The relationship between high-risk human papillomavirus (hrHPV) infection and cervical cancer has become evident from epidemiological and functional studies.1, 2, 3 hrHPV DNA has been detected in almost all cervical squamous cell carcinomas and adenocarcinomas. This finding has led to the widely accepted concept that hrHPV infection is a necessary cause of cervical cancer.1, 2, 3 The close association between hrHPV and cervical cancer has resulted in the use of hrHPV testing on cervical scrapes for the detection of cervical cancer and its precursor stages. Several studies have shown that addition of hrHPV testing to cervical cytology improves the sensitivity and negative predictive value for high-grade cervical intraepithelial neoplasia (CIN) and cervical cancer (≥CIN 2) compared with cytology alone.4, 5 However, the positive predictive value of the hrHPV test is relatively low because many hrHPV positive women will clear the virus. Especially, hrHPV positive women with cytomorphologically normal smears fall in this category. As a consequence, several groups have studied viral parameters that can predict which hrHPV positive women will have or have not cervical cancer. Recent studies suggest that not the presence of viral DNA per sé, but the amount of viral DNA in a cervical scraping (i.e. viral load) would be a potentially relevant determinant for risk assessment of ≥CIN 2.6 In this context, it has been demonstrated for HPV 16 infections that increased viral loads would be associated with an increased risk of ≥CIN 2 whereas reduced amounts of viral DNA reflects the absence of CIN lesions or viral clearance, which is associated with regression of CIN lesions.7 Still, data in favor of viral load being informative for risk of ≥CIN 2 are mainly based on HPV 16, whereas those of hrHPV types other than HPV 16 are limited and inconsistent,8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19 obviously because of the generally low numbers of women with non-HPV 16 types studied so far. Moreover, the potential value of viral load assessment is quite likely to be dependent on the method used to define HPV positivity since its impact seems lower when the hybridization-based hybrid capture 2 (hc2) is used to define HPV positivity when compared with PCR-based assays, which generally display a higher analytical sensitivity.6, 7, 11, 13, 18 Nevertheless, also amongst HPV 16 positive women with normal cytology defined by hc2, semiquantitative viral load assessment could better predict those women who are more likely to have prevalent ≥CIN 3 lesions.19 Still, most studies had in common a substantial overlap of viral load values amongst women with and without high-grade CIN, especially in the range of high viral loads. This precludes setting a cut-off value for high-grade CIN on the basis of high viral loads. Instead, distinguishing a subset of hrHPV positive women with clinically irrelevant HPV infections who will not have or develop high-grade CIN on the basis of low viral loads seems more feasible.6 The latter, however, requires viral load data to be gathered from a large population of hrHPV positive women without underlying high-grade CIN.
Here, we performed a comprehensive viral load analysis for the 4 common hrHPV types HPV 16, 18, 31 and 33 on cervical scrapings of a large reference group of women with normal cytology participating in a population-based cervical screening trial (i.e. POBASCAM; 18, 19) and of a group of women with underlying histologically confirmed CIN. In this way, we succeeded in setting viral load threshold levels for these 4 hrHPV types below which underlying CIN 3 could be excluded.
Material and methods
From the 44,102 women participating in the population-based cervical screening trial POBASCAM,20, 21 a total of 674 with normal cytology and a single type HPV infection of HPV 16 (n = 338), HPV 18 (n = 107), HPV 31 (n = 169) or HPV 33 (n = 60) at baseline, as determined by GP5+/6+-PCR followed by reverse line blot (RLB) genotyping,21, 22 were included in this study. In addition, women with a single type HPV infection by GP5+/6+-PCR RLB having a histologically confirmed cervical intraepithelial neoplasia (CIN) lesion (n = 162) were included for viral load analysis of their cervical scrapes.23 The latter comprised 72 women with HPV 16, 26 with HPV 18, 36 with HPV 31 and 28 with HPV 33 who had been referred for colposcopy-directed biopsy because of an abnormal cervical smear. This group included 125 women with ≥CIN 2 lesions that were selected from a series of 183 ≥ CIN 2 lesions of which another 20 (11%) harbored high-risk human papillomavirus (hrHPV) types different from HPV 16, 18, 31 and 33, and the remaining 38 cases (21%) contained multiple hrHPV infections. Twenty-five (65%) of the latter contained HPV 16.
