Variants of human papillomaviruses 16 and 18 and their natural history in human immunodeficiency virus-positive women

Transcription

1 Journal of General Virology (2005), 86, DOI /vir Variants of human papillomaviruses 16 and 18 and their natural history in human immunodeficiency virus-positive women Nicolas F. Schlecht, 1 Robert D. Burk, 1 Joel M. Palefsky, 2 Howard Minkoff, 3 Xiaonan Xue, 1 L. Stewart Massad, 4 Melanie Bacon, 5 Alexandra M. Levine, 6 Kathryn Anastos, 1 Stephen J. Gange, 7 D. Heather Watts, 8 Maria M. Da Costa, 2 Zigui Chen, 1 Ji Yon Bang, 1 Melissa Fazzari, 1 Charles Hall 1 and Howard D. Strickler 1 Correspondence Nicolas F. Schlecht nschlech@aecom.yu.edu 1 Albert Einstein College of Medicine, Department of Epidemiology and Population Health, 1300 Morris Park Avenue, Bronx, NY 10461, USA 2 University of California, San Francisco, San Francisco, CA, USA 3 Maimonides Medical Center, Brooklyn, NY, USA 4 Southern Illinois University School of Medicine, Springfield, IL, USA 5 The Henry M. Jackson Foundation, Rockville, MD, USA 6 University of Southern California, Los Angeles, CA, USA 7 Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA 8 National Institute of Child Health and Human Development, Bethesda, MD, USA Received 24 March 2005 Accepted 16 June 2005 Highly oncogenic human papillomavirus (HPV) 16 and 18 variants might be expected to be particularly aggressive in HIV-positive women. The association of HPV16 and 18 variant lineages with race, human immunodeficiency virus (HIV) coinfection, CD4 + T-cell count, HIV-RNA level, time-to-clearance of HPV infection and presence of squamous intraepithelial lesions (SIL) among women in the Women s Interagency HIV Study was studied. Subjects were followed semi-annually with Pap smear and cervicovaginal lavage (CVL). HPV DNA was detected in CVLs using MY09/11 L1 PCR assay. Specimens positive for HPV16/18 underwent E6 PCR and sequencing to determine the variant present. Specimens from 195 HPV16- and 162 HPV18-positive women were classified into variant lineages based on sequencing results. African variants of HPV16 and HPV18 were significantly more prevalent among African-Americans than among Caucasians [42 versus 14 % (P=0?001) and 60 versus 13 % (P<0?001), respectively]. However, it was not possible to detect associations between the HPV16 or 18 variant lineages and other factors studied. African variants of HPV16/18 were more common in women of African descent living outside Africa, which could reflect mixing behaviours and/or immunogenetic factors. However, in a large population of HIV-infected women, the variant of HPV16 or 18 was unrelated to persistence of infection or presence of SIL. If non-european variants are more oncogenic, the effect may involve a late stage in cervical tumorigenesis. INTRODUCTION Establishment of a persistent infection by oncogenic human papillomavirus (HPV) is thought to be central to the development of most cases of cervical cancer. Over 95 % of cervical tumour specimens harbour HPV DNA. However, HPV types (defined by their genotypes) vary greatly in the strength of their association with cervical cancer. Of the more than 90 HPV types that have been officially classified, more than 40 commonly infect cervical epithelial cells, but only a subset of these cervical HPVs, termed high-risk or oncogenic HPV types, account for almost all cervical cancers. Two HPV types, HPV16 and 18, alone account for approximately 50 and 14 % of cervical cancers, respectively (Bosch et al., 1995; Walboomers et al., 1999). These two HPV types can be further categorized by genotype to identify intra-type variants that may have different oncogenic risks (Londesborough et al., 1996). Several epidemiological studies have found that certain HPV16 and 18 variants may be more strongly associated G 2005 SGM Printed in Great Britain 2709

2 N. F. Schlecht and others with cervical cancer (Burk et al., 2003; Hildesheim et al., 2001; Ho et al., 1993; Londesborough et al., 1996; Nindl et al., 1999; Villa et al., 2000; Xi et al., 1995, 1997; Zehbe et al., 1998, 2001b). By convention, variants are defined as HPVs that differ by <5 % of the nucleotide sequence in non-coding regions and <2 % of nucleotides in coding regions (Bernard et al., 1994). HPVs with genotype differences greater than this are considered to be different HPV types. HPV16 and 18 variants are typically classified into European, African, Asian-American, North-American or Amerindian lineages according to their population of origin (Villa et al., 2000; Yamada et al., 1995). The most consistent observation has been a higher prevalence of non-european variants of HPV16 in cervical cancer and high-grade cervical lesions than would be expected based on their relative prevalence in normal tissues. Taken as a whole, recent studies suggest that there is a two- to threefold greater association of non-european variants with high-grade cervical lesions than is found for European variants of HPV16 (Hildesheim et al., 2001; Xi et al., 2002). There have been far fewer studies of HPV18, but the limited data available suggest that, as for HPV16, non-european variants of HPV18 may be more common than expected in cancer specimens (Burk et al., 2003; Hecht et al., 1995; Lizano et al., 1997; Terry et al., 1997) and high-grade cervical lesions (Villa et al., 2000). Confirmation of these results in large, well-designed studies is greatly needed, since the number of HPV18-positive specimens genotyped in earlier investigations was small, and the algorithms for classifying the non-european variant groups varied (Burk et al., 2003; Villa et al., 2000). Since almost all studies to date have been cross-sectional or case-control investigations, very little is known regarding the natural history of HPV16 and 18 variants to help explain the possible differences in their associations with cervical cancer. These differences could involve events early in the course of the multistage process of cervical tumorigenesis, such as differences in viral persistence and the risk of developing cervical lesions. Alternatively, differences in the risk of cancer between HPV16 and 18 variants could relate to later events, such as risk of transition from high-grade lesions to cancer. Additionally, highly oncogenic HPV16 and 