Search tips
Search criteria 


Logo of plosonePLoS OneView this ArticleSubmit to PLoSGet E-mail AlertsContact UsPublic Library of Science (PLoS)
PLoS One. 2010; 5(6): e11170.
Published online 2010 June 23. doi:  10.1371/journal.pone.0011170
PMCID: PMC2890415

Social Networks Shape the Transmission Dynamics of Hepatitis C Virus

Bradley S. Schneider, Editor


Hepatitis C virus (HCV) infects 170 million people worldwide, and is a major public health problem in Brazil, where over 1% of the population may be infected and where multiple viral genotypes co-circulate. Chronically infected individuals are both the source of transmission to others and are at risk for HCV-related diseases, such as liver cancer and cirrhosis. Before the adoption of anti-HCV control measures in blood banks, this virus was mainly transmitted via blood transfusion. Today, needle sharing among injecting drug users is the most common form of HCV transmission. Of particular importance is that HCV prevalence is growing in non-risk groups. Since there is no vaccine against HCV, it is important to determine the factors that control viral transmission in order to develop more efficient control measures. However, despite the health costs associated with HCV, the factors that determine the spread of virus at the epidemiological scale are often poorly understood. Here, we sequenced partial NS5b gene sequences sampled from blood samples collected from 591 patients in São Paulo state, Brazil. We show that different viral genotypes entered São Paulo at different times, grew at different rates, and are associated with different age groups and risk behaviors. In particular, subtype 1b is older and grew more slowly than subtypes 1a and 3a, and is associated with multiple age classes. In contrast, subtypes 1a and 3b are associated with younger people infected more recently, possibly with higher rates of sexual transmission. The transmission dynamics of HCV in São Paulo therefore vary by subtype and are determined by a combination of age, risk exposure and underlying social network. We conclude that social factors may play a key role in determining the rate and pattern of HCV spread, and should influence future intervention policies.


Hepatitis C virus (HCV) is a major cause of hepatitis and liver cancer globally, chronically infecting nearly 3% of the world's population [1]. During the second half of the 20th Century, the global HCV pandemic worsened due to widespread viral transmission by blood or blood-derived products, and by unsafe medical practices. Today, needle sharing among injecting drug users (IDUs) is a major risk factor in industrialized countries [2] and prevalence among IDUs ranges from 30 to 90% [3], [4], with between 5.6 to 36% of Brazilian drug users infected [5]. Approximately 15–40% of infected persons clear the virus during the acute phase. Of the remainder, nearly 80% result in chronic infections which may lead to subsequent viral transmission [6]. It has been estimated that 0.8% to 3.5% of the Brazilian population is HCV-positive, while prevalence among those that are also HIV infected may reach 25% [7].

HCV is classified into six major genotypes and many subtypes, varying in geographical distribution, transmission route and treatment response [1], [8], [9]. Genotype 1, and especially subtypes 1a and 1b, are the most prevalent worldwide, although 1b is mostly found among older members of the population who have a history of blood transfusion [3], [10], [11]. HCV-1a and 3a appear to have emerged recently and are spreading rapidly in many global regions, with a high prevalence among young IDUs [3], [10], [11], [12]. Coarse-grained phylogenetic studies suggest that HCV subtypes 1a and 1b reached a plateau globally in the 1980's after previously experiencing exponential growth [13]. Despite the increasing importance of HCV for human health, no vaccine is currently available, treatment is suboptimal in many countries and, alarmingly, up to 40% of the transmission events are of unknown cause [14]. Moreover, there is a severe lack of donor organs, causing tens of thousands of patients with decompensate liver disease and portal hypertension to await orthotopic liver transplantation [15].

The lack of an effective vaccine highlights that prevention is currently the main strategy available to control viral transmission. As a consequence, detailed studies of the molecular epidemiology of HCV, particularly at the scale of local epidemics, are needed to enhance our understanding of the transmission dynamics of this major public health problem. Herein, we investigated the population dynamics of HCV in the most populous region of Brazil – São Paulo state – with the aim of identifying the epidemiological factors that influence viral transmission.

Materials and Methods

Experimental design

Partial NS5b gene sequences were obtained from blood samples collected from 591 patients chosen irrespective of their risk associations between 1997 and 2006 as part of the Viral Genetic Diversity Network (VGDN) program [16]. Patients were sampled from four different regions within São Paulo state; Ribeirão Preto, São José do Rio Preto, São Bernardo do Campo and from two reference centers for HCV surveillance and treatment in São Paulo city (Clínicas Hospital and Reference and Training Center- RTC). After RNA extraction from blood samples and cDNA synthesis, a 361 nt region of NS5b was amplified. Questionnaire data from patients such as age, gender, sex, and marital state were collected at the time of sampling. Relevant clinical information on transmission risk factors was gathered, including: (i) self-declared drug use profile: injected (IDU) or non-injected drugs (crack and cocaine) (NIDU), (ii) sexual behavior, including the number of casual and regular sexual partners during their entire life and, (iii) blood transfusion (BT) before and after 1993 when HCV blood screening started in Brazil, were also obtained. The NS5b sequences were typed by phylogenetic inference with the inclusion of worldwide HCV reference sequences available in GenBank ( and Los Alamos HCV database ( All sequences generated here have been deposited in GenBank and assigned accession numbers GQ490493–GQ491027.

