Validation of the SPF₁₀ LiPA Human Papillomavirus Typing Assay Using Formalin-Fixed Paraffin-Embedded Cervical Biopsy Samples

Abstract

Lower levels of performance of human papillomavirus (HPV) typing assays in studies using formalin-fixed paraffin-embedded (FFPE) tissue compared to those using exfoliated cervical cells have been reported. The interpretation of current studies is limited by bias in inclusion criteria, sample matching, and methods of cell collection. We aimed to validate FFPE tissue for typing by the use of the SPF₁₀ LiPA assay, comparing cervical scrapings to punch and cone biopsy specimens. We examined 165 paired cervical scraping and FFPE punch biopsy samples, and 66 paired FFPE punch and cone biopsy samples. HPV typing was performed using the SPF₁₀ LiPA assay. Kappa statistics were used to measure interrater agreement. The overall agreement with respect to HPV status was 100%. For 74.5% of subjects (kappa = 0.6147), the same numbers of HPV types were detected in scraping and biopsy specimens. The overall positive typing agreement was 95.4% (range, 93.4 to 97.3) for 441 out of 484 individual HPV type analyses. Agreement was good for HPV-39, -42, -43, and -70 (kappa = 0.6506 to 0.7166), excellent for HPV-6, -16, -18, -31, -33, -35, -40, -51, -52, -56, -58, and -66 (kappa = 0.8499 to 0.9665), and absolute for HPV-11, -44, -45, -53, and -68. In 43.9% of cases (kappa = 0.247), the same numbers of HPV types were found in punch and cone biopsy specimens. Overall positive agreement for typing was 86.8% (range, 82.5 to 91.1) for 204 out of 266 individual HPV type analyses. More infections by HPV-18, -33, -51, and -52 were detected in cone specimens. HPV typing by SPF₁₀ LiPA performed equally well for cervical scraping specimens and standard pathological material. Some viral types are preferentially detected in cone specimens, likely reflecting better sampling of diseased epithelium and endocervix tissue.

The etiologic role of human papillomavirus (HPV) in the genesis of cervical cancer and squamous epithelial lesions at different sites has been strongly established (11a). More than 100 HPV types, at least 30 of which are found in the cervix, are known (23). The distribution of HPV types varies greatly worldwide depending on the target population, the severity of cervical lesions, and the geographical area (3).

The clinical emphasis on HPV typing in the management of cervical intraepithelial neoplasia (CIN) has been mainly in distinguishing low- from high-risk infections; therefore, most assays use pooled analysis of multiple HPV types (12). However, as there is important variability in the risk of CIN, depending on the specific HPV types encountered, assays able to discriminate multiple types in a single test give an added value for patient stratification (5, 14, 21). Furthermore, as HPV immunity is essentially specific with respect to type (28), the distribution of HPV types guides the selection of candidate viruses for vaccination (22). Finally, a reduction of the incidence of infections by targeted viruses represents a surrogate endpoint in trials of vaccine efficacy (9).

Analytical detection of specific HPV types usually employs PCR amplification of HPV sequences by either single or multiple consensus primers followed by type discrimination with specific probes. SPF₁₀ LiPA is a broad-spectrum, short-fragment PCR assay based on the amplification of a 65-bp region of the L1 open reading frame (15, 16). Relevant features of the assay are its potential to amplify up to 54 individual HPV types, its proven clinical significance, and its high performance compared to other tests in terms of sensitivity, reproducibility, and coverage of HPV types (4, 17, 24, 26, 30, 31).

Rigorous quality control of both analytical and clinical performance is critical in the use of any HPV typing test (18). The present study was designed to validate formalin-fixed paraffin-embedded (FFPE) cervical biopsy samples for analyses by the use of SPF₁₀ LiPA. The primary aim was to establish the accuracy of HPV DNA detection and typing using FFPE punch biopsy specimens compared to exfoliated cells concurrently obtained by cervical scraping. A secondary aim was to assess the reproducibility of HPV typing for punch biopsy sampling compared to the corresponding cone biopsy sampling technique.

MATERIALS AND METHODS

Patients.

