Maia A. Rabaa

Home , AMV (TV station)

The Pennsylvania State University

The Graduate School

Eberly College of Science

THE MOLECULAR EVOLUTION AND PHYLOGEOGRAPHY OF DENGUE VIRUSES IN ASIA

A Dissertation in

Biology

Maia Anita Rabaa

Submitted in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

December 2012

The dissertation of Maia A. Rabaa was reviewed and approved* by the following:

Andrew F. Read Professor of Biology and Entomology Chair of Committee

Edward C. Holmes Professor of Biology Dissertation Advisor

Isabella M. Cattadori Assistant Professor of Biology

Timothy Reluga Assistant Professor of Mathematics and Biology

Douglas R. Cavener

Professor and Head of Biology

*Signatures are on file in the Graduate School

ABSTRACT

The global burden of dengue and its impact on the health and economics of tropical countries is vast and continues to grow. Outbreaks of severe dengue disease are increasingly common in regions where dengue was thought to be rare only a decade ago. Our knowledge of the epidemiological and evolutionary processes by which dengue viruses (DENV) invade new populations and become established as persistent viral lineages has been limited by a lack of data relating virus to host and vector populations, as well as the immune landscapes in which they spread over time.

Virus lineages enter new populations regularly, but become established only occasionally and are generally detected long after establishment has occurred. The introduction of novel viral lineages has been shown to perturb the immune landscape in endemic environments, often causing massive epidemics and long-term changes in clinical manifestations within the host population. The use of large sets of spatially- and temporally- related DENV genome and envelope gene (E) sequences to investigate these lineage establishment events allows us to trace the timing and routes of viral movement between host populations, including the evolutionary dynamics of the viral lineages themselves. Determining whether trends in viral introduction and establishment can be distinguished within DENV phylogenies, and what these trends may be, is integral to understanding and controlling the spread of disease at local, regional, and global scales.

In this thesis, I explore the spatio-temporal dynamics of DENV dispersal to identify the routes by which dengue invades new populations, characterize trends in viral migration, and identify the factors that affect the speed and scope of DENV migration at various scales and in heterogeneous transmission environments. In hyperendemic settings, I determined that DENV iii

diversity is maintained in densely populated urban areas, frequently spreading through urban and suburban populations and into rural areas, and following a predictable pattern of human movement that suggests a limited role for the vector in the rapid dispersal of an emerging lineage across the population. Once arriving in rural areas, transmission is remarkably spatially and temporally focal. Where climate conditions are suitable, these populations may show persistence of viral populations over multiple seasons, but the likelihood of lineage fade-out here is increased as a result of lower population densities and seasonal bottlenecks in DENV populations.

Hyperendemic urban areas further seed more distant populations at rates that appear to be dependent upon transmission intensities in the donor population and the connectedness of the human host populations, while the establishment of viral populations may show a greater dependence upon the immune landscape and the climate into which the virus is introduced.

This is also true at the largest spatial and temporal scales, where I identify three geographic regions in tropical Asia that act as sources of global DENV diversity and suggest that immune- mediated competition allows for the DENV populations in these areas to be maintained largely in isolation of one another. These analyses are the first to indicate a primary role for South Asia in the evolution and maintenance of global DENV diversity. Not only does South Asia act as a primary source of DENV, but it also appears to be the most frequent source of novel DENV diversity (i.e., new lineages) for populations outside of Asia. Thus, characterizing the diversity of DENV in this region and the processes by which it is maintained and dispersed into new populations is integral to identifying the factors responsible for their emergence, persistence, and expansion into new landscapes across the globe.

TABLE OF CONTENTS

LIST OF FIGURES ...... v

LIST OF TABLES ...... vi

ACKNOWLEDGEMENTS ...... vii

Chapter 1 Introduction ...... 1

Ecology and origins of DENV ...... 1 Molecular biology of DENV ...... 5 Immune response to DENV infection ...... 11 Global dengue epidemiology and ecology ...... 17 Evolution of DENV ...... 23 Spatial dynamics and phylogeography of dengue ...... 29 Phylogenetic and phylogeographic analysis of DENV ...... 32 Aims of this thesis ...... 38

Chapter 2 Frequent in-migration and highly focal transmission of dengue viruses among children in Kamphaeng Phet, Thailand ...... 40

Abstract ...... 40 Introduction ...... 41 Results and Discussion ...... 43 Materials and Methods ...... 57

Chapter 3 Phylogeography of recently emerged DENV-2 in Viet Nam ...... 65

Abstract ...... 65 Introduction ...... 66 Results ...... 68 Discussion ...... 76 Materials and Methods ...... 77 Supplementary Figures ...... 88

Chapter 4 Dengue virus in sub-tropical northern and central Viet Nam: population immunity and climate shape patterns of viral invasion and maintenance ...... 92

Abstract ...... 92 Introduction ...... 94 Results ...... 97 Discussion…………………………………………………………………………….110 Materials and Methods………………………………………………………………115 Supplementary Tables………………………………………………………………121

Chapter 5 Modern global dengue diversity: out of Asia…………………………122

Abstract……………………………………………………………………………….122 Introduction…………………………………………………………………………...123 Results and Discussion……………………………………………………………..126 Conclusions…………………………………………………………………………..139 Materials and Methods……………………………………………………………...140 Supplementary Tables………………………………………………………………145

Chapter 6 Discussion…………………………………………………………………..150

REFERENCES…………………………………………………………………………….157

Appendix A The emergence of rotavirus G12 and the prevalence of enteric viruses in hospitalized pediatric diarrheal patients in southern Vietnam…….185

Appendix B The evolutionary consequences of blood-stage vaccination on the rodent malaria Plasmodium chabaudi…….…….…….…….…….…….…….…….194

LIST OF FIGURES

Figure 1-1: Phylogenetic relationships among dengue viruses...... 4

Figure 1-2: Dengue virus genome and virion structures...... 5

Figure 1-3: Model of antibody-mediated enhancement of viral infection...... 15

Figure 1-4: Global distribution of DENV...... 25

Figure 2-1: Kamphaeng Phet Provincial Hospital data of all reported and PCR confirmed dengue cases from Muang District, Kamphaeng Phet during 2004 to 2007...... 43

Figure 2-2: The timing of isolation of DENV lineages circulating in each of five participating sub-districts of Kamphaeng Phet, Thailand during 2004 – 2007 .... 45

Figure 2-3: ML tree of representative E gene sequences of DENV-1 genotype I...... 47

Figure 2-4: ML tree of representative E gene sequences of the Asian I genotype of DENV-2...... 48

Figure 2-5: ML tree of representative E gene sequences of DENV-3 genotype II. .... 49

Figure 2-6: ML tree of representative E gene sequences of DENV-4 genotype I...... 50

Figure 3-1: MCC phylogeny of the DENV-2 Asian I genotype in southern Viet Nam (2003-2008) according to province of sampling...... 70

Figure 3-2: DENV-2 Asian I genotype dispersal across provinces of southern Viet Nam...... 70

Figure 3-3: MCC phylogeny of the DENV-2 Asian I genotype in southern Viet Nam (2003-2008) according to population density (within HCMC) or province of sampling...... 71

Figure 3-4: DENV-2 Asian I genotype dispersal across a population density gradient in HCMC and southern Viet Nam...... 72

Figure 3-5: DENV-2 Asian I genotype dispersal among districts of Ho Chi Minh City, Viet Nam...... 73

Figure 3-6: DENV-2 Asian I genotype dispersal among districts of Ho Chi Minh City and provinces of southern Viet Nam...... 74

Figure 4-1: Regional phylogeography among 705 DENV-1 genotype 1 sequences isolated in Southeast Asia from 1998 to 2009...... 98 vii

Figure 4-2: Branch-specific Markov Jump counts indicating significant migration and establishment of viral lineages across Viet Nam...... 103

Figure 4-3: Local phylogeography in a sample of Vietnamese DENV-1 clades...... 105

Figure 4-4: Differences in relative marginal log likelihood estimates for fitting of the asymmetric phylogeographic model using Akaike’s information criterion through MCMC (AICM)...... 109

Figure 5-1: Number of genotypes for which a significant migration pathway was detected between each of the 12 regions...... 126

Figure 5-2: MCC phylogenies of the DENV-2 American and Cosmopolitan genotypes according to region of sampling...... 130

Figure 5-3: MCC phylogenies of the DENV-2 American-Asian and Asian I genotypes according to region of sampling...... 132

Figure 5-4: Timeline of emergence and circulation of DENV genotypes in South, maritime Southeast, and mainland Southeast Asia...... 135

Figure 5-5: Differences in relative marginal log likelihood estimates for fitting of the asymmetric phylogeographic model using Akaike’s information criterion through MCMC (AICM)...... 137

viii

LIST OF TABLES

Table 1-1: Structural and functional properties of DENV proteins ...... 8

Table 2-1: Phylogeny-trait association tests of phylogenetic structure of geographic, temporal, and clinical traits of DENV isolated in Kamphaeng Phet, Thailand from 2004 – 2007...... 53