Ethics Statements

All procedures adopted in this work were done according to the terms agreed by the Ethics Committee on Human Research of the University of São Paulo and informed consent terms were signed by all patients.

Inferring phylogenetic history

NS5b sequences were aligned using Clustal X [17]. Phylogenetic trees were inferred using the GARLi program (Genetic Algorithm for Rapid Likelihood Inference) [18], which employs an extensive branch-swapping protocol and optimizes the substitution model iteratively during the search. Initial trees were used to determine the position of our sequences in the context of the global HCV pandemic. In particular, these analyses allowed us to identify and eliminate sequences that were directly related to non-Brazilian sequences suggesting that they had been imported from localities outside of Brazil.

HCV phylodynamics

To investigate the population history of HCV in São Paulo state (after excluding potential migrants) we utilized the Bayesian Markov Chain Monte Carlo (MCMC) approach implemented in the BEAST package [19]. First, the rate of nucleotide substitution as well as the Time to the Most Recent Common Ancestor (TMRCA) of each viral subtype was estimated under the best-fit model of nucleotide substitution obtained by comparing the GTR+Γ+I, GTR+Γ+I+SRD112, HKY and HKY+SRD112 models [20] using Bayes factors (BF) (although similar results were obtained under all substitution models). We also utilized both strict and relaxed (uncorrelated lognormal) molecular clocks. To be as conservative as possible we also employed the least constrained Bayesian skyline coalescent prior, which provided similar substitution rates to those obtained with the best demographic model for each subtype. To improve the MCMC search, we set the operator values as a function of the number of taxa in each group, and set the upper value of the ‘population size’ parameter to 170 million, which hypothetically covers the estimated number of infected people in the world. After optimizing the values of the MCMC operators during preliminary runs, up to 10 additional MCMC runs, each consisting of 20 million generations (with a 50% burnin), were undertaken to obtain convergence of parameter estimates. With the substitution rates in place we then estimated the rate of population growth (r) under the demographic model (i.e. constant, exponential population growth or logistic population growth) that best-fit each subtype data set, with model comparisons again undertaken using BF. Similarly, setting the prior on the substitution rate, we estimated Bayesian skyline plots for each data set. In all cases the convergence of parameters during the MCMC was inspected with Tracer v.1.4 [19], with uncertainties depicted as 95% highest probability (HPD) intervals. To provide an independent assessment of the robustness of our analysis of evolutionary dynamics we also estimated the TMRCA of each data set using the Path-O-Gen program (available at, which employs a linear regression of root-to-tip genetic distances against sampling time based on the ML phylogenies described above.

HCV phylogeography

To evaluate how phylogenies are shaped by underlying geography within São Paulo state we utilized the Bayesian approach available in BEAST [19]. Specifically, we performed a phylogeographic analysis using a standard continuous-time Markov chain (CTMC) model with the Bayesian stochastic search variable selection (BSSVS) procedure [21]. The three main geographic locations of sampling (Metropolitan area of São Paulo, Ribeirão Preto and São José do Rio Preto) were coded as multistate characters and the most parsimonious description of geographic spread was obtained by MCMC sampling from the plausible set of trees using the procedures outlined above.

Fitting a power law to the sexual contact data

We estimated the extent of social interaction among the patients in our data set as their total number of sexual contacts [22]. We fitted a pure power law distribution model of the form P(k) [congruent with] k−α to our data and estimated the minimal value of sexual contacts above which the data follow a power law (xmin), and the maximum likelihood estimator of the scaling exponent α using the goodness-of-fit based method of [23] with the R v.2.10.0 program ( The quality of fit was evaluated by a Kolmogorov-Smirnov goodness-of-fit (K–S) test that compares the D-statistic for the observed data to a critical D value estimated for large n [24] from the expected power law distribution.

Estimating R

To allow a broad-scale comparison of growth dynamics across HCV epidemics we obtained approximate estimates of the basic reproductive number (R) for each HCV genotype using the relation R = 1+rD, where r is the population growth rate and D is the mean time of infectiousness [25]. Growth rate (r) estimates were obtained with BEAST from the best-fit demographic model for each data set. Additionally, we estimated r using GENIE ( under the relation r = ρµ, where µ is the nucleotide substitution rate (subs/site/year) and the composite parameter ρ (r/µ) estimated from the ML phylogeny, with branch lengths adjusted for date of sampling with the TipDate program [26]. Parameter D is an estimator of the generation time distribution (w(t)), which is a function that describes the distribution of times between infections in a chain of transmission [27]. Accurate proposals for w(t) depend on several complex factors that are difficult to evaluate for a persistent virus such as HCV, which causes fluctuating viral load levels during chronic infections that may last several years [28], [29]. Moreover, the probability of virus transmission is expected to vary according by risk factor. Therefore, instead of utilizing w(t), R was calculated using a mean time of infectiousness (D) of 2, 5, 10, and 15 years, which appear to cover a reasonable range of values. Importantly, values of R inferred from phylogenetic trees may not relate linearly to the number of infected people, and may also be strongly affected by the variance on the number of potential infectious contacts events. These limitations notwithstanding, agreements between R estimated from viral phylogenies and from epidemiological data have been noted [30].

