Whole genome sequencing and evolutionary analysis of human papillomavirus type 16 in central China.

Abstract

Human papillomavirus type 16 plays a critical role in the neoplastic transformation of cervical cancers. Molecular variants of HPV16 existing in different ethnic groups have shown substantial phenotypic differences in pathogenicity, immunogenicity and tumorigenicity. In this study, we sequenced the entire HPV16 genome of 76 isolates originated from Anyang, central China. Phylogenetic analysis of these sequences identified two major variants of HPV16 in the Anyang area, namely the European prototype (E(p)) and the European Asian type (E(As)). These two variants show a high degree of divergence between groups, and the E(p) comprised higher genetic diversity than the E(As). Analysis with two measurements of genetic diversity indicated that viral population size was relatively stable in this area in the past. Codon based likelihood models revealed strong statistical support for adaptive evolution acting on the E6 gene. Bayesian analysis identified several important amino acid positions that may be driving adaptive selection in the HPV 16 population, including R10G, D25E, L83V, and E113D in the E6 gene. We hypothesize that the positive selection at these codons might be a contributing factor responsible for the phenotypic differences in carcinogenesis and immunogenicity among cervical cancers in China based on the potential roles of these molecular variants reported in other studies.

Full text

Page 1

Whole Genome Sequencing and Evolutionary Analysis of
Human Papillomavirus Type 16 in Central China
Min Sun1., Lei Gao1., Ying Liu1, Yiqiang Zhao1, Xueqian Wang1, Yaqi Pan1, Tao Ning1, Hong Cai1,
Haijun Yang3, Weiwei Zhai2*, Yang Ke1*
1Key Laboratory of Carcinogenesis and Translational Research (Ministry of Education), Peking University Cancer Hospital & Institute, Beijing, China, 2Center for
Computational Biology and Laboratory of Disease Genomics and Individualized Medicine, Beijing Institute of Genomics, Chinese Academy of Sciences, Beijing, China,
3Anyang Cancer Hospital, Anyang, Henan, China
Abstract
Human papillomavirus type 16 plays a critical role in the neoplastic transformation of cervical cancers. Molecular variants of
HPV16 existing in different ethnic groups have shown substantial phenotypic differences in pathogenicity, immunogenicity
and tumorigenicity. In this study, we sequenced the entire HPV16 genome of 76 isolates originated from Anyang, central
China. Phylogenetic analysis of these sequences identified two major variants of HPV16 in the Anyang area, namely the
European prototype (E(p)) and the European Asian type (E(As)). These two variants show a high degree of divergence
between groups, and the E(p) comprised higher genetic diversity than the E(As). Analysis with two measurements of genetic
diversity indicated that viral population size was relatively stable in this area in the past. Codon based likelihood models
revealed strong statistical support for adaptive evolution acting on the E6 gene. Bayesian analysis identified several
important amino acid positions that may be driving adaptive selection in the HPV 16 population, including R10G, D25E,
L83V, and E113D in the E6 gene. We hypothesize that the positive selection at these codons might be a contributing factor
responsible for the phenotypic differences in carcinogenesis and immunogenicity among cervical cancers in China based on
the potential roles of these molecular variants reported in other studies.
Citation: Sun M, Gao L, Liu Y, Zhao Y, Wang X, et al. (2012) Whole Genome Sequencing and Evolutionary Analysis of Human Papillomavirus Type 16 in Central