After a classic cervical smear was made on a slide, cervical scrapes were collected for HPV testing by placing the brush in 5 ml sterile phosphate-buffered saline (PBS, 0.82% (w/v); NaCl, 0.19% (w/v); Na2HPO4·2H2O,·0.03% (w/v); NaH2PO4·2H2O, adjusted to pH 7.4 with HCl) 0.005% merthiolate. Upon arrival in the laboratory, cells were pelleted at 300g for 10 min and resuspended in 1 ml 10 mM Tris-HCl (pH 7.4).
Informed consents were obtained from all participants, and this study was approved by the medical ethics review board.
HPV testing and quantification of HPV DNA and β-globin
For HPV testing and typing by GP5+/6+-PCR, cervical scrape samples were collected and processed as described previously.20, 21
For viral load testing, the DNA of 100 μl of cervical scrape suspensions was purified using an HPPTP kit, according the recommendations of the manufacturer (Roche, Mannheim, Germany). LightCycler-based real-time HPV type-specific as well as a β-globin real-time PCR assays were performed in separate reactions to quantify the number of HPV copies and the number of cells, respectively. LightCycler real-time PCR assays were applied as described by Hesselink et al.24 for HPV 16 and β-globin gene with slight variations depending on the target, which are further detailed below. All primers and probes were selected with the aid of Vector NTI 7.0 software (Informax, Frederick, MD), according to the LightCycler specifications. The HPV type-specific oligonucleotides were selected within the viral E7 open reading frame. Primer and probes sequences for HPV 16 and β-globin gene target were described previously.24 Sequences for HPV 18 (GenBank accession number X05015) oligonucleotides were forward primer: 5′-aagaaaacgatgaaatagatgga-3′(nt position 696–718); reverse primer: 5′-ggcttcacacttacaacaca-3′ (nt position 780–799); donor probe: 5′-catcaacatttaccagcccgacga-3′ (nt position 725–748) and acceptor probe: 5′-ccgaaccaacaacgtcacacaatgt-3′(nt position 750–773). Sequences for HPV 31 (GenBank accession number J04353) oligonucleotides were forward primer: 5′-actgacctccactgttatgagcaa-3′ (nt position 617–640); reverse primer: 5′-cctgctctgtacacacaaacgaa-3′ (nt position 753–775); donor probe: 5′-gatgaggaggatgtcatagacagtccagct-3′ (nt position 656–685) and acceptor probe: 5′-acaagcagaaccggacacatccaat-3′ (nt position 688–712). Sequences for HPV 33 (GenBank accession number M12732) oligonucleotides were forward primer: 5′-gagaggacacaagccaacg-3′ (nt position 575–593); reverse primer: 5′-gccggtccaagccttc-3′ nt position 678–693); donor probe: 5′-atcctgaaccaactgacctatactgctat-3′ nt position 619–647) and acceptor probe: 5′-gcaattaagtgacagctcagatgagg-3′ (nt position 650–675). Donor probes were labeled at the 3′end with a fluorescein-group and acceptor probes at the 5′end with LC-Red 640 and phosporylated at the 3′ end (Eurogentec). Primers and probes were extensively tested in validation experiments, using serial dilutions of cloned HPV plasmid and human DNA.24 (data not shown).
Real-time PCR reactions contained 2 μl FastStart DNA master hybridization probe reaction mixture, 4.0 mM MgCl2, 1.25 μM of each primer and 0.2 μM of each hybridization probe. Additionally, reaction mixtures allocated for the HPV standard curve dilution series were spiked with 100 ng human placental DNA per reaction. Finally, distilled H2O was added to the reaction mixtures to a final volume of 15 μl, after which 5 μl (20 ng/μl) purified sample was added. LightCycler reactions were run in duplicate. The LightCycler PCR conditions for HPV 16 and β-globin gene target and further performance details were described previously24 and those of HPV 18, 31 and 33 were as follows: an initial preincubation step of 10 min at 95°C followed by 45 cycles of 3 sec denaturation at 95°C, 15 sec annealing at 58°C for HPV 18, at 52°C for HPV 31 and at 58°C for HPV 33. Elongation was 8 sec for HPV 18, and 9 sec for HPV31 and 33 at 72°C. After PCR, samples were cooled to 40°C. Transition rate was set at 20°C per second. For quantification of HPV 10-fold dilutions of plasmid DNA containing the full-length genome of the respective HPV types spiked in 100 ng human placental DNA were used ranging from 10 to 100,000 fg plasmid per reaction. The assays for all these 3 HPV types were reproducible and could linearly detect between 10 fg and 100,000 fg plasmid DNA. For HPV 18, 31 and 33, R2 values of the standard curve were 0.994, 0.993 and 0.998, respectively and efficiencies of the standard curves were 1.73, 1.82 and 1.96, respectively.