Paired cervical scrapings and FFPE punch biopsy specimens from 165 cases were used for comparisons of HPV typing profiles. Samples with multiple infections were preferentially selected. Both types of specimens were collected during the same colposcopy session. Sixty-six paired samples of archival FFPE punch and corresponding cone biopsy specimens were also employed. The time interval between the punch biopsy and conization was 4 weeks on average. The main features of the three series of samples are summarized in Table 1.

TABLE 1.

Main clinical features and HPV status for the study base

Characteristic	Category	Value for indicated specimen group
		Paired scraping and punch biopsy			Paired punch and cone biopsy
		Scraping	Punch biopsy	Both	Punch biopsy	Cone biopsy	Both
No. of women				165			66
Mean age (yr)				37.39			38.21
HPV type status	Negative	20	20		0	0
	1	3	5	1	2
	2	20	24	8	7
	3	84	61	25	16
	4	33	44	28	22
	≥4	5	11	4	19
HPV class of risk	Single LRa	0	0		0	0
	Multiple LR	0	0	0	0
	Single HRb	3	5	1	2
	Multiple HR	74	74	32	29
	LR and HR	68	66	33	35
Squamous cervical lesion status	Negative			22			0
	CIN grade 1			33			5
	CIN grade ≥ 2			110			61

^aLR, low-risk HPV strain(s).

^bHR, high-risk HPV strain(s).

All pathological samples included in the study were collected at the Colposcopy Service of the Institute of Obstetrics and Gynaecology (IRCCS-Fondazione Policlinico San Matteo and University of Pavia, Pavia, Italy) over the period 2005 to 2007. The main indications for colposcopy were abnormal cytological findings produced by a PAP test, a previous history of squamous intraepithelial lesions or CIN, suspected or proven vulvovaginal condylomatosis, and vulvodinia (7). Informed consent for HPV testing was obtained from all subjects.

HPV DNA detection and typing.

Cervical scrapings for HPV typing were obtained immediately before the colposcopy procedure. After speculum examination, scrapes were taken with a cervix brush, suspended in ThinPrep-PreservCyt solution (Cytec Corporation, Marlborough, MA), and stored at 4°C. DNA extraction was performed within 1 week of sampling by lysis and digestion with proteinase K. Briefly, pelleted cells from 1.5 ml of PreservCyt solution were washed in phosphate-buffered saline and resuspended in 100 μl of lysis solution (50 mM KCl, 10 mM Tris-HCl [pH 8.3], 2.5 mM MgCl₂, 0.45% Tween 20, 0.45% NP-40, 500 mg of proteinase K/ml) at 56°C for 1 h. Following heat inactivation of proteinase K, 10 μl of the solution was used for PCR amplification of HPV sequences from the L1 region by the use of SPF₁₀ primers in a final reaction volume of 50 μl for 40 cycles.

For DNA isolation from FFPE samples, three to five 10-μm-thick sections were incubated in 200 μl of a lysis solution (1 mg of proteinase K/ml in 50 mM Tris [pH 8.0[-1 mM EDTA-0.45% Tween 20-0.45% Igepal CA-630) for 16 to 24 h at 56°C. The proteinase K was heat inactivated, and the lysates were centrifuged to eliminate wax, extracted with phenol-chloroform, and resuspended in 100 μl of the solution. Ten microliters was used for the PCR. Entire histological sections were used for DNA extraction of punch and cone biopsy specimens; diseased epithelia were not microdissected. Pooled sections from 2 paraffin blocks were used for each cone specimen (which are usually sampled in their entirety using 10 to 12 blocks); blocks were selected based on CIN extension data.

Positive and negative controls were introduced for each set of 12 reactions; these included DNA from Siha and HeLa cell lines for a specified number of HPV copies and blank reagents throughout all steps of the procedure. Concurrent amplification of beta-globin sequences was used as control for DNA adequacy.

HPV type-specific sequences were detected by the line probe, INNO-LiPA HPV genotyping assay, version 2 (Innogenetics N.V., Ghent, Belgium), according to the manufacturer's instructions. The current version of the assay allows the simultaneous and separate detection of 15 high-risk HPV types (HPV-16, -18, -31, -33, -35, -39, -45, -51, -52, -56, -58, -59, -66, -68, and -70) and 10 low-risk HPV types (HPV-6, -11, -34, -40, -42 to -44, -53, -54, and -74). Hybridization patterns were visually interpreted by two independent readers; discrepant interpretations were resolved by consensus reading or, occasionally, by a third reader.