18 variants might be expected to have a particularly aggressive natural history in the presence of diminished host immune status, such as in HIV-positive women (Chaturvedi et al., 2004; Icenogle et al., 1992; Mayrand et al., 2000; Perez- Gallego et al., 2001; Xi et al., 1998). Thus, HIV-positive women represent a unique opportunity to study the effects of host immune status on variant type and behaviour. To date, however, there have been only a few small studies in HIV-positive women, each involving fewer than 40 HPV16- or 18-positive women. These small studies have reported conflicting results regarding the effects of HIV coinfection on the distribution of HPV variants (Chaturvedi et al., 2004; Icenogle et al., 1992; Perez-Gallego et al., 2001), and none have prospectively assessed the effects of variant lineage on the natural history of infection. Therefore, in a large prospective cohort of HIV-positive and -negative women, the Women s Interagency HIV Study (WIHS), we determined the distribution of HPV16 and 18 variant lineages, the associations of variant lineage with race, HIV coinfection and immune status, and whether variant lineages differed in their persistence and/or their associations with squamous intraepithelial lesions (SIL). METHODS Subjects and specimens. Between October 1994 and November 1995, consenting HIV-positive (n=2058) and -negative (n=568) women over 13 years of age were enrolled in the WIHS from clinical and outreach sources in Brooklyn, Bronx, Chicago, Los Angeles, San Francisco and the District of Columbia (Barkan et al., 1998). The WIHS protocol was approved by each local institutional review board (Strickler et al., 2003). WIHS subjects underwent a structured interview as well as a physical and gynaecologic examination every 6 months. Exfoliated cervicovaginal cells for HPV-DNA testing were obtained by cervicovaginal lavage (CVL), followed by a standard Pap smear conducted using a wooden Ayre s spatula and cytological brush. Baseline data showed that HIV-positive subjects in the WIHS had demographic characteristics and risk behaviours similar to those of US women with AIDS nationwide (Barkan et al., 1998). Clinical laboratory data. Pap smears were examined at a single laboratory and categorized using the 1991 Bethesda system for cytological diagnosis. Two independent cytotechnologists examined all Pap smears. Any Pap smear considered abnormal by either cytotechnologist, as well as 10 % of all negative smears, were examined by a cytopathologist. Women were referred for colposcopy for any abnormal cytology result. Treatment decisions were individualized, but the study-wide protocol recommended excision or ablation for women with known or suspected high-grade cervical disease. T-cell subsets were determined by immunofluorescence using flow cytometry in laboratories participating in the AIDS Clinical Trials Quality Assurance Program. Quantification of HIV-1 RNA in plasma was performed using the isothermal nucleic acid sequence-based amplification (NASBA/Nuclisens) method (Organon Teknika) in laboratories participating in the National Institutes of Health (NIH)/ National Institute of Allergy and Infectious Disease (NIAID) Virology Quality Assurance Laboratory proficiency testing program. The lower limit of quantification through September 1997 was 4000 copies ml 21 using a 0?1 ml sample; from October 1997 to December 1998 the lower limit was 400 copies ml 21 using 0?2 ml; after January 1999 the lower limit was 80 copies ml 21 using 1?0 ml. Detection of HPV DNA. Exfoliated cervicovaginal cells were tested for the presence of HPV DNA, following digestion with proteinase K, using a well-established PCR protocol that amplifies a highly conserved 450 bp segment in the L1 viral gene (flanked by primers MY09/MY11/HMB01) (Palefsky et al., 1999). In brief, following proteinase K digestion, 2 10 ml of each cell digest was used in reactions containing 10 mm Tris/HCl, 50 mm KCl, 4 mm MgCl 2, 200 mm of each deoxyribonucleotide triphosphate, 2?5 U Taq DNA polymerase and 0?5 mm of each primer. There were 35 amplification cycles (95 uc for 20 s, 55 uc for 30 s and 72 uc for 30 s, with a 5 min extension period at 72 uc on the last cycle). Primer set PC04/GH20, which amplifies a 268 bp cellular b-globin DNA fragment, was included in each assay as an internal control. The amplification products were probed for the presence of HPV DNA using a generic probe mixture, as well as for HPV DNA on a type-specific basis 2710 Journal of General Virology 86

3 HPV variants and HIV using filters individually hybridized with biotinylated oligonucleotide probes for HPV types 6, 11, 13, 16, 18, 26, 31, 32, 33, 34, 35, 39, 40, 42, 45, 51, 52, 53, 54, 55, 56, 57, 58, 59, 61, 62, 64, 66, 67, 68, 69, 70, 71 (AE8), 72, 73 (PAP238A), 81 (AE7), 83 (PAP291), 82 (W13B/ AE2), 84 (PAP155), 85 (AE5), 89 (AE6), AE9 and AE10. HPV16/18 variant analysis. All specimens found to be positive for either HPV16 or 18 from any of the first four visits were assessed for the variant present. Variant classification was determined by sequencing of the E6 regions of HPV16 and 18. The E6 region for HPV16 was amplified using a previously described protocol and primer set (Wheeler et al., 1997). The primers used in the direct sequencing reactions were specific for HPV16 or 18 and did not amplify other HPV types, thus reducing ambiguity due to mixed infections with other HPV types. The HPV18 E6 region was amplified by single-tube nested PCR using the outer primers AGTAA- CCGAAAACGGTCGGG (forward) and CGGGCTGGTAAATGTT- GATGA (reverse), and inner primers CGGTGTATATAAAAGATG (forward) and TGCTCGTGACATAGAA (reverse). PCR products were then isolated by gel electrophoresis using the Qiagen Gel extraction kit and sequenced on both strands for the entire E6 regions of the HPV16 and 18 genomes. The upstream regulatory region (URR) was sequenced in seven HPV16-positive specimens that could not be sequenced for E6. The URR was similarly amplified by nested PCR and isolated by gel electrophoresis using a separate protocol (Burk et al., 2003). In total, 243 HPV16- and 225 HPV18-positive samples identified by L1 PCR were tested with a success rate for amplification of the E6 region of approximately 75 %, which did not vary by variant group, HPV viral load (as determined by MY09/11 PCR dot-blot intensity) or CD4 + T-cell count. For efficiency, we determined the variant present in only the first HPV16- or 18-positive CVL in each series of sequential positive specimens (e.g. the CVL at visit 3, for a patient who was HPV16- positive at visits 3, 4 and 5). To confirm that this approach did not result in misclassification of the variant present at later visits (e.g. at visits 4 and 5), we determined the variant present in the very last HPV16/18-positive specimen of each series in a random sample of 20 % of subjects (n=82). Only in one instance did we detect a change in variant lineage classification (HPV18 African to European). In an additional subsample of 22 women with HPV16 infection that persisted for three or more visits, we also sequenced the HPV16 present at each and every visit, and found all samples to contain the same variant and sequence at each visit. Based on the consistency of these results, we deemed it appropriate to assume that the same variant was present in sequential visits when HPV16 or 18 infections persisted. E6 variant classification. Variants were grouped according to lineages categorized as previously reported (Ho et al., 1993; Ong et al., 1993; Yamada et al., 1995). HPV16 variant lineages were designated European (E), African-1 (Af-1), African-2 (Af-2), North- American (NA) or Asian-American (AA) (Fig. 1a). HPV18 variant lineages were likewise classified into E, Af-1, Af-2 or Asian/ Amerindian (As/Ai) lineages. By convention, for HPV16, the E variant containing a thymine (T) at nucleotide position 350 was designated the primary reference sequence to which all other HPV16 variants were compared (Yamada et al., 1995), whereas convention designates the As/Ai variant as the HPV18 primary reference sequence (Ong et al., 1993). Our principal comparison, set a priori, was between E variants of HPV16 or 18 and non-e lineages, since, as discussed above, these were the groupings most consistently found to have differences in their associations with cervical cancer. Statistical analyses. Contingency tables were constructed from the data. For cross-sectional analyses, Fisher s exact tests were used to assess the relation between variant group and categorical variables such as race and HIV seropositivity, whereas the Mantel Haenszel score test was used when assessing ordered factors such as CD4 + T-cell count and HIV-RNA level. Additional categorical and ordinal risk factors evaluated included: age, education, smoking history, alcohol consumption, parity, oral contraceptive use, age at first intercourse, number of sexual partners at baseline, number of sexual partners in the 6 months prior to HPV detection, sexual orientation, anti-retroviral therapy, history of sexually transmitted diseases and the HPV16 or 18 viral load determined by the hybridization dot density in MY09/11 PCR. CD4 + T-cell counts and plasma HIV- RNA levels were categorized according to conventional clinical strata shown to be associated with HPV infection in previous studies (Palefsky et al., 1999). The proportion of subjects on highly active anti-retroviral therapy (HAART) was low, as the period of observation initially preceded the introduction of HAART. Multivariable analyses of cross-sectional data were conducted using logistic regression. To study the effects of HPV16 and 18 variant lineage on the persistence of infection we used multivariable Cox proportional hazards regression. Specifically, persistence was measured using time-to-clearance of an HPV16 or 18 infection that was newly detected (studying incidently detected HPV makes it possible to estimate time of onset), with clearance defined using a stringent criterion of at least two sequential negative HPV-PCR results (to minimize concerns that a single falsenegative result might affect the findings). The mid-point calendar date between two consecutive visits was used to estimate the time of event. Because this analysis was restricted to incident infections the data were more limited. Therefore, we used non-oncogenic HPV types as the reference group, with HPV16 and 18 variant lineage included as a variable in each multivariable Cox-model; that is, we calculated the hazard ratio (HR) for time-to-clearance for each variant lineage relative to that of all non-oncogenic HPV types, which includes a broad category of HPV types with many incident and clearance events so as to provide a stable statistical base for estimation of HRs. This approach also made it possible to stratify the baseline hazard by HPV-type viral load (determined in MY09/11 PCR), HIV serostatus and CD4 + T-cell count detected at time of infection, to best control for these important factors in the analysis. Further adjustment for age, ethnic group and other potential confounders (see above) was achieved by including these variables in the multivariable Cox models. The Wald test was used to measure the statistical significance of differences in the HRs between variant groups. Proportional hazards assumptions were checked and no violations were found. Lastly, to study the cross-sectional relationship of HPV16 and 18 variant lineage with SIL, we used multivariable logistic regression controlling for age, HPV-viral load, CD4 + T-cell count, HIV serostatus, ethnic group and whether the HPV infection was prevalent or incident. Restricting observations to incidently detected infections alone did not change the associations. Statistical analyses were performed using Stata-8 and SAS-9. RESULTS One hundred and ninety-five HPV16-positive and 162 HPV18-positive women contributed data to this analysis. Approximately 85 % of these women were HIV-positive, and most were African-American (58 %), while 17 % were Caucasian, 23 % Hispanic, and 3 % Asian, Native-American or of Pacific Island descent (categorized together as Other ). Details regarding CD4 + T-cell count and other immune parameters are discussed below. Among those women genotyped for HPV16 and 18 variants, most (65 %) were heavy smokers of greater than 10 pack-years and 60 % had multiple (five or more) male sexual partners at enrolment. The mean age at enrolment of the study population was

4 N. F. Schlecht and others 2712 Journal of General Virology 86