Finally, the human population growth rate was approximated using census data in time fitted to both exponential and logistic curves, and by estimating the rate coefficients under the model that best fit the data using the SPSS v.11.0.1 program (SPSS Inc. Chicago, IL).


HCV epidemiology and phylogeography

A total of 591 HCV sequences obtained from São Paulo State were successfully amplified, sequenced and typed by phylogenetic analysis using reference sequences obtained from the Los Alamos HCV database. Genotype 1 was the most prevalent (70.5%), comprising 51.3% of subtype 1b and 48.7% of HCV-1a samples. Subtype 3a was the next most prevalent, representing 24.2% of all samples. Genotype 2 included subtypes 2b and 2c and was found at frequencies of 2.1% and 1.9% of the population, respectively. Genotypes 4 and 5, not commonly found in Brazil, were detected in five patients, one infected with the subtype 4a and four with 5a. Subtypes only found in a small number of patients (2, 4 and 5) were discarded for the subsequent analysis. Phylogenies of the most common 1a, 1b and 3a subtypes also suggested different demographic histories (Figure S1). Specifically, the HCV-1b phylogeny possessed three distinct clusters comprising Brazilian sequences, such that the same local epidemic is associated with different 1b lineages. In contrast, there was no evidence for phylogenetically distinct clusters in trees of subtypes 1a and 3a.

To reduce the effect of any phylogeographic structure on our subsequent analysis of transmission dynamics we first excluded all taxa that we suspected to be “migrant” due to their close phylogenetic relationship to samples obtained from outside Brazil (and taken from GenBank). In addition, we performed a Bayesian analysis of phylogeographic structure, dividing our time-stamped sequences into three main geographic sources: (i) the Metropolitan area of São Paulo city, including patients from the Clinicas Hospital, the Reference and Training Center (RTC), and São Bernardo dos Campos (located in the outskirts of São Paulo), (ii) Ribeirão Preto, located in the west of São Paulo state some 317 km from the São Paulo metropolis, and, (iii) São José do Rio Preto in northern São Paulo state, 443 km from the São Paulo city and 205 km from Ribeirão Preto. For each HCV genotype, the root location with the highest posterior probability falls in the metropolitan area of São Paulo city and is shown as black branches in the maximum clade credibility (MCC) trees (Figure 1). Although we found small phylogenetically distinct clusters of sequences spreading in Ribeirão Preto (sectors highlighted in green) and São José do Rio Preto (highlighted in blue) in all three trees, overall there was clearly no segregation by region of sampling. In addition, all region-specific clusters radiated at approximately the same time as those from São Paulo (grey concentric circles in Figure 1) suggesting that they should not impact significantly on overall dynamics.

Figure 1
Phylogeographic and IDU risk analysis.

Time-scale of HCV emergence and dynamics in São Paulo

Key demographic and epidemiological parameters estimated for all HCV subtypes, or collected from patients' questionnaires, are shown in Table 1. Importantly, our estimates for the TMRCA of each subtype obtained under the best-fit demographic model covered essentially the same range as those estimated by the Bayesian skyline analysis, suggesting that these results are robust to demographic history (see supplementary Table S1A to S1F for details of the statistical analysis). Although HCV subtypes 1a, 1b and 3a showed similar rates of nucleotide substitution, there were significant differences among subtypes in both TMRCA and population growth rate (Table 1). To infer how the demographic history of each subtype changed through time we inferred Bayesian skyline plots using a relaxed molecular clock and the substitution rates estimated previously under the best-fitted model of nucleotide substitution (that is, GTR+Γ+I+SRD112 for HCV-1a, and 1b and HKY+SRD112 for HCV-3a). Although the 95% HPD values for the skyline plots under the relaxed clock were too wide for meaningful interpretation, those generated using a strict clock gave a reasonable demographic signal (Figure 2 and S2).

Figure 2
Patterns of population growth for HCV1a, 1b and 3a.
Table 1
Risk factor, demographic and epidemiologic parameters of HCV.

