Human Papillomavirus 18 Genetic Variation and Cervical Cancer Risk Worldwide

ABSTRACT

Human papillomavirus 18 (HPV18) is the second most carcinogenic HPV type, after HPV16, and it accounts for approximately 12% of squamous cell carcinoma (SCC) as well as 37% of adenocarcinoma (ADC) of the cervix worldwide. We aimed to evaluate the worldwide diversity and carcinogenicity of HPV18 genetic variants by sequencing the entire long control region (LCR) and the E6 open reading frame of 711 HPV18-positive cervical samples from 39 countries, taking advantage of the International Agency for Research on Cancer biobank. A total of 209 unique HPV18 sequence variants were identified that formed three phylogenetic lineages (A, B, and C). A and B lineages each divided into four sublineages, including a newly identified candidate B4 sublineage. The distribution of lineages varied by geographical region, with B and C lineages found principally in Africa. HPV18 (sub)lineages were compared between 453 cancer cases and 236 controls, as well as between 81 ADC and 160 matched SCC cases. In region-stratified analyses, there were no significant differences in the distribution of HPV18 variant lineages between cervical cancer cases and controls or between ADC and SCC. In conclusion, our findings do not support the role of HPV18 (sub)lineages for discriminating cancer risk or explaining why HPV18 is more strongly linked with ADC than SCC.

IMPORTANCE This is the largest and most geographically/ethnically diverse study of the genetic variation of HPV18 to date, providing a comprehensive reference for phylogenetic classification of HPV18 sublineages for epidemiological and biological studies.

INTRODUCTION

High-risk human papillomavirus (HPV) types are the etiological agents of cervical cancer. HPV18 was first described in 1984 (1) and is the prototype member of the alpha-7 HPV species. Based upon its enrichment in cervical cancer compared to the level in cytologically normal women (2) and its presence in 16% of cervical cancers worldwide (3), HPV18 is widely accepted as the second most carcinogenic HPV type after HPV16.

HPV18 is known to be present in a higher proportion of cervical adenocarcinomas (ADC) (∼37%) than cervical squamous cell carcinomas (SCC) (∼12%), an attribute that is shared by other members of the alpha-7 species (3). This suggests a phylogenetic trait denoting a tendency to cause ADC, and previous studies have suggested that the association with ADC is driven by particular HPV18 variant lineages (4–6).

Most HPV18 infections are asymptomatic and are cleared by the immune system. Factors that favor a small proportion of HPV18 infections to progress to cervical cancer are poorly understood, but studies have implicated a role for HPV18 genetic variation (7–11).

Based upon common phylogenetic patterns of single-nucleotide polymorphisms (SNPs) in the L1 viral genomic region, HPV18 variants originally were classified as European (E), Asian Amerindian (AA), or African (AFR) (12). This classification has been superseded by a whole viral genome sequencing approach that has defined three major lineages (A, B, and C) and additional sublineages (A1 to A5 and B1 to B3) (13) that can be largely translated to the historical nomenclature (A1 and A2 = AA, A3 to A5 = E, and B/C = AFR) (14). In the whole-genome approach, differences of ∼1.0% define variant lineages, and differences of 0.5 to 0.9% define sublineages.

Using a multicenter case-control study design based upon HPV18-positive samples from the biobank at the International Agency for Research on Cancer (IARC), the aim of the current study was to evaluate the genetic diversity of HPV18 worldwide and to explore the association between HPV18 genetic variants and the risk for cervical cancer in geographically and ethnically diverse female populations.

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 (15–35 and an as-yet-unpublished study from Rwanda). The collection of samples has spanned a period of over 25 years from 1989 until 2014 and 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 HPV type by using a standardized and well-validated protocol (general primer GP5+/6+ PCR-enzyme immunoassay followed by reverse line blot assay) (36) in one centralized laboratory (Molecular Pathology Unit, Department of Pathology, VU University Medical Center, Amsterdam, The Netherlands). All HPV18-positive cervical samples in the IARC biobank were selected for the current analysis, without exclusion, and were categorized into the following regions: northern Africa, sub-Saharan Africa, eastern Asia and Pacific, South/Central Asia, Europe, North America, and South America/Central America.

PCR and DNA sequencing.