5 HPV variants and HIV Fig. 1. (a) E6 nucleotide sequence variations among HPV16 isolates and variant classifications. (b) E6 nucleotide sequence variations among HPV18 isolates and variant classifications. E6 primary reference sequences are shown in the first row of Fig. 1(a, b). Only nucleotide positions for which variations from prototype were detected are shown. Nucleotide substitutions are labelled by the corresponding nucleic acid (A, C, G, T). Positions at which no variation was found relative to the prototype nucleotide are marked with dashes. Forward slashes (e.g. G/T), indicate specimens where two nucleic acids were detected for the same position. Shaded columns demarcate the several nucleotide positions through which branch designations could be completely defined. Lineage classification abbreviations are: E, European; As, Asian; Ai, Amerindian; Af, African; AA, Asian- American; NA, North-American. *, Novel E6 sequences detected. Coding-region numbers in bold type indicate nucleotide positions that likely result in non-synonymous changes in the corresponding amino-acid sequence. 35?6±8?1 years. Additional details on the WIHS cohort have been previously published (Strickler et al., 2003). Fig. 1(a) presents the distribution of HPV16 variants observed in our study and their lineage classification. E variants were the most common HPV16 group detected (64 %, n=124), followed by Af (28 %, n=55) and AA/NA (8 %, n=16) variants. Of the E variants detected, most (n=55) were the primary reference sequence (28 %); 31 (16 %) were similar to the primary reference sequence except for a single nucleotide difference at position 350, a G (guanine) instead of a T (thymine), designated E-350G, which can result in a leucine-to-valine amino-acid change; and 19 (10 %) differed from these two major E subgroups by a single nucleotide. An additional 13 (7 %) were considered Asian variants because of a G-to-T substitution at position 178, though still part of the E lineage. Among the common non-e variants detected, 37 (19 %) were categorized as the Af-1 subgroup reference, 8 (4 %) as the Af-2 subgroup reference and 12 (6 %) as the AA subgroup reference; the remainder had small differences (mostly single nucleotide changes) from these three major sequences. Notably, 11 (6 %) HIV-positive subjects appeared to have coinfection with more than one HPV16 variant, as determined by the detection of two different nucleotides at a single position. Instances of HPV16 coinfection were also more likely to occur in subjects with poor immune status (i.e. seven subjects had a CD4 + T-cell count of <200 cells mm 23 and an HIV-RNA level of > copies ml 21 ). Seven additional HPV16-positive specimens that could not be amplified by E6 PCR were instead amplified and sequenced within the URR (not shown) and found to include six E and one AA isolate. We also identified two novel HPV16 variant sequences: an NA variant with C (cytosine) instead of T at position 109, and an As variant with G instead of T at position 183. Interestingly, the latter substitution also likely results in a non-synonomous (non-silent) amino-acid change: isoleucine to arginine. HPV18 was less heterogeneous than HPV16, with a small group of variants accounting for most HPV18-positive specimens. In addition, unlike HPV16, the most common HPV18 variants were Af (48 %, n=77), followed by E (38 %, n=62) and As/Ai (14 %, n=23) (Fig. 1b), and the vast majority of the variants contained the exact subgroup reference sequence for each lineage. The exception was Af-2, which had a common variant (n=14) containing a T instead of a G at position 342. Coinfection with more than one variant was rare in HPV18-positive women, being detected in just 2/162 (1?2 %) of HPV-positive women. Three E samples were also positive for the T18-7 variant (see Fig. 1b) (de Boer et al., 2005), but with additional (silent) substitutions at positions 149 and 377. Interestingly, we found that all of the major HPV16 and 18 variant groups detected in our population, and even their subgroups, could be distinguished based on the nucleotides found at just a small number of key positions. With respect to HPV16, these positions were 131G, 132C/T, 143/5GT, 178G, 350G and 532G (as shown in Fig. 1a). For example, Asian variants, which are a subgroup of the E lineage, could be distinguished from the E primary reference sequence solely by the presence of a G at position 178. Non-E variants, including AA, NA and Af lineages, could be distinguished from the E primary reference sequence by a T at position 145 (or alternatively at 335). Within the Af lineage, the two main subgroups, Af-1 and Af-2, could be distinguished from the E primary reference sequence by a C-to-G substitution at position 143 and/or either G-to-C (Af-1) or G-to-T (Af-2) substitutions at position 132. An analogous algorithm could be derived to distinguish the four major HPV18 variant groups detected in our population, based on departures from the As/Ai primary reference sequence at three positions in the E6 region: 317C, 548G and 549A. All HPV18 E variants carried a C-to-A (adenine) substitution at position 549, whereas Af variants carried an additional A-to-G substitution at position 548. Af variants could be further divided into Af-1 and Af-2 subgroups, based on a T-to-C substitution at position 317. Table 1 shows the distribution of HPV16 and 18 variants by race. Caucasians and Hispanics/Others had similar distributions of HPV16 variants; the majority of Caucasians (81 %) and those classified as Hispanic/Other (81 %) were positive for HPV16 E variants, and had similar distributions of E variant subgroups (not shown) as well as a similar prevalence of Af and AA/NA variants. African-Americans (including Hispanics of African-American origin), in contrast, had a higher prevalence of HPV16 Af variants (42 %) than either Caucasians (14 %) (P=0?001) or those classified as Hispanic/Other (9 %) (P<0?001). Consistent with this, African-Americans also had a significantly higher prevalence of HPV18 Af variants (60 %) than Caucasians (13 %; P<0?001). However, the distribution of HPV18 variants in

