Human papillomavirus 33 worldwide genetic variation and associated risk of cervical cancer

Abstract

Human papillomavirus (HPV) 33, a member of the HPV16-related alpha-9 species group, is found in approximately 5% of cervical cancers worldwide. The current study aimed to characterize the genetic diversity of HPV33 and to explore the association of HPV33 variants with the risk for cervical cancer. Taking advantage of the International Agency for Research on Cancer biobank, we sequenced the entire E6 and E7 open reading frames of 213 HPV33-positive cervical samples from 30 countries. We identified 28 HPV33 variants that formed 5 phylogenetic groups: the previously identified A1, A2, and B (sub) lineages and the novel A3 and C (sub)lineages. The A1 sublineage was strongly over-represented in cervical cases compared to controls in both Africa and Europe. In conclusion, we provide a classification system for HPV33 variants based on the sequence of E6 and E7 and suggest that the association of HPV33 with cervical cancer may differ by variant (sub)lineage.

Background

A subset of human papillomaviruses (HPV) is considered the necessary cause of cervical cancer and a proportion of other anogenital and head and neck carcinomas (Bouvard et al., 2009). While infection with a high-risk HPV is a common event among sexually active women, the majority of infections are cleared within 2 years (Goodman et al., 2008; Insinga et al., 2009; Plummer et al., 2007; Rosa et al., 2008). There is a need to identify the factors that influence a patient's progression from HPV infection to viral persistence, cellular transformation, and cervical cancer. Current evidence suggests that sequence variations within HPV16 and HPV18 may influence viral persistence and clinical outcome (Gheit et al., 2011; Schiffman et al., 2010; Zuna et al., 2009; Sathish et al., 2005; Villa et al., 2000; Berumen et al., 2001).

HPV33 is a high-risk HPV within the same phylogenetic species (alpha-9) as HPV16 (Bernard et al., 2010; de Villiers et al., 2004) and accounts for approximately 5% of cervical cancer cases worldwide, with some variation in this proportion by geographical region (Guan et al., 2012; Li et al., 2011). For example, while HPV33 is found in 5.4% of cervical cancer cases in Eastern Asia, it is only found in 1.7% of cases in Oceania (Li et al., 2011). HPV33 was originally cloned from an invasive cervical carcinoma (Beaudenon et al., 1986), and the entire viral sequence was described shortly after (Cole and Streeck, 1986). Since then, a few studies have described genetic variation in HPV33 worldwide (Chen et al., 2011; Stewart et al., 1996; Godínez et al., 2013) of which the most comprehensive was based upon the whole genome sequencing of 20 HPV33-positive samples (Chen et al., 2011). HPV33 variants were classified into two major lineages, A and B. The A lineage was further divided into two sublineages, A1, which includes the prototype sequence [M12732.1 (Cole and Streeck, 1986)], and A2. This classification is based upon the definition that the full genome sequence of a major variant lineage differs by approximately 1.0% from another variant lineage of the same HPV type, with differences of 0.5–0.9% defining sublineages (Chen et al., 2011).

Previous studies of HPV33 variants and clinical outcome have tended to be small and focused on a particular geographical region or ethnic group (Khouadri et al., 2006; Bokal et al., 2010; Gagnon et al., 2004; Garbuglia et al., 2007; Ntova et al., 2012; Raiol et al., 2009; Wu et al., 2009; Xin et al., 2001). Hence, the aims of the current study were to further characterize the genetic diversity of HPV33 worldwide and to explore the association of well-defined HPV33 variant lineages with the risk for cervical cancer. To do this, we sequenced and analyzed the entire E6 and E7 open reading frames of HPV33-positive cervical samples stored at the International Agency for Research on Cancer (IARC), including cervical cancer cases and controls collected during more than 20 years of studies on HPV.

