Abstract
Introduction
The aim of the study was to investigate whether biobanked liquid-based cytology (LBC) vaginal samples could be reanalyzed for the biomarkers HPV DNA and mRNA without loss of sensitivity.
Methods
One hundred LBC samples with ASCUS or CIN1 were tested for HPV DNA and mRNA before and after biobanking. DNA analysis targeted the viral genes E6 and E7, 12 high-risk and 2 low-risk HPV types together with the human control gene HBB, using real-time PCR. The Aptima HPV assay was used for mRNA analysis of 14 high-risk HPV types.
Results
With Aptima there was 84% agreement between results before and after biobanking. The sensitivity and specificity were 0.79 (95% CI, 0.68-0.88) and 0.94 (95% CI, 0.80-0.99), respectively. With the DNA-based method, the agreement between results was 87%, the sensitivity 0.85 (95% CI, 0.75-0.92) and the specificity 0.95 (95% CI, 0.77-1.00). Both methods presented a significant difference between positive results before and after biobanking; McNemar test: p = 0.004, p = 0.003, Cohen's kappa: 0.67 (95% CI, 0.53-0.81), 0.68 (95% CI, 0.52-0.84). Cycle threshold values for the DNA method were higher for all genotypes after biobanking, except for HPV-59. Some loss of sensitivity was seen after biobanking but the concordance between HPV detection before and after biobanking was good for both evaluated methods.
Conclusions
Biobanking of LBC vaginal samples offers a good platform for HPV testing and could be extended to further molecular analyses. However, in order to ensure a valid test result a larger portion needs to be analyzed from the biobanked sample.
Introduction
Liquid-based cytology (LBC) has extended the possibility of biobanking into the field of cytology (1). In Sweden, women between 23 and 59 years are invited to cervical cancer screening every third (23-49) to fifth year (50-59) and LBC has since 2009-2012 been the method of choice. In collaboration with the biobanking initiative, the Swedish Biobanking and Molecular Resource Infrastructure (BBMRI), Sweden is in the process of establishing a biobanking facility aimed at complete population-based biobanking of about 100,000 cervical cytology samples each year. The BBMRI.se constitutes a collaboration between 9 Swedish universities and is supported by the Swedish Research Council. The overall objectives of the biobanking facility are to offer a platform for studies of molecular pathology of cervical dysplasia and cancer, and also to perform surveillance of the vaccination program (2).
Infection with the human papilloma virus (HPV) is the cause of cervical cancer (3); however, only a small fraction of all infections will ultimately lead to dysplasia if left untreated. Well over 100 different types of HPV exist, some with a preference for the genital mucosal linings. Twelve types have been classified as high-risk types, where HPV-16 and 18 account for about 70% of all cervical cancers (4, 5). Testing algorithms for cervical cancer screening are under current review in Europe as well as the US. The cytological screening programs have since their implementation reduced the incidence of cervical cancer substantially, but are flawed by limited reproducibility (6). HPV testing has been shown to have higher sensitivity than cytology for detecting high-grade cervical intraepithelial lesions (CIN) but is lacking in specificity (7). Recent data from 4 European randomized trials, however, indicate that HPV-based screening provides up to 70% greater protection against cervical carcinoma compared with cytology (8). In addition to HPV, additional biomarkers such as viral and host methylation as well as overexpression of cellular biomarkers such as p16ink4a may be of clinical use when evaluating HPV status. In Sweden, high-risk HPV triage for atypical squamous cells of undetermined significance (ASCUS)/CIN1 has been in use since 2012 according to national recommendations.
When HPV testing is implemented in primary screening, or as a triage alternative, different tests and alternatives have to be considered. For lesion progression, HPV infection needs to be persistent, with overexpression of the viral oncoproteins E6 and E7. A wide variety of HPV detection methods are on the market, including signal amplification assays as well as many nucleic acid amplification tests (9). PCR-based methods are sensitive and can be based on either consensus primers for the conserved L1 region on the virus or be type-specific using different viral genes (10). Alternative testing includes detection of oncoprotein mRNA (E6 and E7) with the purpose of identifying clinically relevant lesions with deregulated viral gene expression, compared to permissive infections.