6 N. F. Schlecht and others Table 1. Distribution of HPV16 and 18 variant lineages according to race Race HPV16 variants* HPV18 variants* E AfD AA/NA E AfD As/Ai (n=124) (n=55) (n=16) (n=62) (n=77) (n=23) Caucasian 29 (81 %) 5 (14 %) 2 (6 %) 14 (58 %) 3 (13 %) 7 (29 %) African-American 52 (49 %) 45 (42 %) 9 (8 %) 29 (30 %) 59 (60 %) 10 (10 %) Hispanic/Other 43 (81 %) 5 (9 %) 5 (9 %) 19 (48 %) 15 (38 %) 6 (15 %) *Percentage values in rows may not add up to 100 % due to rounding errors. DUsing the distribution of E variants as referent, the prevalence of HPV16 Af variants was higher in African-Americans compared to Caucasians and Hispanic/Other (P<0?05 by Fisher s exact test). women classified as Hispanic/Other differed from Caucasians. The prevalence of HPV18 Af (38 %), As/Ai (15 %) and E (48 %) in the Hispanic/Other group was intermediate between African-Americans and Caucasians. Adjustment for potential confounders, including HIV serostatus, CD4 + T-cell count, number of sexual partners, HPV-viral load and study recruitment site did not alter the strong associations between race and the distribution of HPV16 and 18 variants. The distribution of variants between the different racial groups varied little across the different recruitment sites (Brooklyn, Bronx, Washington DC, Chicago, Los Angeles and San Francisco) or across the other demographic characteristics listed previously. Similar findings (not shown) were observed in analyses that excluded Other ethnicities from the Hispanic/Other group (i.e. a group composed of only Hispanics) and compared HPV16 and 18 variant distributions by race. Table 2 presents data regarding HIV serostatus and other immune parameters, and their associations with the distribution of HPV16 and 18 variants. Although HPV16 Af variants appeared to be more common among HIV-positive subjects and less common in those with a low CD4 + T-cell count or high HIV-RNA level, after adjustment for race, these associations were not statistically significant. HPV18 Af variants also had an apparent positive association with HIV serostatus that was not significant after adjustment for race. Overall, none of the risk factors studied, which additionally included age, study site, lifetime number of sexual partners, education, smoking history, alcohol Table 2. Distribution of HPV16 and 18 variant lineages according to HIV infection and other immune factors Descriptive characteristic HPV16 variants* HPV18 variants* E Af AA/NA E Af As/Ai (n=124) (n=55) (n=16) (n=62) (n=77) (n=23) HIV serostatus Negative 11 (69 %) 3 (19 %) 2 (13 %) 7 (58 %) 3 (25 %) 2 (17 %) Positive 113 (63 %) 52 (29 %) 14 (8 %) 55 (37 %) 74 (49 %) 21 (14 %) CD4 + T-cell countd > (61 %) 10 (32 %) 2 (6 %) 4 (25 %) 8 (50 %) 4 (25 %) (59 %) 27 (34 %) 5 (6 %) 28 (42 %) 34 (51 %) 5 (7 %) < (69 %) 11 (19 %) 7 (12 %) 21 (36 %) 27 (46 %) 11 (19 %) HIV-RNA leveld < (58 %) 11 (33 %) 3 (9 %) 13 (35 %) 22 (59 %) 2 (5 %) (45 %) 13 (45 %) 3 (10 %) 11 (42 %) 12 (46 %) 3 (12 %) (62 %) 15 (30 %) 4 (8 %) 13 (35 %) 18 (49 %) 6 (16 %) > (74 %) 12 (20 %) 4 (7 %) 16 (40 %) 16 (40 %) 8 (20 %) HIV therapyd No 29 (55 %) 21 (40 %) 3 (6 %) 13 (34 %) 20 (53 %) 5 (13 %) Yes 84 (67 %) 31 (25 %) 11 (9 %) 41 (37 %) 53 (48 %) 16 (15 %) *Percentage values in rows may not add up to 100 % due to rounding errors. DHIV-positive subjects only Journal of General Virology 86

7 HPV variants and HIV Table 3. Time-to-clearance of HPV infection according to HPV16 and 18 variant lineages HPV type and variant lineage* HR for clearance of incident infections (95 % CI)D No. cleared incident infections (%) P value for difference in HR between variantsd HPV16 variants E 1?67 (0?98 2?84) 53 (47 %) Referent Af 2?90 (1?46 5?75) 32 (44 %) 0?167 AA/NA Insufficient data 4 (0 %) 2 HPV18 variants E 2?25 (1?23 4?13) 30 (57 %) Referent Af 1?67 (0?83 3?36) 31 (35 %) 0?486 As/Ai 0?57 (0?13 2?44) 11 (27 %) 0?078 Other oncogenic types 1?98 (1?46 2?69) 448 (41 %) 2 (excluding HPV16/18) Non-oncogenic types 1?0 (Referent) 300 (39 %) 2 *HPV types grouped by oncogenic potential: any HPV16 variants, any HPV18 variants (without HPV16), any other oncogenic HPV types (31, 33, 35, 39, 51, 52, 56, 58, 59, 68 and 73); and only non-oncogenic HPVs (6, 11, 13, 26, 32, 34, 40, 42, 45, 53 55, 57, 61, 62, 64, 66, 67, 69, 70 72, 81 85, 89, AE9 and AE10). DHR and 95 % CI adjusted for HPV viral load, HIV serostatus/cd4 + T-cell count, race and treatment for cervical neoplasia. dp values determined using the Wald test. consumption, parity, oral contraceptive use, age at first intercourse, lifetime number of sexual partners at baseline, number of sexual partners in the previous 6 months, sexual orientation or history of sexually transmitted diseases, were associated with the variant detected for either HPV16 or HPV18 (data not shown). To study whether the variant of HPV16 and 18 detected was associated with the persistence of infection, we assessed the relative time-to-clearance for each variant lineage using multivariable Cox regression models, adjusted for HPVviral load, HIV serostatus/cd4 + T-cell count, race and treatment for cervical neoplasia (Table 3). The overall followup time for the study population was similar for HIVpositive and -negative women (mean 51?4±20?2 and 53?4±17?4 months, respectively). Because this analysis was restricted to incident infections, the data were more limited (n=89 HPV16 and n=72 HPV18). For statistical considerations, therefore, as a means of optimally estimating HRs (see Methods), the time-to-clearance of each HPV16 and 18 variant lineage was set relative to all non-oncogenic HPV types combined. Unexpectedly, most HPV16 and 18 variants had a higher clearance rate than non-oncogenic HPV types. However, there were no significant differences in the HR estimates between variant groups. Only HPV18 As/ Ai variants approximated a significant difference in clearance rate relative to E variants (P=0?078). Table 4 shows the associations between HPV16 and 18 variant lineages and the presence of SIL. Given the need to stratify the HPV data (by presence or absence of SIL), we again used all non-oncogenic HPV as a common reference group to stabilize the estimated effects of variant lineage. Although most HPV16 and 18 variant lineages were, as expected, significantly more associated with SIL than nononcogenic HPV, only the Af HPV18 variants had an association with SIL that was significantly different (and weaker) than that of the E lineage (P=0?009). Due to the small number of high-grade