Materials and methods

Origin of clinical specimens

The IARC has coordinated cervical cancer case series, cervical cancer case-control studies, and population-based HPV prevalence surveys in a large number of countries around the world (Bosch et al., 1995; Muñoz et al., 2003; Franceschi et al., 2003; Clifford et al., 2005; Li et al., 2006; Bardin et al., 2008; Dondog et al., 2008; Keita et al., 2009; Sherpa et al., 2010; Alibegashvili et al., 2011; De Vuyst et al., 2012; Aruhuri et al., 2012; Sideri et al., 2009) and as yet unpublished studies from Fiji, Bhutan, and Senegal. The collection of samples has spanned a period of over 20 years, but predates the introduction of HPV vaccines. Informed consent was obtained from all participants, and the studies were approved by the IARC Ethical Review Committee. Cervical samples (exfoliated cells or tissue biopsy specimens) derived from these studies have been comprehensively genotyped for 37 HPV types by using a standardized and well-validated protocol (General Primer GP5+/6+PCR–EIA followed by reverse line blot assay) (van den Brule et al., 2002) in one centralized laboratory (Molecular Pathology Unit, Department of Pathology, VU University Medical Center, Amsterdam, The Netherlands). All HPV33-positive cervical samples in the IARC biobank were selected for the current analysis, without exclusion, and were categorized into the following regions: Africa, Asia and Oceania, Europe, and South America. Country specific details are noted in Table 1.

Table 1.

Geographic distribution of 213 HPV33-positive cervical samples

Region/country	No.
AFRICA	56
Algeria	2
Guinea	17
Kenya	3
Mali	1
Morocco	3
Nigeria	4
Senegal	13
South Africa	10
Tanzania	1
Uganda	2
ASIA and OCEANIA	87
Bhutan	21
China	6
Fiji	3
India	13
Korea	5
Mongolia	20
Nepal	3
Thailand	8
Vanuatu	1
Vietnam	7
EUROPE	36
Georgia	7
Italy	5
Poland	10
Spain	14
SOUTH AMERICA	34
Argentina	3
Brazil	6
Chile	7
Colombia	5
Paraguay	11
Peru	2
TOTAL	213

PCR and DNA sequencing

DNA extraction from stored samples was performed using the High Pure PCR Template Preparation Kit (Roche, Mannheim, Germany), and DNA isolates were subjected to β-globin PCR to assess sample quality, as described previously (Hesselink et al., 2009). Sequencing of the entire HPV33 E6 and E7 region (nucleotides 109–866) was performed as described previously (Godínez et al., 2013) using a series of HPV33 specific primer pairs that were designed to amplify overlapping regions of the HPV33 E6 and E7 open reading frames in order to cover the entire E6 and E7 region.

To reveal single nucleotide polymorphisms (SNPs), sequences of the specimens were aligned to the prototype HPV33 sequence (NCBI accession number M12732) using multalin software (http://multalin.toulouse.inra.fr/multalin/). SNPs that were observed in only one sample were confirmed by re-examination of the sequence traces. Isolates that did not classify into existing lineage categories were confirmed by manual re-examination of the sequencing traces and with additional sequencing, where necessary. Multiple sequence traces for each sample were compiled to provide one sequence encompassing the entire HPV33 E6 and E7 region. All sequences were submitted to GenBank (accessions KC862070–KC862080, KC881011–KC881020, and KF536962–KF536968).

Phylogenetic analysis

Unrooted consensus trees were built using the Phylogeny Inference Package (PHYLIP), version 3.69 (Felsenstein, 1989). This included generating 10,000 bootstraps using the F84 model of DNA distances, clustering with the unweighted pair group method with arithmetic mean (UPGMA), and applying the majority rule extended, or greedy, method of consensus. Trees created with a maximum-likelihood method showed similar results and are not described further. For the principal tree analysis, all unique sequence variants found in IARC samples were included. In a subsequent analysis, IARC variants were supplemented by other unique E6/E7 sequences reported in the literature, including 2 from Garbuglia et al. (2007), 6 from Khouadri et al. (2006), 1 from Wu et al. (2009), and 4 from Chen et al. (2011) and personal communication. The IARC samples that contained a variant that appeared to belong to a newly described (sub)lineage were full-genome sequenced as described previously (Chen et al., 2011) to establish the phylogenetic classification and nomenclature.

