Biases in human papillomavirus genotype prevalence assessment associatedwith commonly used consensus primers

Abstract

Consensus primers targeting human papillomaviruses (HPVs) have biases in sensitivity toward certain HPV types. We applied 3 primer sets (GP5+/6+, MY09/11, PGMY09/11) in parallel on 120 Chinese cervical cancer specimens. GP5+/6+ exhibited a poor sensitivity for HPV52, for which the prevalence among squamous cell cervical cancer was underestimated from 14.6% to 0%. The fact that HPV52 should rank second in prevalence among squamous cell cervical carcinoma in Hong Kong could be missed if GP5+/6+, a worldwide commonly used primer set, was selected for HPV detection. Biases in HPV type-specific sensitivity may result in misprioritization of vaccine candidates. © 2005 Wiley-Liss, Inc.

Cervical cancer is the second most common cancer in women worldwide. Strong and consistent evidence accumulated over the past 2 decades confirms the aetiologic role of human papillomaviruses (HPVs) and provides a strong impetus for developing HPV vaccines to prevent cervical cancer. More than 100 HPV types have been identified, and at least 30 have been found in cervical cancers.1 Given the diversity of HPV types and the largely type-specific immunity after natural infections,2, 3, 4 it is important to delineate the prevalence of different HPV types found in cervical cancers so as to guide the selection of vaccine candidates. Most studies on HPV have employed consensus primers with an intention to cover a broad spectrum of HPV types. With the more than 10% sequence variation between HPV types, biases in sensitivity of a given primer set toward certain types could happen. In our study, we examined the influence of detection methods on assessing the prevalence of HPVs and thus their priority as cervical cancer vaccine candidates.

Material and methods

A total of 120 cervical cancer specimens (105 fresh frozen and 15 paraffin embedded; 89 squamous cell carcinoma, 26 adenocarcinoma, 4 adenosquamous carcinoma and 1 lymphoepitheliod carcinoma) collected from Hong Kong Chinese aged 26–84 years (mean 55 years; SD 13.9) were examined. Total DNA was extracted by the QIAamp DNA mini kit (QIAGEN, Hilden, Germany) and with the quality of extracted preparations confirmed by beta-globin PCR.5 The clinical materials were collected with a written informed consent. Our study was approved by the local institutional ethics committee, and the human experimentation guidelines of the local institute were followed in the conduct of our study.

HPV detection and typing was accomplished by 3 different methods in parallel. In the first method, HPV DNA was amplified by the GP5+/6+ primers that target an approx. 150 bp fragment of the L1 region.5, 6 The HPV type was identified by direct sequencing of PCR amplicons. In the second method, the MY09/11 primers that target an approx. 450 bp fragment of the L1 region was used for PCR.7, 8 HPV type was identified by restriction fragment length polymorphisms (RFLPs) using endonucleases RsaI and DdeI as previously described.8 Ambiguous RFLPs were resolved by direct sequencing of PCR amplicons. The third method was based on the premarket version of the HPV Consensus PCR/Line Blot Genotyping Kit (Roche Molecular Systems, Alameda, CA). This kit uses the PGMY09/11 primers and is followed by hybridization with 27 HPV type-specific probes immobilised on a strip.9

Results

Altogether 99.2% (119/120) of the specimens had HPV DNA detected by one or more methods. The 5 most common HPV types found from cervical cancers in Hong Kong as revealed by a combination of the 3 methods were HPV16 (50.0%), HPV18 (21.7%), HPV52 (12.5%), HPV58 (8.3%) and HPV33 (5.0%). HPV18 was the most prevalent (61.5%) found among adenocarcinoma (Table I). The PGMY09/11 method was most superior in sensitivity with a positive rate of 95.8% compared to 84.2% of the MY09/11 and GP5+/6+ methods (p = 0.006 by χ² test) (Table II). Among all the HPV types identified, a statistically significant difference in the positive rate as revealed by the 3 detection methods was only observed for HPV52. The GP5+/6+ method failed to detecte HPV52 from all of the 15 specimens that were shown to be HPV52-positive by other methods. The MY09/11 method also had a lower positive rate for HPV52 when compared to the PGMY09/11 method, although a statistically significant difference was not observed (HPV52-positive rate and 95% confidential interval [CI] 0.09 (0.04–0.14) for the MY09/11 method; 0.13 (0.07–0.18) for the PGMY09/11 method)(Table II).

