Our data suggest that the method of DNA purification from oral rinse samples has a potentially large impact upon the ability to detect HPV genomic DNA in these samples by PCR amplification. Our data further suggest that amplification of a human gene present in high copy number, such as β-globin, may not be appropriate for categorization of the sample as adequate for detection of small-copy-number nonhuman sequences. This conclusion is supported by our observation that samples processed by some of the methods were uniformly β-globin positive, and yet up to half of the oral HPV infections still failed to be detected (Table ).
Puregene purification preceded by thorough RNA and protein digestion resulted in the detection of the greatest numbers of HPV-positive subjects and HPV infections in matched samples compared to the numbers detected by all other methods evaluated. The inability to detect HPV genomic DNA in proteinase K-digested samples could be attributed to the presence of PCR inhibitors in the oral rinse samples, as previously observed by our group (16
) and others (25
). Stepwise modifications to the digestion procedure revealed that the cause of PCR inhibition from sample to sample was quite heterogeneous and could include total nucleic acid overload, protein inhibition, and specific inhibition by RNA. We recognize that ERV-3 assay-based estimates may be inaccurate in the presence of PCR inhibitors because this is also a PCR-based assay. However, the use of serial 10-fold dilutions overcame this potential problem and allowed evaluation of the effects of DNA digestion and the purification procedures on DNA amplification.
It would appear that PCR inhibition is a larger problem in oral samples than samples from other anatomic sites prone to HPV infection. However, some studies report an inability to amplify β-globin from more than 5% of cervical swab samples (20
) and more than 15% of anal swab samples (3
), suggesting that PCR inhibition may play a role in the amplification of HPV DNA from samples from other anatomic sites. We have shown than DNA purification can eliminate this PCR inhibition. The relative loss of human DNA during purification (e.g., purification with the QIAGEN kit and phenol-chloroform) also appeared to affect the ability to detect HPV genomic DNA in oral samples. Puregene provided high DNA purity while preserving human DNA yield, factors which likely affected our improved results with Puregene. Our data indicated that a significant relative loss of human DNA (QIAGEN kit or phenol-chloroform versus Puregene) did not affect the detection of a high-copy-number human gene (β-globin) but did significantly affect HPV detection (Table ). We acknowledge that other DNA purification methods not evaluated as part of this study design may be equally successful in reducing PCR inhibition and in detecting oral HPV infection. Furthermore, we did not evaluate the effects of the DNA purification methods on the detection of HPV in samples from other anatomic sites, such as cervical swab samples. The variability in oral HPV detection observed with different DNA purification methods in this study suggests that evaluation of the effect of the DNA purification method on HPV DNA detection in samples taken from other anatomic sites may be warranted.
This research has direct relevance to the interpretation of existing and future studies, suggesting that HPV prevalence may be underestimated in studies that report results for unpurified oral exfoliate samples. According to our data, oral HPV studies that performed no further DNA purification after protein removal (1
) may have underestimated the oral HPV prevalence by as much as sixfold, while studies that used only ethanol precipitation or phenol-chloroform extraction (17
) may have underestimated the prevalence by 40 to 75%. Our data underscore the importance of DNA purification to avoid the misclassification of HPV status (false-negative results) in oral exfoliate samples and suggest that misclassification is highly dependent on the purification method used. While this research focused on the effect of optimization of laboratory assays for HPV DNA detection, misclassification may also occur due to factors that affect sampling in the oral region; potential sources of sampling variation include behavioral factors, such as tooth brushing; clinical factors, such as treatment for cancer; procedural differences, such as use of a swab compared to use of a rinse and whether a gargle is done; and variation in viral shedding (which may differ in cases and controls).
This study was intentionally performed with HIV-positive men with CD4-cell counts <200 because the high oral HPV infection prevalence in this population increased the power of the study to detect differences in HPV detection between methods. We acknowledge that this study did not include women or immunocompetent subjects. However, the effect of DNA purification method may be equally important and potentially more important in these populations. In this study, the HPV-16 and the HPV-18 viral loads were associated with the ability to detect infection at both visits. The improved sensitivity of the assay may therefore be more important in immunocompetent individuals who may have lower oral HPV viral loads.
