Human papillomavirus (HPV) is a family of over 200 different double-stranded DNA viruses about 8 kbp1 in size. HPV primarily infects basal epithelial cells of the skin or inner lining of mucosal tissues such as the mouth, the respiratory tract, or the anogenital epithelium.2 Persistent infection with HPV is necessary for the transformation from normal epithelial cells of the cervix uteri to invasive squamous cell carcinoma of the cervix. Because of the high prevalence of this disease, HPV has been extensively studied.3 Depending on the risk of inducing transformation to cervical cancer, HPV types were classified into different risk groups: low risk, probable high risk, and high risk.4 However, this classification has not yet been definitively established, and to date there is no generally accepted risk assessment scheme for HPV subtypes. One of the most common classification system is that proposed by Muñoz et al,4 although there are also others that classify HPV with slight differences5; 3 HPV subtypes (69, 71, and 74) remain unclassified. The HPV genome can be functionally divided into 3 regions by which the DNA sequence is differentially conserved between the HPV types. The noncoding, so-called long-control region, shows the highest degree of variation. The second early region is involved in viral replication, and the third late region encodes for 2 structural proteins for the capsid, which are more conserved.2 This sequence homology is important for the design of primers used in diagnostic testing. If the homology is too high, the risk of cross-reaction with other HPV types increases, but if the homology of the targeted region is low, it is not possible to use consensus primers due to the unacceptably high number of primers required.
As primer design is the primary factor in establishing a specific HPV test, and HPV classification is not definitively established, it is evident that the diagnosis, and especially the genotyping, of HPV patients is critical. Our study focused on the genotyping capacity of 4 well-established and commercially available tests, namely, HPV SIGN, Linear Array, High-Risk HPV, and INNO-LiPA.
MATERIALS AND METHODS
We collected from our daily routine practice 181 cytology specimens from cervical smears in which HPV genotyping was to be performed following internal guidelines. The majority of samples were taken from women with atypical squamous cells of uncertain significance (ASCUS) or low-grade squamous intraepithelial lesion (LSIL). This comparative study was done using the same DNA extraction for each sample.
The 4 analytical methods of this study are broadly used in routine diagnostics and have well-established experimental procedures, thus we will simply summarize the principle of each method and provide the Internet link for the detailed protocols.
The study was approved by the local ethics committee EKNZ (Ethikkommission Nordwest und Zentralschweiz).
HPV SIGN PQ Genotyping (Diatech) test is based on 2 consecutive analyses. The aim of the first analysis is to detect the presence or absence of HPV DNA using a multiplex real-time polymerase chain reaction (PCR) containing primers targeting a hypervariable region of L1 ORF and the human β-globin gene. The melting curve analysis obtained with the EVA Green chemistry provides a semiquantitative signal of the presence of β-globin by a peak at around 87°C, and HPV amplicons by 1 or more peaks in the range of 78 to 81°C. Only positive samples were further analyzed by 4 different sequencing primers’ mixtures and processed on the Pyromark Q24 (Qiagen). The sequences obtained were compared by multialignment on the Identifier software (Biotage, Uppsala Sweden), and its genotyping was scored by percentage of identity to references sequences. A list of the major HPV genotypes identifiable through HPV SIGN PQ is provided, and includes 16 HPV high risk (16, 18, 31, 33, 35, 39, 45, 51, 52, 54, 56, 58, 59, 68, 73, 82), 12 HPV low risk (6, 11, 34, 40, 42, 43, 44, 61, 70, 72, 81, 89), and 7 probable high risk (in the instruction manual called HPV intermediate risk) (26, 53, 66, 67, 84, 90, 91). However, further typing HPV by sequencing should be possible as the results are generated from a HPV library containing all variants available in public databases (HPV1 to HPV117 including JEB2, RTRX7, SIBX3, SIBX9) for a 45 bp hypervariable region of a highly conserved HPV gene. This method also distinguishes between the sequence variations within the same HPV genotype, for example, HPV16, HPV16 African 1, and HPV16 African 2.
Linear Array HPV Genotyping (Roche) is an adaptation of the MY09/11 system by Gravitt et al6 called PGMY09/11. HPV sequences of about 450 bp are amplified from the L1 ORF region by multiplex PCR. Probes for the following 37 HPV types are fixed on a membrane strip: 13 HPV high-risk genotypes (13, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68) and 24 low-risk genotypes (6, 11, 26, 40, 42, 53, 54, 55, 61, 62, 64, 66, 67, 69, 70, 71, 72, 73, 81, 82, 83, 84, IS39, and CP6108). The PCR product containing the biotin-labeled primer is hybridized to the strip. Streptavidin-horseradish peroxidase conjugate is linked to biotin and the presence of HPV is detected visually by addition of tetramethylbenzidine as substrate.
Real-Time High-Risk HPV (Abbott)
The Abbott Real-Time High-Risk (HR) HPV assay is a qualitative in vitro PCR assay that utilizes homogenous target amplification and detection technology for the detection of high-risk HPV DNA in cervical cells collected in liquid cytology media. The Abbott Real-Time HR HPV assay is intended to detect 14 high-risk HPV genotypes: 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, 68 and to partially genotype 16, 18 from other 12 high-risk genotypes.