China. PLoS ONE 7(5): e36577. doi:10.1371/journal.pone.0036577
Editor: Zhi-Ming Zheng, National Institute of Health-National Cancer Institute, United States of America
Received November 11, 2011; Accepted April 10, 2012; Published May 4, 2012
Copyright: ? 2012 Sun et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted
use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work is supported in part by Natural Science Foundation of China 30872937, 30930102, 31000957 and 91131011, "973" Project of National Ministry
of Science and Technology Grant 2011CB504301, 2012CB316505, "863" Key Projects of National Ministry of Science and Technology Grant 2006AA2Z467, Charity
Project of National Ministry of Health 200902002, and Natural Science Foundation of Beijing 7100001. The funders had no role in study design, data collection and
analysis, decision to publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: keyang@bjmu.edu.cn (YK); weiweizhai@big.ac.cn (WZ)
. These authors contributed equally to this work.
Introduction
Human papillomaviruses (HPVs) are common and are clinically
important pathogens [1]. Infection with high risk types of HPV is a
necessary factor for the development of precancerous lesions and
cervical cancer [2,3,4]. Of those that can infect human beings,
over 120 different types have been isolated, and among these
around 20 types are classified as high-risk HPV types (HR-HPV)
based on their established association with cancer [1,5,6]. Among
these high risk HPV types, HPV16 has been found to be the most
prevalent and shows the strongest association with invasive
cervical cancer [7,8].
It is now generally accepted that HPV has co-existed with its
human host over a very long period of time and has evolved into
multiple evolutionary lineages [9,10]. Intratypic variants of
HPV16 have been identified from different geographic locations
and are classified according to their host ethnic groups as
European (including prototypes and Asian types), Asian American,
African and North American [11,12]. Through epidemiological
and in-vitro experimental studies, natural variants of HPV16 have
shown substantial differences in pathogenicity, immunogenicity
and tumorigenicity. These variants may reflect the evolution of the
viral population as it has adapted to local human ethnic groups
[13]. By studying molecular evolution of the viral genomes,
patterns of this evolutionary history can be identified and
important molecular variants responsible for viral pathogenicity
and carcinogenesis may be characterized [14].
There has been a paucity of HPV 16 population studies in
China. Most previous studies have focused on studying the two
majorviraloncogenes
[15,16,17,18,19,20,21,22,23,24]. The major goal of these studies
has been to explore existing variants in the viral population.
Although cataloging extant mutations is a necessary step in
understanding HPV16 evolution, prioritizing the functional
importance of these identified changes by examining their
evolutionary pattern is potentially much more informative. In this
work, we want to expand upon previous studies by characterizing
the genome wide pattern of genetic diversity, and more
importantly we want to pinpoint major genes/variants that are
driving the adaptation of the virus to the human populations in
central China. These evolutionarily important mutants may be
used for further epidemiological and experimental studies where
the functional consequences associated with these variants may be
investigated and vaccines targeting these sites can be developed.
E6 and E7
PLoS ONE | www.plosone.org1 May 2012 | Volume 7 | Issue 5 | e36577