SIL, however, we could not evaluate the relative association between variants and lesion severity. DISCUSSION We determined the HPV16 and 18 variants present in a large cohort of HIV-positive and -negative women to investigate: (i) the effects of HIV coinfection and immune status on the distribution of HPV16 and 18 variants; (ii) other factors, such as race, that may also affect the distribution of variants; and (iii) whether the natural history of infection differs between HPV16 and 18 variants in HIV-positive women. This investigation is, to our knowledge, the largest study of HPV16 variants and cervical neoplasia in HIV-positive women to date, and the first in HIV-positive women to examine variants of HPV18. A number of different HPV16 and 18 variants were detected in this population of women at high risk of HPV infection. Most of these variants were categorized as having one of a limited number of common sequences that have been previously identified (Ong et al., 1993; Yamada et al., 1995). There was a low prevalence of HPV16 AA variants compared to E and Af variants (Yamada et al., 1997), but this was not unusual given the small number of Hispanic subjects found in our population (Berumen et al., 2001; Casas et al., 1999;

8 N. F. Schlecht and others Table 4. Prevalence of SIL according to HPV16 and 18 variant lineages HPV type and variant lineage* Odds ratio (OR) for any SIL (95 % CI)D No. subjects (% SIL)d P value for difference in OR between variants HPV-16 variants E 6?27 (3?5 11?3) 75 (51 %) Referent Af 5?56 (2?5 12?2) 36 (44 %) 0?777 AA/NA 20?01 (4?0 101) 11 (82 %) 0?135 HPV-18 variants E 10?91 (4?6 26?0) 31 (65 %) Referent Af 2?59 (1?1 5?9) 36 (31 %) 0?009 As/Ai 18?86 (4?6 78?0) 12 (75 %) 0?498 Other oncogenic types 4?02 (2?7 5?9) 563 (37 %) 2 (excluding HPV16 and 18) Non-oncogenic types 1?0 (Referent) 453 (10 %) 2 *HPV types grouped by oncogenic potential: any HPV16 variants, any HPV18 variants (without HPV16), any other oncogenic HPV types (31, 33, 35, 39, 51, 52, 56, 58, 59, 68 and 73); and only non-oncogenic HPVs (6, 11, 13, 26, 32, 34, 40, 42, 45, 53 55, 57, 61, 62, 64, 66, 67, 69, 70 72, 81 85, 89, AE9 and AE10). DOR and 95 % CI by logistic regression adjusted for age, HPV viral load, HIV serostatus/cd4 + T-cell count, race and whether the HPV infection was prevalent or incident. dbaseline sample defined as the first detected HPV infection (as determined by the detection of HPV DNA by MY09/11 PCR analysis in CVL samples). P values derived by Wald test. Chaturvedi et al., 2004; del Refugio Gonzalez-Losa et al., 2004; Picconi et al., 2003; Villa et al., 2000). More surprising was the detection of a T-to-G substitution at position 183 in 12 HPV16 E variants which has not been reported among E variants except in a small number of cases of in situ or invasive cervical cancer (Burk et al., 2003; Matsumoto et al., 2000; Zehbe et al., 1998). Additionally, we detected two novel HPV16 variant sequences: an NA variant with a T-to- C substitution at 109, and an As variant with a T-to-G substitution at 183, the latter of which also likely results in a non-synonymous amino-acid change. Furthermore, three HPV18 E samples were positive for the T18-7 variant that has been identified previously (de Boer et al., 2005; Ong et al., 1993) but with additional (silent) substitutions at positions 149 and 377. Of special note, we found that a small number of key nucleotide positions could be used to distinguish all of the major variant lineages and perhaps even their subgroups. With respect to HPV16, these positions were 131/2, 143/5, 178, 350 and 532, and for HPV18, the relevant nucleotide positions were 317, 548 and 549. In addition to these, a polymorphic site within the HPV16 E6 gene (at position 276) has also been identified in an E subgroup variant found in Asian and European populations (de Boer et al., 2004; Hu et al., 2001a, b; Kammer et al., 2002; Matsumoto et al., 2000; Zehbe et al., 2001a). If confirmed by other studies, it may be possible to exploit these key nucleotide positions to develop simplified high-throughput methods for identifying the variants present in cervical specimens (Gemignani et al., 2004; Wheeler et al., 1997). The greatest predictor of variant lineage was race. In particular, Af variants of HPV16 and 18 were significantly more common in African-Americans than in Caucasians. While this may initially seem an obvious finding, it is of scientific interest to note that even outside Africa, Af variants of HPV16 and 18 commonly infect women of African descent. The reason for this is not concretely known. Although it is likely to reflect race preference in sexual mixing behaviours (i.e. women of African descent may be more likely to have sexual partners of African descent), immunogenetic (e.g. human leukocyte antigen haplotype) and other factors may also play a role (Giannoudis & Herrington, 2001; Terry et al., 1997; van Duin et al., 2000; Zehbe et al., 2001a). We observed no association between the HPV16 and 18 variant lineage and HIV coinfection or markers of immune status (i.e. CD4 + T-cell count and HIV-RNA level) among HIV-positive women. Prior data from the literature are limited. Among the few studies that included HIV-positive women, two small studies (n<25 in both) reported nonsignificant increases in detection of non-e variants for HPV16 (Icenogle et al., 1992; Perez-Gallego et al., 2001), and one showed a significantly greater prevalence of 350G variants (Chaturvedi et al., 2004). While the authors of the latter study controlled for age, no adjustments were made for immune status (i.e. CD4 + T-cell count and HIV- RNA level) or race, and 35 % (n=33) of the women had SIL, of which two-thirds were high-grade. An earlier study of HPV16 variants in anal specimens from HIV-positive and -negative men and women in San Francisco, which involved 2716 Journal of General Virology 86