Statistical analysis.

The concordance between HPV types identified in scrapings and those identified in punch and cone biopsy samples was evaluated by the computation of kappa-statistic measurements of interrater agreement. A kappa value of 0.75 or more suggests excellent agreement, and a value of 0.40 or less indicates poor agreement (8). For each HPV type, we also reported the percentages of agreement between groups. The percentage of agreement is the proportion of readings that are placed in the same category by the test. For the overall population, we also computed the percentage of positive agreement (concordance of positive readings) together with 95% confidence intervals (CIs). Chi square tests was used to determine linear trends across ordered categories.

RESULTS

A total of 165 paired cervical scraping and punch biopsy specimens were selected from a large cohort of over 2,000 patients referred to a tertiary colposcopy service who underwent routine HPV typing using exfoliative cytology. There were 20 HPV-negative samples and 145 HPV-positive samples that accounted for 484 individual infections by 22 different HPV genotypes (Table 1). Multiple HPV infections were present in 98% of samples; 26% had four or more infecting types.

The overall agreement for HPV status was 100%. The same numbers of viral types were detected in 74.48% of cases (kappa = 0.6147; P < 0.00001), with slightly better performance shown by biopsy results (Table 2). Sixteen HPV types were detected only in scraping specimens and 27 only in biopsy specimens. Notably, biopsy was more effective in the detection of HPV-18 (49 versus 44 infections) and -39 (10 versus 5 infections), whereas no specific HPV type was preferentially detected in scraping specimens. The overall percentage of positive agreement for typing was 95.4% (95% CI, 93.4 to 97.3), corresponding to 441 concordant HPV types out of 484 individual infections. Observed figures for crude agreement were high for all HPV types, ranging from 96.55% to 100% (Table 2). Interrater agreement shown by kappa-statistic determinations was good for HPV-39, -42, -43, and -70 (kappa = 0.6506 to 0.7166), excellent for HPV-6, -16, -18, -31, -33, -35, -40, -51, -52, -56, -58, and -66 (kappa = 0.8499 to 0.9665), and absolute for HPV-11, -44, -45, -53, and -68 (kappa = 1.0000).

TABLE 2.

Comparison^a of HPV typing results obtained for 165 paired cervical scraping and punch biopsy specimens of all genotypes

HPV type	No. of positive samples in indicated specimen category			Agreement (%)	Expected agreement (%)	Kappa	SE	P
	Scraping and punch biopsy	Scraping	Punch biopsy
6	35	1	1	98.62	62.67	0.9630	0.0830	<0.0001
11	16	0	0	100.00	80.37	1.0000	0.0830	<0.0001
16	100	1	2	97.93	58.00	0.9507	0.0830	<0.0001
18	44	0	5	96.55	56.37	0.9210	0.0828	<0.0001
31	72	1	3	97.24	50.01	0.9448	0.0830	<0.0001
33	8	1	0	99.31	88.96	0.9375	0.0829	<0.0001
35	6	1	1	98.62	90.81	0.8499	0.0830	<0.0001
39	5	0	5	96.55	90.13	0.6506	0.0778	<0.0001
40	20	1	1	98.62	75.23	0.9443	0.0830	<0.0001
42	4	2	1	97.93	92.70	0.7166	0.0827	<0.0001
43	1	0	1	99.31	97.95	0.6636	0.0782	<0.0001
44	7	0	0	100.00	90.81	1.0000	0.0830	<0.0001
45	7	0	0	100.00	90.81	1.0000	0.0830	<0.0001
51	41	2	0	98.62	58.84	0.9665	0.0830	<0.0001
52	36	1	3	97.24	61.31	0.9287	0.0830	<0.0001
53	8	0	0	100.00	89.57	1.0000	0.0830	<0.0001
56	13	0	2	98.62	82.54	0.9210	0.0826	<0.0001
58	9	2	0	98.62	87.15	0.8927	0.0826	<0.0001
59	0	1	0	99.31	99.31
66	3	1	0	99.31	95.29	0.8537	0.0822	<0.0001
68	3	0	0	100.00	95.95	1.0000	0.0830	<0.0001
70	3	1	2	97.93	93.98	0.6561	0.0825	<0.0001
Totalb	441	16	27

^aCalculated by kappa statistic measurement of interrater agreement.