Page 2

The nonsynonymous to synonymous rate ratio dN/dSin protein
coding regions has provided an important means for studying
molecular evolution of genes, and the use of this method has
gained increasing popularity in recent years [25]. The basic
rational of this method is that synonymous mutations do not
change the underlying protein coding sequences and are not
affected by natural selection. The synonymous substitution rate dS
provides a natural measurement for the rate of evolution under
neutral processes [26]. Since nonsynonymous mutations alter the
underlying protein sequences and can be affected by natural
selection, the relative magnitude of the nonsynonymous substitu-
tion rate dNto the synonymous rate dSprovides a good means for
studying natural selection [27]. Specifically, dN/dS.1 represents
positive selection, dN/dS=1 indicates neutral evolution, and dN/dS
,1 implies there is purifying selection (or negative selection).
Thus, the nonsynonymous to synonymous rate ratio dN/dS
provides a proxy for studying natural selection acting on coding
genes, and many statistical methods have been developed to look
for genes which are under the influence of natural selection,
particularly Darwinian positive selection [28,29].
Recent development of codon based substitution models has
provided a natural extension of previous methods by allowing
different codons to have different dN/dSvalues [30,31]. Statistical
methods such as the likelihood ratio test can be employed to
determine whether patterns of molecular evolution at a certain
gene can be explained with models without invoking positive
selection [32,33]. Upon rejecting the null hypothesis in favor of the
alternative model where positive selection is explicitly allowed,
special codon positions under adaptive evolution can be identified
using a Bayesian based approach [33,34]. These methods have
been widely applied to many datasets, including some multiple
whole genome sequences [35].
In this work, we took a whole genome approach and sequenced
76 HPV16 isolates from Henan Province, China (located in
central China, see Figure S1). We wanted to determine whether
any of the genes in the HPV16 genome is driven by positive
selection. In addition, we sought to identify those codon positions
and associated amino acid changes responsible for the adaptive
evolution in this viral population.
Materials and Methods
Sample collection
HPV viruses often have low concentrations in normal tissues
and are difficult to amplify. In this study, ninety four paraffin-
embedded blocks of cervical cancer samples were collected to
extract the viral genomes from the human population. Of these
ninety four samples, seventy six tested HPV16 positive and were
used for subsequent sequence analysis. These tissue specimens
were collected from women with cervical cancers during their
primary treatment between 2005 and 2007 at Anyang Cancer
Hospital, Henan province, China (Figure S1). All the patients
received no chemotherapy before the surgery. The tumor samples
in this study were a small proportion of the patient samples from
this hospital where surgery (i.e. removing uterus) was chosen as an
effective treatment. Later stage cancer patients will directly go to
radiation therapy without surgery. The clinical stage and
associated age information for these patients were presented in
the supplementary information (Table S1). Official approval from
the Institutional Review Board of Peking University School of
Oncology, and an informed consent was signed by each patient
before sample collection.
DNA preparation