When implementing a biobanking platform for reuse of collected samples, one appropriate concern is whether stored cytology samples can be reanalyzed for biomarkers without loss of sensitivity. In the process of transferring samples to the biobanking plates for long-time storage, cells are collected in a 2-step manner, reducing the sample volume considerably. The object of this study was therefore to address these issues through a validation study where HPV mRNA and DNA were tested before and after the biobanking process.
Material and Methods
One hundred consecutive collected cervical cytology samples with a diagnosis of ASCUS or CIN1 were included in the study. Samples were collected from the Department of Laboratory Medicine at Örebro University Hospital in 2012. ThinPrep vials holding 20 mL, from either primary screening samples (n = 44) or follow-up samples after a previous diagnosis (n = 56), were used. All included women were aged 35 years or older and their samples were forwarded for HPV testing according to the national recommendations. Upon inclusion in the validation study, all samples were deidentified and given a serial number that could not be connected to either personal data or clinical evaluation. No informed consent was obtained since this was a purely methodological study using anonymized patient samples without any connection to patient data or patient identity. DNA and mRNA analysis was performed on all cases in parallel before and after biobanking.
Samples were transferred from ThinPrep vials to 96-well plates using the Freedom EVO platform (Tecan) according to Perskvist et al (11). In short, primary sedimentation of samples was performed and 4 mL was pipetted to an intermediate tube. After a second sedimentation step, a cell volume of 500 µL was transferred to the repository plate. Plates were kept in a −20° freezer for 2 to 6 months and thawed before secondary DNA or mRNA analysis.
Dna Analysis
DNA extraction was performed before and after biobanking. 250 µL of sample was used for DNA extraction before biobanking and 20 µL of biobanked cell material + 230 MQ was used after biobanking using the QIAamp DNA Minikit (Qiagen) according to the instructions of the manufacturer. The quality and quantity of DNA was spectrometrically estimated (NanoDrop Technologies, USA). Eluates were diluted to a concentration of 10 ng/µL. Samples with concentrations below 10 ng/µL were analyzed undiluted.
Targeting the viral genes E6 or E7, detection of 12 high-risk HPV types (HPV-16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58 and 59) and 2 low-risk types (HPV-6 and 11) together with the human control gene beta-globin, HBB, was performed with real-time PCR on the 7900 HT Real-Time PCR System (Thermo Fisher Scientific). The targeted segments of the E6/E7 region were found by Lindh et al (12) to be type specific and also showed good agreement with the Roche Linear Array. Reactions of 20 μL included 1 or 2 primer-probe pairs, each holding 0.9 µM forward and reverse primers together with 0.2 µM probe and 1x Taqman® Universal Mastermix. Primers and probes (Thermo Fisher Scientific) were according to Lindh et al (12) with some alterations, as previously reported (13). PCR reactions underwent an initial step at 50°C for 2 minutes followed by a denaturation step at 95°C for 10 minutes before repeated cycling for 40 cycles at 95°C for 15 seconds followed by 60°C for 60 seconds. Each run of samples included positive, negative and non-template controls. Results were analyzed with the SDS 2.4 software for ABI 7900 HT (Thermo Fisher Scientific). Each curve was manually surveyed and samples with a cycle threshold (Ct) less than 35 were considered as positives. A threshold of 0.1 was used for all detectors. The method has been used in clinical routine settings and in several research projects (13-14-15).
mRNA Analysis
One milliliter of sample was used for the Aptima HPV assay (Gen-Probe) on the Panther platform (Hologic) before biobanking and 20 µL of biobanked material was used in the post analysis. The Aptima HPV assay is a qualitative method for detection of mRNA from 14 different high-risk HPV types (HPV-16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66 and 68) using transcription-mediated amplification of E6 and E7 after target capture. The HPV mRNA is isolated from the sample by means of capture oligomers that are linked to magnetic microparticles. HPV mRNA amplification uses TMA, a transcription-based nucleic acid amplification method with 2 enzymes, MMLV reverse transcriptase and T7 RNA polymerase. First, a DNA copy of the mRNA sequence is generated by reverse transcriptase. The DNA copy holds a promoter sequence for T7 RNA polymerase that is further used to produce multiple copies of the RNA HPV target. RNA amplicon detection is done by probe hybridization.
