Evaluation of Seegene Anyplex II Performance for Detection of Human Papillomavirus Genotypes in Formalin-Fixed, Paraffin-Embedded Cervical Cancer Specimens Open Access

Samples used in this study were drawn from the Australian Cervical Cancer Typing Study (ACCTS)⁷  of 847 cancers (551 [65.1%] squamous, 243 [28.7%] adenocarcinoma, 36 [4.3%] adenosquamous, and 17 [2.0%] other) collected between the years 2005 and 2015. The ACCTS study demonstrated the prevalence of HPV types in this cohort of cervical cancers. The protocol for processing and HPV typing undertaken through the ACCTS study has previously been described.⁷  Briefly, all whole tissue sections (WTSs) were stained with hematoxylin-eosin from before and after the WTS sample and were evaluated by a trained histopathologist. A 9-μm WTS was deparaffinized and processed for DNA extraction using the MagNA Pure 96 DNA and Viral NA Small Volume Kit (MP96, Roche Molecular Systems Inc, Pleasanton, California) and eluted in 100 μL of the Roche elution buffer.⁷  All specimens underwent testing by SPF10, which detects 25 HPV genotypes, including 13 high-risk genotypes (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68). HPV typing was performed according to manufacturer’s specifications, with a specific PCR assay used to identify HPV 68 and 73 genotypes, as reported previously.⁸  DNA extracts were stored at −30°C for approximately 5 years before the current study. For the current study, 248 extracts were selected based on their positivity for HPV types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68/73 determined by SPF10.

SPF10 does not have an internal control (IC); hence, sample assessability was first reevaluated by the detection of a 110-bp fragment from the β-globin gene, as described previously.⁸  Three samples that failed detection were deemed unassessable and were excluded from further analyses. Seegene Anyplex II has incorporated an IC into the test setup for validating the results. The results with a weak (+) or negative IC signal are considered invalid by Anyplex II. In this study, regardless of the IC value, when a sample was positive for at least 1 HPV genotype it was deemed “assessable” and was included in our further analyses.

Anyplex II HR HPV assay detects 14 high-risk HPV genotypes (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68) and an IC in 1 reaction well. The Anyplex II assay reports the results with a crossing point (Cq) range, 31 or fewer cycles (+++), 31 to 39 cycles (++), and more than 40 cycles (+). This Cq range was used to assess the validity of the results. Samples were deemed invalid if the IC had a Cq greater than 40 and was negative for HPV. To analyze samples on the Anyplex II, 5 μL of stored nucleic acid extract was tested according to the manufacturer’s instructions. Samples with HPV-negative results on Anyplex II were retested by SPF10.

HPV typing results were assessed for the frequency of oncogenic genotypes and any genotype bias between the 2 used assays. Overall test agreement (the number of concordant samples divided by the total number of samples), as well as negative and positive percent agreements (NPA and PPA, respectively) were calculated for any HPV genotype, and for each HPV individual genotype detection. Comparisons were analyzed using MedCalc (MedCalc Software Ltd, Ostend, Belgium) statistical software and Cohen κ statistic.^25,26 

Cohen κ statistic was assessed as 0 or lower as indicating no agreement and 0.01 to 0.20 as none to slight, 0.21 to 0.40 as fair, 0.41 to 0.60 as moderate, 0.61 to 0.80 as substantial, and 0.81 to 1.00 as almost perfect agreement.²⁶ 

The ACCTS study was approved by the Austin Hospital and Mercy Health, Heidelberg, Australia (LNR13Austin170, R13–60); The Royal Women’s Hospital, Parkville, Australia (13/25); South Eastern Area Laboratory Services, New South Wales Health Pathology, Australia (13/305); Hunter Area Pathology Service, New South Wales Health Pathology, Australia (13/305); Pathology West, Western Sydney Local Health District, Australia (LNR/13/POW/617); Royal Prince Alfred Hospital, Camperdown, Australia (X13–0445); and Royal Brisbane and Women’s Hospital, Herston, Australia (14/QRBW/3). The current study is a substudy of the ACCTS study covered by the above approvals, and so it did not require additional approval.