Case-control analysis

Samples were classified as either controls [including normal, atypical squamous cells of unknown significance (ASCUS), low-grade intraepithelial lesion (LSIL), or cervical intraepithelial neoplasia (CIN) 1] or cases [squamous cell carcinoma, adenocarcinoma, adenosquamous cell carcinoma, or unspecified invasive cervical cancer]. Samples from population-based HPV prevalence studies for which histology and cytology were unavailable were also classified as controls (n = 15). Samples reported as CIN2, CIN3, or high-grade squamous intraepithelial lesion (HSIL) were excluded from the case-control analysis (n = 15), but were included in the previously described phylogenetic analysis. Region-specific associations between variant (sub)lineage and case-control status were assessed by 2-sided p-values arising from Fisher's exact test without combining (sub)lineages. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated for the A1 sublineage vs. the combination of all other (sub)lineages. All statistics were calculated with SAS version 9.3 (SAS Institute, Cary, NC, USA). Because of the strong regional heterogeneity, regions were never combined in the statistical analyses.

Results

Sequencing

The entire E6 and E7 genes were sequenced in a total of 213 HPV33-positive cervical samples from 30 countries, including 56 samples from 10 countries in Africa, 87 samples from 10 countries in Asia/Oceania, 36 samples from 4 countries in Europe, and 34 samples from 6 countries in South America (Table 1).

A total of 36 SNPs were identified across the E6 and E7 open reading frames. The observed combinations of these SNPs resulted in 28 unique sequences, which will be called variants (Table 2). In E6, there were 24 SNPs, 16 resulting in amino acid changes. In E7, there were 12 SNPs, 6 resulting in amino acid changes. No SNPs were observed in the 14 nucleotide region between the E6 and E7 open reading frames. Thus, 5.3% and 4.1% of the nucleotides in E6 and E7, respectively, were subject to variation. The maximum nucleotide pairwise difference was approximately 2% for both E6 and E7.

Table 2.

HPV33 variants based on the sequence of the E6 and E7 regions of 213 HPV33-positive cervical samples.