9 HPV variants and HIV a subset of women from the WIHS and men seen in the same clinical centre, found no association between HPV16 non-e variants and HIV serostatus (Da Costa et al., 2002). Interestingly, we observed a slightly higher frequency of coinfection with more than one HPV16 variant than in some previous studies (Wheeler et al., 1997), which in our population was largely restricted to HIV-positive women with poor immune status (i.e. CD4 + T-cell count <200 cells mm 23 and HIV- RNA level > copies ml 21 ). If mixed HPV-variant infections are in fact higher among immune-compromised HIV-infected individuals, this observation may potentially support a model for immunological protection. However, such infections represented only a fraction of HPV-positive subjects in our study (less than 6 %), and further investigation into their potential association with immune status and effect on HPV persistence is needed. Taken as a whole, there are currently little data to suggest that the distribution of cervical HPV16 or 18 variants is affected by HIV coinfection or host immune status. There are also scant prior data regarding the association of HPV16 and 18 variant lineage with the natural history of infection in either HIV-positive or -negative women. We detected no significant differences in time-to-clearance among HPV16 or 18 variant lineages. While we also unexpectedly found non-oncogenic HPV types tended to persist longer than either HPV16 or 18 types, the relative time-to-clearance of oncogenic and non-oncogenic HPV has varied between studies (Franco et al., 1999; Giuliano et al., 2002; Liaw et al., 2001; Richardson et al., 2003). Our finding is not unique (Ahdieh et al., 2001; Moscicki et al., 2004) and may relate to subset analysis (HPV16 and 18 alone) (Rousseau et al., 2001). In HIV-positive adolescents, Moscicki et al. (2004) found no differences in persistence of HPV16/18 and non-oncogenic types. The few published studies that have compared the persistence of HPV16 between variant lineages report conflicting findings (Bontkes et al., 1998; Londesborough et al., 1996; Villa et al., 2000; Xi et al., 2002). A common limitation to these typically small studies was that they combined both prevalent and incident infections (Bontkes et al., 1998; Londesborough et al., 1996; Villa et al., 2000), even though prevalent HPV infections are more likely to persist than are incidently detected HPV infections, and the time of onset cannot be estimated with prevalent infections (Ho et al., 1998). Consistent with our results, an earlier study by Xi et al. (2002), which also measured persistence of incident HPV16 infections (n=62), found no differences in persistence between reference-like variants (including E variants) and non-reference-like ( non-e-like ) variants. The current study is to our knowledge the first to assess the persistence of different HPV18 variant lineages, so there are no relevant data for comparison. We observed a higher risk of persistence (i.e. lower clearance rate) of HPV18 As/Ai variants compared to E variants that approximated statistical significance. However, there is no a priori biologic reason we are aware of to have anticipated this result, and the finding should be considered cautiously until confirmed by other independent studies. Reports of a stronger association of cervical cancer with non-e than E variants of HPV16 is the main reason for the current interest in studying HPV16 lineages (Berumen et al., 2001; Burk et al., 2003; Hildesheim et al., 2001; Londesborough et al., 1996; Zehbe et al., 1998). However, an association between HPV16 variant lineage and low-grade lesions has not been established. In fact, a recent prospective study found no association between HPV16 variant lineage and risk of any SIL [relative risk=1?0, 95 % confidence interval (CI) 0?4 2?4], whereas the incidence of cervical intraepithelial neoplasia-2/3 (CIN-2/3) was increased 3?5- fold (95 % CI 1?0 11?8) in the presence of a non-e versus E variant (Xi et al., 2002). Consistent with these data, we found no significant differences in the association of SIL (mainly low-grade SIL) with HPV16 variant lineage. For HPV18, we observed a significantly reduced odds of SIL associated with Af relative to E variants. Unfortunately there are few prior studies of HPV18 variant lineage and cervical neoplasia for purposes of comparison. Villa et al. (2000) reported an increased odds of SIL for non-e HPV18 variants relative to E variants, although they combined these with non-e HPV16 variants in their analyses (i.e. they did not study HPV18 separately), and only one subject harboured an HPV18 Af variant. Burk et al. (2003) found no association between HPV18 variant lineage and squamous cell carcinoma, but the investigators combined Af and E lineages in their analysis. De Boer et al. (2005) did not detect any Af variants in adenocarcinoma samples collected from women from three international populations (Surinamese, Dutch and Indonesian), but observed a higher prevalence of Ai/AA variants in adenocarcinoma compared to squamous cell carcinoma. It would therefore seem that individual HPV18 variant lineages might carry differential risks for cervical cancers. Whether the HPV18 Af lineage is associated with a lower risk of cervical neoplasia, as observed in the current study, will need to be confirmed in independent populations. This investigation has several important limitations that should be noted. First, even though this study represents the largest of its kind in HIV-positive women, we were often limited in our inability to study variants except in a grouped fashion (i.e. by lineage), rather than by specific variant subgroup. Secondly, 75 % of, rather than all, specimens positive for HPV16 or 18 by MY09/MY11 PCR were successfully amplified and sequenced for E6. The E6 amplification rates did not vary by HPV viral load estimated by dot-blot intensity in the MY09/MY11 PCR assay, or by immune status or other parameters, but still we cannot be certain how the excluded samples might have affected our results. Lastly, cytology was used to assess clinical end points. Although both cytology and histology have similar levels of inter-rater agreement, histological confirmation of the presence and grade of neoplasia would have added certainty to the diagnoses