Table 2-2: Phylogeny-trait association tests of phylogenetic structure of geographic and temporal traits of DENV-1 and DENV-4 populations isolated via school-based surveillance in Kamphaeng Phet, Thailand from 2004 – 2007...... 55

Table 2-3: Number of DENV sequences obtained from school children and community clusters in five sub-districts of Kamphaeng Phet, Thailand from 2004 – 2007. .. 61

Table 2-4: Number of DENV sequences obtained via school absence-based surveillance within a childhood cohort and village cluster studies activated by illness within the cohort in Kamphaeng Phet, Thailand from 2004 – 2007...... 61

Table 3-1: Phylogeny-trait association tests of phylogeographic structure of DENV-2 in southern Viet Nam...... 69

Table 4-1: Viral migration patterns across the complete data set (full tree) and in Vietnamese clades only...... 99

Table 4-2: Estimates of the Time to the Most Recent Common Ancestor (TMRCA) and the time of the last viral isolate of clusters circulating in central and northern Viet Nam...... 101

Table 5-1: Inferred ancestral locations and times to the most recent common ancestor (TMRCA) of each of the major DENV genotypes...... 134

Table 5-2: Description of DENV sequences used in this study...... 143

PREFACE: CONTRIBUTIONS TO THIS THESIS

All of the chapters included in this thesis were designed, analyzed, and written by myself with the guidance of my advisor, Eddie Holmes, who provided invaluable direction and technical insight. Chapters 2, 3, and 4 involved collaborations with researchers from multiple institutes, who provided sequences and epidemiological expertise on the study populations. The individuals who contributed to this thesis are as follow:

Chapter 2. Chonticha Klungthong, In-Kyu Yoon, Piyawan Chinnawirotpisan, Butsaya

Thaisomboonsuk, Darunee Tannitisupawong, Ananda Nisalak, Mammen P. Mammen Jr.,

Robert V. Gibbons, and Richard G. Jarman of the Armed Forces Research Institute of Medical

Sciences (AFRIMS) in Bangkok, Thailand, Anon Srikiatkhachorn of the Department of Medicine at the University of Massachusetts Medical School, Alan L. Rothman of the Institute for

Immunology and Informatics at the University of Rhode Island, Jared Aldstadt of the Department of Geography at the University at Buffalo, Suwich Thammapalo of the Bureau of Vector-Borne

Disease in the Thailand Ministry of Public Health, Timothy Endy of the Department of Infectious

Diseases at the State University of New York at Syracuse, and Thomas W. Scott of the

Department of Entomology at the University of California at Davis all participated in the design of the cohort and cluster studies from which these viruses came, performed study visits, or sequenced viruses. In-Kyu Yoon, Mammen P. Mammen, Robert V. Gibbons, Alan L. Rothman,

Timothy Endy, Thomas W. Scott, and Richard G. Jarman assisted in the writing and advised on epidemiological aspects of this study.

Chapter 3. This study was made possible by collaboration with the dengue group at the Oxford

University Clinical Research Unit in Ho Chi Minh City, Viet Nam. Particularly integral to this

study were Vu Thi Ty Hang, who isolated virus samples, Bridget Wills and Jeremy Farrar, who identified and treated dengue cases and participated in study design, and Cameron P.

Simmons, who designed and organized the clinical study, laboratory investigations, and the sequencing of virus isolates at the Genome Sequencing Center for Infectious Diseases at the

Broad Institute.

Chapter 4. Mai Quynh Le and Thu Thuy of the National Institute of Hygiene and Epidemiology,

Hanoi, Viet Nam, and Hai Yen Le and Thanh Xuyen Nguyen of the Military Institute of Hygiene and Epidemiology, Hanoi, Viet Nam collected viruses and sequenced viruses. Cameron P.

Simmons of the Oxford University Clinical Research Unit, ho Chi Minh City, Viet Nam, Annette

Fox of the Oxford University Clinical Research Unit, Hanoi, Viet Nam, Robert V. Gibbons of the

Armed Forces Research Institute of Medical Sciences (AFRIMS) in Bangkok, Thailand, and

John G. Aaskov of the School of Biomedical Sciences, Queensland University of Technology provided viral sequences, epidemiological data, and technical expertise.

ACKNOWLEDGEMENTS

This thesis would not have been possible without the knowledge, support and advice of many people. I would first like to thank my advisor, Eddie Holmes, for challenging me to work, write, and think more efficiently and effectively, and for the support and confidence that you’ve shown in me when times were especially strange.

Additionally, I am grateful for continued advice, guidance, and thoughtful conversations with Derek Cummings of the Bloomberg School of Public Health, Baltimore, who helped to spark my interest in dengue dynamics and who continually challenges me to ask bigger questions. I would like to thank Cam Simmons, Jeremy Farrar, and the dengue group at the

Oxford University Clinical Research Unit in Ho Chi Minh City, Viet Nam for sharing their vast expertise in all things dengue, as well as Steve Baker, Maciej Boni, Katie Anders, Jane Chun, and My Phan Vu Tra for insightful work chats and for making me feel at home in Viet Nam. I am grateful to the members of my thesis committee, Andrew Read, Isabella Cattadori, and Tim

Reluga for their input into my work and helpful conversations and words of encouragement along the way, and to the National Science Foundation and the Paul and Harriet Campbell

Distinguished Graduate Fellowship for their financial support.

I would also like to thank the members of the Holmes Lab, past and present, who have shared their knowledge, insight, and support during my time at Penn State. In particular, I would like to thank Cadhla Firth and Mang Shi for helping me hone my ideas and deal with my many computer issues, and for all of the late night laughs.

Finally, I would like to thank my family and friends In Ohio, Baltimore, and Seoul for their support over many years of study and travel.

xii

CHAPTER 1

INTRODUCTION

Ecology and origins of DENV

Dengue viruses (DENV) are members of genus Flavivirus within family Flaviviridae. All flaviviruses are single-stranded, positive-sense RNA viruses, and most require hematophagous arthropods (ticks or mosquitoes) to complete their natural horizontal transmission cycle. The flaviviruses can infect various vertebrate species and arthropods, and are responsible for disease in birds, domesticated and wild animals, and primates (Lindenbach et al., 2007).

Infection with flaviviruses varies from asymptomatic to lethal. Of all known flaviviruses, more than 50% are associated with human disease, and generally manifest as an influenza-like illness characterized by sudden fever onset, arthralgia, myalgia, retro-orbital headaches, maculopapular rash, leukopenia, vascular leakage, and/or encephalitis (Gould and Solomon,

2008; Lindenbach et al., 2007; Seligman, 2008).

Viruses in genus Flavivirus fall into three primary phylogenetic groupings, largely corresponding to their vector or lack thereof. The no-known-vector group includes flaviviruses that have been found to infect bats and rodent species. The tick-borne flaviviruses are comprised of phylogenetically distinct bird- and mammal-infecting clades, and include the tick- borne encephalitides (TBEV). DENV falls within the mosquito-borne viral lineage, which is further divided according to vector species: Culex spp. [including human disease-causing West

Nile Virus (WNEV), St. Louis encephalitis (SLEV), and Japanese encephalitis (JEV) viruses] and Aedes spp. [Yellow Fever (YFV), Zika (ZIKV), DENV] (Jenkins et al., 2001; Vasilakis and

Weaver, 2008).

Dengue viruses comprise an antigenic complex of four phylogenetically and phenotypically distinct viruses (also referred to as serotypes) based on their antigenic cross- reactivities (cross-neutralization). These form a well-supported clade within the flaviviruses but diverge by approximately 30% across the genome, indicating that the four DENV serotypes are no more similar to one another than they are to other flavivirus species (Kuno et al., 1998).

Molecular-clock studies suggest that distinct serotypes began to diverge no more than 2000 years ago while still in a primarily non-human primate-mosquito infection cycle (Dunham and

Holmes, 2007; Twiddy et al., 2003). The sylvatic cycle is assumed to be ancestral, originating in

African or Asian primates, because a critical community size of 10,000 to 1 million is believed to be required for efficient human-to-human transmission (Gubler, 1997).

The unique antigenic profiles of each serotype suggest that DENV diversification may have been driven by selection pressures to evade neutralizing cross-reactive immune responses in the primate population. Some epidemiological models have further suggested that the existence of four serotypes may be an immunological phenomenon in which viremia- enhancing (and thus transmission-enhancing) cross-reactivity would select for viral populations that are divergent enough to evade cross-protective immunity but still similar enough to induce a cross-reactive immune response, thus allowing them to circulate sympatrically (Adams and

Boots, 2006; Cummings et al., 2005; Ferguson et al., 1999). However, if the cross-reactive immune response exerted such selective pressures in the primate population, it is likely that competitive exclusion would have limited the early divergence of multiple DENV serotypes from a single ancestral population (Ferguson et al., 1999). Consequently, it has been proposed that four serotypes evolved from their common ancestor in separate ecological niches (i.e. geographic [allopatric] isolation, distinct primate host populations, different vector species). The modern diversity in host and vector species suggest that these are not limiting ecological niches for DENV (Diallo et al., 2005; Vasilakis et al., 2007), thereby supporting the hypothesis that geographic isolation of ancestral DENV and primate host populations may have resulted in 2

allopatric evolution of the four DENV serotypes. Furthermore, cross-reactive enhancement has not been demonstrated in non-human primates. The epidemiological record suggests that cross-reactive enhancement may be a relatively new phenomenon resulting from the introduction of antigenically distinct DENV populations into new human populations in the last

100 years (Gubler, 1997; Kuno, 2009) as discussed below.