Viral load values obtained from the duplicate tests were averaged for calculations. Mann-Whitney tests were used to determine differences in viral load between women with normal cytology (reference group) and women with CIN (test group). Type-specific viral load thresholds were defined by using the values discriminating the lowest 25th, 33rd, 50th, 67th and 75th percentiles of viral load in the reference group of women with normal cytology. These values were subsequently used to determine the sensitivity and 95% confidence intervals for underlying ≥CIN 2 or CIN 3. Student's t test was used to determine whether the age differed between women with viral load levels above or below the most discriminative viral load threshold (i.e. 33rd percentile). The level of statistical significance was set at 0.05 and SPSS version 12 was used for all analyses.
Type-specific viral load in women with normal cytology and CIN
Type-specific viral load data obtained from analysis of cervical scrapes of women with normal cytology and underlying CIN displaying single infections with HPV type 16, 18, 31 or 33 are summarized in Table I. The median viral load in women with normal cytology was 0.61 (range: minimum–maximum): <0.01–456.08) copies per cell (c/c) for HPV 16 (n = 338), 0.62 (range: 0.01–361.05) c/c for HPV 18 (n = 107), 3.27 (range: 0.02–3979.08) c/c for HPV 31 (n = 169) and 10.95 (range: <0.01–623.16) c/c for HPV 33 (n = 60). For all these 4 HPV types, the viral load was significantly higher in scrapes of women with ≥CIN 2 when compared with those of women with normal cytology (Table I). The median viral load values for women with ≥CIN 2 were 11.62 (range: 0.25–530.68) c/c for HPV 16 (n = 58, p < 0.001), 10.37 (range: 0.10–4694.10) c/c for HPV 18 (n = 17, p < 0.001), 23.40 (range: 0.45–1180.80) c/c for HPV 31 (n = 27, p < 0.001) and 83.70 (range: 1.70–5807.72) c/c for HPV 33 (n = 23, p < 0.001). When compared with women with normal cytology, women with CIN 1 also displayed higher viral loads, although this was only statistically significant for HPV types 16 and 18. Higher viral copy numbers were found for HPV 16, 31 and 33 in women with ≥ CIN 2 when compared with those with CIN 1 but in none of these cases this difference was statistically significant.
Table I. Type-Specific Relationship between HPV DNA Loads in Women with Normal Cytology and CIN
When comparing the viral load values between the different HPV types, it was evident that the viral load was lowest for type 16 and 18 and highest for HPV 33 both in women with normal cytology and those with ≥CIN 2 (Tables I and II). In women with normal cytology, the viral load for HPV 16 and 18 was significantly lower than for HPV 31 (p < 0.001, for both type 16 vs. 31 and type 18 vs. 31) and HPV 33 infections (p < 0.001, for both type 16 vs. 33 and type 18 vs. 33). In addition, women containing HPV 31 had significantly less viral copy numbers than those harboring HPV 33 (p = 0.001). Also in women with ≥CIN 2, viral copy numbers per cell were significantly higher in HPV 33 positive women when compared with those positive for HPV 16 (p < 0.001), HPV 18 (p = 0.006) and HPV 31 (p = 0.038).
Table II. Comparisons between Viral Copy Numbers Per Cell Per HPV Type*
Mann-Whitney tests were used to calculate the p-values. p-values representing significant differences are indicated in italics. See Table I for viral load data.
HPV 16 versus
HPV 18 versus
HPV 31 versus
Assessment of sensitivity of different viral load threshold values for ≥CIN 2 and CIN 3
All HPV 16 positive women with ≥CIN 2 had equal to or more than 0.25 viral c/c in their cervical scrapes (Table I). When comparing this value to percentiles of lowest viral load values deduced from the values measured in the 338 HPV 16 positive women with normal cytology (Table III), it appeared that this amount is higher than thresholds for the 25th (threshold: 0.11 c/c) and 33rd (threshold: 0.20 c/c) percentiles, but lower than that of the 50th percentile (threshold: 0.61 c/c). The sensitivities for underlying ≥CIN 2 were 100% (95% CI 93.8–100%) in case of the 25th and 33rd percentile thresholds and 91.4% (95% CI 81.0–97.1%) in case of the 50th percentile threshold (Table III). The sensitivity for ≥CIN 2 decreased gradually when using thresholds of higher percentiles. Similar sensitivity values were obtained when only HPV 16 positive women with underlying CIN 3 were taken into account. These figures indicate that for HPV 16 infections a viral load threshold could be set at the 33rd percentile threshold (i.e. 0.20 c/c) of women with normal cytology to increase the specificity of HPV testing for prevalent ≥CIN 2 or ≥CIN 3 without loosing sensitivity. In practice, this would mean that the use of this threshold allows distinguishing about 33% of HPV 16 infected women with normal cytology within a population-based screening cohort without ≥CIN 2 or ≥CIN 3.