DNA extraction from stored samples was performed using a High-Pure PCR template preparation kit (Roche, Mannheim, Germany) or a QIAamp DNA minikit (Qiagen, Hilden, Germany). PCR was performed using a series of M13-tagged HPV18-specific primer pairs that were designed to amplify overlapping regions of the HPV18 long control region (LCR) and E6 open reading frame in order to cover the entire region (see Table S1 in the supplemental material). If a set of three overlapping primers was not able to amplify a sample, then a set of five overlapping primers creating smaller amplicons was used. Each sample underwent two rounds of PCR for each primer pair: 2 μl of DNA in a 10-μl PCR followed by 2.5 μl of PCR product in a 25-μl PCR using the Qiagen multiplex PCR master mix (Qiagen, Hilden, Germany). PCR success was evaluated on a 1.5% agarose gel, and successful reactions were Sanger sequenced in both directions at GATC Biotech (Constance, Germany) using universal M13 sequencing primers, the sequences of which had been incorporated into the PCR products.

The forward and reverse sequence traces of three or five amplicons for each sample were compiled to provide one sequence encompassing the entire HPV18 LCR and E6 open reading frame (nucleotides 7137 to 581) using Geneious v 6.1.8 (Biomatters, Auckland, New Zealand) and compared to the corresponding region of the HPV18 reference sequence (NCBI GenBank accession number X05105 (37), revised as previously described (38) and downloaded from PaVE (pave.niaid.nih.gov) (39). Single nucleotide polymorphisms (SNPs) were confirmed with examination of the sequence chromatograms. All unique sequences were compared pairwise by clustalw2 (http://www.ebi.ac.uk/Tools/msa/clustalw2) using default settings.

Case-control analysis.

Samples were classified as either controls (including normal cells [n = 171], atypical squamous cells of undetermined significance [ASCUS, n = 4], low-grade intraepithelial lesion [LSIL, n = 42], or cervical intraepithelial neoplasia grade 1 [CIN1, n = 6]) or cases (squamous cell carcinoma [n = 339], adenocarcinoma [n = 56], adenosquamous cell carcinoma [n = 25], or unspecified invasive cervical cancer [n = 33]). Samples from population-based HPV prevalence studies for which histology and cytology were unavailable also were classified as controls (n = 13). Samples reported as cervical intraepithelial neoplasia (CIN) grade 2 or 3 or high-grade squamous intraepithelial lesion (HSIL) were excluded from the case-control analysis (n = 19), as were three samples for which sublineage was not able to be determined (two cases and one control). 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 sublineages. Region-specific odds ratios (ORs) and 95% exact confidence intervals (CIs) for the A1 sublineage were calculated against the combination of all other A sublineages.

Case-case analysis.

ADC cases included samples diagnosed as adenocarcinoma (n = 56) or adenosquamous cell carcinoma (n = 25). Two SCC cases were matched by country to each ADC case and, as close as possible, by age. For two of the ADC cases, there was only one SCC case available for matching. The average age was 45.8 and 46.7 years for ADC and SCC cases, respectively. Statistical analysis was performed as described above for the case-control analysis.

Nucleotide sequence accession numbers.

Sequences determined in the course of this work are available from GenBank under accession numbers KP749485 to KP749680.

RESULTS

Sequencing.

The entire LCR and E6 open reading frame were sequenced for a total of 711 HPV18-positive cervical samples from 39 countries, including two countries in northern Africa, nine in sub-Saharan Africa, nine in eastern Asia and the Pacific, five in South/Central Asia, three in Europe, two in North America, and nine in South/Central America (Table 1).

TABLE 1.