Sample volume was manually transferred to an Aptima specimen transfer tube and automatically analyzed on the Panther system. In each run, a positive and negative calibrator was included. An internal control was used to validate the different steps of the analysis. mRNA detection was measured as photon signals (called relative light units, RLU) and the results were interpreted based on the analyte signal-to-cutoff value (S/CO). The cutoff was calculated from the median RLU of the negative and positive calibrators and the signal corresponded to the RLU value of the sample. A sample was considered positive when S/CO was above 0.5 and the test results were determined automatically by the assay software. The analytical sensitivity, the concentration of HPV mRNA that yields a positive result in 95% of cases for the included genotypes, ranged between 20.9 (HPV-39) and 410.2 (HPV-52) copies per reaction. No human control gene was included in the analysis and the HPV result was expressed as either positive or negative, without genotype information. Clinical evaluation of the test has shown a sensitivity similar to that of DNA testing but higher specificity for samples >CIN2, and the test was CE-marked in 2008 (Aptima HPV Assay, Gen-Probe AW-11141-1601 Rev. 003).
Statistics
Sensitivity and specificity with 95% confidence intervals (CI) (viral accuracy) were calculated using GraphPad Prism version 6. Results were compared by means of McNemar's test and Cohen's kappa values using the SPSS package, version 22. Wilcoxon's signed-rank test was used to compare the median differences of Ct values before and after biobanking. Also, test positivity ratios of DNA versus RNA were calculated before and after biobanking.
Results
DNA Method Compared with mRNA Method before and after Biobanking
Prior to biobanking, 67 (67%) of the samples were high-risk HPV positive with the Aptima HPV assay and 78 (78%) were high-risk HPV positive with the DNA real-time PCR (Cohen's kappa: 0.68; 95% CI, 0.52-0.84). The difference was significant according to the McNemar test (p = 0.003).
Among the nonconcordant samples between the 2 methods, the most frequent genotype found in the DNA analysis was HPV-51 (n = 3) followed by HPV-18 (n = 2) and HPV-39 (n = 2). Single infections of HPV-16, HPV-52 and HPV-59 were also present as well as 2 cases of double infections: HPV-18/HPV-31 and HPV-45/HPV-51. Six of these genotyping results were detected at a Ct value below 30, indicating a substantial viral load, while the remaining 8 results were detected between 30 and 34 PCR cycles.
One of the 100 samples had a negative DNA result but a positive APTIMA result. The sample was confirmed to contain HPV-53 with a third method: MGP 5+/6+ consensus primers (16) followed by pyrosequencing (13).
After biobanking, 55 (55%) of the samples were high-risk HPV positive with the Aptima HPV assay and 67 (67%) were high-risk HPV positive with the DNA real-time PCR (Cohen's kappa: 0.62; 95% CI, 0.46-0.78). The difference was significant according to the McNemar test (p = 0.002). Five samples analyzed with the real-time DNA method had no valid human control gene, indicating insufficient cells for analysis.
The test positivity ratios, DNA versus RNA, were 1.16 (78/67) before biobanking and 1.22 (67/55) after biobanking.
RNA Results before and after Biobanking
For the Aptima HPV test there was 84% agreement before and after biobanking (Tab. I). The sensitivity and specificity were 0.79 (95% CI, 0.68-0.88) and 0.94 (95% CI, 0.80-0.99), respectively. Before biobanking, 67 samples were positive with the APTIMA and after the procedure 14 results were lost. The lost negative samples had initial RLU values between 106,917 and 3,157,937. The comparable DNA results for the same samples were consistent in 7 of cases (before and after biobanking), but the result differed in the other 7.
Concordance between investigated assays (real-time PCR for DNA and Aptima HPV assay for mRNA) before and after biobanking
Calculations were made from positive and negative results. The sample result before biobanking was used as the reference. For real-time PCR, 2 calculations are provided, the first based on the exclusion of samples with an invalid human control gene (n = 95) and the second including all samples (n = 100), where samples are classified according to the HPV result only.
Sensitivity = proportion of positive samples after biobanking among the positive samples before biobanking; specificity = proportion of negative samples after biobanking among the negative samples before biobanking.
Also, 2 samples had an opposite turnout: they were initially negative and after biobanking found to be positive. Both samples had positive DNA results before and after biobanking and were genotyped to HPV-51. After further evaluation of the Aptima results for these 2 samples, one had an S/CO just below the cutoff (0.489).
A significant difference between RNA-positive results before and after biobanking was found (McNemar test: p = 0.004). Cohen's kappa was calculated as 0.67 (95% CI, 0.53-0.81).