10 N. F. Schlecht and others To summarize, we detected a broad range of HPV16 and 18 variants in our population of HIV-positive and -negative women, and found that a small number of key nucleotide positions could be used to distinguish all of the major variant groups and perhaps even their subgroups. Whether this approach to differentiating HPV16 and 18 variants can be generalized needs to be examined in independent populations. HIV coinfection and immune status (i.e. CD4 + T- cell count and HIV-RNA level) were not associated with the distribution of variants, but we observed a strong association of HPV16 and 18 variant lineage with race. In particular, the data showed that even outside Africa, Af variants of HPV16 and 18 are more common in women of African descent, most likely due to race preference in sexual mixing behaviours and/or immunogenetic factors. We did not, however, find evidence to suggest that non-e HPV16 variant lineage, which is reported to be high-risk for cervical cancer, was associated with either longer time-to-clearance of infection or greater risk of SIL (primarily low-grade SIL). Similarly, no association of HPV18 lineage with time-toclearance of infection, and only an inverse association between HPV18 Af variants and SIL was observed. Thus, if it is correct that non-e HPV16 or 18 variant lineage is a risk factor for cancer, our data, as well as the results of a previous prospective study of HPV16, collectively suggest that the effect is most likely at a later stage in the multistage process of cervical tumorigenesis (e.g. a greater risk of progression following establishment of cervical neoplasia). ACKNOWLEDGEMENTS HPV-DNA testing was funded through R01 CA All specimens and other data in this manuscript were collected by the WIHS, with centres (Principal Investigators) at New York City/Bronx Consortium (Kathryn Anastos); Brooklyn, NY (Howard Minkoff); Washington, DC, Metropolitan Consortium (Mary Young); the Connie Wofsy Study Consortium (Ruth Greenblatt); Los Angeles County/Southern California Consortium (Alexandra Levine); Chicago Consortium (Mardge Cohen); Data Coordinating Centre (Stephen J. Gange). The WIHS is funded by the NIAID, with additional supplemental funding from the National Cancer Institute, the National Institute of Child Health and Human Development (NICHD), the National Institute on Drug Abuse, the National Institute of Dental Research, the Agency for Health Care Policy and Research, and the Centers for Disease Control and Prevention. U01-AI-35004, U01-AI , U01-AI-34994, U01-HD (NICHD), U01-AI-34993, U01-AI-42590, M01-RR and M-01-RR REFERENCES Ahdieh, L., Klein, R. S., Burk, R. & 7 other authors (2001). Prevalence, incidence, and type-specific persistence of human papillomavirus in human immunodeficiency virus (HIV)-positive and HIV-negative women. J Infect Dis 184, Barkan, S. E., Melnick, S. L., Preston-Martin, S. & 7 other authors (1998). The Women s Interagency HIV Study. WIHS Collaborative Study Group. Epidemiology 9, Bernard, H. U., Chan, S. Y., Manos, M. M., Ong, C. K., Villa, L. L., Delius, H., Peyton, C. L., Bauer, H. M. & Wheeler, C. M. (1994). Identification and assessment of known and novel human papillomaviruses by polymerase chain reaction amplification, restriction fragment length polymorphisms, nucleotide sequence, and phylogenetic algorithms. J Infect Dis 170, Berumen, J., Ordonez, R. M., Lazcano, E. & 7 other authors (2001). Asian-American variants of human papillomavirus 16 and risk for cervical cancer: a case-control study. J Natl Cancer Inst 93, Bontkes, H. J., van Duin, M., de Gruijl, T. D. & 11 other authors (1998). HPV 16 infection and progression of cervical intra-epithelial neoplasia: analysis of HLA polymorphism and HPV 16 E6 sequence variants. Int J Cancer 78, Bosch, F. X., Manos, M. M., Munoz, N. & 7 other authors (1995). Prevalence of human papillomavirus in cervical cancer: a worldwide perspective. International biological study on cervical cancer (IBSCC) Study Group. J Natl Cancer Inst 87, Burk, R. D., Terai, M., Gravitt, P. E. & 10 other authors (2003). Distribution of human papillomavirus types 16 and 18 variants in squamous cell carcinomas and adenocarcinomas of the cervix. Cancer Res 63, Casas, L., Galvan, S. C., Ordonez, R. M., Lopez, N., Guido, M. & Berumen, J. (1999). Asian-american variants of human papillomavirus type 16 have extensive mutations in the E2 gene and are highly amplified in cervical carcinomas. Int J Cancer 83, Chaturvedi, A. K., Brinkman, J. A., Gaffga, A. M., Dumestre, J., Clark, R. A., Braly, P. S., Dunlap, K., Kissinger, P. J. & Hagensee, M. E. (2004). Distribution of human papillomavirus type 16 variants in human immunodeficiency virus type 1-positive and -negative women. J Gen Virol 85, Da Costa, M. M., Hogeboom, C. J., Holly, E. A. & Palefsky, J. M. (2002). Increased risk of high-grade anal neoplasia associated with a human papillomavirus type 16 E6 sequence variant. J Infect Dis 185, de Boer, M. A., Peters, L. A., Aziz, M. F., Siregar, B., Cornain, S., Vrede, M. A., Jordanova, E. S., Kolkman-Uljee, S. & Fleuren, G. J. (2004). Human papillomavirus type 16 E6, E7, and L1 variants in cervical cancer in Indonesia, Suriname, and The Netherlands. Gynecol Oncol 94, de Boer, M. A., Peters, L. A., Aziz, M. F., Siregar, B., Cornain, S., Vrede, M. A., Jordanova, E. S. & Fleuren, G. J. (2005). Human papillomavirus type 18 variants: histopathology and E6/E7 polymorphisms in three countries. Int J Cancer 114, del Refugio Gonzalez-Losa, M., Laviada Mier y Teran, M. A., Puerto-Solis, M. & Garcia-Carranca, A. (2004). Molecular variants of HPV type 16 E6 among Mexican women with LSIL and invasive cancer. J Clin Virol 29, Franco, E. L., Villa, L. L., Sobrinho, J. P., Prado, J. M., Rousseau, M. C., Desy, M. & Rohan, T. E. (1999). Epidemiology of acquisition and clearance of cervical human papillomavirus infection in women from a high-risk area for cervical cancer. J Infect Dis 180, Gemignani, F., Landi, S., Chabrier, A., Smet, A., Zehbe, I., Canzian, F. & Tommasino, M. (2004). Generation of a DNA microarray for determination of E6 natural variants of human papillomavirus type 16. J Virol Methods 119, Giannoudis, A. & Herrington, C. S. (2001). Human papillomavirus variants and squamous neoplasia of the cervix. J Pathol 193, Giuliano, A. R., Harris, R., Sedjo, R. L. & 7 other authors (2002). Incidence, prevalence, and clearance of type-specific human papillomavirus infections: The Young Women s Health Study. J Infect Dis 186, Hecht, J. L., Kadish, A. S., Jiang, G. & Burk, R. D. (1995). Genetic characterization of the human papillomavirus (HPV) 18 E2 gene in 2718 Journal of General Virology 86