Table III. Type-Specific Threshold Values for Detection of High-Grade CIN
In contrast to HPV 16, sensitivities for ≥CIN 2 did not reach 100% when using viral load thresholds down to the 25th percentiles for HPV types 18, 31 and 33 (Table III). Using the 25th percentile thresholds for these types, the sensitivity for ≥CIN 2 varied from 91.3% (for HPV 33) to 96.3% (for HPV 31). However, when considering only women with CIN 3 who were infected with one of these types, sensitivities of 100% (with 95% CI 59.0–100% for HPV 18 and 95% CI 75.3–100% for both HPV 31 and 33) were obtained in case the thresholds of up to the 33rd percentiles were used, similar as what was found for HPV 16. The 33rd thresholds were 0.30 c/c, 1.03 c/c and 5.13 c/c for HPV 18, 31 and 33, respectively.
Overall, when combining the data of all 4 HPV types, the sensitivity for ≥CIN 2 was 96.8% (95% CI 92.0–99.1%) when the type-specific thresholds were set at the 25th percentile, 95.2% (95% CI 89.9–98.2%) at the 33rd percentile and 87.2% (95% CI 80.0–92.5%) at the 50th percentile. Restricting the analyses to women with CIN 3, these figures reached a sensitivity of 100% (95% CI 93.9–100%) for all 4 HPV types at both the 25th and 33rd percentiles thresholds and 89.8% (95% CI 75.0–94.0%) at the 50th percentile threshold (Table III). There was no significant age difference between the 33% women with normal cytology displaying the lowest viral load when compared with the women having a viral load higher than the 33rd percentile (38.3 years vs. 38.0 years; p = 0.702).
In this study, we aimed to assess the discriminative power of viral load thresholds of HPV 16, 18, 31 and 33 to distinguish women with single infections for one of these types without high-grade CIN by making use of a large population-based screening cohort of women with normal cytology. For all HPV types analyzed, we demonstrated a statistically significant difference in median viral load per cell between women with normal cytology and those with high-grade CIN. In addition, for all these types, a 100% sensitivity for underlying CIN 3 was obtained when using viral load thresholds that equal the type-specific viral load values marking the 33rd percentile values of HPV positive women with normal cytology within a population-based screening setting. Hence, although the absolute viral load levels at the 33rd percentiles differed between the different types, the use of type specific viral load thresholds can exclude prevalent CIN 3 in 33% of the women with normal cytology participating in a population-based screening program, who are GP5+/6+-PCR positive for HPV 16, 18, 31 or 33.
Our findings support the idea that increased viral loads are generally associated with an increased risk of high-grade CIN, even in case of infections with hrHPV types different from HPV 16. A plausible explanation for this concept would be that decreased viral loads are a consequence of an effective immune surveillance. This finds support by the observation that HPV 16 and 18/45 positive carriers of commonly reported protective human histocompatibility leukocyte antigen class II (HLA II) alleles displayed lower viral loads together with short term HPV infections and a decreased risk of cervical carcinoma in situ.25, 26
Although our findings for non-HPV 16 types are in line with some studies,11 it contradicts others.12 Variables such as differences in methods to preselect hrHPV positive women6 and sample size may have contributed to inconsistent data for non-HPV 16 types. In our study, we preselected women with single type infections. However, meanwhile we have collected preliminary data of cervical scrapings containing multiple infections including HPV 16, 18 and/or 31 and corresponding biopsies of some women with underlying CIN lesions. It appeared that in this biopsy series only one of the HPV types present in the paired scrape could be detected by GP5+/6+-PCR. Interestingly, the type detected in the biopsies was the type that displayed the highest viral DNA in the corresponding scrapings (unpublished data). From this pilot, we anticipate that in multiple infections the HPV type with the highest viral load is clinically most relevant. Therefore, in multiple infections, the far majority of which contain HPV 16, 18, 31 and/or 33,21 additive viral load analysis of the predominant HPV type is quite likely to be informative as well in terms of absence of prevalent CIN 3 lesions. In addition, it is noteworthy that we used women with normal cytology as reference group and considered these women as having no underlying ≥CIN 2 lesion. Still, this group may include a small proportion of women with prevalent high-grade CIN lesions missed by cytology. Since we anticipate that such cases would display increased viral copy numbers that fall in the higher viral load percentiles,19 we consider the sensitivity of the 33rd percentile thresholds set in this study not over represented. It is also noteworthy that total DNA can be derived from many types of cells from the cervix, including inflammatory cells that are not HPV host cells. This may affect the relationship of HPV to cell number in a variable manner. Nevertheless, this apparently does not influence the outcome for sensitivity of ≥CIN 2 when the 33rd percentile thresholds are used. Therefore, we have the impression that the influence of DNA of the non-HPV host cells on viral load levels in general practice is limited.