Together, these evolutionary analyses suggest that HCV-1b was the first subtype to enter São Paulo and which experienced rapid population growth in the 1970's, possibly spread by blood transfusion and unsafe medical practices (the frequency of each risk factor is given in Table 1). HCV-1b was followed by HCV-3a in the mid 1960s (Table 1). HCV-1a was most likely the last subtype to emerge and started growing exponentially in 1990's, after the end of paid blood donation and the onset of obligate blood-donor screening (Figure 2), attaining the highest growth rate of all (Table 1). Subtypes HCV-1a and 3a are known to dominate IDU-associated infections in most industrialized countries. However, we found only 27% of HCV-1a and 20% of HCV-3a infected individuals in our study to be at risk of exposure by IDU, and a similar proportion, albeit with lower numbers, of NIDU (Table 1). This last result is of relevance since people that use non-injecting drugs (NIDU) are also at risk of HCV infection, mainly due to crack and cocaine [31]. Indeed, approximately half of our NIDU patients that reported sharing cocaine paraphernalia also reported nasal bleeding (76 patients). Interestingly, most of our NIDU used cocaine exclusively, with a small percentage (6%) also reporting smoking crack, indicating that crack may not be particularly important for HCV transmission among our study population. Finally, because HCV prevalence in IDUs can reach 90% [32], it is possible that IDUs act as reservoirs that disseminate the virus through the population as a whole.

We therefore explored the impact of injecting drug use on the HCV transmission network. To do this we reconstructed the ancestral mode of transmission across our subtype phylogenies using “IDU”, “non-IDU”, and “unknown status” as character states. This revealed a non-IDU root in all cases, suggesting that HCV transmission did not originate exclusively from IDU (Figure 1). However, we did observe six small IDU clusters in the HCV-1a phylogeny and one cluster in the HCV-1b tree (highlighted red sectors in Figure 1). Moreover, since IDU occurrence was higher in HCV-1a than HCV-3a and HCV-1b, we also tested if IDU could explain the differences in growth rates we observed. By excluding the IDUs from HCV-1a data set, we observed that the exponential growth rate only increased from 0.51 to 0.56 and with overlapping 95% HPDs. We also observed that 28 to 38% of our patients had unknown risk factors for HCV transmission, which suggested that additional host-associated factors should be considered as determinants of transmission dynamics.

Dynamics associate with age structure of HCV patients

An important observation from our analysis of HCV population dynamics was that viruses that emerged more recently – subtypes 1a and 3a –also experienced higher rates of population growth. Strikingly, the year of birth of people infected by HCV-1a and 3a was uni-modal, with modes of around 1954 and 1964 for HCV-3a and HCV-1a, respectively (Figure 3). In marked contrast, the year of birth structure of HCV-1b carriers was clearly multi-modal, with three peaks roughly spaced according to decade-sized intervals. This tri-modal distribution may be explained by the fact that subtype HCV-1b has likely been circulating in São Paulo since the beginning of the 20th century, and because carriers survive for long time periods following infection. Therefore, people infected several decades ago were sampled during our study.

Figure 3
Viral and host age structure.

Host population size and social networks impact on HCV transmission

The population dynamics of infectious agents are greatly influenced by a variety of epidemiological factors, including host population size and density, the spatial distribution of the host, and the rate of contact between individuals [33], [34]. As an initial exploration of how these factors may impact HCV transmission we investigated census population growth concurrent with the spread of HCV in São Paulo. This analysis revealed a positive relationship between human population increase and the growth rates of each HCV subtype according to their time of origin (Figure S3).

Since the rate of contact between infected and susceptible hosts (i.e., the nodes of a transmission network) are crucial determinants of epidemic dynamics [35], [36], the number of connections per highly-connected node should increase as a function of host population density and size. Moreover, it reflects the intensification of social interaction as measured by the number of sexual contacts [22]. We therefore examined the dependency of the cumulative frequency distribution P(k) on the total number of sexual partners (k) obtained from questionnaires available for 492 of the 591 patients in our study group. Strikingly, we observed a negative relationship of P(k) for increasingly higher values of the number of reported sexual contacts (k), showing stereotypically sequential decay on the number of sexual contacts (Figure 4A). We then fitted the data to a pure power law distribution and estimated the value of xmin = 10 and the maximum likelihood value of the scaling exponent α = 2.15. Although other heavy-tailed distributions may explain our data [23], it conformed well to a power law as shown by the significance of the K–S test statistic D = 0.0254 (p<0.001, n = 217, D-critical[congruent with]0.0913), expounding the scale-free nature of this HCV network. In general, values of k for most of the cumulative frequency values of P(k) indicated a higher number of sexual contacts and connectivity for HCV-1a, followed by HCV-3a and finally HCV-1b, and most strongly so at the heavy-tail side of the curves (Figure 4A).

Figure 4
Scale-free sexual contact network of HCV carriers.

Since our patients did not identify sexual partners among others in the cohort, we were not able to infer the exact topology of the HCV network. Nevertheless, we examined the social, risk-related and clinical factors at the two ends of the overall distribution (i.e., for all HCV carriers taken together to obtain robust parameter estimates). By dividing all patients into two extreme log-binned groups with “low-connectivity”, defined as 5 or less sexual partners (40% HCV-1b, 37% 1a and 23% 3a), and “high-connectivity”, reflecting 50 or more sexually partners (56% HCV-1a, 30% 3a and 14% 1b), we found that greater sexual activity was also associated with increased exposure to known HCV risk factors (STD, drug use, tattooing and imprisonment) and specific behaviors, such as unprotected sex (Figure 4B).