5 mm paraffin sections of formalin-fixed tissue were de-
paraffinized in xylene, and washed with 100%, 95%, and 75%
ethanol. The tissue was pelleted, air dried and digested with
proteinase K (200 mg/ml) at 55uC overnight. 200 ul of this
material was isolated using an H.Q. & Q. Tissue DNA Kit (U-
GENE BIOTECHNOLOGY CO., LTD, Anhui, China). DNA
was re-suspended in a final volume of 100ul 10 mM Tris. The
DNA concentration was determined by use of a Nano-Drop
(NanoDrop Technologies, Wilmington, Delaware USA). A full
description of sample processing and DNA extraction were
presented in great detail in supplementary materials (Text S1).
Quality Control
Our experimental work followed strict quality control to avoid
possible contamination from lab environments. As presented in
great detail in a previous study [36], DNA extraction, PCR
reaction and DNA electrophoresis were done in separate rooms
and specimens moved only in one direction. Laboratory personnel
were instructed to wear gloves when handling the samples and the
experimental area were regularly cleaned before beginning work.
In addition, a routine procedure of inspecting the experimental
area surfaces (cotton bud was first applied to various of surfaces,
e.g. lab benches, subsequently they were soaked into deionized
water overnight. HPV detection was applied to the supernatant.
Experiments were allowed only when negative results were
observed). Additionally, we also used a mouse liver tissue as an
internal control together with the cancer samples. Experiments
were preceded only negative results were observed from these
internal controls (Text S1) and also our previous study [36].
HPV DNA Detection and HPV16 DNA identification
A modified set of primers, SPF1/GP6+, which amplify an L1
fragment of approximately 184 bp were used. The polymerase
chain reaction was carried out as follows. Qiagen Hot Start Taq
DNA polymerase mixture was used with 4 mM MgCl2, and
10 pmol of each primer. The activation of the enzyme was carried
out at 95uC for 15 minutes, followed by 40 amplification cycles at
95uC for 40 seconds, 49uC for 50 seconds, 72uC for 30 seconds,
and a final extension at 72uC for 5 minutes.
The presence of HPV16 DNA in the L1 positive samples was
evaluated by type-specific PCR which amplified a 335bp (nt231 to
565) fragment of HPV16 E6. PCR was performed at 95uC for
15 minutes, followed by 40 amplification cycles at 95uC for
40 seconds, 57uC for 40 seconds, 72uC for 40 seconds, and a final
extension at 72uC for 5 minutes. The experimental conditions and
amplification regions are presented in the supplementary materials
(Text S1 and Table S2, S3 and S6).
PCR and Sanger sequencing
PCR primers were designed to cover the HPV genome.
Platinum Taq DNA polymerase High Fidelity (Invitrogen Co.,
Carlsbad, CA, USA) was used for PCR experiments (Table S5
and S6). PCR products were purified using a PCR clean-up gel
extraction column (MACHEREY-NAGEL GmbH & Co, Du ¨ren,
Germany) according to the manufacturer’s instructions and were
directly sequenced using a capillary sequencer (ABI Prism 3100).
For this study, in addition to the quality control listed above, five
specimens were chosen to repeat the experimental procedures
(including sample processing, Text S1). A different primer sets
were used to amplify the HPV genome (Table S4). The PCR
products were purified and ligated into the pEASY-T1 vector
(Transgen Biotech Co. LTD, Beijing, China) and 3-5 colonies per
Adaptive Evolution of HPV16 in Central China
PLoS ONE | www.plosone.org2 May 2012 | Volume 7 | Issue 5 | e36577