^bOverall positive agreement, 95.4% (95% CI, 93.4 to 97.3).

Sixty-six pairs of punch and cone biopsy samples were also examined to assess the reproducibility of HPV typing using different FFPE specimens. The sample accounted for 266 individual HPV infections by 20 different HPV genotypes (Table 1). Multiple HPVs were present in 98% of samples, and 48% (31 samples) had four or more infecting types: either multiple high-risk types (9 samples) or high-risk and low-risk types (22 samples).

The same numbers of viral types were found in 43.9% of cases (kappa = 0.247; P < 0.001), with better performance exhibited by cone biopsy specimens. The overall percentage of positive agreement for typing was 86.8% (95% CI, 82.5 to 91.1), corresponding to 204 concordant HPV types out of 266 individual infections (Table 3). Twenty HPV types were detected only in punch biopsy specimens and 42 only in cone specimens. More infections by HPV-33 (7 infections versus 1 infection) and HPV-52 (25 infections versus 19 infections) were detected in cone specimens, whereas no specific type was preferentially detected in punch biopsy samples. The agreement of the results for the two specimen categories for HPV detection was not uniform across the HPV types. Agreement was poor for HPV-33 (kappa = 0.3592), good for HPV-6, -18, -44, -52, -66, and -79 (kappa = 0.4884 to 0.0.7298), excellent for HPV-16, -31, -35, -39, -40, -43, -45, -51, -53, -56, and -58 (kappa = 0.7843 to 0.9474), and absolute for HPV-11 and -68 (Table 3).

TABLE 3.

Comparison^a of HPV typing results obtained for 66 paired cervical punch and cone biopsy specimens of all genotypes

HPV type	No. of positive samples in indicated specimen category			Agreement (%)	Expected agreement (%)	Kappa	SE	P
	Punch and cone biopsy	Punch biopsy	Cone biopsy
6	14	4	5	86.36	59.64	0.6621	0.1230	<0.0001
11	7	0	0	100.00	81.04	1.0000	0.1231	<0.0001
16	47	1	2	95.45	61.02	0.8834	0.1230	<0.0001
18	18	5	4	86.36	55.05	0.6966	0.1230	<0.0001
31	35	2	2	93.94	50.73	0.8770	0.1231	<0.0001
33	2	1	5	90.91	85.81	0.3592	0.1113	0.0006
35	4	0	2	96.97	85.95	0.7843	0.1202	<0.0001
39	5	1	1	96.97	83.47	0.8167	0.1231	<0.0001
40	11	1	0	98.48	71.21	0.9474	0.1229	<0.0001
43	2	1	0	98.48	92.70	0.7925	0.1204	<0.0001
44	4	0	3	95.45	84.62	0.7045	0.1176	<0.0001
45	7	0	2	96.97	78.65	0.8581	0.1218	<0.0001
51	14	1	3	93.94	63.22	0.8352	0.1227	<0.0001
52	18	1	7	87.88	55.14	0.7298	0.1205	<0.0001
53	3	0	1	98.48	89.94	0.8493	0.1217	<0.0001
56	5	1	1	96.97	83.47	0.8167	0.1231	<0.0001
58	5	0	2	96.97	83.43	0.8172	0.1210	<0.0001
66	1	0	2	96.97	94.08	0.4884	0.1058	<0.0001
68	1	0	0	100.00	97.02	1.0000	0.1231	<0.0001
70	1	1	0	98.48	95.55	0.6598	0.1157	<0.0001
Totalb	204	20	42

^aCalculated by kappa statistic measurement of interrater agreement.

^bOverall positive agreement, 86.8% (95% CI, 82.5 to 91.1).