HIV-1 and HBV co-infection

HCV, HIV-1, and hepatitis B virus (HBV) share similar routes of transmission and HCV infects 25 to 30% of HIV-infected persons [37]. Although our sampling protocol was designed to minimize the inclusion of persistent viral co-infections, we in fact sampled 36 HIV-1 positive out of 878 HCV carriers. This 4.1% prevalence was six times higher than the HIV-1 prevalence of 0.7% in the general Brazilian population []. From the seventeen co-infected samples sequenced here, there were nine patients infected by HCV-1a, four by 1b and four by 3a. From these, 2.7% where at the low-connectivity end of the HCV network and 6.1% were located at the high-connectivity end, which also had several other evident risk factors associated other than HIV-1 (Figure 4B). Therefore, some small proportion of the increase in growth rate we see in 1a and 3a compared to 1b could have been due to an increased rate of HCV transmission due to HIV co-infection (Figure 4B). Similarly, we found two HCV patients from a total of 494 to be HBV positive (0.4% prevalence, which was 10 times lower than prevalence of 3.7% in metropolitan areas of Brazil [38]. Both these individuals had medium connectivity (from 6 to 49 sexual partners) in the HCV network.


Our study reveals that co-existing yet distinct epidemics of different HCV subtypes circulate in São Paulo and among discrete age-structured groups that are differently exposed to risk factors for viral transmission. Subtypes HCV-1a and 3a also grew faster than 1b during the last 20 years, and not reaching the plateau observed in other localities [13]. Since our results are based on the analysis of a large data set of randomly sampled sequences from a one single regional epidemic, and backed by data on risk factors and exposure behaviors, we believe that they are potentially a more accurate reflection of HCV epidemic dynamics. We found that the metropolitan area of São Paulo city was the epicenter of HCV spread in the State, coinciding with it being also the epicenter of human population growth during the 20th century. Perhaps because HCV-1b reached high prevalence earlier in the 20th century it has been able to establish sustained transmission among generations born from the 1940's to the 1970's. In addition, subtype 1b was mainly transmitted through contaminated blood transfusion, which is not obviously associated with an age-specific risk group, and the range of patients ages is expected to be large in this case. This is reflected in the fact that 47% of HCV-1b carriers born in the 1970's (i.e., members of the last peak) received blood transfusion at an early age. In contrast, HCV-3a and subsequently HCV-1a colonized increasingly larger waves of susceptible hosts. These results agree with previous studies of HCV epidemics using smaller sample sizes showing that HCV-1a has grown faster than 1b in many regions, including Brazil [12], [39], [40], [41], [42]. While the nucleotide substitution rate we obtained for HCV-1a (Table 1) was slightly higher than those estimated for HCV in other localities [12], it is possible that this signifies an increased strength of natural selection in this case, in turn reflecting an elevated epidemic growth rate. Indeed, tree-based estimates of the basic reproductive number (R) for the HCV-1a and HCV-3a epidemics in São Paulo were higher than those observed among IDUs in the United Kingdom [12], which also exhibits a lower rate of census population growth during the last century (see Figure S4 A and C). However, because there was no clear distinction in risk association between subtypes HCV-3a and HCV-1a, and because their growth rates were associated with initial time of spread, it is clear that additional factors are needed to fully explain the differing population dynamics of these subtypes.

We therefore propose that the association between age modes and times of subtype origin (horizontal arrows in Figure 3) is largely explained by social factors that might confine HCV transmission within specific age groups. Both an age group restriction on transmission and the control measures implemented in Brazil in 1980 and 1993 could help explain the slowdown in HCV-1b transmission that occurred after 1985 (Figure 2). In addition, age structure may explain more than just the epidemics in São Paulo since the lower number of young adults (between 20 to 39 years of age) in the UK compared with São Paulo is compatible with their difference in viral growth rates (Figure S4 B). It is likely that age-structured waves of HCV carriers transmit the virus among age-related susceptible hosts, perhaps due to behavioral generation exclusion (i.e. a “generation gap”). This accords with the observation that people generally only have significant contact with others like themselves (“the homophylic principle”), which governs the emergence of assortative mixing in social networks [43]. When assortative mixing occurs among highly connected individuals (which can be thought of as “hubs”), such that nodes with many connections tend to be connected to other nodes with many connections (i.e. preferential attachment), then scale-free networks emerge such as the one we document here [44], [45]. Although blood transfusion and IDU enhance HCV transmission, it has been also shown that “survival” sex work among young people was the strongest predictor of elevated HCV incidence in Canada [46]. Importantly, this justifies our sexual-contact network analysis and also agrees with our results on age stratification (Figure 3) and risk factors among highly connected patients (Figure 4).