Page 3

ligation were subsequently sequenced with the using the same
Sanger method. The PCR primers and reaction conditions are
presented in the supplementary materials (Text S1, Table S4 and
S6).
Sequence alignment and phylogenetic reconstruction
For each sample, sequence segments across the HPV genome
were concatenated into a single genome. Alignment software
MUSCLE [37] was used to align these genomes against each other
with default parameters. The corresponding annotation informa-
tion was extracted by comparing the sequence alignment to the
HPV 16 reference genome [38,39] (see data availability).
Phylogenetic relationships were built using PhyML with the
General Time Reversible (GTR) model and gamma distributed
rate variation among sites [40]. Local phylogeny for each gene was
also constructed with the same procedures. In order to access the
confidence in the phylogenetic relationship, non-parametric
bootstrap analysis was also carried out using PhyML and
summarized with the SUMTREES package [41].
Population analysis
We carried out a sliding window analysis on the genetic diversity
along the HPV genome using custom written python scripts
(available upon request). The genetic diversity was estimated based
on both the Watterson and the Tajima methods [42,43]. For a
focal window, Watterson’s estimator of genetic diversity uses
information from number of polymorphic sites. Specifically,
Watterson’s estimator of genetic diversity is
S
P
n{1
i~1
1
i
, where S is
the number of polymorphic (or segregating) sites, and n is the
sample size [43]. Likewise, Tajima’s estimator of genetic diversity
is the average pairwise differences between two sequences taken at
random from the sample [42]. Both estimators capture the genetic
diversity within a sample gathered from a population with slightly
different weight on sites at different frequencies. For a standard
equilibrium population, these two estimators should obtain similar
values. Differences between estimated values implies either non-
equilibrium populations (e.g. past population growth or bottleneck)
or occurrence of natural [selection 42].
PAML analysis
CODEML from the PAML package was used to look for the
signal of positive Darwinian selection across the HPV genome. In
particular, we used the M1a/M2a model and the M7/M8 model
to construct the likelihood ratio test for detecting positive selection
[44]. In brief, in the M1a model (null model), there are two
categories of sites with different omega (the nonsynonymous rate
to synonymous rate ratio or dN/dS) values. One category has an
omega value between zero and one, representing the set of codons
evolving under purifying selection. The second category has an
omega value of 1.0, corresponding to those sites under neutral
evolution. In the alternative model (M2a), an extra category of
sites with omega values greater than one is added. If the alternative
model provides significant improvement in the likelihood in
supporting the alternative model (Likelihood Ratio Test or LRT),
the gene under study is said to have sufficient statistical support for
existing of positive selection [32,33].
In the M7 model, omega values are constructed to follow a beta
distribution between zero and one. In the M8 model, one extra
category of omega with value greater than one is added to the
model to allow for positive selection. The likelihood ratio test can
also be constructed with the M7/M8 models to test for positive
selection. In general, nested test between M1a/M2a is more
robust/less powerful than M7/M8 comparisons, even though
most of the time, they give very similar results [44].
In the likelihood ratio test, twice the log likelihood difference
between the two models is compared with the chi-square
distribution wherein the degree of freedom is equal to the
differences in the number of free parameters between the two
models. In both the M1a/M2a and M7/M8 comparisons, two
degree of freedoms should be used (both the omega parameter and
proportion of sites for the extra category). Upon rejecting the null
hypothesis in favor of the alternative model with positive selection,
Bayesian Emprical Bayes (BEB) procedures can be used to identify
the set of sites under positive selection [34].
Results
HPV infection and PCR amplification
With the carefully designed PCR primers, we were able to
detect HPV in 87 (92.5%) of the 94 cancer samples. Of these 87
samples, 80 cases were positive for HPV16. In addition to type 16,
other HPV types were also detectable at low frequencies (Text S1).
Of the 80 cases that were HPV16 positive, we were able to extract
HPV 16 DNA sequences from 76 samples. HPV concentrations in
four other cases appeared to be too low for efficient PCR
reactions.
Following the PCR reactions, Sanger sequencing was conducted
for all the PCR products. In order to check the quality of the
experiments, five specimens were chosen to repeat the lab
procedures with clone sequencing (instead of direct PCR/Sanger
sequencing). Three to five colonies per ligation were cloned and
subsequently sequenced from both ends using traditional Sanger
methods. Based on regions which overlapped by more than one
PCR region and multiple colonies per ligation, consistent results
were found for each sample/region and match with our results
from direct PCR/Sanger sequencing. This suggested that type 16
HPV was the predominant type in these cancer patients and
diversity within hosts was not substantial. Except for three regions
(475 bp and 309 bp in E1 in all 76 samples and 928 bp in E2 in 31
samples), all other parts of the HPV genome were successfully
amplified and sequenced. The undetectable dosages in parts of the
E1/E2 regions could either due to PCR failures (see discussions)
and might also reflect the viral genome integration typically found
in cancer samples [45,46] and the low concentration of the
episomal form of the viral particles in the samples. The technical
aspects of amplification and sequencing are presented in detail in
the supplementary materials (Text S1).
Phylogentic relationship
After retrieving these sequences, the computer software PhyML
was used to reconstruct the phylogenetic relationships among the
76 samples under a General Time Reversible (GTR) model of
nucleotide substitution and gamma distributed rate variation
among sites. The resolved maximum likelihood tree is shown in
Figure 1. This figure illustrates the set of sequences which are
grouped into two major clades including European prototype
(E(p)) and the Asian (E(As)) type, with the strong statistical support
of a bootstrap value of 100% among the 500 replicates. The
observed frequency of the Asian type is 43.4% (33 out of 76) which
is within the range of values observed in previous studies within
China [18,19,21,23,24].
In addition most of the differences within each variant (E(p) or
E(As)) are quite small as compared with the divergence between
variants (Figure 1). This pattern indicates a population history
wherein two viral subpopulations shared some evolutionary history
in the distant past and subsequently diverged from each other [12].
Adaptive Evolution of HPV16 in Central China
PLoS ONE | www.plosone.org3 May 2012 | Volume 7 | Issue 5 | e36577