Geographic distribution of 711 HPV18-positive cervical samples

Region and country	No. of samples
	Case	Control	HSIL/CIN2/3	Total
Northern Africa	48	5	0	53
Algeria	26	4	0	30
Morocco	22	1	0	23
Sub-Saharan Africa	75	41	7	123
Benin	1	0	0	1
Guinea	2	9	1	12
Kenya	18	0	2	20
Mali	16	1	0	17
Nigeria	1	12	2	15
Rwanda	0	18	2	20
South Africa	18	1	0	19
Tanzania	12	0	0	12
Uganda	7	0	0	7
Eastern Asia and Pacific	184	90	7	281
China	0	13	5	18
Fiji	0	11	0	11
Indonesia	24	0	0	24
South Korea	8	4	0	12
Mongolia	6	24	0	30
Philippines	76	2	0	78
Thailand	69	15	0	84
Vanuatu	1	12	2	15
Vietnam	0	9	0	9
South/Central Asia	31	74	5	110
Bhutan	5	53	3	61
India	24	16	2	42
Iran	1	1	0	2
Nepal	0	4	0	4
Pakistan	1	0	0	1
Europe	24	14	0	38
Georgia	10	9	0	19
Poland	9	5	0	14
Spain	5	0	0	5
North America	9	0	0	9
Canada	8	0	0	8
United States	1	0	0	1
South/Central America	84	13	0	97
Argentina	8	7	0	15
Bolivia	2	0	0	2
Brazil	15	3	0	18
Chile	7	2	0	9
Colombia	3	0	0	3
Cuba	3	0	0	3
Panama	11	0	0	11
Paraguay	26	0	0	26
Peru	9	1	0	10
Total	455	237	19	711

A total of 189 variations were identified among the 1,302 bases in the LCR and E6 region of the HPV18 genome (14.5% variable nucleotide positions), resulting in 209 unique sequences, or variants, of which 196 were not previously described (see Table S2 in supplemental material). In the LCR there were 126 SNPs, 12 deletions (only three of which were present in more than one sample), and one duplication of 82 bp (present in one sample only). In E6 there were 50 SNPs, 23 of which could result in amino acid changes. The maximum pairwise difference of the LCR and E6 open reading frame sequence between any two variants was 2.72%.

Phylogenetic analysis.

The variants clustered into three main groups (Table 2) corresponding to the previously described lineages A, B, and C (13). In the A lineage, 35 variants, representing 248 samples, were of the same A1 sublineage as the HPV18 reference sequence. Four variants, representing six samples, corresponded to the previously reported A2 sublineage. A3 and A4 sublineages accounted for 86 variants representing 303 samples. A3 appeared to be uniquely defined by the SNP C7486T and A4 to be uniquely defined by the SNP A41G (see Table S2 in supplemental material), but this observation is not consistent with the previously described A3 whole-genome sequence GQ180786 (13), which has 7468C and 41G. Hence, A3 and A4 sublineage variants were combined into one sublineage for the course of this analysis. Twenty-three variants, representing 38 samples, corresponded to the previously reported A5 sublineage.

TABLE 2.

Description of HPV18 variant (sub)lineages based on distinguishing positions in the LCR and E6^a

^aOnly SNPs that are consistent for one or more sublineages are included. Nucleotide positions that are diagnostic for a lineage are shaded dark gray. SNPs that are shaded light gray discriminate the sublineage from other sublineages within the same lineage. Boldface indicates an SNP resulting in a change of amino acid: N129K at nucleotide position 491 for the C-to-A SNP and H133Y at nucleotide position 501. Three samples of undetermined sublineage are not shown.

^bThe deletions (Del) are positions 7162 to 7168 and 7245 to 7251.

^cWe did not observe a T at position 7551 for A1 isolates or an A at position 7857 for B1 isolates, but both are reported in the literature (13).

In the B lineage we observed 24 variants, representing 46 samples, that corresponded to the previously reported B1 sublineage; 16 variants, representing 37 samples, corresponded to the previously described B2 sublineage; and eight variants, representing 18 samples, corresponded to the previously described B3 sublineage. In addition, we observed five variants representing six samples that appeared to form a new candidate, sublineage B4 (Table 2). This sublineage was seen in samples from Kenya, Rwanda, Cuba, and Brazil and is approximately 1% different from the other B sublineages in the LCR and E6 region.

We observed five variants representing six samples in the C lineage, with no clear indication of sublineages. Only three variants, each represented by one sample, did not meet the classification described above. Two variants appeared B-like in LCR but A-like in E6 (possibly the amplification of two different variants present in the same sample), and a third variant shared sequence similarity with both the A5 and the A3/A4 sublineages (see Table S2 in supplemental material).