The flowchart of sample selection and processing is presented in Figure 1. Of 847 WTSs of FFPE cervical samples with HPV-positive test results by SPF10,⁸  total nucleic acid extracts from 248 samples were randomly selected. Three samples failed quality control (β-globin detection)⁸  and were excluded from further analysis. Of the 245 remaining samples, 210 (86%) were positive for HPV and the remaining 35 were negative by Anyplex II. To rule out any negative effects of the time interval between the 2 assays on HPV results, we retested the 35 Anyplex II negative samples by SPF10. Of the 35 retested samples, 2 (HPV 45–positive samples) were HPV negative and excluded from further analysis, whereas 1 contained fewer HPV genotypes (HPV 45 versus HPV 33 and 45). The remaining samples showed 100% concordance with the original SPF10 test results.

Of the remaining 243 samples with detectable β-globin, 32 samples (13%) had an invalid IC (Cq >40), of which 17 (7%) were negative for any HPV genotype, deemed unassessable, and were excluded. The remaining 226 assessable samples were designated group A and were used for individual HPV genotype agreement analysis. Of the 226 samples, 15 (6.6%) were HPV positive with an invalid IC, and the remaining 211 with a valid IC (Cq >40) formed group B, which was analyzed separately. Overall, 14 HPV genotypes were detected by both platforms (Supplemental Table 1; see the supplemental digital content, containing a table and a figure at https://meridian.allenpress.com/aplm in the March 2024 table of contents).

An assessment of the agreement between the 2 assays was undertaken. Using Anyplex II, oncogenic HPV genotypes were detected in 210 of the 243 SPF10 HPV-positive samples (86.4%; 95% CI, 81.5–90.5). A similar prevalence for each individual genotype was detected by both platforms (Figure 2). HPV 16 and 18 were the most frequently detected genotypes (79 versus 88 and 63 versus 59, for SPF10 and Anyplex II, respectively). Anyplex II and SPF10 showed very high agreement for the detection of the 2 most important oncogenic genotypes HPV 16 (219 of 226; 96.9%; 95% CI, 93.7–98.75) and HPV 18 (221 of 226; 97.8%; 95% CI, 94.9–99.3). The overall agreement for any HPV genotype detection was 88% (221 of 251, which included 8 samples with 1 type on SPF10 and 2 types on Seegene; 95% CI, 83–92; Table 1). The PPA varied between genotypes, with high values for the most frequently detected types in cancer lesions of HPV 16 (81 of 81; 100%), HPV 18 (59 of 64; 92%), and HPV 33 (21 of 23; 91%), whereas they were slightly lower for HPV 45 (16 of 20; 80%). κ values were high for the nonavalent vaccine types (ranging from 0.88 to 0.95), except for HPV 52 (0.69). The lowest PPA values were obtained for HPV 66 and 68, which were only present in small numbers. The NPA was 100% for all genotypes except HPV 16 (95%) and 52 (99%). For 201 samples that had a single type detected by both assays, there was 100% genotype agreement.

A high overall assay agreement of 96% to 100% was reported for most individual genotypes. SPF10 and Anyplex II results were fully concordant at genotype level for 194 of 226 samples (86%); the discordant results included a single genotype missing/addition, and no different genotypes were detected in corresponding samples. The remaining 14 samples (7%) with detection of multiple types were partially concordant (Table 2). Notably, there were 6 samples where Anyplex II detected both HPV 16 and 18 genotypes, but SPF10 only detected HPV 18. In 1 case, Anyplex II failed to detect HPV 18 from an SPF10 HPV 16/18–positive sample.