Inno-Lipa HPV Genotyping Extra (Fujirebio) is a line probe assay using the SPF10 primer system. Short HPV sequences of about 50 to 65 bp are amplified from the L1 ORF region by multiplex PCR. In addition, a set of primers for amplification of the human HLA-DPB1 gene was added to monitor sample quality and extraction. Probes for the following 28 HPV types are fixed on a membrane strip: 15 high-risk (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 68, 73 and 82), 3 probable high-risk (26, 53, 66), 7 low-risk (6, 11, 40, 43, 44, 54, 70), and 3 unclassified (69, 71, and 74) genotypes. The PCR product containing the biotin-labeled primer is hybridized to the strip. Streptavidin-horseradish peroxidase conjugate is linked to biotin and the presence of HPV is detected visually by addition of 5-bromo-4-chloro-3-indolylphosphate (BCIP) and nitroblue tetrazolium chloride (NBT) as substrate. Interpretation of the result can be done by direct visualization or using the software LIRAS for LiPA HPV.
HPV Risk Classification Used for This Study
We used the classification of HPV types associated with cervical cancer proposed by Muñoz et al,4 Varnai et al,5 or HPV SIGN PQ instruction manual:
From Muñoz et al
High-risk HPV (HR): 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 68, 73, and 82.
Probable high risk (pHR): 26, 53, 66.
Low risk (LR): 6, 11, 40, 42, 43, 44, 54, 61, 70, 72, 73, 81, CP6108.
From Varnai et al
Probable high risk (pHR): 9, 30, 67, 34.
Low risk (LR): 6, 11, 32, 40, 42, 43, 54, 61, 62, 70, 72, 74, 81, 83, 84, 86, 87, 91, and CP6108.
From the HPV SIGN Instruction Manual
High-risk HPV (HR): 16, 18, 31, 33, 35, 39, 45, 51, 52, 54, 56, 58, 59, 68, 73, and 82.
Probable high risk (pHR, also described as IR): 26, 53, 66, 67, 84, and 90.
Low risk (LR): 6, 11, 34, 40, 42, 43, 44, 61, 70, 72, 81, and 89.
As these classification schemes are arbitrary, the risk stratification of a given HPV was chosen corresponding to the study with the highest number of studied cases (higher probability of accuracy). In cases of doubt, the higher risk class was chosen for a given HPV subtype. With 1 exception, HPV subtype 54 was indicated as high-risk HPV by the HPV SIGN instruction manual, but was classified as low risk according to the 2 other hallmark papers of HPV risk classification, and was therefore always classified as low-risk HPV in our study.
As no “gold standard” exists for HPV typing, to compare the capacity of the different methods to correctly evaluate detection accuracy of HPV status, we developed a scoring system (Table 1) based on the same score used in external quality control for HR HPV. Basically, the scoring system “rewards” the absence of clinical consequences; for example, 2 points are given for the detection of confirmed HR HPV as HR and −1 point if the method failed to detect a confirmed HPV HR. We did not differentiate between negative and low-risk (LR) HPV, both of which belong to the same category for sample scoring. One point was given in the case of misclassification between high-risk and potential high-risk samples, misclassification of any sample status with an unknown risk (?R) result. Correct identification of a negative/low-risk sample status as negative/low-risk result was scored 2 points, invalid results 0 points.
Of 181 samples initially analyzed by 2 methods (HPV SIGN and Linear Array), 120 showed concordant results, among which 42 were positive for HPV and the following genotypes were detected: HPV16 (n=16), HPV42 (n=5), HPV18 (n=4), HPV81 (n=3), HPV6, 31, 56 (n=2), and only once HPV39, 43, 45, 54, 58, 62, 67, and 73, respectively.
Sixty-one samples showed unclear HPV status and were therefore considered discordant. These 61 cases were further analyzed with both High-Risk HPV (Abbott) and INNO-LiPA (Fujirebio) methods (Table 2).
A wide range of discrepancies was observed, from undetected HPV to detection of other HPV types, including a shift from high-risk HPV to low-risk HPV or vice versa (Table 3). Undetected low-risk HPV and negative results were noted separately in our results, but due to external quality control standards they were considered as equal for the scoring system (Table 1).
On the basis of the scoring system described above, every method was assessed with a total score comprising the results of all 61 samples (Table 3). The highest ranking was achieved by the INNO-LiPA method with 102 points (of 122 possible points), followed by Real Time HR with 72 points, HPV SIGN with 66 points, and Linear Array with 60 points, respectively. However, with the Real Time HR method, the distinction between pHR and HR was not possible, therefore the result Other was always adjusted to the corresponding sample status (pHR or HR) leading to the maximum possible points.
As the number of possible detectable HPV subtypes is rather high, we looked for representations of different types in the results of each method among the 61 cases (Table 2). We found imbalances in HPV16 detection 10 times with HPV SIGN, once with Linear Array, 5 times with Real Time HR, and 7 times with INNO-LiPA, respectively.