Table I. Distribution of Human Papillomavirus (HPV) Types Among 120 Hong Kong Cervical Cancer Specimens According to Histologic Types

HPV type1	No. (%) of HPV-positive specimens
	Squamous cellcarcinoma(n = 89)	Adeno-carcinoma(n = 26)	Adenosquamouscell carcinoma(n = 4)	Lympho-epitheloidcarcinoma(n = 1)	Overall (n = 120)
1 HPV detected by any of the 3 methods. 2 Regardless of single or multiple infection. Specimens are counted more than once when multiple infections are detected.
Any HPV2	89 (100)	25 (96.2)	4 (100)	1 (100)	119 (99.2)
HPV16	48 (53.9)	9 (34.6)	2 (50.0)	1 (100)	60 (50.0)
HPV18	9 (10.1)	16 (61.5)	1 (25.0)	0	26 (21.7)
HPV26	1 (1.1)	0	0	0	1 (0.8)
HPV31	4 (4.5)	0	0	0	4 (3.3)
HPV33	6 (6.7)	0	0	0	6 (5.0)
HPV35	4 (4.5)	0	0	0	4 (3.3)
HPV39	1 (1.1)	0	0	0	1 (0.8)
HPV45	2 (2.2)	0	0	0	2 (1.7)
HPV51	1 (1.1)	0	0	0	1 (0.8)
HPV52	13 (14.6)	2 (7.7)	0	0	15 (12.5)
HPV53	1 (1.1)	0	0	0	1 (0.8)
HPV58	9 (10.1)	0	1 (25.0)	0	10 (8.3)
HPV59	2 (2.2)	0	0	0	2 (1.7)
HPV66	1 (1.1)	0	0	0	1 (0.8)
HPV68	2 (2.2)	0	0	0	2 (1.7)
Any multiple infection	9 (10.1)	2 (7.7)	0	0	11 (9.2)
HPV16, 18	0	1 (3.8)	0	0	1 (0.8)
HPV16, 59	1 (1.1)	0	0	0	1 (0.8)
HPV16, 68	1 (1.1)	0	0	0	1 (0.8)
HPV18, 52	2 (2.2)	1 (3.8)	0	0	3 (2.5)
HPV33, 52	1 (1.1)	0	0	0	1 (0.8)
HPV16, 18, 52	1 (1.1)	0	0	0	1 (0.8)
HPV16, 39, 52	1 (1.1)	0	0	0	1 (0.8)
HPV31, 35, 58	1 (1.1)	0	0	0	1 (0.8)
HPV18, 31, 35, 58, 66	1 (1.1)	0	0	0	1 (0.8)

Table II. Distribution of Human Papillomavirus (HPV) Types Among 120 Hong Kong Cervical Cancer Specimens According to Detection Methods

HPV type	No. (%) of HPV-positive specimens (n = 120)			p-value1
	GP5+/6+	MY09/11	PGMY09/11
1 By χ^2 test to assess difference in positive rates obtained by the 3 methods. 2 Regardless of single or multiple infections. Specimens are counted more than once when multiple infections are detected.
Overall2	101 (84.2)	101 (84.2)	115 (95.8)	0.006
HPV16	58 (48.3)	53 (44.2)	54 (45.0)	0.791
HPV18	23 (19.2)	20 (16.7)	26 (21.7)	0.616
HPV26	0	0	1 (0.8)	0.367
HPV31	3 (2.5)	1 (0.8)	4 (3.3)	0.409
HPV33	6 (5.0)	6 (5.0)	5 (4.2)	0.940
HPV35	2 (1.7)	2 (1.7)	4 (3.3)	0.599
HPV39	0	0	1 (0.8)	0.367
HPV45	2 (1.7)	0	2 (1.7)	0.364
HPV51	0	0	1 (0.8)	0.367
HPV52	0	11 (9.2)	15 (12.5)	<0.001
HPV53	0	1 (0.8)	1 (0.8)	0.605
HPV58	7 (5.8)	8 (6.7)	10 (8.3)	0.740
HPV59	0	0	2 (1.7)	0.134
HPV66	0	0	1 (0.8)	0.367
HPV68	0	0	2 (1.7)	0.134
Multiple infections	0	1 (0.8)	9 (7.5)	Not done
HPV16, 18	0	0	1 (0.8)	Not done
HPV16, 52	0	1 (0.8)	0	Not done
HPV18, 52	0	0	3 (2.5)	Not done
HPV33, 52	0	0	1 (0.8)	Not done
HPV39, 52	0	0	1 (0.8)	Not done
HPV16, 18, 52	0	0	1 (0.8)	Not done
HPV31, 35, 58	0	0	1 (0.8)	Not done
HPV18, 31, 35, 58, 66	0	0	1 (0.8)	Not done