Scraping, punch biopsy, and cone specimens of the same lesion were available for 51 subjects (Table 4). Comparison of typing results obtained with the three different types of specimens confirmed that cone and punch biopsy specimens detected more HPV types overall than scraping, with levels of 93.5% (172/184), 82.6% (152/184), and 77.7% (143/184), respectively (chi square value for trend = 17.5; P < 0.001). Punch and/or cone biopsy samples were more effective than scraping in the detection of HPV-18 (22 versus 15). For cone specimens versus punch biopsy and/or scraping specimens, infections by HPV-33 (5 versus 1), HPV-51 (18 versus 15), and HPV-52 (26 versus 19) were preferentially detected.

TABLE 4.

Comparison^a of HPV typing results obtained for 51 paired cervical scraping, punch biopsy, and cone specimens for selected genotypes

HPV type	No. of positive samples in indicated specimen category							Kappa	P
	All	Scraping	Punch biopsy	Cone biopsy	Punch biopsy and scraping	Punch and cone biopsy	Cone biopsy and scraping
6	12	0	0	5	4	0	0	0.7507	<0.0001
11	6	0	0	0	0	0	0	1.0000	<0.0001
16	36	1	1	1	0	0	0	0.9128	<0.0001
18	14	0	3	4	1	0	0	0.7738	<0.0001
31	25	0	2	2	0	1	0	0.8853	<0.0001
33	0	0	0	4	0	0	1	0.4201	0.0006
39	2	0	0	1	0	3	1	0.5967	<0.0001
40	10	0	0	0	0	0	0	0.9711	<0.0001
51	13	0	0	3	0	1	1	0.8486	<0.0001
52	15	0	0	7	0	3	1	0.7156	<0.0001
Total	133	1	6	27	5	8	4

^aCalculated by kappa statistic measurement of interrater agreement.

DISCUSSION

The reproducibility, the accuracy, and the sensitivity of HPV typing assays have relevant implications for the establishment of individual HPV transmission, the monitoring of virus persistence, and the management of diseases of women with cytological abnormalities (18).

Analysis of HPV types by the use of FFPE cervical tissue has several potential fields of application (11), including epidemiological studies of HPV circulation in different geographical areas and time periods, cancer screening, and vaccination trials. Targeted analysis of cervical lesions may be a more clinically relevant outcome in assessing vaccine efficacy than cytology, as, in principle, it rules out vaginal infections or transient infections due to recent sexual activity (9).

The present results indicate that HPV typing by SPF₁₀ LiPA performs equally well with cervical scraping specimens and with FFPE punch biopsy samples. Good to excellent agreement was observed in the assessment of the numbers and the types of multiple HPV infections, as shown by the reporting of kappa values systematically above 0.65. More discrepancies were found in the comparison between punch and cone biopsy samples. A total of 172 out of 184 HPV types (93.5%) were found in cone biopsy specimens compared to 152 (82.6%) in punch biopsy specimens. In particular, cone biopsy was more sensitive in the detection of HPV-33 and -52. Overall, the use of FFPE specimens resulted in detection of more HPV-18, -33, -51, and -52 infections than the use of cervical scrapings.

Only a few studies have critically evaluated the use of FFPE tissue specimens in HPV typing compared to the use of exfoliated cell samples collected by washing or scraping. Quint et al. (25) used SPF₁₀ LiPA to compare 174 paired washing and biopsy samples that represented a spectrum of cervical lesions similar to those examined in the present study. However, a large proportion of their biopsy samples were obtained an average of 100 days after scraping, cone biopsy samples were not examined, and only 40% of patients had multiple infections. Those authors reported results that indicated excellent agreement in the assessment of HPV status but reduced performance of FFPE specimens in the detection and typing of multiple infections.

A recent report by Gravitt et al. (11) indicated a lower level of agreement of the results of HPV typing of paired biopsy and lavage specimens and a systematic underperformance of assays using FFPE tissue. That study had major biases in design and execution: (i) the authors used different HPV typing assays for lavage and biopsy specimens; (ii) 25% of the tissue samples were excluded because of a loss of CIN lesions; (iii) the exfoliated cells were collected by lavage and not by cervical scraping, which allows a better sampling of the transformation zone; (iv) the lavage and biopsy specimens were not collected concurrently; and (v) 60% of the infections represented a single HPV type. Most of the differences detected between lavage and biopsy in the study of Gravitt et al. (11) were likely related to the different levels of performance of the assays rather than to the types of specimens.