The presence of human and sylvatic cycles in both Africa and Asia has resulted in some debate over the geographic origins of DENV. However, a lack of resolution within the greater flavivirus phylogeny precludes determination of the viral origins of the DENV (Holmes and

Twiddy, 2003; Kuno et al., 1998). The geographic distribution of viruses within the mosquito- borne flaviviruses alone is vast, including pathogens from Africa, Asia, and the Americas, and yields no clue as to the origins of DENV. The detection of all four DENV serotypes circulating in the primate population in isolated forests of Malaysia, via virus isolation (Rudnick, 1965;

Rudnick et al., 1967) or evidence of seroconversion (Rudnick, 1978), suggests that early diversification may have occurred in geographically isolated primate populations in Asia and later expanded into overlapping areas. Sylvatic DENV isolates (DENV-1, -2, -4) from Asia generally fall at highly divergent, basal positions in the DENV phylogenies, suggesting that these viruses have circulated in Asia for at least hundreds of years (Figure 1-1) (Wang et al.,

2000; Weaver and Vasilakis, 2009; Wolfe et al., 2001). In contrast, an African primate cycle has been detected only for the DENV-2 serotype, and these isolates show <100 years of divergence. Thus, the existence of a primate cycle in both Africa and Asia may be the result of movement of infected humans, vectors, or primates from Asia to Africa. In addition, Southeast

Asia shows the highest DENV diversity and prevalence in human populations.

BF_DakAr2039_1980 CI_DakArA1247_1980 BF_DakArA2022_1980 CI_DakAr578_1980 5% sequence MRCA ~ 120 CI_DakAr510_1980 divergence GN_PM33974_1981 years ago SN_DakAr141070_1999 SN_DakAr141069_1999 1 SN_DakArD75505_1991 Sylvatic NG_IBH11664_1966 NG_IBH11208_1966 1 NG_IBH11234_1966 MRCA ~ 320 SN_DakHD10674_1970 years ago SN_DakArD20761_1974 1 MY_DKD811_2008 MY_P8-1407_1970 DENV-2 VN_MD1272_2004 TH_BID_V3501_1995 1 TH_16681_1964 TL_NewGuineaC_1944 1 CN_44_1989 ID_1022DN_1975 JM_N1409_1983 CU_Cuba13_1997 Human VN_MD863_2002 SG_05K3330DK1_2005 1 BF_1349_1983 CN_FJ11_99_1999 VE_Ven2_1987 MX_BID_V3354_1983 PE_IQT2913_1996 1 PF_3056_1990 ID_Sleman78_1978 ID_TB55i_2004 MRCA ~ 100 PH_H87_1956 years ago CN_80_2_1980 BR_RO1_02_2002 Human MQ_IMTSSA_MART1243_1999 DENV-3 1 IN_ND143_2007 LK_IMTSSA_SRI1266_2000 1 TH_BID_V3360_1973 TW_98TW364_1998 VN_BID_V1891_2007 Sylvatic ID_A88_1988 1 PF_FP0705_2001 MRCA ~ 120 USA_HawO3663_2001 years ago MM_31459_1998 VN_BID_V4018_2008 Human DENV-1 JP_Mochizuki_1943 1 MX_BID_V3703_2007 PR_BID_V1162_1998 MM_Myanmar23819_1996 MY_P72_1244_1972 Sylvatic TH_ThD4_0087_1977 MRCA ~ 200 TH_ThD4_0485_2001 PH_H241_1956 years ago PR_BID_V2446_1999 1 VE_BID_V2164_1998 Human DM_814669_1981 TH_ThD4_0476_1997 DENV-4 1 1 TH_ThD4_0017_1997 MY_P73_1120_1973 1 MY_P75_514_1975 Sylvatic MY_P75_215_1975

Figure 1-1. Phylogenetic relationships among dengue viruses. A phylogenetic tree of DENV strains from all 4 serotypes inferred by Bayesian analysis. Horizontal branches are scaled according to the number of substitutions per site. Bayesian probability values are shown for key nodes. Virus strains are coded by abbreviated country of collection/strain name/year of collection (Adapted from Weaver & Vasilakis 2009).

Molecular biology of DENV

Dengue virus is a single-stranded, positive-sense RNA virus. Its positive polarity renders it infectious and capable of synthesizing proteins upon entry into a suitable host cell without the assistance of a complementary RNA intermediate. The ~10.8 kb genome consists of a single open reading frame (ORF) encoding for a single polyprotein, flanked by two untranslated regions (5’ and 3’ UTR; ~100 and 400-700 nucleotides, respectively) (Figure 1-2a).

The UTRs form conserved secondary structures that mediate genome circularization and play important roles in genome replication (Chambers et al., 1990). Within the ORF, the N-terminal of the genome encodes three structural proteins (capsid [C], envelope [E], and either premembrane [prM] in immature virions or membrane [M] in mature virions), which form the structure of the virion. The remainder of the genome encodes seven nonstructural proteins

(NS1, NS2A, NS2B, NS3, NS4a, NS4B, and NS5) that play essential roles in virus replication

(Lindenbach et al., 2007).

a Figure 1-2. Dengue virus genome and virion structures. a. Dengue virus genome, encoding three structural (capsid [C], membrane [M], and envelope [E]) and seven nonstructural proteins (NS1, NS2a, NS2b, b NS3, NS4a, NS4b, NS5) (Adapted from Guzman et al. 2010). b. Cross-sectional view of the virion structure of dengue virus. (Source: ViralZone:www.expasy.org/viralzone, Swiss Institute of Bioinformatics).

The mature DENV virion is approximately 50 nm in diameter and is comprised of a host- derived lipid bilayer and the C, E, and M structural glycoproteins, which carry the RNA genome

(Figure 1-2b). Each of these plays integral roles in viral structure and synthesis. The primary

role of the capsid (C) protein is to encapsulate the viral genome. Intracellularly, a hydrophobic

C-terminal region of the C protein functions as a signal peptide for translocation of the prM to the endoplasmic reticulum (ER) but is cleaved from the C protein by an NS3/NS2B viral protease during viral maturation (Lobigs, 1993). The prM protein is the intracellular precursor to the M protein and is closely associated with the E protein during viral replication and maturation as discussed below (Mukhopadhyay et al., 2005).

The envelope (E) glycoprotein is the primary protein exposed on the surface of the

DENV particle. As such, it functions in binding to host cells, fusion to the host cell membrane during viral attachment and entry, and virion assembly and budding. Its role in antigenicity makes the E protein particularly important for studies of DENV evolution and thus, the many functions of the E gene warrant detailed discussion. In the intracellular environment, E glycoproteins take on a trimeric conformation in the acidic endosome and protrude from the viral surface, thus allowing fusion of the viral and endosomal membranes and releasing the viral genome into the cell cytoplasm (Chu and Ng, 2004; Gollins and Porterfield, 1985, 1986). In the alkaline extracellular environment, however, E protein molecules exist in a dimeric conformation and lay flat on the surface of the virion in a characteristic herringbone pattern (Kuhn et al., 2002;

Mukhopadhyay et al., 2005). In this position, they are available for binding and uptake by receptors on the host cell surface, but are also exposed to neutralizing antibodies during the primary immune response to viral infection.

Thus, the E gene encodes the primary antigenic structure of DENV, and host immunity to dengue infection is mediated by neutralizing antibodies to the E glycoprotein. Flavivirus- specific, complex-specific, and DENV serotype-specific antigenic determinants were initially defined based on cross-reactivity using the hemagglutination-inhibition assay (Casals and

Brown, 1954; Sweet and Sabin, 1954); reactivity was later determined to be based on antigenicity within the E gene (Trent, 1977; Trent et al., 1976). The amino acid (and nucleotide)

E gene sequences yield further information on the differences and similarities among the four 6

DENV serotypes. Within each serotype, the amino acid sequences of the E gene are well- conserved, showing a minimum of 90 to 97% sequence similarity between genotypes

(determined using data described in chapter 5). Between serotypes, however, amino acid sequence similarity ranges from 63 to 67%. Those amino acids that differ among serotypes are distributed across all three of the E protein’s structural domains, even though these differ in function, but tend to be in located in particularly exposed positions on the viral surface (Modis et al., 2005). Together, these observations suggest that the four DENV serotypes may have evolved in order to escape host immune pressures exerted by neutralizing antibodies directed broadly at exposed surfaces of the virion.

Table 1-1. Structural and functional properties of DENV proteins.