	E6																								E7												No. of Isolates	Variant ID	GenBank accession
AA position	3	3	6	9	18	35	36	40	46	55	56	69	73	83	86	93	104	113	122	124	138	139	142	147	12	12	19	45	50	55	62	63	65	87	93	97
Prototype AA	Q	Q	E	P	A	K	P	S	A	R	E	F	I	V	N	K	I	Q	N	R	A	C	S	T	V	V	P	A	A	V	N	T	V	V	T	Q
Variant AA	E	Q	E	P	V	N	T	F	V	R	K	L/F	L	L	H	N	M	R	S	R	V	C	S	T	I	V	T	E/V	A	V	N	A	V	V	S	L/P
Nucl. Position	115	117	126	135	161	213	214	227	245	273	274	315	325	355	364	387	420	446	473	480	521	525*	534	549	606	608	627	706	722	737	758	759	767	833	850	862
Prototype Nucl.	C	A	G	A	C	A	C	C	C	A	G	C	A	G	A	A	T	A	A	A	C	T	C	T	G	T	C	C	T	A	C	A	T	G	C	A
A1	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	88	1	KC881011
	-	-	-	-	-	-	A	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	2	2	KC862070
	-	-	-	-	-	-	-	T	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	1	3	KF536962
	-	-	-	-	-	-	-	-	T	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	1	4	KC862071
	-	-	-	-	-	-	-	-	-	-	A	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	1	5	KC862072
	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	G	-	-	-	-	-	-	-	-	-	-	1	6	KC862073
	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	T	-	-	-	-	-	-	-	-	3	7	KC862074
	-	-	-	-	-	C	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	13	8	KC881012
	-	-	-	-	-	C	-	-	-	-	-	-	-	-	-	C	-	-	-	-	-	-	-	C	-	-	-	A	-	-	-	-	-	-	-	-	1	9	KC881013
	-	-	-	-	-	C	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	G	-	-	-	-	-	T	1	10	KC881014
A2	-	-	-	-	-	C	-	-	-	-	-	-	-	-	C	-	-	-	-	T	-	-	-	-	-	-	-	-	-	G	-	-	-	-	-	T	50	11	KC881015
	G	-	-	-	-	C	-	-	-	-	-	-	-	-	C	-	-	-	-	T	-	-	-	-	-	-	-	-	-	G	-	-	-	-	-	T	1	12	KF536964
	-	G	-	-	-	C	-	-	-	-	-	-	-	-	C	-	-	-	-	T	-	-	-	-	-	-	-	-	-	G	T	-	-	-	-	T	1	13	KF536963
	-	-	-	-	-	C	-	-	-	-	-	G	-	-	C	-	-	-	-	T	-	-	-	-	-	-	-	-	-	G	-	-	-	-	-	T	1	14	KC862075
	-	-	-	-	-	C	-	-	-	-	-	-	-	-	C	-	-	-	-	T	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	T	7	15	KC881016
	-	-	-	-	-	C	-	-	-	-	-	-	-	-	C	-	-	-	-	T	-	-	-	-	-	-	-	-	-	-	-	-	C	-	-	T	2	16	KC862076
	-	-	-	-	-	C	-	-	-	-	-	-	-	-	C	-	-	-	-	T	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	3	17	KC862077
A3	-	-	-	-	-	C	-	-	-	G	-	-	-	-	C	C	-	G	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	T	2	18	KC862078
B	-	-	-	-	-	-	A	-	-	-	-	T	C	T	-	C	-	-	-	-	T	-	T	-	-	-	-	-	C	-	-	-	-	T	G	T	16	19	KC881019
	-	-	-	-	-	-	A	-	-	-	-	T	C	T	-	C	-	-	G	-	T	-	T	-	-	-	-	-	C	-	-	-	-	T	G	T	4	20	KF536965
	-	-	-	-	T	-	A	-	-	-	-	T	C	T	-	C	-	-	-	-	T	-	T	-	-	-	-	-	C	-	-	-	-	T	G	T	1	21	KF536968
	-	-	-	-	-	-	A	-	-	-	-	T	C	T	-	C	-	-	-	-	T	-	T	-	A	-	-	-	C	-	-	-	-	T	G	T	1	22	KC881018
	-	-	-	-	-	-	-	-	-	-	-	T	C	T	-	C	-	-	-	-	T	-	T	-	A	-	-	-	C	-	-	-	-	T	G	T	4	23	KC881017
	-	-	A	-	-	-	-	-	-	-	-	T	C	T	-	C	-	-	-	-	T	-	T	-	A	-	-	-	C	-	-	-	-	T	G	T	1	24	KC862079
	-	-	-	-	-	-	A	-	-	-	-	T	C	T	-	C	-	-	-	-	T	-	T	-	-	-	-	-	-	-	-	-	-	T	G	T	2	25	KC881020
	-	-	-	-	-	-	A	-	-	-	-	T	C	T	-	C	-	-	-	-	T	-	T	-	-	-	A	-	-	-	-	-	-	T	G	T	1	26	KF536967
	-	-	-	-	-	-	-	-	-	-	-	T	C	T	-	C	-	-	-	-	T	-	T	-	A	-	-	-	-	-	-	-	-	T	G	T	2	27	KF536966
C	-	-	-	G	-	-	-	-	-	-	-	-	-	T	-	-	G	-	-	-	T	C	T	-	-	-	-	-	-	-	-	G	-	T	G	C	2	28	KC862080