In contrast to those two previous studies, the present investigation indicated that the use of FFPE specimens, in particular, in the cone biopsy assays, enabled the detection of more HPV types than scraping. The enrichment of pathological tissue allowed by targeted biopsies and the more effective sampling of the endocervix, especially in the case of infections by HPV-18 (19), are logical explanations of these results. Differences in setting and design, in particular concerning the selection of women, may explain the discrepancies observed between the results of the studies. Other variables that impact the performance of the typing assay, such as the modalities of sample collection and storage, the methods of DNA extraction, and the protocols of PCR amplification and detection (10, 17), may also have contributed. For all the above-named reasons, the present results cannot be transferred to other typing assays that have analytical features different from those of SPF₁₀ LiPA and that use different amplicons and primer systems (6, 17).

A possible bias of the present study derives from the selection of women with abnormal cytological findings and infections with multiple HPV types who are not representative of the general population of HPV-infected subjects. This type of patient population sampling, however, reflects the current indications for cervical biopsy procedures that are usually not performed for patients with normal cytology or colposcopy indications. Infections with multiple HPV types also tend to occur more frequently among patients with CIN, and such multiple infections tend to correlate with the severity of lesions (7).

Several factors support the consistency of the results. The SPF₁₀ LiPA assay has been widely validated in the setting of clinical and epidemiological studies and in population screenings (4, 24, 26). This assay is particularly suitable for the study of FFPE material, as it targets the smallest amplicon of any HPV DNA typing system available (15). In fact, the efficiency of the use of primer sets in the amplification of HPV sequences from archival samples is inversely correlated to the length of the amplicon, because DNA extracted from FFPE tissue is usually at a low concentration and fragmented (2, 13). The analytical sensitivity of SPF₁₀ LiPA is the highest among HPV typing assays, and it is likely to be orders of magnitude above the sensitivity required in clinical practice (17, 27). Reduced sensitivity toward certain genotypes has been reported for SPF₁₀ LiPA compared to other assays, especially for HPV-42, -16, -18, -39, -56, -58, -59, -66, and -59 (4, 17, 30), but, except for HPV-18 (see above), none of these types showed lower agreement with respect to the results obtained in our study. Furthermore, discrepant typing results were not correlated to CIN grade or to the number of infecting viruses (data not shown).

The performance of HPV typing tests based on consensus PCR mainly depends on the number of infecting viruses and their relative concentrations (30). SPF₁₀ LiPA has been shown to underestimate the presence of multiple infections compared with type-specific PCR because of primer competition between sequences, especially when a low-copy-number virus is markedly underrepresented compared to other types (29). As 98% of our samples had multiple infections, the observed agreement values should represent conservative estimates. We have previously determined HPV DNA titers for 130 FFPE biopsy specimens by real-time PCR with SPF₁₀ primers and found no differences attributable to numbers of infecting types, CIN severity, and duration of storage (B. Dal Bello, A. Spinillo, P. Alberizzi, S. Cesari, B. Gardella, and E. M. Silini, submitted for publication).

Validation of SPF₁₀ LiPA for HPV typing of FFPE specimens is a prerequisite for the use of archival tissue in epidemiological and clinical studies. Based on the study of archival CIN biopsy specimens collected over the last 3 decades, we have recently shown that HPV type distribution has significantly changed in our area, providing the proof of principle of time course changes in viral circulation (Dal Bello et al., submitted).

Analysis of targeted biopsy specimens may also help to clarify some of the controversies about the clinical significance of multiple HPV infections. Results showing detection of multiple HPV types in exfoliated cell samples have been disputed, as they might reflect unselected cervical areas or vaginal infections, whereas targeted biopsies would allow a more precise topographic link between CIN and specific HPV types. The present results indicate that cervical scraping and biopsy are equivalent with respect to HPV typing, in spite of lesion targeting. This observation challenges the current notion that only a few viral types play the main characters in cervical carcinogenesis whereas all others have “walk-on parts” (20, 22). Any further refinement of the targeting of type-specific HPV infections would require demanding techniques capable of spatial resolution such as “in situ” hybridization (10).