Page 4

This may reflect earlier divergence in HPV 16 viral populations as
they were adapting to different human ethnic groups.
Genetic diversity
As compared with many RNA virus such as HIV or influenza
virus, the HPV 16 population shows relatively low diversity across
the viral genome (summarized in Table 1 and also Figure S2).
Mean pairwise differences between any two sequences across the
HPV genome are often less than 0.01 (see Figure 2). As compared
with many RNA viruses, this level of diversity is quite small [47].
In addition, the genome-wide variation in genetic diversity
correlates well with previous observations based on datasets
gathered from other geographic locations [12,14]. The highest
genetic diversity was found to be around the E4/E5/NCR regions
and in the URR/E6/E7 regions. After splitting into subgroups,
European Prototype group shows slightly higher genetic diversity
than European Asian types (Figure 2).
Comparing the two measurements of genetic diversity,
Watterson’s estimate based on the number of segregating sites is
close to Tajima’s estimate which is based on pairwise differences.
Similar measurements of these two suggest a genealogical pattern
that is typically found in an equilibrium population. This implies
that the effective viral population size in the recent past has been
quite stable [42,48]. This is consistent with the observation in
human population that despite the world wide human census
population size has increased quite rapidly in recent history, the
effective human population size is relatively stable.
Signals of positive selection
There are two major sources of adaptive forces which act on
viral genes over their life history. One source of driving force is
host immune surveillance. Mutations that lead to escape from
immune recognition by the host immune system (class I and II
molecules) often provide viruses with increased fitness (higher
surviving rate). The second source is viral gene function.
Mutations that result in viruses with increased functional ability
(e.g. increased enzymatic activity or binding affinities on down-
stream targets) are often adaptive. These two forces represent the
‘‘attack and defense’’ aspects of viral life history and embody two
sides of the same coin. In reality, these two sources often overlap
due to the pleiotropic effects of viral genes.
The nonsynonymous to synonymous rate ratio (often denoted as
omega, dN/dS or Ka/Ks) measures the relative ratio of the
nonsynonymous to synonymous evolutionary rate and is a good
Figure 1. Maximum Likelihood phylogenetic tree for 76 HPV16 samples from Anyang. This tree is constructed using whole genome
sequences and bootstrap scores larger than 70% are displayed. The tree itself is shown in bold, the sample IDs are linked through dashed lines. EPs
are the European Prototypes and EAs are the European Asian types.
doi:10.1371/journal.pone.0036577.g001
Adaptive Evolution of HPV16 in Central China
PLoS ONE | www.plosone.org4 May 2012 | Volume 7 | Issue 5 | e36577

Page 5

indicator of selection. In particular, positive selection favoring
nonsynonymous change will lead to an elevated nonsynonymous
substitution rate relative to the synonymous rate. other words,
positive selection will result in dN/dSvalues greater than one. On
the other hand, negative selection (or purifying selection) acting to
preserve certain amino acid positions will give dN/dSvalues of less
than one. If nonsynonymous substitutions are relatively neutral,
the dN/dSvalue will be close to one. The CODEML software
package was used to apply a wide range of statistical models to test
for positive selection. The results for the likelihood ratio test using
two sets of models for evaluation of all HPV genes are listed in
Table 2, and most of these genes showed only a very weak signal
or no signal for positive selection (Table 2). The only gene for
which evidence of positive selection reached statistical significance
is the oncogenic E6 gene. Gene E1 showed marginal significance
at the level of 0.1 (p value=0.09). A few interesting codons were
identified in this analysis in a few other genes, even though these
genes did not reach statistical significance due to the small number
of changes (Table 1). These results will be discussed in detail with
regard to their functional significance in the following section.
Discussion
Functional significance of the genes and codons
Several of the amino acid positions identified in the E6 gene are
of significant interest. During the host immune response, viral
peptides are often presented to the host immune system (e.g.
cytotoxic T-Lymphocyte CTL) through antigen presenting func-
tions mediated by Major Histocompatibility (MHC) molecules.
Previous studies have found that mutations in the amino acid
position L83V may play an important role in cancer progression.
For example, the L83V polymorphism located within the epitope
which binds to MHC molecules was found to be associated with
cervical tumor development [49,50]. This variant can also
Figure 2. Sliding window plot of genetic diversity across the HPV genome. Window size is set as 250 bp and step size=50 bp. The solid
arrows indicate the two regions in the E1 gene that failed in the PCR reactions. The dashed arrow indicates the region in E2 gene that failed in some
of the specimens. Theta_w is the Watternson’s estimate of genetic diversity which is based on the number of polymorphic sites [43], and theta_piis
the Tajima’s estimate of genetic diversity relying on average pairwise differences [42].
doi:10.1371/journal.pone.0036577.g002
Adaptive Evolution of HPV16 in Central China
PLoS ONE | www.plosone.org5May 2012 | Volume 7 | Issue 5 | e36577