Misclassification of oral HPV exposure (false-negative results) due to nonoptimal DNA processing may also contribute to underestimates of the cancer risk associated with oral HPV infection. Two of the three studies of oral HPV infection predicted to have the least DNA extraction-related misclassification based on the data explored above reported increased odds of HNSCCs in individuals with oral HPV (27
), while the study that did not find a relationship used posttreatment cases (26
) (see below). The other widely cited study that did not report an increased oral cancer risk used purification methods subject to more inhibition (14
). Whether DNA purification-related misclassification of oral HPV exposure in cases and controls is nondifferential has not been evaluated but would bias the estimation of cancer risk toward the null. There are currently no data comparing HPV viral loads in oral exfoliates of cases and controls, so the direction of any possible differential misclassification is unknown. Two widely cited studies of oral HPV infection (18
) used posttreatment cases, the results for which may be more likely than those for controls to be misclassified as false negative due to lower viral loads as a result of therapy; this could underestimate the risk of oral HPV infection.
We chose to amplify target HPV genomic DNA with PGMY primer pools because of their relatively equal amplification sensitivities across HPV types, an increased type-specific HPV prevalence (11
), and an improved ability to detect multiple concurrent infections (11
) commonly observed in the oral cavity. We recognize, however, that we may have underestimated the frequency of oral HPV infection in the study population by limiting detection to the HPV types present on the Roche line blot assay. Although there is substantial heterogeneity in primer choice for PCR amplification, most molecular epidemiological studies of HPV infection rely on PCR amplification of HPV genomic DNA. Therefore, we believe that our results could extend to any HPV detection method that relies on PCR amplification of the target.
There were some samples that were positive for HPV-16 and HPV-18 by quantitative PCR that were negative for HPV-16 or HPV-18 by line blot analysis. The majority of these samples had viral loads less than 50 copies per PCR mixture, consistent with the expected lower sensitivity of consensus compared to that of type-specific PCR. However, there were some samples with viral loads as high as ~1,800 copies that were negative by line blot assays. In every case, the sample was also positive for two or more other HPV types. This is likely due to outcompetition for primers by other HPV types with higher viral loads in the sample: the presence of infections with multiple types is known to decrease multiplex assay sensitivity.
In our study design, cell number-standardized samples were created from dilutions of the stock Puregene samples that were used for both ERV-3 quantitation and volume standardization. Analysis of volume-standardized samples was performed last. Eight of 197 (~4%) samples available for volume-standardized analysis were β-globin negative on line blotting, whereas all of these samples were previously β-globin positive on cell number-standardized analyses, and all performed well in the ERV-3 amplification. Because of the nature of the study design, the samples had undergone repeated manipulation, and it is likely that repeated freezing-thawing of the samples may have caused some DNA degradation.
We observed very high rates of agreement between HPV detection results using volume-standardized versus cell number-standardized samples that were comparable to the 91% overall agreement in HPV infection status observed with cervical samples collected by different methods on the same day (9
) as well as the same cervical samples tested on different days (2
). The strong agreement with repeat testing of the same samples provides evidence that the variability observed between visits is not due to limitations in the reproducibility of the HPV detection assay. By contrast, the lower concordance observed for oral exfoliate samples collected 1 week apart either may be ascribed to sampling differences on the 2 days (e.g., the time since the subject last ate or brushed his teeth) or may indicate that oral HPV infection is quite dynamic in this immunocompromised patient population. HPV DNA detection at all sites is subject to variability in sampling and viral load, and no method that reproducibly detect HPV infection all of the time exists. We conclude that the optimized oral rinse method described above can reproducibly categorize an individual as HPV infected and is capable of reproducibly detecting type-specific oral HPV infections. Longitudinal studies sampling the oral cavity at short intervals are now needed to determine the incidence and clearance of oral HPV infection and how frequently sampling is required to monitor the natural history of these infections.
Our data show an increased consistency in HPV detection in patients with higher HPV viral loads because of problems with the reproducibility of the assay at the lower limit of detection (21
). Individuals with discordant HPV status were also more likely to have low HIV viral loads, suggesting that high systemic HIV viral loads may result in higher oral HPV viral loads in those who are infected. This association was independent of the CD4-cell count, suggesting that a high systemic HIV viral load may either directly or indirectly effect local oral mucosal immunity.