To further determine the performance characteristics of this assay, specimens were tested by PCR in combination with HPV type-specific RFLP analysis. Based on the HPV typing by PCR-RFLP results, we concluded that the HPV-CARD assay was able to detect 12 high-risk HPV types. Comparison of these 2 assays reveled 91% concordance between the HPV-CARD assay and PCR/RFLP for the detection of the high-risk viruses. Our assay missed 2 cases infected with HPV-16, 1 case with HPV-39, and 1 case with combined HPV-39 and HPV-72 infection. There are 2 possible explanations for missing these cases: either the HPV copy number per cell is below the detection limit of the HPV assay, or the specimens had a very low number of positive cells. In 2 of the cases, the cell density of the slides was below 10,000 cells, compared with the average cell density per slide for the entire study of 23,000 cells.
The concordance between PCR-negative samples and samples negative by HPV-CARD was 79%. One of the explanations for this discrepancy is in methodologic differences between the approaches. The HPV-CARD assay can detect a few HPV-positive cells, which contain a few hundred or more copies of HPV DNA, among tens of thousands of noninfected cells. Such samples would contain a low total number of HPV copies per extracted DNA sample, possibly below the detection limit by PCR (103 copies per samples) or HC2 (104 copies per sample).26 Indeed, in the majority of discrepant cases the number of HPV-infected cells was very low (1–10 cells). However, in several cases, the number of cells stained positive by our assay was relatively high (50–100 cells) and the reason for a negative by PCR result is not clear. Perhaps, as some studies have reported, an increased concentration of human DNA or blood in the samples may reduce the sensitivity or inhibit the PCR reaction.27
A substantial accumulation of cervical epithelial cells infected with HPV was observed with the increasing severity of the lesions. This is made evident through both the mean and median values for the LSIL and HSIL categories. The average number of HPV-infected cells was more than 3 times higher in the HSIL group compared with the LSIL group (450 cells vs 127 cells) and nearly twice as high when comparing median values (74 cells vs 46 cells). When expressed as a percentage of the total quantity of epithelial cells on a slide, the proportion of HPV-infected cells in the ASCUS group was found to be 0.08%, 0.5% in the LSIL group, and 2% in the HSIL group. For the HSIL samples there were instances where 10% of the cells on the slide were infected with HPV.
We observed a good correlation between the staining pattern of HPV and the severity of cervical dysplasia. With advancement of the disease, the relative proportion of HPV-positive cells with a punctate staining pattern increased from 1 in 4 in ASCUS specimens to 1 in 3 in LSIL specimens. Cells with a punctate or mixed HPV pattern represented nearly half of all HPV-positive cells found in HSIL specimens. A number of groups have analyzed HPV integration at various stages of cervical dysplasia and cancer. Some studies have reported that integration of HPV is an important molecular event and is associated with the transition to invasive carcinoma.28–30 Other studies have reported the presence of integrated HPV in premalignant lesions with a histologic classification of cervical intraepithelial neoplasia (CIN) type 1 and CIN2/CIN3, which is in agreement with our current findings.11, 31, 32 It is clear that more studies are needed to determine the stage of disease at which integration is initially detected. The use of a standardized assay to determine HPV integration might be useful in clarifying some of the current controversies.
In conclusion, we designed an optimized HPV-CARD assay based on fluorescent in situ hybridization technology and obtained novel information regarding the proportion and physical state of HPV-infected epithelial cells in different cytologic categories of cervical specimens. We demonstrated that the developed assay possesses high analytical sensitivity; produces low background, high signal-to-noise ratio; and allows quantification of HPV-infected epithelial cells and distinction of HPV physical states. Overall, the results of the current study demonstrate a positive correlation between the severity of the lesion, the quantity of HPV-positive epithelial cells, and accumulation of cells with integrated HPV. Although analysis of a larger population is required to investigate the clinical utility and to establish positive and negative predictive values of this assay, we believe the current study demonstrates the feasibility of using this assay for the detection of HPV in cervical cytology specimens.
We thank Dr. Nancy Kiviat, Dr. Kevin Halling, Benjamin Kipp, and Dr. Elizabeth Unger for valuable suggestions and clinical specimens.