Protein-

Protein coding region Primary functions Reference

(bps)

Structural

forms nucleocapsid encasing genomic RNA, C 297-300 (Sangiambut et al., 2008) nuclear localization

interacts with E to prevent membrane fusion prM 498 (Li et al., 2008) during maturation

Attachment to host cell receptors and fusion E 1479-1485 (Nayak et al., 2009) with the endosome

Nonstructural

NS1 1056 unknown, possible role in pathogenesis (Chen et al., 2009)

(Mackenzie et al., 1998; Munoz- NS2a 654 IFN α/β blocking, anchors replication complex Jordan et al., 2003)

NS2b 390 proteolytic processing (Falgout, Pethel, Zhang, & Lai, 1991)

proteolytic processing, enhances NS4b (Falgout et al., 1991; Umareddy et NS3 1854-1857 helicase serine protease activity, nucleotide al., 2006) triphosphatase

membrane rearrangement, anchors the (Lindenbach and Rice, 1999; Miller et NS4a 450 replication complex, interacts with NS1 al., 2007; Munoz-Jordan et al., 2003)

anchors the replication complex, blocks (Mackenzie et al., 1998; Munoz- NS4b 735-744 interferon α/β, colocalizes with NS3 and dsRNA Jordan et al., 2003)

NS5 2697-2700 viral RdRp, N-terminal methyltransferase (Egloff et al., 2007; Yap et al., 2007)

In addition to the three primary structural proteins, roughly two-thirds of the DENV genome is comprised of seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4a, NS4B, and NS5). The putative primary functions of DENV genomic proteins are indicated in Table 1-1.

A number of these remain poorly understood. The best-characterized among these are NS5 and NS3. NS5 is the largest and most conserved of the flavivirus proteins and acts as the viral

RNA-dependent RNA polymerase (RdRp) (Nomaguchi et al., 2003; Sugrue et al., 1997). It has

also been shown to induce the production and secretion of interleukin-8, a neutrophil chemoattractant with primary function in DENV pathogenesis (Medin et al., 2005).

NS3 is the second largest protein in the DENV genome and exerts various functions in polyprotein synthesis and RNA replication. As the catalytic domain of the NS2B-NS3 protease complex, the N-terminal domain of NS3 is required for polyprotein processing and cleaves

NS2B, NS3, NS4A, and NS5 (Chambers et al., 1990; Gorbalenya, Donchenko, Koonin, &

Blinov, 1989). The C-terminal region of NS3 exhibits nucleoside triphosphatase (Bartelma and

Padmanabhan, 2002; Benarroch et al., 2004) and helicase functions (H. Li, Clum, You, Ebner, &

Padmanabhan, 1999) that are required for efficient synthesis of viral RNA. In addition to its role in replication, DENV NS3 is known to induce apoptosis in mammalian cells via caspase 3 and caspase 8 activation (Shafee and AbuBakar, 2003).

Notably, the role of NS1 remains poorly understood but is believed to function in RNA replication. NS1 localizes to RNA replications sites (Mackenzie et al., 1996a), and mutation of the glycosylation sites of NS1 of yellow fever virus (YFV) was found to drastically reduce RNA replication and virus infectivity (Muylaert et al., 1996). In other flaviviruses and possibly in

DENV, NS1 is secreted from infected cells along with E protein, and thus may function in intracellular virion transport or release from the host cell (Lee, Crooks, & Stephenson, 1989).

Secreted NS1 is an important target of humoral immunity and may play a role in pathogenesis as discussed below. Additionally, due to its association with the plasma membrane and expression on the cellular surface, NS1 elicits a complement-mediated antibody response and may confer partial protection against virus challenge (Schlesinger et al., 1990).

Infection & replication in humans

In humans, DENV enters the bloodstream following a bite from an infected mosquito. The infection and replication cycle of DENV in human cells is initiated by the attachment of the viral 9

particle to its target cell through interaction between the E surface glycoprotein and receptors on the host cell surface as described above. Several putative mammalian host cell receptors for

DENV have been proposed (Chen et al., 1997, 1999; Jindadamrongwech et al., 2004; Reyes-

Del Valle et al., 2005; Tassaneetrithep et al., 2003; Thepparit and Smith, 2004), but among these, the interaction between DENV and a dendritic cell-specific intercellular adhesion molecule (ICAM-3)-grabbing nonintegrin (DC-SIGN) is the best characterized (Lozach et al.,

2005; Pokidysheva et al., 2006; Tassaneetrithep et al., 2003). Recent evidence suggests that glycosylation sites on neighboring E proteins bind to a single carbohydrate recognition site of

DC-SIGN (Pokidysheva et al., 2006). Attachment to DC-SIGN may signal the recruitment of a secondary, uncharacterized high-affinity receptor to mediate viral entry (Pokidysheva et al.,

2006). The cell types in which DENV attachment and replication occur remain a topic of investigation. Monocytes, macrophages, and dendritic cells have been implicated as the primary targets for DENV replication in vivo (Jessie et al., 2004), although controversy remains as to whether numerous other human cells may be involved based on findings from in vitro studies (Anderson, 2003).

Following attachment, DENV enters the cell via receptor-mediated endocytosis, moving through clathrin-coated pits into a low-pH endocytic compartment (Acosta et al., 2008; Krishnan et al., 2007; van der Schaar et al., 2008). E glycoprotein trimers at the viral surface fuse with the endosomal membrane to release the viral genome into the intracellular space (Chu and Ng,

2004; Gollins and Porterfield, 1985, 1986). The 5’ UTR then directs the RNA to the ribosomes, where the ORF is translated into a precursor polyprotein. Following initial translation of the viral genome, RNA-dependent RNA polymerase (NS5) produces negative-sense, single-stranded

RNA that serves as a template for the production of new positive-sense, single-stranded RNA genomes and protein translation (Clyde et al., 2006). Synthesized RNA is then packaged by the

C protein to form a nucleocapsid, which is directed to the ER. Here, the nucleocapsid is surrounded by a lipid envelope on which prM-E protein heterodimers are attached. Immature 10

virions are trafficked through the ER lumen and into the Golgi complex. In this immature state, prM caps the fusion peptide of the E protein, possibly stabilizing it to prevent premature intracellular membrane fusion prior to virion release. Prior to the release of the virions from the cell, the prM protein is cleaved by furin (Yu et al., 2008; Zhang et al., 2003; Zybert et al., 2008), generating a membrane-associated M protein and a pr peptide and rendering the newly-mature virions infectious (Rodenhuis-Zybert et al., 2010b). Finally, mature viruses are released into the extracellular space.

Infection & replication in mosquitoes

Following ingestion of DENV by Aedes spp. (primarily Aedes aegypti and Aedes albopictus) in a bloodmeal, the virus enters the midgut, where E proteins bind to receptors on midgut epithelial cells and virions enter the host cell via endocytosis (Black et al., 2002). The receptors that bind to DENV in mosquito cells are yet unknown, but recent research has shown that DENV surface proteins interact with heat-shock protein (HSP70) (Reyes-Del Valle et al.,

2005), R80, R67 (Mercado-Curiel et al., 2006), and a 45-kDa protein (Yazi Mendoza et al.,

2002). Uncoating, transcription, and translation precede virus maturation and release.

Infectious viral particles then disseminate from midgut epithelial cells and infect secondary target organs, eventually reaching the salivary glands where it is capable of infecting humans upon feeding (Black et al., 2002). As in human hosts, a number of barriers to infection and successful transmission into a new host exist within the vector and may affect the local and global evolution of DENV, including differences in susceptibility within vector species and in local populations (Gubler and Rosen, 1976, 1977), susceptibility of the vector to an infecting strain (Anderson and Rico-Hesse, 2006; Gubler and Rosen, 1977; Hanley et al., 2008), barriers to midgut infection and escape (Bennett et al., 2005; Gubler et al., 1979a; Tabachnick et al.,

1985), and the modulation of DENV infection by the normal flora in the mosquito system (Xi et

al., 2008). Notably, replication in the mosquito vector may be significantly affected by environmental temperature (Kay et al. 1989; Watts et al. 1987; Kramer & Ebel 2003), resulting in shorter extrinsic incubation periods (EIP) under higher but not extreme temperatures.

Immune response to DENV infection

Upon the bite of an infected mosquito, DENV is believed to initially infect interstitial dendritic cells (DCs), initiating the production of interferons (IFNs) by these DCs within a few hours post-infection via the interaction of virus with pathogen-recognition receptors (PRRs)

(Fernandez-Garcia et al., 2009; Trinchieri and Sher, 2007). In vivo and in vitro studies have shown that the actions of type I (α, β) and type II IFN (γ) play critical roles in the innate host immune response against DENV infection (Chen et al., 2008b; Diamond et al., 2000; Ho et al.,

2005; Johnson and Roehrig, 1999; Shresta et al., 2004b), and indicate that early activation of natural killer (NK) cells, the primary producers of IFNγ, is important in the clearance of DENV infection (Azeredo et al., 2006; Shresta et al., 2004a). Secreted IFN binds to receptors present on infected and surrounding uninfected cells, resulting in the activation of the JAK/STAT pathway and a cascade of protein expression (Clemens and Williams, 1978; Pavlovic et al.,

1990; Rodenhuis-Zybert et al., 2010b). This induces alterations in cell metabolism to inhibit viral infection (Chareonsirisuthigul et al., 2007; Chen et al., 2008b; Shresta et al., 2005). However, studies have shown that the viral NS2A, NS4A, NS4B, and NS5 of DENV may suppress the

IFN-mediated innate immune response by reducing STAT activation and thus blocking IFN signaling (Ashour et al., 2009, 2010; Ho et al., 2005; Munoz-Jordan et al., 2003). Interestingly, the suppression of IFN signaling appears to be strain- but not serotype-specific, which suggests that this function may have arisen independently in multiple serotypes (at least in DENV-2 and

DENV-4) to evade the host immune response (Umareddy et al., 2008).