In total, there were 38 nucleotide positions in the LCR and E6 region that could discriminate at least two sublineages from each other (Table 2). Only two of these resulted in an amino acid change (N129K at nucleotide position 491 and H133Y at nucleotide position 501). At the lineage level, nine SNPs were diagnostic, that is, consistently present and unique (dark gray shaded nucleotide positions in Table 2). Each lineage had at least one diagnostic SNP (e.g., C at 7161 for A lineage, G at 7323 for B lineage, and A at 437 for C lineage). Most sublineages had one or more positions that discriminated a sublineage from the other sublineages within the same lineage (light gray positions in Table 2), but only A1 (A at position 7567), A2 (G at position 7511), and candidate B4 (deletion from 7162 to 7168, 500A and 501T) sublineages had diagnostic SNPs that were unique for only that sublineage.

Many other positions were polymorphic within one or more (sub)lineages but did not appear to define phylogenetic subgroups (see Table S2 in supplemental materials). Some potential phylogenetic stratification was observed within A1 (four variants representing 21 samples, all with A89C and A218T), A3/A4 (13 variants representing 72 samples, all with C7164A and C7184T; 21 variants representing 48 samples, all with T7657C), and A5 (seven variants representing nine samples, all with A536C and T553G).

Geographical distribution.

The distribution of HPV18 (sub)lineages varied by geographical region (Fig. 1), with a predominance of the A lineage in most regions except sub-Saharan Africa, where the B lineage predominated. Sub-Saharan Africa showed the broadest range of (sub)lineages. The B lineage also accounted for a small fraction of samples from northern Africa and South/Central America. The C lineage was specific to Africa.

FIG 1.

Distribution of HPV18 (sub)lineages by geographical region, irrespective of cervical diagnosis.

In the A lineage, the A1 sublineage predominated in eastern Asia and the Pacific, whereas the A3/A4 sublineage strongly predominated in Europe but also in northern Africa, South/Central Asia, and South/Central America. The majority of A5 isolates were detected in Africa.

Case-control analysis.

The distribution of HPV18 variant (sub)lineages was compared between invasive cervical cancer cases (n = 453, all histologies) and controls (n = 236) after exclusion of two cases and one control of undetermined sublineage. Within each region, the distribution of the variant lineages (A, B, and C) did not differ significantly between the cases and controls (Table 3).

TABLE 3.

Distribution and statistical comparison of HPV18 lineages between invasive cervical cancer cases and controls

Region	No. of samples by lineage										P valuea	ORb (95% CI), A1 vs A2-A5
	Case					Control
	A1	A2-A5	B	C	Total	A1	A2-A5	B	C	Total
Northern Africa	1	43	2	1	47	2	3	0	0	5	1.000	0.0 (0.0–1.0)
Sub-Saharan Africa	1	15	56	2	74	0	9	29	3	41	0.501	∞ (0.0-∞)
Eastern Asia and Pacific	133	48	3	0	184	45	45	0	0	90	1.000	2.8 (1.6–4.9)
South/Central Asia	4	27	0	0	31	32	41	0	0	73		0.2 (0.0–0.6)
Europe	2	22	0	0	24	1	13	0	0	14		1.2 (0.1–75.0)
North America	4	5	0	0	9	0	0	0	0	0
South/Central America	15	58	11	0	84	2	11	0	0	13	0.351	1.4 (0.3–14.5)
Total	160	218	72	3	453	82	122	29	3	236	NAc	NAc

^aComparing the distribution in the three lineages (A, B, and C) in cases and controls. Fisher's exact test was used.

^bExact confidence intervals.

^cNA, not appropriate due to strong heterogeneity in lineage distribution by region.

At the sublineage level, A1 variants were overrepresented in cases compared to controls in eastern Asia and the Pacific (OR, 2.8; 95% CI, 1.6 to 4.9) (Table 3). However, the opposite was true in South/Central Asia (OR, 0.2; 95% CI, 0.0 to 0.6), driven by a significant overrepresentation of A3/A4 variants in cases versus controls (see Table S3 in supplemental material). No marked differences in the distribution of sublineages were observed by case-control status in other regions (see Table S3).

Results of the case-control analysis were not materially affected upon the reinclusion of the 19 excluded HSIL/CIN2/3 as cases or in sensitivity analyses restricting to cases with single HPV18 infection or SCC histology only.

Case-case analysis.