Phylogenetic analysis

Twenty-six of the 28 unique variants clustered into three groups in the phylogenetic tree (Fig. 1A, Table 2) that corresponded to the previously described (sub)lineages A1, A2, and B (Chen et al., 2011). Nine variants, representing 24 samples (variant IDs 2–10 in Table 2), were of the same A1 sublineage as the prototype variant (NCBI accession number M12732, variant ID 1 in Table 2, n = 88). Seven variants, representing 65 samples (variant IDs 11–17 in Table 2), corresponded to the previously reported A2 sublineage, and 9 variants, representing 32 samples (variant IDs 19–27 in Table 2), corresponded to the previously reported B lineage. Two samples (variant ID 18 in Table 2) were similar in sequence to the variants in the A2 sublineage, but had three distinct sites of variation that set them apart from the A2 sublineage. Two samples (variant ID 28 in Table 2) were similar in sequence to, but also showed considerable differences from, variants in the B lineage (Table 2). Samples from these two unusual variants were whole-genome sequenced. The comparison with other HPV33 genomes (Chen et al., 2011) indicated that variant 18 represents the novel A3 sublineage, and variant 28 represents the novel C lineage.

Fig. 1.

Phylogenetic trees of the HPV33 E6 and E7 variants in the present study only (A) or with additional variants published in the literature (B). The numbers at the end of the branches correspond to variants listed in Table 2 and Table S1. The prototype sequence is variant 1 in the A1 sublineage.

The tree structure remained unchanged when additional unique sequence variants from the literature (Table S1) were added (Fig. 1B). The 3 main clusters corresponding to (sub)lineages A1, A2, and B remained obvious. Additional support for a true A3 sublineage came from a similar variant reported by Wu et al. (2009) (EU918766.1) for which the whole genome sequence was found to differ by 0.8% and 0.6% from publicly available whole genome sequences of representatives of the A1 and A2 sublineages, respectively.

Variants clustering in each of the (sub)lineages A1, A2, A3, B, and C showed a specific “core” pattern of SNPs. There were 13 and 3 nucleotide positions in E6 and E7, respectively, that distinguished at least one (sub)lineage from another (dark gray background for nucleotide positions in Table 2). Further, 9 SNPs were “diagnostic” (i.e., consistently present and unique), for one (sub) lineage (light gray background for nucleotide bases in Table 2), but no such diagnostic SNP existed in E6 or E7 for the A1 sublineage. A minimum combination of 3 nucleotide genotypes in E6 (e.g., 355, 364, and 387) is required for the correct classification of a isolate into one of five (sub)lineages, based on the current data.

There was an unequal distribution of HPV33 (sub)lineages across the regions (Fig. 2). Although the A1 and A2 sublineages were found across all regions, their relative frequency varied. The B lineage was specific to Africa. The two isolates in our study that are of the A3 sublineage are both from Asia/Oceania (China and Vanuatu), and the two isolates of the C lineage are both from Africa (Mali and Nigeria).

Fig. 2.

Geographical distribution of HPV33 (sub)lineages shown as a proportion of the total number of HPV33-positive samples collected from each region.

Case-control analysis

The distribution of HPV33 variant (sub)lineages was compared between invasive cervical cancer cases (n = 81) and controls (n = 117) (Table 3). For this analysis, the samples that were diagnosed as HSIL, CIN2, or CIN3 were removed (n = 15). The distribution of variant lineages differed significantly between cases and controls in both Africa and Europe. These differences were driven by an overrepresentation of the A1 sublineage in the cases from both regions. The relative risks for the A1 sublineage compared to all non-A1 lineages were 9.4 (95% CI = 1.9–51.4) and 43.2 (95% CI = 4.0–1945) in Africa and Europe, respectively. No significant difference in the risk for the A1 sublineage was observed in Asian/Oceania (OR = 0.6, 95% CI = 0.1, 2.8). The small number of controls from South America precluded a similar analysis.