Five to seven days after the bite of an infected mosquito, the primary humoral immune response is mounted by antibodies directed at the E and prM viral surface glycoproteins

(Cardosa et al., 2002; Lai et al., 2008) and surface-expressed and secreted NS1 (Libraty et al.,

2002; Shu et al., 2000). The antibody response to NS1 induces complement-mediated lysis in infected cells (Henchal et al. 1988; Kurosu et al. 2007; Costa et al. 2006), which acts to reduce

infection but may also contribute to disease pathogenesis. The response to E and prM modifies the infectiousness of DENV particles, but has the potential to both neutralize and enhance infectiousness. Neutralization of virus in this case is only effective when the number of antibodies attached to the virion reaches a threshold (Gromowski and Barrett, 2007), which differs according to the attacking antibody. Antibodies with stronger neutralization actions are generally directed against domain III of the E protein and are serotype- or strain-specific (Lisova et al., 2007; Rajamanonmani et al., 2009); these require binding at only a small fraction of accessible viral surface epitopes for neutralization or inhibition to occur. Weaker antibodies directed at the less accessible domain II of the E protein or the fusion peptide, which is highly conserved across DENV, do not bind as well and necessitate attachment in larger quantities

(Lai et al., 2008).

Early research on immunity to DENV infection demonstrated that infection with one serotype resulted in lifelong or long-term protection against re-infection with the same serotype

(homotypic immunity) and short lived (<9 months) protection against the other serotypes

(heterotypic immunity) (Sabin, 1950). The strongly neutralizing antibodies function to prevent future homotypic infection by inhibiting the natural route of cell entry and redirecting virus to immune cells, where they are subsequently lysed. The transient nature of heterotypic immunity lies in the weak antibodies directed at E protein domain II. Over time following infection, these antibody concentrations decline and the individual is then susceptible to infection with other

DENV serotypes.

Fundamental to any discussion of DENV is the increased risk of severe illness in heterotypic secondary infection or primary infection of infants born to dengue-immune mothers, a phenomenon that was first documented in Southeast Asia in the 1960s (Halstead et al., 1970) and has since been confirmed in setting across endemic and epidemic transmission settings

(Sangkawibha et al. 1984; Burke et al. 1988; Halstead et al. 1970; Kliks et al. 1988; Chau et al.

2009; Hammond et al. 2005; Guzman et al. 2002). These clinical and epidemiological

observations have led to the (now) widely accepted hypothesis of ‘antibody-dependent enhancement’ (ADE) of disease. This model postulates that cross-reactive antibodies from a previous infection or subneutralizing serotype-specific antibodies interact with DENV without neutralizing the virus, and instead bind the viral particles and direct them to Fc receptor-bearing cells such as macrophages, monocytes, and DCs, the natural targets for DENV infection. The

DENV-antibody complex is thought to both enhance the attachment of DENV to the cell surface and facilitate cell entry via Fc receptor-mediated endocytosis (Halstead, 2003; Mathew and

Rothman, 2008) (Figure 1-3). Critically, the uptake of DENV-antibody complexes may lead to both a higher number of infected cells and an increased viral yield produced per infected cell

(Gollins and Porterfield, 1985; Goncalvez et al., 2007; Halstead, 1979; Libraty et al., 2002;

Stephenson, 2005; Vaughn et al., 2000; Wang et al., 2003). Thus, the ADE phenomenon may act to not only enhance the severity of infection, but also to enhance viremia and thus disease transmission, which implies that it may play an important role in shaping the dynamics of disease transmission. In addition to ADE, differences in human leukocyte antigen (HLA) haplotypes (Loke et al., 2001; Polizel et al., 2004; Zeng et al., 1996), age, gender, and comorbidities (Figueiredo et al., 2010; Limonta et al., 2008), as well as viral genetic factors

(discussed below), have all been associated with the risk of severe disease.

Figure 1-3. Model of antibody- mediated enhancement of viral infection. Virus particles opsonized with high affinity antibodies at low occupancy can be internalized via the virus- and Fc receptor, after which the unbound E proteins mediate membrane fusion from within endosomes. Low affinity antibodies, irrespective of the occupancy, dissociate from the virus particle in the acidic endosome such that viral fusion can occur. Antibodies against prM enhance viral infection but the mechanism remains elusive (Adapted from van der Schaar et al. 2009).

Interestingly, antibodies against prM have also been implicated in ADE (Dejnirattisai et al., 2010; Huang et al., 2006; Rodenhuis-Zybert et al., 2010a). DENV-infected cells secrete large numbers of immature prM-containing particles, and prM antibodies are detected in patient sera in both primary and secondary infections (Cardosa et al., 2002; Murray et al., 1993; Se-

Thoe et al., 1999). These prM antibodies were found to be capable of rendering immature, non- infectious particles nearly as infectious as wild-type virus (Dejnirattisai et al., 2010; Rodenhuis-

Zybert et al., 2010a). Although this phenomenon acts to enhance the infectivity of immature virions, it does not appear to enhance the number of virions produced per infected cell and thus may not play role in enhancement (Figure 1-3).

Although virus-specific cytotoxic T lymphocytes have been shown to function in recovery from a number of viral infections, their role in DENV infection is poorly understood, and these too may play dual roles in protection against and enhancement of infection. CD4+ and CD8+ T cell responses are also mounted by the host immune system and appear to be directed at multiple viral proteins, particularly NS3. Serotype-specific and cross-reactive CD8+ cells have been detected in DENV-immune individuals (Kurane et al., 1991) and a recent study has suggested that they play a protective role (Yauch et al., 2009). However, activation of memory

CD8+ T cells during heterologous secondary infection appears to result in the overproduction of cytokines and immune-modulating activity observed in severe hemorrhagic dengue cases, and may play a role in immune enhancement of disease (Mathew and Rothman, 2008; Simmons et al., 2005; Zivna et al., 2002). The lack of a good animal model of disease has hindered the study of the T cell response and other immune responses to dengue, and further research in this area is needed to fully elucidate the processes by which dengue infection are neutralized or enhanced under different immunological scenarios.

Global dengue epidemiology and ecology

Dengue is considered the most important arboviral disease in the world, owing to its massive disease burden and rapid expansion. Disease incidence has increased 30-fold in the last three decades alone, expanding into new countries and moving further into rural territories in endemic areas. Approximately 50 million DENV infections occur annually, mostly among the

2.5 billion people currently living in dengue-endemic countries in the tropics and sub-tropics

(World Health Organization, 2009a). While DENV serotypes and genotypes are phylogenetically and antigenically distinct, they are not clinically distinguishable in terms of disease manifestations, and all carry the potential to result in a range of infection outcomes, from asymptomatic infection to severe disease and death. Asymptomatic or very mild illness appears to occur in the majority of infections, with symptomatic to asymptomatic ratios ranging from 1:1 to 1:8 (Balmaseda et al., 2006; Burke et al., 1988; Endy et al., 2002a, 2011; Graham et al., 1999; Porter et al., 2005; Sangkawibha et al., 1984; Thein et al., 1997). Dengue fever (DF), a self-limiting febrile illness lasting for 2 to 10 days, is the second most likely outcome. In these cases, hospitalization may be necessary only if complications or warning signs of more severe illness arise (World Health Organization, 2009a). These may be signs of progression to severe dengue, also known as dengue hemorrhagic fever (DHF), which can include severe plasma leakage, severe hemorrhage, and/or severe organ impairment and may result in death as symptoms progress to shock. A shift in the clinical nomenclature to replace use of DHF with severe dengue of grade I, II, or III reflects knowledge gained from decades of dengue treatment as well as changes in dengue epidemiology and clinical manifestations (Deen et al., 2006;

World Health Organization, 2009a). Due to the very recent acceptance of this change in terminology, severe dengue and DHF are used interchangeably in the work presented in this thesis.

Dengue-like illness was first described in China as early as the Third, Seventh and Tenth

Centuries (Gubler, 1997) but absent from historical records for several centuries until a similar illness was described in the West Indies and Panama in the 1600s. The first reports indicating a possible pandemic indicated epidemics in Batavia (now Jakarta), and Cairo, Philadelphia, and

Seville and Cadiz, Spain from 1779-1788, followed by a nearly global pandemic that moved through Africa, India, the Middle East, Oceania, Southeast and East Asia, and the Americas from 1823 to 1916 (Brown, 1977; Gubler, 1997). Epidemics lasted for three to seven years in each region, with concurrent epidemics often occurring oceans apart. It is impossible to determine that these outbreaks were all caused by DENV, and if so, which serotype(s) were involved, but given the speed, timing, and coastal and riverine locations of these epidemics, it seems likely that these could have been caused by the same serotype or a sequence of previously isolated serotypes being shuffled around the globe as the shipping trade increased global trade and human movement. With the movement of humans and cargo, so moved the

African mosquito vector, Aedes aegypti, which rapidly invaded urban areas across the tropics

(Godding, 1890; Gordon Smith, 1956a). The invasion of this anthropophilic vector coupled with rapid urbanization in parts of Southeast Asia and India likely brought DENV out of the forests and rural areas and into large, dense populations where endemicity had clearly been established by the end of the 19th century. For the next several decades, DENV activity proceeded largely undetected in native populations of endemic Southeast Asia, with occasional outbreaks occurring in adult travelers or migrants who were likely being exposed to dengue for the first time (Gordon Smith, 1956b; Meade, 1976). Surveys of rural populations suggested that transmission intensity was apparently high enough that nearly all children had been infected at an early age, but infection caused such mild illness in children that it was thought that natives of

Southeast Asia were immune to the disease.