Upon comparison of 81 ADC and 160 matched SCC cases, the distribution of variant lineages did not differ between ADC and SCC (Table 4), either overall or within any geographical region. Furthermore, no significant differences were observed when variants were classified at the sublineage level or when the 56 adenocarcinomas and 25 adenosquamous cell carcinomas were analyzed separately (data not shown).

TABLE 4.

Distribution and statistical comparison of HPV18 lineages between ADC and SCC: a matched case-case analysis

Region	No. of samples by lineage								Fisher's exact test P value
	Adeno				Squamous
	A	B	C	Subtotal	A	B	C	Subtotal
Northern Africa	4	0	0	4	7	0	1	8	1.000
Sub-Saharan Africa	2	10	0	12	5	18	0	23	1.000
Eastern Asia and Pacific	40	2	0	42	84	0	0	84	0.109
South/Central Asia	6	0	0	6	11	0	0	11
Europe	4	0	0	4	8	0	0	8
North America	2	0	0	2	4	0	0	4
South/Central America	9	2	0	11	19	3	0	22	1.000
Total	67	14	0	81	138	21	1	160	0.629

DISCUSSION

Among 711 HPV18-positive cervical samples, with a high proportion of isolates from Africa and Asia, we observed 209 unique variants in the LCR and E6 region, which is greater than that of any previous report (13). In addition to previously described lineages and sublineages (A1 to A5, B1 to B3, and C), we saw evidence of a candidate B4 sublineage. We also saw potential sublineages within the A lineage, the most common of these being 13 A3-type variants (representing 72 samples), all with C7164A and C7184T, a combination of SNPs that also has been recorded in A3-type variants elsewhere (10, 40, 41).

The LCR was confirmed to contain much more phylogenetic information than E6 and distinguished nine (sub)lineages (A1, A2, A3/A4, A5, B1, B2, B3, candidate B4, and C) without the requirement for E6. However, even LCR and E6 combined did not allow us to clearly distinguish A3 from A4 variants. A3 appeared to be defined by C7486T and A4 by A41G, but this classification did not hold true for the previously described A3 whole-genome sequence GQ180786 (13). For epidemiological studies based on the detection of SNPs in E6 and/or the LCR, we propose a practical classification of variant lineages using 38 nucleotide positions (29 and nine positions in LCR and E6, respectively) that can each distinguish at least two of the nine (sub)lineages described above from each other (Table 2). Nevertheless, there is much redundancy in this classification, so that not all of these positions are required for classification. Indeed, there are a smaller number of diagnostic SNPs that are specific for a given (sub)lineage.

Our LCR/E6 classification fits with other HPV18 sequences reported in the literature (10, 12, 40–50) and is consistent with lineage definitions based on the gold standard approach of HPV18 whole-genome sequencing (13). Although based on LCR and E6 only, our analysis included many novel variants, particularly among previously underrepresented A5 and B (sub)lineages, which would warrant sequencing the whole genomes in order to strengthen the full picture of HPV18 genetic evolution.

As we have shown previously for HPV16 (51), HPV33 (52), and HPV45 (53), the distribution of HPV18 variant (sub)lineages around the world was confirmed to be geographically/ethnically specific (Fig. 1). Consistent with previous multiregional reports (5, 12, 13), we saw a predominance of B lineages (previously named African variants [14]) in sub-Saharan Africa, whereas there was a strong predominance of A3/A4 lineages (previously named European) in Europe. Consistent with the previous nomenclature of A1/A2 lineages as Asian-Amerindian variants (14), the A1 sublineage was found predominantly in eastern Asia and the Pacific and also was frequent in South/Central Asia and the Americas. We additionally showed that C lineage variants were highly specific for Africa, although they were rarely detected in our series and appear to have been present in only a few cases in previous studies (12, 13, 46, 47). Of note, the A5 sublineage, which has only recently been reported in a single isolate from Thailand (13, 54), was found in a higher proportion of samples from northern Africa than elsewhere. The fact that the LCR of A5 sublineage variants share many features with B and C lineage variants suggests that northern African HPV18 isolates represent a previously understudied branch of HPV18 evolution, as observed for HPV16 (51).

Due to the geographic heterogeneity of variant (sub)lineages, we performed case-control comparisons stratified by region. Using this approach, we were not able to observe any difference in cervical cancer risk between HPV18 variants at the lineage level (Table 3).