The above analysis of group A (Figure 1) included 15 samples that tested “invalid” by Anyplex II IC but positive for HPV. These 15 samples had HPV genotyping results in complete agreement with the corresponding SPF10 results. Performing a sensitivity analysis excluding these 15 samples (ie, limiting analysis to group B; Supplemental Figure 1) had minimal impact on the test concordance, with PPA, NPA, and overall agreement varying by less than 2% for overall detection and for the detection of types (except for HPV 68) compared with the analysis with the full 226 samples (Supplemental Table 1).

This study evaluated the performance of Anyplex II for the detection and genotyping of HPV from FFPE cervical cancer samples.⁸  High levels of agreements on HPV genotyping were observed between Anyplex II and SPF10, a manufacturer-validated method for HPV genotyping using FFPE samples. Overall genotyping assay agreement was 91% (range, 87%–94%), and for individual vaccine-preventable HPV types it was 96% to 100% (Table 1). However, Anyplex II displayed a lower analytic sensitivity than SPF10 because 13.6% of SPF10-positive samples remained undetected by Anyplex II.

Surveillance of the HPV genotypes causing cervical cancer is necessary to ensure prevention measures continue to be evidence based. This can inform both the development of new vaccines and the refinement of HPV-based cervical screening programs. Cervical cancer samples are routinely taken for histopathology evaluations, and these FFPE specimens are stable for long-term storage at room temperature. Surveillance can be achieved through HPV testing of the collected FFPE cancer samples, which requires a robust, efficient, and inexpensive HPV typing assay. FFPE cervical specimens are taken directly from the lesions and can be stored at room temperature for many years with minimum maintenance requirements. The manufacturer-validated SPF10 assay is widely used for HPV genotyping FFPE samples. However, relative to quantitative PCR–based tests, such as Anyplex II, SPF10 is expensive, more labor intensive, and requires additional equipment, such as an automated blotting device, to ensure accurate and reproducible hybridization results. Anyplex II detects 14 high-risk HPV genotypes in 1 well and no additional equipment is required, with a shorter turnaround time than SPF10 (4 hours versus 8 hours). The assay is compatible with large-scale DNA extraction platforms, such as MP96.²⁵  The current study validated Anyplex II for FFPE specimens and further demonstrated that it can function as a strong candidate for HPV genotyping in a clinical setup.

Other approaches to HPV typing of FFPE samples include RNAscope technology, which detects HPV RNA from FFPE tissue specimens in situ and retains the tissue architecture.²⁷  This assay has been used for simultaneous detection of 7 genotypes (16, 18, 31, 33, 35, 52, and 58), with a turnaround time of about 8 hours.²⁸  New versions of the RNAscope HPV assay use a pool of multiple genotype-specific probes to detect all 13 high-risk HPV, but these assays are unable to assess the presence of individual HPV types.³⁹ 

Another approach to HPV typing of FFPE samples is amplicon sequencing using the GP5+/6+ primers combined with Illumina sequencing technology.²⁹  This is a high-throughput sequencing approach for assessing relatively large numbers of FFPE samples. This has some drawbacks, including the need to batch analysis of samples using complex computational algorithms and the high cost of Illumina sequencing, and the use of broad-spectrum primers GP5+/6+ may miss some HPV types with nucleotide mismatches at primer binding sites.

In the current study, some discrepancies were identified between Anyplex II and SPF10. Although more samples were deemed HPV 18 positive by SPF10 (n = 64 versus n = 59), more genotype 16 was detected by Anyplex II (n = 79 versus n = 88); notably, this most often happened in samples where HPV 18 was codetected by Anyplex II. We did not observe any great difference between the 2 tests detecting other oncogenic genotypes, including those included in HPV nonavalent vaccine (31, 33, 45, 52, and 58), nor did we see a bias toward certain types.