The impact of World War II in the Pacific Theatre caused new ecologic and demographic changes in the region. Uncontrolled urbanization, the destruction of water distribution and 18

sewage infrastructure, and the abandonment of war materials created ideal breeding grounds for Aedes aegypti. The movement of troops through the area offered a constant replenishment of susceptibles and a vehicle by which local DENV strains and mosquito populations were likely dispersed to distant populations (Brown, 1977; Halstead, 1992; Hotta, 1952; Sabin, 1952). It was in this setting that dengue hemorrhagic fever (DHF) emerged as a major public health problem in Southeast Asia. While hemorrhagic manifestations in dengue were not previously unheard of, these were generally rare. In 1953-1954, however, an epidemic of a newly described “epidemic hemorrhagic fever” ravaged Manila, Philippines, with a second epidemic sweeping through the city two years later (Halstead and Yamarat, 1965; Quintos et al., 1954). A third epidemic of similar clinical description occurred in Bangkok in 1958, although sporadic cases had occurred there since 1950 (Halstead, 1980; Hammon, 1973). It was during the

Bangkok outbreak that the etiology of this severe disease was finally determined and the disease was termed ‘dengue hemorrhagic fever’ [DHF] (Hammon et al., 1960). By 1963, DHF epidemics had swept through cities and towns in the Philippines, Thailand, Malaysia, India,

North and South Vietnam, and Singapore (Chew et al., 1961; Halstead et al., 1965; Hammon et al., 1960; Quintos et al., 1954; Ramakrishnan et al., 1964).

Clinical studies in the 1960s first noted an association between secondary heterotypic dengue infection and DHF in Southeast Asia (Halstead et al., 1967, 1973; Hammon, 1973;

Johnson et al., 1967), prompting the first description of antibody-dependent enhancement

(ADE) hypothesis in dengue infection. The role of secondary infection as the primary risk factor for severe dengue disease was confirmed as a global phenomenon over the following decades

(Burke et al., 1988; Guzman et al., 1987; Marchette et al., 1973). The circulation of multiple serotypes in a region, either concurrently (hyperendemicity) or sequentially, therefore is responsible for the rapid appearance of DHF epidemics across Southeast Asia and the resultant morbidity and mortality. While multiple serotypes were likely present in some locations in

Southeast Asia prior to these detected epidemics, the movement of people, vector, and virus in 19

the middle of the 20th Century certainly played a primary role in disseminating foreign lineages and serotypes to new areas.

Following the first epidemics, a cyclical pattern of regular epidemics occurring every three to five years was established in most of Southeast Asia, each larger than the last as the geographic distributions of virus and vector expanded within each country (Gubler, 2002;

Halstead, 1965; Hammon et al., 1960). Serotype-specific surveillance data indicate that multiannual fluctuations in the incidence of dengue disease reflect cyclical oscillations in the dominant serotype (Cummings et al., 2005; Nisalak et al., 2003; Recker et al., 2009; Rodriguez-

Barraquer et al., 2011). These patterns have been linked to the strong neutralizing immune response to homotypic infection, as well as the weak neutralizing and potentially cross- enhancing response to heterotypic infection (Adams et al., 2006; Cummings et al., 2005; Recker et al., 2009). Large multiannual epidemics tend to be caused by one (or occasionally two) dominating serotype. Subsequent to an epidemic, the large proportion of the population immune to the epidemic serotype reduces the number of susceptibles available for infection with that serotype, and transient cross-neutralizing reactions may further reduce local DENV transmission for up to two years (Salje et al., 2012), possibly resulting in population bottlenecks that could impose strong selection pressures on viral populations.

Hyperendemic DENV transmission (all four serotypes) has been maintained in populations throughout Southeast Asia since at least the 1950s. The burden of dengue is the highest here, where large DHF epidemics now occur every 2-4 years, along with year-round but highly seasonal transmission during interepidemic periods (Cummings et al., 2004, 2009; Goh,

1995; Huy et al., 2010; Thai et al., 2010a). Transmission intensities are sufficiently high that most people experience at least two DENV infections before the age of 20 (Anders et al., 2011;

Chan, 1987; Chareonsook et al., 1999; Eram et al., 1979; Shekhar and Huat, 1992; Vong et al.,

2010). The average age of infection tends to be higher in regions outside of Southeast Asia and reflects lower overall transmission intensities even in hyperendemic environments (Chungue et 20

al., 1992; Hayes et al., 1996; Messer et al., 2002; Rodriguez-Barraquer et al., 2011). South

Asia also shows a long history of hyperendemic transmission. Although the endemic disease burden here appears to be somewhat lower than that in Southeast Asia, epidemics in this region are becoming more frequent and widespread (Baruah and Dhariwal, 2011; Chaturvedi et al.,

1970; Dorji et al., 2009; Gupta et al., 2006; Lucas et al., 2000; Malla et al., 2008). This is also true in Oceania, where lower levels of endemic activity occur and epidemics are spurred on by the introduction of new lineages to isolated populations (Condon et al., 2000; Glaziou et al.,

1992; Hirshman, 1976; Mackenzie et al., 1996b; Rosen, 1967). While populations in the

Americas were largely sheltered from dengue prior to the 1960s due to regional vector control campaigns, there was a resurgence following the introduction of DENV-1, -2, and -4 in the late

1970s and early 1980s (Reiter, 1996). The introduction of specific DENV genotypes may have contributed to the rapid invasion and establishment of novel DENV lineages in the Americas as discussed below (Barbosa da Silva et al., 2002; Das and Shapiro, 2002; Guzman and Kouri,

2003; Guzman et al., 1991, 2002; San Martin et al., 2010). Comparatively little is known about the epidemiology of dengue in Africa. At least three serotypes circulate here and reports of dengue fever epidemics from the continent have increased in recent years, yet epidemics of

DHF have rarely been reported (Diallo et al., 2008; Durand et al., 2000; Franco et al., 2010;

Vazeille-Falcoz et al., 1999). Limited surveillance and reporting, coupled with a potential host genetic component that protects against severe infection (de la C Sierra et al., 2007; Hansen,

1990; Kouri et al., 1987), make detection difficult and contribute to a lack of knowledge of the dynamics in this area.

The distribution of all four DENV serotypes throughout the tropics and sub-tropics is nearly complete. Models suggest that DENV will continue to spread to new geographic areas under global climate change scenarios, although the impact of this spread on global incidence is unclear (Jetten and Focks, 1997; Patz et al., 1998; Zhang et al., 2008). Recent outbreaks in the

United States and Europe and the spread of the vector species, Aedes albopictus, suggest that 21

the spread of DENV will continue into semi-temperate areas if effective control measures are not implemented (Gould et al., 2010; Radke et al., 2012; Seyler et al., 2009). The current lack of an effective vaccine renders vector control a high priority to halt the spread of DENV and reduce morbidity and mortality in currently endemic areas.

Evolution of DENV

Evolutionary change, including that in viruses, is driven by five primary forces: mutation, natural selection, recombination, genetic drift, and migration. Like all RNA viruses, DENV replication is dependent on the notoriously error-prone RNA-dependent RNA polymerase, which lacks a proofreading mechanism and thereby has the potential to contribute novel mutations of varying fitness during every replication cycle (Drake and Holland, 1999; Moya et al., 2004).

Indeed, RNA-dependent RNA polymerases have estimated error rates of ~10-3-10-4 per cycle, resulting in ~1 error per round of genome replication, with levels of intra-host genetic variation suggesting that equivalent values occur in DENV (Holmes, 2003; Lin et al., 2004; Thai et al.,

2012).

Despite the potential to produce abundant genetic variation, because DENV exists in a human-vector cycle and must be capable of infecting both host types for effective transmission, the viruses appear to be subject to strongly purifying selection (Bennett et al., 2006; Twiddy et al., 2003; Zanotto et al., 1996). Accordingly, most amino acid changes are deleterious, reduce fitness in the human host, vector, or both, and thus are purged from the population. The most compelling evidence supporting this hypothesis comes from estimation of the relative rates of non-synonymous (dN) to synonymous (dS) substitutions per site, in which dS > dN is suggestive of purifying selection and, alternatively, dS < dN is indicative of positive selection. Studies show that purifying selection is very strong in DENV (dS >> dN), and indicate significant differences in dN/dS within and between hosts. Within individual hosts, dN/dS is often close to 1, a value expected when all mutations are equally likely and selection pressures have not been imposed; between hosts, this is generally < 0.1 (Holmes, 2003; Thai et al., 2012). This difference suggests that the majority of amino acid variants arising within the host are deleterious and are likely purged when entering the vector population (and when viruses from vectors pass to humans). Similarly strong purifying selection has been shown for a number of other arboviruses

(Beaty and Bishop, 1988; Coffey et al., 2008; Strauss and Strauss, 1988; Woelk and Holmes,

2002).