Other than one small study that detected a nonsignificant difference of lineage for controls versus CIN3+ cases in Spain (45), previous studies have compared HPV18 variants and clinical outcome using the older E/AA/AFR nomenclature. No significant difference in CIN3+ risk was observed between E/AA and AFR variants during 7 years of follow-up of 221 HPV18-positive women from Costa Rica (55), and no differences were observed in the distribution of HPV18 E/AA versus AFR lineages between 47 controls and 51 cancers in the United States (40) or in two smaller comparisons of cancers and controls (4, 46).

A U.S. study reported that HPV18 variants are more likely to be detected (56), persist (56), and progress to CIN3+ (11) in a host whose race indicates an ancestral geographic distribution that was once shared with that of the variant. However, such a race-risk interaction was not apparent in Brazil (57), and we saw no clear evidence that B/C lineages were more likely to cause cancer in Africa or that A lineages were more likely to cause cancer in Europe. However, there was a significant overrepresentation of A1 variants in cancer in eastern Asia and the Pacific. As A1 variants account for the large fraction of the formerly named AA lineage (together with the rarely detected A2), this is consistent with the U.S. study that reported the risk of developing CIN3+ to be significantly higher for AA than AFR variants (11). However, we did not see this pattern in other regions, and A1 variants actually were significantly underrepresented in cases from South/Central Asia. Such regional differences could be considered consistent with the type of race-associated risk differences described above. Alternatively, these differences may be chance findings driven by small numbers at the sublineage level and the fact that there is residual heterogeneity between the countries grouped together by region (see Table S3 in supplemental materials). Whatever the underlying cause, such apparent regional differences reveal an inherent complexity in studies of HPV variants and cervical cancer risk and give a warning about the extent to which data can be pooled across countries/regions.

Previous studies suggested that the distribution of HPV18 variants differ between ADC and SCC (4–6). However, these studies either included small numbers of HPV18-positive cancers, for which interpretations were based largely on extrapolations of findings on HPV16 (4–6), or were based upon pooling of cases from multiple continents (5). In our much larger comparison, including 81 ADCs, each carefully matched by country and age to two SCCs, we observed no differences in HPV18 variant distribution, either overall or in any individual region. This finding is consistent with a U.S. study in which E/AA/AFR variants were equally distributed among 15 ADC and 10 SCC (40) and suggests that the overrepresentation of HPV18 in ADC (3) is not due to the specific glandular tropism of any HPV18 genetic variant lineage.

We did not observe any changes in amino acids in E6 that are predicted to be critical for zinc binding (cysteine residues 32 and 35, 65 and 68, 105 and 108, and 138 and 141) or the suspected LXXLL binding motif involving Y56 and I130 (see Table S1 in supplemental materials). There were no variants with SNPs at the E6* splice sites involving nucleotides 229 to 233 and 416 to 420 (58), but SNPs at the neighboring nucleotide positions 226 and 227 were observed in 5 samples and 1 sample, respectively. Interestingly, there were 31 samples (12 variants) in the A3/A4 sublineages for which the PDZ binding motif RETQV at amino acid residues 154 to 158 was changed to RATQV (SNP A568C). However, the change from glutamic acid (E) to alanine (A) in the second position of the motif is not predicted to be critical (reviewed in reference 59). Lastly, a C104T SNP in the E6 promoter that has been linked with significantly reduced YY1 transcription factor binding and tumor recurrence (9) was commonly present in A3/A4 and always in A5 variants (Table 2).

Our study was limited by the number of HPV18-positive controls rather than invasive cervical cases, given that it mainly included geographical regions with few or no cervical screening programs. Nevertheless, the reliance on 20 years of IARC studies in a uniquely diverse set of populations from around the world meant that the number of HPV18-positive invasive cervical cancers and controls in the current study was by far the largest studied to date.

In conclusion, the present study provides a practical approach for phylogenetic classification for use in epidemiological studies of the natural history and carcinogenicity of HPV18 genetic variants worldwide. Although some possible differences in carcinogenicity between A sublineages were identified in different regions of Asia, our results suggest that HPV18 variants are not useful for discriminating cancer risk (at least not at the lineage level) and cannot explain why HPV18 is more strongly linked with ADC than SCC.