Table 3.

Comparison of HPV33 (sub)lineages between cases and controls

Lineage	Cases						Controls						Fisher's Exact Test p-value	OR (95% CI) A1 vs. non-A1
	A1	A2	A3	B	C	Total	A1	A2	A3	B	C	Total
Africa	11	0	0	6	1	18	4	0	0	23	1	28	0.001	9.4 (1.9, 51.4)
Asia and Oceania	4	7	0	0	0	11	34	36	2	0	0	72	0.646	0.6 (0.1, 2.8)
Europe	18	5	0	0	0	23	1	12	0	0	0	13	<0.001	43.2 (4.0, 1945)
South America	26	3	0	0	0	29	4	0	0	0	0	4	1.000	0.0 (0.0, 14.2)
Total	59	15	0	6	1	81	43	48	2	23	1	117

Removing the cases for which multiple HPV types were detected did not substantially change the point estimates, although the confidence intervals did increase due to the loss of 7 cases in Africa (4 were from the A1 sublineage, 2 from the B lineage, and 1 from the C lineage) and 6 cases in Europe (5 from the A1 sublineage and 1 from the A2 sublineage; results not shown). However, despite the reduction in the number of samples, the results remained significant with a p-value of 0.004 for Africa and <0.001 for Europe.

When the samples classified as CIN2/3 or HSIL were included with the cases, the associations became stronger and more precise for Africa as 6 cases were added to the A1 sublineage, 1 case was added to the A2 sublineage, and 3 cases were added to the B lineage (results not shown). There were no samples classified as HSIL or CIN2/3 from Europe. The distribution of cases, controls, and HSIL/CIN2/3 within each region, country, and variant (sub) lineage are shown in Table S2.

Discussion

The IARC has one of the largest and most diverse collections of HPV-genotyped cervical specimens in the world and, therefore, is uniquely situated to study genetic variations found within the genome of high-risk HPV types and their geographic and ethnic distribution as well as their association with cervical cancer risk. Based on 213 HPV33-positive sequenced samples from 30 countries in Africa, Asia, Oceania, Europe, and South America, we were able to identify all previously described HPV33 variant (sub) lineages (i.e., A1, A2 and B) as well as provide additional levels of phylogenetic stratification with the identification of an A3 sublineage and a C lineage. This phylogenetic tree structure was robust even when the analysis was expanded to include additional sequences from the published literature (Chen et al., 2011; Khouadri et al., 2006; Garbuglia et al., 2007; Wu et al., 2009). Our analysis was particularly informative with respect to describing the variants of HPV33 in Africa and Asia/Oceania because samples from these areas were better represented here than in previous studies. Indeed, the amount of genetic variation in E6 and E7 captured among our samples (5.3% and 4.1% of nucleotides, respectively) was higher than in any previous study of HPV33 variants (Chen et al., 2011; Stewart et al., 1996; Godínez et al., 2013; Khouadri et al., 2006; Bokal et al., 2010; Gagnon et al., 2004; Garbuglia et al., 2007; Ntova et al., 2012; Raiol et al., 2009; Wu et al., 2009; Xin et al., 2001).

The distribution of major HPV33 variant lineages around the world was confirmed to be highly geographically/ethnically specific. The A1 sublineage was observed to be distributed throughout the world, although the relative frequency varied by region. The A2 sublineage was rarely detected in Africa and South-America while the B lineage was specific for Africa. The C lineage was only found in samples from Africa, and the A3 sublineage was specific for Asia/Oceania, as supported by two additional reports from China (Wu et al., 2009) and Japan [based on E6 only, (Xin et al., 2001)]. However, the observed rarity of the A3 and C (sub) lineages and the limited number of samples from the Americas must be considered when interpreting these results.