Despite these selective constraints, viral gene sequence data offer an alternative method with which to infer the evolutionary and epidemiological histories of DENV in humans.

Molecular-clock analyses suggest that all four serotypes independently entered the human population and established an exclusively human-vector cycle of transmission only in the last

100-350 years (Holmes and Twiddy, 2003; Twiddy et al., 2003; Zanotto et al., 1996), and to some extent corroborate the epidemiological record. These dates may indicate the timing of the major cross-species transmission events from primates to humans and/or of the spread and establishment of anthropophilic Aedes aegypti as the primary DENV vector. Certainly, multiple spillover events between the sylvatic and human cycles likely occurred prior to and since the emergence of the four serotypes (Cardosa et al., 2009; Vasilakis et al., 2007, 2008), but only a single lineage of each has become established in the human population. Notably, in the three serotypes for which sylvatic strains have been isolated (DENV-1, -2, -4), emergence in the human cycle was accompanied by changes in the primary receptor-binding domain III of the E protein, which suggests adaptation to infect humans or new vectors associated with human hosts (Wang et al., 2000).

The emergence of the four DENV serotypes into human endemic/epidemic cycles was followed by the rapid radiation and spread of genetic diversity in the last 100 to 200 years. This recent diversity is structured in a number of genotypes within each serotype. Intra-serotype variability within DENV serotypes was first characterized in 1990 using a 240-bp fragment of the

E/NS1 gene region of DENV-1 and DENV-2 (Rico-Hesse, 1990) and identified distinct

“genotypes” (or “subtypes”), arbitrarily defined as having no more than 6% sequence divergence. Using E gene sequences, phylogenetic studies corroborate the existence of these genotypic clusters and suggest a geographic component to the diversity observed within serotypes (Holmes and Twiddy, 2003; Twiddy et al., 2002a). This is exemplified by the naming 24

of the six genotypes of DENV-2 (Lewis et al., 1993; Twiddy et al., 2002a). “Asian I” and “Asian

II” are found almost exclusively in Asia, while the “Cosmopolitan” genotype is distributed across the tropics and subtropics. The “American” genotype circulated in the Americas from the 1940s until the 1990s and its origin was unknown upon the naming of the genotypes. This issue is resolved in this work (see Chapter 5). The “American-Asian” genotype appears to originate in southern Southeast Asia and was introduced to the Americas in the late 1970s or early 1980s, eventually replacing the “American” genotype throughout the region. The last of these is the sylvatic genotype, which diverged from the human DENV-2 lineage prior to its diversification.

The sylvatic DENV-2 genotype contains both Southeast Asian and African viruses, the latter of which comprise all African sylvatic isolates. Genotypes among the remaining three serotypes are not named by region, but simply with number (I-V); DENV-1 includes five genotypes

(Goncalvez et al., 2002; Rico-Hesse, 2003); DENV-3, five genotypes (Lanciotti et al., 1994;

Rico-Hesse, 2003; Wittke et al., 2002), and DENV-4, two human genotypes and a Malaysian sylvatic genotype (Chungue et al., 1995; Lanciotti et al., 1997) (Figure 1-4). The geographic distributions and global dispersal patterns of the major lineages are discussed further and in some cases redefined in Chapter 5 of this work.

Figure 1-4. Global distribution of DENV. Broad-scale geographic distributions of DENV serotypes and SE Asia (north) DENV-1 (I, II, III) genotypes are indicated: DENV-2 (Am/As, As I, As II, Cos) South Asia DENV-3 (II) DENV-1 (II, III) DENV-4 (I, III) Latin America & Caribbean (I)-(IV), genotypes I-IV; DENV-2 (Am, Cos) DENV-1 (III) DENV-3 (II, III) DENV-2 (Am, Am/As, Cos) DENV-4 (I) (As I) and (As II), Asian I Africa SE Asia (south) DENV-3 (III, IV) DENV-1 (III) DENV-1 (II, III) DENV-4 (II) and II; (Am/As), DENV-2 (Cos) DENV-2 (As I, As II, Cos) DENV-3 (III) DENV-3 (I,II) American-Asian; (Am), DENV-4 (I, II)

American; (Cos), Oceania DENV-1 (II) Cosmopolitan (Adapted DENV-2 (As I, Cos) DENV-3 (III) from Holmes, 2009). DENV-4 (II)

Asia appears to contain the greatest genetic diversity in DENV, supporting the hypothesis for a viral origin of human DENV in the Asian sylvatic cycle. However, the nearly global distribution of the DENV-2 Cosmopolitan genotype compared to limited distributions of the Asian I and II genotypes, for example, suggests that some viral lineages may be inherently more prone to invading and becoming established in new populations than others (i.e. they have increased “epidemic potential”), compatible with an increased fitness. Studies of selection pressures undertaken to date in DENV have revealed showed a moderate signal for positive selection in the E gene sequences of DENV-3 and -4 but none in DENV-1 (Twiddy et al., 2002a,

2002b). Among positively selected sites within in the E gene, most were located in or near B- and T-cell epitopes, which suggests that selection pressures may be related to host immune evasion (and hence that immune responses are not completely protective). Genotype-specific analyses of DENV-2 suggested that the Cosmopolitan genotype is under stronger selection pressure than other DENV-2 genotypes, which may make it a more successful invader than other lineages (Twiddy et al., 2002a). Alternatively, differences in the spread of DENV lineages may be determined by factors related to features of the source populations from which they radiate and their connectivity to other regions, as well as the immune status of populations to which they are introduced (discussed in Chapter 5).

Another potential signature of natural selection acting on the diversity of DENV are the lineage replacement and extinction events that have been documented in multiple endemic populations. DENV genotypes and sub-genotype lineages appear to undergo ongoing processes of birth and death that may relate to the relative fitnesses of the viral lineages involved. In Thailand, for example, a turnover of DENV-2 strains (subgenotype) was observed between 1980 and 1987 (Sittisombut et al., 1997), and of DENV-1 (Zhang et al., 2005) and

DENV-3 (Wittke et al., 2002) strains in the 1990s. Also in the 1990s, the DENV-2 Asian-

American genotype was replaced by the Asian I genotype in Thailand; complete replacement and apparent extinction of the Asian-American lineage had occurred across mainland Southeast 26

Asia by the mid-2000s (Hang et al., 2010). The replacement and subsequent extinction of

DENV-1 Genotype 1 lineages in the 1990s has been attributed to selection for enhanced transmission in the vector (Lambrechts et al., 2012), while the rapid replacement of DENV-2 genotypes was linked to increased viremia caused by the Asian I strain in human infection, with little difference in infectivity in mosquitoes. The Asian-American genotype had previously invaded the Americas in the 1980s, causing a widespread increase in severe disease and replacing the previously dominant American genotype (Bennett et al., 2006; Rico-Hesse et al.,

1997). Interestingly, in Viet Nam, the invasion of Asian I DENV-2 into a predominantly

American-Asian genotype population was associated with an increase in DENV-2 incidence, such that DENV-2 rapidly became the dominant circulating serotype. Three years later, DENV-

2 incidence dropped dramatically, coinciding with the extinction of the American-Asian genotype in Viet Nam (Hang et al., 2010); a very similar pattern was observed during a clade replacement event in DENV-1 in Thailand (Zhang et al., 2005). These recent observations of selection for enhanced transmissibility represent major changes in our understanding of the means by which

DENV diversity and phenotypic changes arise.

Viral genetic factors also appear to affect disease outcomes. In addition to the enhancement of secondary infection due to ADE, the sequence of infection was purported to play a role in the risk of severe disease, with a greatly increased risk of DHF, for example, upon primary infection with DENV-1, -3, or -4 followed by infection with DENV-2 compared to a different sequence of infections (Halstead, 1982; Monath, 1994b; Sangkawibha et al., 1984).

Later, differences in disease severity, viremia, and potentially transmission fitness were detected between two lineages of the same serotype, supporting the hypothesis that viral genetics, host immunity, and pathogenesis are intertwined (Armstrong and Rico-Hesse, 2001;

Cologna and Rico-Hesse, 2003; Cologna et al., 2005; Hang et al., 2010; Leitmeyer et al., 1999;

Rico-Hesse et al., 1997). More recently, studies in a Nicaraguan cohort showed an increase in the severity of secondary DENV-2 infections over several seasons (Hammond et al., 2005; 27

OhAinle et al., 2011). This change was attributed to the replacement of a DENV lineage of the

American-Asian genotype (NI-1) with a different lineage that was also of the American-Asian genotype (NI-2). An increase in viremia accompanied the severe disease observed in the later period of the study, thus suggesting that a fitter virus (NI-2) had emerged within the population.