Remarkably, median viral load values differed considerably between HPV 16 and 18 infections and infections with HPV 31 and HPV 33. Particularly, HPV 33 loads were outlying, and at least 7-fold higher than those of HPV 16/18, both in women with normal cytology and in women with CIN. It is unlikely that these differences reflect type-specific differences in real-type PCR assays since a recent study, using a different real-time PCR procedure, also revealed a markedly higher viral load of particularly HPV 33 in comparison to HPV 16 and 18 in a control population of women without cervical disease.27 It is still questionable what would be the underlying mechanism for this type-specific difference in viral load.
It should be realized that the figures obtained in this study are based on cross-sectional analyses that included women with prevalent high-grade CIN having abnormal cervical smears. Recent longitudinal studies performed on archival smears suggest that the risk of developing incident cervical carcinoma in situ or carcinoma in women with normal cervical smears increases with higher viral load for most hrHPV types, but calculated odds ratios may differ considerably between different HPV types.27, 28 Further follow-up data of the HPV 16, 18, 31 or 33 positive women with normal cytology in the POBASCAM study, as are currently being gathered, will ultimately yield essential information to what extent the defined viral load thresholds need some adaptation to also ensure a 100% sensitivity for incident ≥CIN 3 lesions.
A major challenge is now to translate the viral load information collected herein into clinical practice. The defined viral load cut-off values were determined after stratification for the amount of input DNA. Given the variations in DNA yield of cervical scrapings,12, 27 measurements of the DNA content and the use of standard DNA input amounts in real-time HPV tests seem to be required for a reliable application of a viral load cut-off. Because the agreement between semiquantitative viral load measures of GP5+/6+-PCR and type specific real time viral load data is poor,24 we propose not to use the set threshold for a modified cut-off of the GP5+/6+-PCR. Instead, type-specific real-time PCR as a follow-up test for women with normal cytology who are GP5+/6+-PCR positive for HPV 16, 18, 31 and/or 33 seems more reliable. Given the fact that the combined data from several recent studies have revealed that the HPV types 16, 18, 31 and 33 display an excess risk of high-grade CIN,21, 29, 30 we advocate to group together women with normal cytology harboring hrHPV types different from HPV 16, 18, 31 and 33 with those containing HPV 16, 18, 31 and/or 33 at viral load levels below the defined 33rd percentile thresholds into a category of women who need less aggressive follow-up. Since single and multiple infections with HPV 16, 18, 31 and/or 33 together comprise about 62% of hrHPV infections in women with normal cytology, this approach would result in the distinction of about 58% of all hrHPV positive women with normal cytology participating in a population-based screening program (i.e. one third of the 62% women with infections involving these 4 common types with viral load levels below the 33rd percentiles plus 38% of women who test positive for other hrHPV types) who would need a less aggressive management (i.e. follow-up by cytology and HPV testing after 2 years). This increased specificity of HPV testing for prevalent CIN 3 can be reached without the cost of a marked reduction in sensitivity.
In practice, our data open possibilities to test women first with a clinically validated hrHPV test, such as hc2 or GP5+/6+-PCR followed by typing. For those that test hrHPV positive, reflex cytology can be performed to detect women with abnormal cytology who need direct referral for colposcopy directed biopsy (i.e. women with > borderline or mild dyskaryosis (BMD)) or repeat testing after 6 months (i.e. women with BMD).31,32 Remaining hrHPV positive women with normal cytology who are positive for HPV 16, 18, 31 and/or 33 are subjected to type-specific real-time PCR for viral load analysis. Those with HPV 16, 18, 31 and/or 33 viral load values above the mentioned 33rd percentile thresholds could then be subjected to a more aggressive management, i.e. follow-up by cytology and HPV testing after 1 year.29 This scenario is depicted in Figure 1. In this way, almost 58% of hrHPV positive women can be excluded from extensive follow-up, which results in a significant increase in specificity and positive predictive value of hrHPV testing for ≥CIN 3 in cervical screening programs.
The authors are grateful to M. Lettink and M. Verkuyten for technical assistance.