Upon knowing the difference in performance for HPV52, 2 follow-up experiments were performed. First, the analytic sensitivity was assessed by the method of limiting dilution. The results indicated that both the MY09/11 and PGMY09/11 systems achieved a lower detection limit of 50 copy equivalents of HPV52 per 10 μL inoculum, which was similar to those for HPV16 and HPV18. Second, the HPV52-positive samples were reamplified and sequenced to confirm the typing result as well as to detect possible sequence variations in the GP5+ and GP6+ primer binding sites. The primer pair PGMY09-G and PGMY11-B was selected from the PGMY primer pool for this purpose as they were designed for amplifying HPV52. The results showed that all the HPV52 isolates collected in our study had a synonymous nucleotide substitution at position 6764 T→C compared to the prototype (Genbank accession no. NC_001592). This substitution resulted in an additional mismatch at the sixth position of the 3′ end of GP6+ primer.

Of the 120 specimens, 11 (9.2%) had multiple infections. The PGMY09/11 method detected most of them (9/11, 81.8%), MY09/11 detected 2/11 (18.2%), whereas the GP5+/6+ method detected none. The inability of the GP5+/6+ method to detect multiple infections was at least partly due to subsequent typing by sequencing that had difficulties in revealing multiple types.

When the prevalence rates of HPV types among cervical cancers of all histologic types were ranked respectively by the 3 methods, a similar result was observed for PGMY09/11 and MY09/11 (HPV16 > HPV18 > HPV52 > HPV58 > HPV33). However, due to the poor sensitivity for HPV52, a different ranking was observed when the GP5+/6+ method was being used (HPV16 > HPV18 > HPV58 > HPV33 > HPV31). The GP5+/6+ method resulted in an underestimation of the prevalence of HPV52 from 12.5% to 0% when all histologic types of invasive cervical cancers were considered. When the analysis was confined to invasive squamous cell cervical cancers, HPV52 should account for 14.6% of cases and was second only to HPV16 in prevalence ranking.

Discussion

GP5+/6+ and MY09/MY11 are the most commonly used primers for detecting genital HPVs,10 whereas the PGMY09/11 primers was first introduced in 2000.11 HPV prevalence data generated by GP5+/6+ and MY09/MY11 primers constituted a major portion in international analyses on the risk association and vaccine priority for HPVs.1, 10, 12 Although in some studies the identification of HPV types was accomplished by hybridization with multiple type-specific probes that may increase the sensitivity, our results indicated that the GP5+/6+ PCR by itself has suboptimal sensitivity for HPV52. This is in line with a previous study reported by Qu et al.,13 where the GP5+/6+ system also had missed all of the four samples in which HPV52 was detected by the MY09/11 method. In our current study, we found that the poor sensitivity could partly be explained by the circulation of a variant with additional sequence mismatch over the primer binding site. The extent of distribution and the risk association of this variant is currently unknown. HPV52 ranked sixth in prevalence among cervical cancers worldwide when a combination of results generated from MY09/11 and GP5+/6+ was analyzed,1, 14 and thus was regarded as having lower priority. The magnitude of underestimation associated with the GP5+/6+ system depends on the true prevalence of HPV52, which is likely to have geographic and ethnic variations.15 In turn, the implication of these underestimations depends on the true oncogenicity of HPV52. Munoz et al.14 estimated the odds ratio for squamous cell cervical cancer of HPV52 to be 200 (95% CI 67.8–590.1), which is about half of that for HPV16, and ranked fifth among high-risk HPV types. The study also indicated that when HPV16 existed as coinfections with another high-risk type, the odds ratio increased by about one-third. Among cervical cancer patients in Hong Kong, a substantial proportion (40%) of HPV52 existed as coinfections with one or more high-risk HPV types. The oncogenic risk attributed to HPV52 in these coinfected cases, in particular its possible augmenting effect on the risk of the other coexisting HPV, deserves further investigation.

Our results showed that coinfection with multiple HPV types is not uncommon in cervical cancer cases. Our study design was not able to provide a fair comparison on the ability of the 3 primer sets to detect multiple infections. Nevertheless, the potential problem of certain HPV detection systems in underestimating coinfections should be considered when the assessing oncogenic risk and the priority for a vaccine candidate, as well as when projecting the cost benefit of future cervical cancer vaccines.