While the geographic heterogeneity of the variant sublineages hampers the possibility to compare their carcinogenic potential across all regions, by using a multicenter case-control comparison stratified by region we were able to identify significant associations between HPV33 variant sublineages and cervical cancer risk. Our data suggested an overrepresentation of the A1 sublineage variants in cervical cancer cases in Europe and Africa. Although we did not observe such an effect in Asia/Oceania, an increased risk of CIN3 and invasive cervical cancer has been reported for A1 lineages in a study from Japan (Xin et al., 2001).

We did not observe any changes in amino acids at critical positions related to known biological functions. In E6, there were no changes in the amino acid sequences that are critical for zinc binding, TP53 binding (Y54), E6AP binding (I128), nor PDZ binding (RETAL at positions 145–149). In E7, there were no mutations in the prototypical LXCXE RB1 binding site at positions 22–26, in the two CKII phosphorylation sites at serines 31 and 32, nor in the zinc-binding motif. The lack of mutations in biologically relevant positions mirrors that which was seen in the analysis of HPV16 E6 variants (Cornet et al., 2012, personal communication).

However, this does not preclude the possibility of significant biological effects caused by changes in amino acids that have not yet been mapped to a specific biological or oncogenic function. In fact, the HPV33 E6 variant containing the combination of I73L, V83L, K93N, A138V, and silent mutations at F69 and S142, has been found to be better at degrading P53 and MAGI-3 than the prototype HPV33 E6 (Ainsworth et al., 2008). Similarly, there is the possibility of linked mutations in other regions of the HPV33 genome that have a role in the oncogenic potential of the virus. Therefore, differences in HPV33-associated cervical cancer rates by population might be explained by genetic heterogeneity within the HPV33 variant sublineages around the world. Alternatively or additionally, host genetic factors, which differ by population, could have a role in the association between a particular HPV33 variant and cervical cancer development, as has been suggested for HPV16 (Zehbe et al., 2001).

To facilitate epidemiological studies comparing the clinical behavior of HPV33 genetic variants and their eventual pooling for sample size requirements, we highlight a practical classification of all HPV33 variant sublineages based upon a selection of 16 discriminatory nucleotide positions in E6 and E7, that can each distinguish at least one (sub)lineage described above from another. Nevertheless, there is redundancy in this classification, so that a minimum of 3 nucleotide genotypes in E6 (e.g., positions 355, 364, and 387) is required for correct lineage classification. Although there is a small number of diagnostic SNPs that are specific for a given sublineage, none exist for the most common A1 sublineage.

Of note, the use of these SNPs for HPV33 sublineage classification fits with all other HPV33 sequences published in the literature (Table S1) and is consistent with previous lineage definitions based on complete HPV33 genome sequencing (Chen et al., 2011). Thus, while whole genome sequencing remains the gold standard for defining HPV phylogeny, this simple classification based on E6 and E7 may be useful in situations where sample quality, quantity, or lack of access to appropriate technical platforms is an obstacle to achieving a complete sequence.

The major limitation of the current study was the limited sample sizes particularly for HPV33-positive cases in Africa and Asia/Oceania and HPV33-positive controls in South America. The very high ORs that we report should, therefore, be interpreted with caution because of the very broad CIs and the possibility that there is heterogeneity between the countries grouped together by region given that the distribution of cases and controls was not balanced by country (Table S2). Nevertheless, the reliance on 20 years of IARC studies on HPV from around the world meant that the number of HPV33-positive cases and controls in the current study was by far the largest studied to date.

In conclusion, we provide an updated classification of HPV33 phylogeny for use in epidemiological studies and suggest that HPV33 variant (sub)lineages may be associated with different risks of cervical cancer. Although our results need to be confirmed in a large series of cases and controls drawn from the same population, the current study provides information on the genetic basis of differences in the carcinogenicity of HPV33 variant lineages and may help us unravel important biological and/or immunological interactions between the virus and the host that could lead to better tools to control HPV infection and its malignant consequences.