Importantly, although we do not currently understand to which extent HCV is sexually transmitted, the association of high connectivity and risk behavior in the HCV network, agrees with genotype composition and age structuring, since 29% of the low-connectivity patients were older than the high-connectivity group. This also fits with the observed trend of higher rates of viral population growth among younger age groups. The exception to this was the obviously increased risk of blood transfusion among the low-connectivity group (with a particularly high prevalence in HCV-1b carriers), which also had a more even sex ratio compared to the male-dominated high-connectivity end of the spectrum. Moreover, both extreme groups had marked differences in other factors, such as STD, IDU partners, tattooing, MSM, imprisonment, IDU, NIDU, not using condoms, etc., which could impact on the rates of contact for HCV transmission (Figure 4B). These data, together with the age-stratification data shown in Figure 3, are strongly suggestive of assortative mixing.

In sum, the patterns obtained from clinical and sociologic data suggest that each HCV subtype analyzed here is characterized by different social networks. We therefore propose that increasing population growth, with its knock-on effects on increased social network connectivity and assortative mixing in urban settings, has played a major role in increasing the transmission rate of HCV in São Paulo and perhaps more widely.

Supporting Information

Table S1

Bayes factor values obtained for substitution models comparison.

(0.07 MB DOC)

Figure S1

Maximum likelihood phylogenetic trees of HCV subtypes. Phylogenies were inferred using 335 sequences from HCV-1a, 499 sequences from HCV-1b and 252 HCV-3a sequences (i.e. a combination of those viruses sampled here and those collected from GenBank). Sequences sampled in this work are shown in phylogenies by colored branches, which indicate the four distinct localities in the State of São Paulo: Ribeirão Preto (red), São Bernardo do Campo (green), São Paulo (blue) and São José do Rio Preto (yellow). There was an evident lack of geographical structure in the data. Black branches correspond to global reference sequences collated from GenBank. As Brazilian sequences that fell within clusters of non-Brazilian sequences likely signify recent migration they were excluded from the subsequent phylodynamic analyses.

(1.03 MB TIF)

Figure S2

Bayesian skyline plots of the main subtypes found in São Paulo: HCV-1a, HCV-3a and HCV-1b. Skyline plots describe the mean change (bold line) in genetic diversity under both the relaxed uncorrelated molecular clock (A, C, E) and strict molecular clock (B, D and F) with values of the 95% of high posterior density (HPD) shown in each case (light line).

(0.88 MB TIF)

Figure S3

Relationship between population size in São Paulo and viral dynamics. The graph shows the exponential population growth in São Paulo State in the period 1920 to 2005 (left y-axis) in millions of people and the increase of viral growth rate (estimated using BEAST) during the same time period (right y-axis). For comparison, the TMRCA for each subtype was obtained also with Path-O-Gen, using the best ML phylogeny inferred with GARLi. A summary of the change in time of HCV growth rates and population size in São Paulo is shown in the graph inserted in the upper left of the figure. Dots indicate the growth rate values estimated for each subtype (shown with same colors as in Figures 2 and and3)3) versus the log of the human population size at the time of the most recent common ancestor (TMRCA) of the subtype.

(0.52 MB TIF)

Figure S4

Higher human population growth correlates with higher rates of HCV growth. (A) Census population growth in São Paulo (red dots) and the UK (blue dots) during the 20th Century. The best-fit curve for both populations indicate linear growth in the UK and exponential growth in the State of São Paulo. (B) By partitioning the population of Brazil and the UK into age groups, it appears that most of the growth in Brazil is taking place among the age group that we found to be at a higher risk of infection by HCV-1a and 3a (20–39 years-old). (C) The population growth rate of HCV- 1a and HCV-3a was consistently higher in Brazil than in the UK, suggesting that human population growth accelerates HCV transmission.

(0.62 MB TIF)


We thank Andrew Rambaut for comments and suggestions on both the text and the analytical methodology used, and Cosma Shalizi for providing R scripts for heavy-tailed distribution analysis.

Viral Genetic Diversity Network (VGDN) consortium members. Adriana Tateno, Alberto José da Silva Duarte, Ana Luiza Martinelli, Cláudio Sérgio Pannuti, Edimo Garcia de Lima, Edison Luiz Durigon, Eduardo Massad, Fernanda de Mello Malta, Fernando Lopes Gonçales Júnior, Flair José Carrilho, Gabriela Peçanha Lopes, Jansen de Araújo, José Eduardo Levi, José Fernando Garcia, José Luiz Caldas Wolff, João Manuel Grisi Candeias, Leonardo José Richtzenhain, Maria Inês de Moura Campos Pardini, Mariliza Henrique da Silva, Maurício Nogueira, Michele Mesquita Gomes-Gouvêa, Mirthes Ueda, Mônica Yamaguchi, Paula Rahal, Sang Won Han, Suzete Notaroberto, Vanda Dolabela de Magalhães, Viviane Botosso, Juliana Aliano Bloch, Rosamar E.F. Rezende.


Competing Interests: The authors have declared that no competing interests exist.