Moreover, we observed an absence of detection for HPV51, HPV52, and HPV53 by the method HPV SIGN, whereas these types were detected 4, 4, and 8 times by Linear Array, respectively, and 6, 13, and 8 times by INNO-LiPA method, respectively (Table 2).
The HPV SIGN method detected only a few double infections, and no triple or quadruple infections, whereas Real Time HR, Linear Array, and INNO-LiPA detected multiple infections more frequently (Table 4).
Samples that were not applicable due to failure of HPV detection or that did not allow detection of a pHR or HR HPV were analyzed in correlation with their DNA concentrations (Fig. 1). In all methods, the majority of inapplicable results appeared in samples with low DNA content.
The detection and genotyping of HPV is important because there is a strong link between high-risk HPV infection and the development of squamous cell carcinoma of the cervix,3 as well as anogenital and head and neck SCC.7 Assessment of HPV status in cervical cytology classified as ASCUS or LSIL helps to refine clinical decision making on follow-up examinations and further investigation.8–11 For this reason, we compared the benefits and limitations of 4 commercially available PCR-based detection tests.
In our diagnostic HPV testing routine, only 66% (120 of 181) of results from Linear Array and HPV SIGN were identical. This is quite a low concordance rate and resulted mainly from misses of HPV (21 cases) or differently detected HPV subtypes, representing a clinically important shift in risk stratification in a subfraction of 13 cases.
In the absence of a gold standard method for assessing the HPV status of cytologic specimens, it is difficult to be certain of the actual HPV status/subtype of a given sample. To shed more light on the real HPV status of a given sample, about which nobody knows the fact, we decided to perform a discrepancy analysis with 2 additional methods, also in the knowledge of possibly overestimating the tests that were added after the first analytic run.12 We are aware that there might be false-positive (eg, clinically irrelevant HPV presence or methodical bias) and false-negative (HPV not detected despite presence) results in this group. However, all LSIL were in 1 or more method positive for HPV, meaning that in this subgroup negative results are most probably false-negative results. For the whole group there are only 2 cases that were positive for HPV only, with 1 of the 4 methods suggesting that false-positive results (eg, due to methodical bias) are quite infrequent. However, the goal of our work was not to compare 4 different test methods, but to show significant differences in the results obtained from the same DNA extraction of ambiguous cases and to discuss possible reasons for that.
We used a scoring system similar to that used in external quality schemes to estimate the performance of the different methods in the defined discrepant cases. INNO-LiPA showed a high efficiency in detecting high-risk HPV and multiple infections with a low rate of nonevaluable specimens (8 cases). This performance is likely due to the short size of the HPV amplicons generated (65 bp), which also allows detection of HPV with low DNA content and suboptimally preserved DNA. For instance, HPV type 52 was detected in 13/61 cases by the INNO-LiPA methodology, whereas all other methods detected this type in only 4 cases or not at all (Table 2). Of course, the possibility of an overestimation of this type, for example, due to cross-reactions with other HPV types cannot be excluded.
Linear Array showed the highest number of possible detectable genotypes, and the lowest rate of HPV16 detection (only 1 of the discrepant cases), whereas the other methods detected this genotype 10, 7, and 5 times, respectively (HPV SIGN, INNO-LiPA, and Real time HPV, respectively). In all these cases, a low DNA concentration was detected, meaning that the quality was probably insufficient for the Linear Array technology, which works with a rather high amplicon length of 450 bp, requiring sufficient DNA of good quality. In addition, the high number of unevaluable cases (8, due to lack of human DNA) is probably the result of the requirement for good quality and large amount of DNA to reach an amplicon length of 450 bp. This would also be problematic in formalin-fixed and paraffin-embedded tissues where the length of achievable DNA amplicon is normally below 300 bp.
The detection of multiple infections continues to be challenging for detection methods (Table 4). Only 2 cases showed multiple infections of the 61 selected cases per the HPV SIGN method. Linear Array and INNO-LiPA showed 9 and 14 multiple infections, respectively. Real Time HR is unable to detect multiple infections other than combinations of 16, 18, or Other. Combinations of high-risk genotypes other than 16 and 18 would be reported as other (of note: not all high-risk HPV genotypes are represented in this assay), and low-risk HPV is not detectable as the method is designed to detect high-risk DNA. This lack of detection of multiple infections with the HPV SIGN method is a result of the sequencing approach with comparison with the BLAST database not being considered in its design, as multiple HPV infections would create overlapping peaks in the chromatogram leading to noninterpretable results. In addition, the HPV SIGN method did not reveal any cases with detected HPV51, 52, or 53 genotypes in our study cohort, whereas Linear Array or INNO-LiPA detected these genotypes at least 4 times each. These cases were not only multiple infection, but primarily single infections that went undetected. A possible explanation might be a detection weakness due to primer design problems.
The multiplex amplification of HPV DNA in 4 certified methods showed significant differences in the results obtained from the same sample/DNA extraction, confirming the need for a gold standard, as indicated in other studies.13–17 The weaknesses observed and discussed in this study might encourage companies to improve their tests, and prompt users to be cognizant of possible pitfalls in result interpretation.