DNA Results before and after Biobanking
In the DNA assay, the human gene HBB was included to verify the amount of DNA for otherwise negative samples. In 1 of the 100 samples analyzed before biobanking there was no amplification of HBB, but instead an HPV result at a low Ct value was found, indicating excessive viral DNA for that sample (HPV-59 at Ct 17). Five samples in the post-biobanking analysis were also without HBB and otherwise negative. Therefore, the DNA results (Tab. I) were further divided into 2 groups, where group 1 excluded the 5 samples without the human control gene (n = 95) after biobanking for otherwise negative samples, and group 2 was based purely on the HPV result regardless of the HBB result (n = 100). The agreement of the results before and after biobanking was 89% (group 1) and 87% (group 2) (Tab. I). The sensitivity for sample results before and after biobanking was 0.88 (95% CI, 0.78-0.94) and 0.85 (95% CI, 0.75-0.92), respectively, while the specificity for both groups was 0.95 (95% CI, 0.75-1.00 and 0.77-1.00).
A significant difference between DNA-positive results before and after biobanking was found for groups 1 and 2 (McNemar test: p = 0.021 and p = 0.003). Cohen's kappa was calculated as 0.72 (95% CI, 0.57-0.88) and 0.68 (95% CI, 0.52-0.84), respectively.
The Ct values before and after biobanking according to genotype are shown in Table II. All detectable Ct values are included, also those above Ct 35. The Ct range for the human control gene HBB was between 22 and 28 before biobanking, with a mean of 24.24 (SD ±0.809), and 24-40 after biobanking, with a mean of 28.28 (SD ±3.636). All 12 HPV genotypes were represented in this material both before and after biobanking. The most abundant types before biobanking were HPV-16 (n = 18) and HPV-52 (n = 18). After biobanking, HPV-16 was still found in 16 cases, making it alone the most frequent genotype. The highest range of Ct values was recorded for HPV-56 both before biobanking (range 20) and after biobanking (range 21). When comparing the median differences of Ct values before and after biobanking, using the Wilcoxon signed-rank test, a statistically significant difference was found for HBB (p = 0.000), HPV-16 (p = 0.000), HPV-18 (p = 0.003), HPV-45 (p = 0.017), HPV-51 (p = 0.034), HPV-52 (p = 0.002), and HPV-58 (p = 0.017). Before biobanking, 19 samples showed double infections with HPV, and multi-infection with several HPV types (more than 2 genotypes) was present in 13 samples. Of the multi-infected samples, 7 were double-infected after biobanking, 1 was single-infected, and 1 was negative. Of the double-infected samples, 3 were single-infected after biobanking and 2 were negative.
Ct values in DNA testing
Mean Ct values (HPV genotypes and human control gene) before and after biobanking. Wilcoxon's signed-rank test was used to compare the median differences (Ct values between before and after biobanking). The Ct results include all detectable curves. A significance level of <0.05 was used.
Ct = cycle threshold; n = samples with HBB or HPV result; SD = standard deviation.
Twelve samples that were positive for HPV DNA prior to biobanking had a negative HPV result after biobanking. Their Ct values for HBB were between 23 and 28 before biobanking and between 24 and 35 after biobanking (Tab. III), and for 3 samples the HBB result after biobanking was also lost. Initial Ct values for the HPV genotypes varied between 16 and 34. When setting aside the positive sample cutoff of Ct 35, 5 of the samples presented the initial genotype after biobanking but at a higher threshold (see Tab. III for details).
Detailed genotyping results for discordant samples (positive-to-negative) in DNA analysis
Cycle threshold numbers are given for the human control gene HBB and the different HPV genotypes.
Ct = cycle threshold.
For 1 sample, a negative result was found before biobanking which changed to a positive result for HPV-56 after biobanking. After reevaluation of the initial result, HPV-56 was found to be present at a Ct above 35, which led to its negative classification.
Discussion
In this study, we have assessed the biobanking process for HPV DNA and mRNA detection. Also, comparative data on baseline DNA and mRNA results are provided.