Funding: This project was made possible by the Viral Genetic Diversity (VGDN) program funded by the Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) ( project number 00/04205-6 and by the Conselho Nacional de Pesquisa (CNPq) (, Brazil. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.


1. WHO. Global surveillance and control of hepatitis C. Report of a WHO Consultation organized in collaboration with the Viral Hepatitis Prevention Board, Antwerp, Belgium. Journal of Viral Hepatitis. 1999;6:35–47. [PubMed]
2. Trépo C, Pradat P. Hepatitis C virus infection in Western Europe. Journal of Hepatology. 1999;31(Suppl 1):80–83. [PubMed]
3. Pawlotsky JM, Tsakiris L, Roudot-Thoraval F, Pellet C, Stuyver L, et al. Relationship between hepatitis C virus genotypes and sources of infection in patients with chronic hepatitis C. . J INFECT DIS. 1995;171:1607–1610. [PubMed]
4. Matheï C, Van Dooren S, Lemey P, Van Damme P, Buntinx F, et al. The epidemic history of hepatitis C among injecting drug users in Flanders, Belgium. Journal of Viral Hepatitis. 2008;15:399–408. [PubMed]
5. Lopes CL, Teles SA, Espirito-Santo MP, Lampe E, Rodrigues FP, et al. Prevalence, risk factors and genotypes of hepatitis C virus infection among drug users, Central-Western Brazil. Rev Saude Publica. 2009;43(Suppl 1):43–50. [PubMed]
6. Villano SA, Vlahov D, Nelson KE, Cohn S, Thomas DL. Persistence of viremia and the importance of long-term follow-up after acute hepatitis C infection. Hepatology. 1999;29:908–914. [PubMed]
7. Reiche EMV, Bonametti AM, Morimoto HK, Morimoto AA, Wiechemann SL, et al. Epidemiological, immunological and virological characteristics, and disease progression of HIV-1/HCV-co-infected patients from a southern Brazilian population. INT J MOL MED. 2008;21:387–395. [PubMed]
8. Simmonds P, Mellor J, Sakuldamrongpanich T, Nuchaprayoon C, Tanprasert S, et al. Evolutionary analysis of variants of hepatitis C virus found in South-East Asia: comparison with classifications based upon sequence similarity. J GEN VIROL. 1996;77 (Pt 12):3013–3024. [PubMed]
9. Simmonds P. Genetic diversity and evolution of hepatitis C virus—15 years on. J GEN VIROL. 2004;85:3173–3188. [PubMed]
10. Dubois F, Desenclos JC, Mariotte N, Goudeau A. Hepatitis C in a French population-based survey, 1994: seroprevalence, frequency of viremia, genotype distribution, and risk factors. The Collaborative Study Group. Hepatology. 1997;25:1490–1496. [PubMed]
11. Silini E, Bono F, Cividini A, Cerino A, Maccabruni A, et al. Molecular epidemiology of hepatitis C virus infection among intravenous drug users. Journal of Hepatology. 1995;22:691–695. [PubMed]
12. Pybus OG, Cochrane A, Holmes EC, Simmonds P. The hepatitis C virus epidemic among injecting drug users. Infec Gene Evol. 2005;5:131–139. [PubMed]
13. Magiorkinis G, Magiorkinis E, Paraskevis D, Ho SY, Shapiro B, et al. The global spread of hepatitis C virus 1a and 1b: a phylodynamic and phylogeographic analysis. PLoS Med. 2009;6:e1000198. [PMC free article] [PubMed]
14. Karmochkine M, Carrat F, Dos Santos O, Cacoub P, Raguin G. A case-control study of risk factors for hepatitis C infection in patients with unexplained routes of infection. Journal of Viral Hepatitis. 2006;13:775–782. [PubMed]
15. Brown KA. Liver transplantation. Curr Opin Gastroenterol. 2005;21:331–336. [PubMed]
16. Pardini MIdMC, Jamal LF, Durigon EL, Massad E, Perez JF, et al. Boosting virology in Brazil. PLoS Biol. 2008;6:e57–e57. [PMC free article] [PubMed]
17. Thompson JD, Gibson TJ, Plewniak F, Jeanmougin F, Higgins DG. The CLUSTAL_X windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Research. 1997;25:4876–4882. [PMC free article] [PubMed]
18. Zwickl DJ. Austin: The University of Texas at Austin; 2006. Genetic algorithm approaches for the phylogenetic analysis of large biological sequence datasets under the maximum likelihood criterion.
19. Drummond AJ, Rambaut A. BEAST: Bayesian evolutionary analysis by sampling trees. BMC Evol Biol. 2007;7:214–214. [PMC free article] [PubMed]
20. Shapiro B, Rambaut A, Drummond AJ. Choosing appropriate substitution models for the phylogenetic analysis of protein-coding sequences. Mol Biol Evol. 2006;23:7–9. [PubMed]