The Anyplex II system uses an IC to validate sample quality. In this study, the genotyping results of 32 of 243 extracts were called “invalid” because of insufficient IC detection; however, HPV genotypes were detected in 15 of these samples (Cq >30). The results for these 15 samples were completely concordant with SPF10 results, and sensitivity analysis showed that including HPV-positive, IC invalid samples had no profound effect on assay comparison. The manufacturer’s instructions on sample validity were developed for cervical screening applications. Our results suggest redefining “sample assessability” for FFPE samples to include IC “invalid” samples where at least 1 HPV genotype is detected at a level of “+” or above, with an appropriate sensitivity analysis to identify potential bias. Also, our results highlight the need for further optimization of Anyplex II to improve its sensitivity during the detection of its IC cellular gene. The discrepancy between host and viral DNA detection by Anyplex II could be partly explained by the fact that (1) host DNA decays faster than HPV DNA in FFPE samples,³⁰  (2) viral DNA, relative to β-globin, may be present at high copy number, and (3) the assay has a higher sensitivity for the detection HPV genome than β-globin.

SPF10 and Anyplex II systems have been used for the detection and genotyping of HPV in FFPE clinical samples^31–33  but have not previously had a head-to-head comparison. Lillsunde Larsson et al³¹  compared the performance of Anyplex II and 3 other HPV genotyping assays for the detection of HPV from 99 FFPE samples from a variety of anatomic sites. The results suggested (1) a high agreement between Anyplex II and an in-house reference real-time PCR, targeting the E6 or E7 genes of 12 high-risk (types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, and 59) and 2 low-risk HPV (types 6 and 11), and (2) that Anyplex II is sensitive and can be successfully used for an extended HPV genotyping. Interestingly, in contrast to our results, Rollo et al³⁴  reported only 1 invalid result for Anyplex II when they tested 160 oropharyngeal cancer FFPE samples. This discrepancy could be due to the size of biopsies, tissue types (oropharyngeal cancer versus cervical cancer), and the different DNA extraction methods (MP96 versus QIAamp DNA FFPE Tissue Kit from Qiagen). Also, they showed that Anyplex II was more sensitive than SPF10 INNO-LiPA Extra, which is a less sensitive version of the RHA kit HPV SPF10-LiPA25 used in our study.³⁵ 

This study has limitations. Thirteen percent of samples that were positive for HPV on SPF10 were negative on Anyplex II. These discrepant results may be due to sample degradation during storage, although efforts were made to account for gross sample degradation by retesting of Anyplex II negatives on SPF10. It is acknowledged that the HPV detection/typing results from 3 of 35 samples that were HPV positive in the original SPF10 assessment were not reproduceable following storage. This may suggest that some DNA degradation occurred across the panel of samples. However, retesting the 35 Anyplex II negative samples by SPF10 suggests that long-term storage does not substantively affect HPV detection using SPF10 but may have done so for the Anyplex II depending on the assay limit of detection or stochastic performance. This could also be explained by the fact that limit of detection of SPF10 is lesser than that of Anyplex II (eg, SPF10 detects 1 to 2 copies of genotypes 16 and 18 genomes, but the Anyplex II detection limit for these 2 genotypes is 50 copies).^36,37  The results indicate that further optimization of Anyplex II (ie, increasing the assay detection level) may improve the performance of this assay using FFPE samples. The agreement values for HPV types less strongly associated with cervical cancer were impacted by small numbers, resulting in wide CIs.

In conclusion, the Anyplex II assay performed with high agreement with SPF10 for detection of HPV genotypes in FFPE samples. Of note, reference assay SPF10 appeared to have a higher analytic sensitivity for the detection of HPV from FFPE samples, although to fully evaluate analytic sensitivity, the analysis of HPV-negative samples and those with shorter archival time would be required. Consistent with Meijer criteria (90% positivity),³⁸  a close correlation (86%) between SPF10 and Anyplex II indicates that both tests can be used for testing FFPE tissue samples for the presence of HPV genotypes. Anyplex II has the advantage of convenient setup, being cheaper, having a shorter turnaround time, and requiring fewer resources and less expertise.