Our results show that more samples (11/100, 11%) were positive for HPV using DNA detection compared to mRNA detection prior to biobanking as well as after biobanking (12/100, 12%). Similar sensitivity and higher specificity for the Aptima HPV assay compared to DNA-based methods (HC2) were found by other authors (17, 18), indicating that more clinically relevant infections are targeted where increased oncogene expression is found. For this methodological study, no follow-up for high-grade lesions was done, and thus no clinical evaluation on specificity is available. Of the DNA-positive/mRNA-negative samples before biobanking, HPV-51 was the most abundant genotype, followed by HPV-18 and HPV-39. According to Bruni et al (4), HPV-51 is the second most common genotype among women with low-grade lesions (10%). For women with high-grade lesions, the proportion of HPV-51- positive cases decreases to 5.5%. Hence the detected HPV-51-positive/mRNA-negative samples may represent infections that may resolve spontaneously. The compared methods have some differences that may also have reflected on the outcome. HPV-66 and HPV-68 are only included in the Aptima assay and not in the real-time DNA method, and the Aptima has also been found to have cross-reactivity for the HPV genotypes 26, 67, 70 and 82. We also report an additional example of this cross-reactivity, since 1 Aptima-positive sample (DNA-negative) was sequenced to hold HPV-53. Since no specific genotyping result is provided with the Aptima assay, we are not able to compare genotype outcomes between methods.
The concordance between HPV detection before and after biobanking was good for both evaluated methods, as for most samples the positive result could be repeated after biobanking. Some loss of sensitivity was, however, seen after biobanking, where 14 RNA results and 12 DNA results were lost, and a difference was also observed between the 2 methods in that the DNA-based real-time PCR method had higher sensitivity (0.85) than the Aptima test (0.79). This is also confirmed by the higher positive ratio (HPV DNA-positive versus RNA-positive samples) after biobanking compared with before biobanking, potentially also indicating that RNA is more vulnerable to biobanking than DNA. The loss of HPV detection in both methods could also be a result of the small volume (20 µL) that was used in the assays; this could be increased for improved reproducibility. Additional explanations could be the manual pipetting that was performed, and also the possibility that some samples had a low cell count. In the DNA real-time PCR, the human control gene HBB was included to confirm any negative sample results, a control procedure that is absent in the mRNA assay. A missing control gene result can indicate sampling mistakes as well as a low cell count, which will otherwise not be acknowledged. Interestingly, some positive results emerged after biobanking of negative samples for both methods. Although the samples are concentrated in the biobanking process, DNA analysis shows that mean Ct values are generally higher after biobanking than before. Re-evaluation of the results before biobanking revealed that 2 of them were initially borderline positive. Calculations of sensitivity and specificity were done on single tests results, and no replicates were performed. The Aptima HPV assay has been CE marked for use in Europe and has been evaluated for assay precision. The assay showed low signal variability in high-positive samples, but for samples containing few copies the variability in S/CO was substantially higher, as could be expected. However, analyte levels are not necessarily associated with S/CO values (i.e., the mRNA level is not necessarily correlated with the magnitude of a positive assay signal). With regard to the DNA method, Lindh et al (12) have shown that the method has good reproducibility, especially for samples with a Ct below 35. Notably, both tests have been used in clinical routine diagnostics, where samples are run as single tests. If samples had been assessed twice, a wider variation would be expected, especially for samples with a viral mRNA or DNA load near the limit of detection. As a consequence, a lost mRNA or DNA result after biobanking may either reflect degradation of the mRNA or DNA, or the expected variation for a low-copy sample.
We acknowledge that our study has some limitations. Since our evaluation was limited to predominant and known forms of HPV, it could not be extended to non-tested genotypes. Also, since samples were tested only once, any assay variation may have influenced the final results, at least for low mRNA and DNA copy samples.
Based on our data we conclude that when testing for a biomarker in biobanked samples, assays may need to be optimized in terms of sample volume and method thresholds. The sensitivity of the analysis of choice needs to be considered. As in the case of real-time PCR, additional cycles and new thresholds for cutoff values may have to be estimated. Examples of this are shown in Table III, where 5 of the samples analyzed with the DNA real-time PCR were classified as negatives (cutoff Ct 35) but found to have positive signals above the cutoff threshold. This study, however, also shows that biobanking of LBC vaginal samples offers a good platform for HPV testing of high-risk genotypes and could possibly be extended to further molecular analyses. In order to ensure a valid test result, a larger sample portion needs to be withdrawn from the biobanked sample. The long-time effects of biobanking need to be further explored.
Footnotes
Financial support: No grants or funding have been received for this study.
Conflict of interest: None of the authors have any financial interest related to this study to disclose.