21. Lemey P, Rambaut A, Drummond AJ, Suchard MA. Bayesian phylogeography finds its roots. PLoS Comput Biol. 2009;5:e1000520–e1000520. [PMC free article] [PubMed]
22. Liljeros F, Edling CR, Amaral LA, Stanley HE, Aberg Y. The web of human sexual contacts. Nature. 2001;411:907–908. [PubMed]
23. Clauset A, Shalizi CR, Newman MEJ. Power-Law Distributions in Empirical Data. Siam Review. 2009;51:661–703.
24. Miller LH. Table of Percentage Points of Kolmogorov Statistics. J Am Stat Assoc. 1956;51:111–121.
25. Pybus OG, Charleston MA, Gupta S, Rambaut A, Holmes EC, et al. The epidemic behavior of the hepatitis C virus. Science. 2001;292:2323–2325. [PubMed]
26. Rambaut A. Estimating the rate of molecular evolution: incorporating non-contemporaneous sequences into maximum likelihood phylogenies. Bioinformatics. 2000;16:395–399. [PubMed]
27. Grassly NC, Fraser C. Mathematical models of infectious disease transmission. Nat Rev Microbiol. 2008;6:477–487. [PubMed]
28. Chevaliez S, Pawlotsky JM. Hepatitis C virus: virology, diagnosis and management of antiviral therapy. World J Gastroenterol. 2007;13:2461–2466. [PubMed]
29. Sherlock S. European livers. Lancet. 1993;342:1127–1128. [PubMed]
30. Mondini A, de Moraes Bronzoni RV, Nunes SH, Chiaravalloti Neto F, Massad E, et al. Spatio-temporal tracking and phylodynamics of an urban dengue 3 outbreak in Sao Paulo, Brazil. PLoS Negl Trop Dis. 2009;3:e448. [PMC free article] [PubMed]
31. Martinez A, Talal AH. Noninjection drug use: an under-appreciated risk factor for hepatitis C virus transmission. Liver Int. 2008;28:757–760. [PubMed]
32. Aceijas C, Rhodes T. Global estimates of prevalence of HCV infection among injecting drug users. Int J Drug Policy. 2007;18:352–358. [PubMed]
33. Anderson RM, May RM. Infectious diseases of humans. Oxford: Oxford University Press; 1992. A framework for the population biology of infectious diseases. pp. 665–661.
34. Olsson GE, White N, Ahlm C, Elgh F, Verlemyr A-C, et al. Demographic factors associated with hantavirus infection in bank voles (Clethrionomys glareolus). Emerg Infect Dis. 2002;8:924–929. [PMC free article] [PubMed]
35. Friedman SR, Aral S. Social networks, risk-potential networks, health, and disease. J Urban Health. 2001;78:411–418. [PMC free article] [PubMed]
36. Kiss IZ, Green DM, Kao RR. Disease contact tracing in random and clustered networks. Proc Biol Sci. 2005;272:1407–1414. [PMC free article] [PubMed]
37. Alter MJ. Epidemiology of viral hepatitis and HIV co-infection. Journal of Hepatology. 2006;44:S6–9-S6-9. [PubMed]
38. Andrade AFB, Oliveira-Silva M, Silva SGC, Motta IJF, Bonvicino CR. Seroprevalence of hepatitis B and C virus markers among blood donors in Rio de Janeiro, Brazil, 1998-2005. Memórias Do Instituto Oswaldo Cruz. 2006;101:673–676. [PubMed]
39. de Oliveira T, Pybus OG, Rambaut A, Salemi M, Cassol S, et al. Molecular epidemiology: HIV-1 and HCV sequences from Libyan outbreak. Nature. 2006;444:836–837. [PubMed]
40. Nakano T, Lu L, He Y, Fu Y, Robertson BH, et al. Population genetic history of hepatitis C virus 1b infection in China. J Gen Virol. 2006;87:73–82. [PubMed]
41. Nakano T, Lu L, Liu P, Pybus OG. Viral gene sequences reveal the variable history of hepatitis C virus infection among countries. J Infect Dis. 2004;190:1098–1108. [PubMed]
42. Pybus OG, Drummond AJ, Nakano T, Robertson BH, Rambaut A. The epidemiology and iatrogenic transmission of hepatitis C virus in Egypt: a Bayesian coalescent approach. Mol Biol Evol. 2003;20:381–387. [PubMed]
43. McPherson M, Smith-Lovin L, Cook JM. Birds of a Feather: Homophily in Social Networks. Annu Rev Sociol. 2001;27:415–444.
44. Barabasi, Albert Emergence of scaling in random networks. Science. 1999;286:509–512. [PubMed]
45. Newman MEJ. The Structure and Function of Complex Networks. Siam Review. 2003;45:167–256.
46. Shannon K, Kerr T, Marshall B, Li K, Zhang R, et al. Survival sex work involvement as a primary risk factor for hepatitis C virus acquisition in drug-using youths in a canadian setting. Archives of Pediatrics and Adolescent Medicine. 2010;164:61–65. [PubMed]

Articles from PLoS ONE are provided here courtesy of Public Library of Science