Finally, the role of recombination as a driver of DENV evolution remains unclear.

Phylogenetic evidence of recombination has been shown in all four serotypes, but not among serotypes, and always relatively infrequently (Aaskov et al., 2007; AbuBakar et al., 2002; Chen et al., 2008a; Craig et al., 2003; Tolou et al., 2001; Twiddy and Holmes, 2003; Uzcategui et al.,

2001). The mechanism of recombination in this case, as in other RNA viruses, is most likely a copy-choice process whereby the polymerase switches between multiple infecting virus genomes during replication (Lai, 1992). The finding that recombination occurs in DENV is not surprising given that Aedes aegypti is a nervous feeder that may bite multiple hosts in a single blood meal and that mixed infections have been documented in humans and mosquitoes

(Aaskov et al., 2007; Craig et al., 2003; Thai et al., 2012).

Spatial dynamics and phylogeography of dengue

The complexity of DENV transmission dynamics over space and time necessitate a comprehensive epidemiological and evolutionary characterization of these processes at various spatial and temporal scales. Spatial and temporal dynamics of DENV transmission are driven by multiple factors, including virus genetics (Cologna et al., 2005; Endy et al., 2004; Hang et al.,

2010), host immune status (Endy et al., 2004), the age structure of the population (Anders et al.,

2011; Cummings et al., 2009; Nagao et al., 2008), the mosquito vector, and environmental variables such as temperature and rainfall (Chowell et al., 2011; Colon-Gonzalez et al., 2011;

Hii et al., 2009). Human movement has also been suggested to contribute to local viral transmission dynamics (Adams and Kapan, 2009; Balmaseda et al., 2010; Rabaa et al., 2010;

Raghwani et al., 2011; see also Chapters 3 and 4), and is clearly responsible for the complex distribution of DENV at the global scale (discussed in Chapter 5). Understanding local patterns of DENV transmission in endemic dengue populations is integral to the rational deployment of vector control activities, health system management, and the design and monitoring of vaccine trials, and provides a context in which to test the effects of host, vector, and viral factors on

DENV transmission and dispersal; understanding these dynamics at the regional and global levels allows us to investigate the routes by which DENV move into new populations and to identify conditions that allow for the invasion and establishment of vector-borne diseases in new environments.

Globally, DENV incidence continues to rise and outbreaks are being reported in previously DENV-free environments. The spatial distribution of DENV generally overlaps with that of its primary vector, Aedes aegypti, although a significant role for Aedes albopictus has been proposed under climate change scenarios (Gubler et al., 2001; Patz et al., 1998; Rogers et al., 2006). In almost all locations, DENV transmission is highly seasonal, with most infections occurring during periods of the year when temperature and precipitation are highest (Hurtado-

Diaz et al., 2007; Lu et al., 2009; Nakhapakorn and Tripathi, 2005; Rosa-Freitas et al., 2006; Wu et al., 2007). Evidence suggests that the extrinisic incubation period, vector biting rate, and vector mortality, all of which vary with ambient temperature, determine the timing of the ‘dengue season’ (Bartley et al., 2002b; Focks et al., 1993; Lambrechts et al., 2011; Pant et al., 1973;

Watts et al., 1987). While a link between rainfall and vector-borne diseases is intuitively pleasing, evidence that precipitation influences the dengue season in endemic areas is generally observational and anecdotal.

The tracking of viral populations through phylogenetic analysis has shown that viral populations can be distinguished at the national level, with neighboring countries generally maintaining their own viral populations and in situ evolution, and occasional movement of viral populations across national-level boundaries (A Nuegoonpipat et al., 2004; Araujo et al., 2008;

Carrington et al., 2005; Diaz et al., 2006; Hang et al., 2010; Holmes et al., 2009; Raghwani et al., 2011). This is also true at the regional level and is most obvious in the Americas, which has experienced seeding of multiple DENV populations from Asia that have been maintained autochthonously (i.e. in situ transmission/evolution) in the region for decades (Allicock et al.,

2012; Foster et al., 2003, 2004). Of the primary phylogeographic patterns of RNA viruses proposed by Holmes (2008), this suggests a sink-source transmission pattern with Asia as a source, with population subdivision at the country level (Holmes, 2008).

While seasonality is largely determined by climate, it appears that the cycling of population immunity plays an important role in the multiannual periodicity (generally two to four years) of DENV epidemics in endemic populations. Mathematical models of directly transmitted diseases such as measles indicate that multiannual cycles in incidence are consistent with a predator-prey interaction between the pathogen and susceptible humans (Grenfell et al., 2001).

These models predict that large booms in incidence decrease the number of susceptible individuals such that an epidemic fades-out and a period of time must elapse to repopulate the pool of susceptibles via birth or waning immunity. The natural history of the pathogen and the 30

timescale at which susceptibles are reintroduced, therefore, dictate the timescale of oscillations in disease incidence. Opportunities for stochastic fade-out of viral populations, particularly in less densely populated areas, may arise during these interepidemic periods as well as during annual transmission minima. Time-series analyses of DENV transmission at a national level suggest that urban areas, which tend to show the highest genetic diversity year-round (Nisalak et al., 2003; Wu et al., 2009), may play a role in reseeding dengue to rural areas (Cummings et al., 2004; Rodriguez-Barraquer et al., 2011). Phylogeographic studies in Viet Nam support this hypothesis and suggest that a gravity model transmission pattern (Holmes, 2008) may dictate the dispersal of DENV to areas within and outside of the urban setting (Rabaa et al., 2010a;

Raghwani et al., 2011).

Locally, DENV transmission is highly spatially and temporally focal. Epidemiological studies in Latin America and Southeast Asia have indicated a heterogeneous risk of DENV infection at the neighborhood level in both urban and rural settings, with increased risk of infection when cases occurred within <1 km of the home (Endy et al., 2002b; Mammen et al.,

2008; Salje et al., 2012; Schmidt et al., 2011; Thai et al., 2010b). While it is difficult to decouple this risk from neighborhood features such as socioeconomic status and mosquito densities using epidemiological data, phylogenetic analyses indicate that these patterns of spatial clustering are generally caused by focal transmission of local viral populations (Balmaseda et al., 2010; Jarman et al., 2008; Schreiber et al., 2009), which may be maintained for several years even in rural environments (see Chapter 2).

Given the inherent complexity of the spatial dynamics of DENV transmission, it is critical to further characterize the spatial structure of DENV within endemic populations, the rate at which DENV lineages diffuse through space (particularly in the face of a partially immune population) and the relative contributions of the host, vector, and virus on the transmission of

DENV at various spatial scales.

Phylogenetic and phylogeographic analysis of DENV

The central topics in the evolutionary biology and epidemiology of dengue virus include:

1) its population history, including its viral and geographic origins, the processes by which the four serotypes diverged and independently entered the human population, the means and locations in which the viral populations persist in interepidemic periods, and the population level interactions of the four serotypes; 2) the basis for the present genetic diversity of DENV and the context and means by which new lineages emerge, and 3) the factors that determine disease outcomes and epidemic potential. Using phylogenetic methods, the lineage dynamics and aspects of the population history of DENV can be relatively easily inferred. The central challenge in using these methods is connecting the viral sequence data to disease data for the purposes of epidemiological inference, given that the four dengue serotypes and the diversity within them show complex immune interactions at both the human and viral population levels, which we are only now beginning to understand.

Dengue viruses evolve rapidly over time, generating extensive diversity detectable in nucleotide sequences while amino acid sequences remain largely conserved. The tempo of

DENV evolution means that viruses isolated only a few weeks or months apart may exhibit measurable genetic differences. Because the ecological and evolutionary dynamics of RNA viruses occur on the same approximate timescale, biologists can utilize sequence data to trace ecological processes and population dynamics that were not directly observed (Pybus and

Rambaut, 2009). Serially-sampled viral genetic sequence data contain important information about the phylogenetic relationships between viruses, times since divergence, signatures of past evolutionary events, and rates of population growth (Holmes, 2009). The distribution of branching events in a phylogeny and the characteristics associated with those events can be used to estimate features such as: to what degree a virus population has changed in size over time, the timing of viral origin or divergence events between strains, and the tempo and direction

of viral dispersal over space and time (Drummond et al., 2003, 2002, 2005; Lemey et al., 2009a;

Pybus and Rambaut, 2009).

Phylogenetic analysis

A number of models can be used to infer phylogenetic histories from a set of observed viral sequences, including parsimony, distance-based, maximum likelihood, and Bayesian estimations. These models are used to reconstruct the evolutionary history of a population of gene sequences based on sequence similarity in the form of a phylogenetic tree, with individual nodes on a tree representing a hypothetical most recent common ancestor of the observed gene sequences located at the tips of the tree. The simplest of these methods are the parsimony and distance-based methods. Parsimony methods are based on the principle that evolutionary change is rare, and therefore the most accurate tree minimizes the number of character changes required to explain differences among sequences (Felsenstein, 1982).

Distance-based methods, which include the widely-used neighbor-joining method, sequentially join least-distant pairs in order to minimize the sum of all branch lengths (Saitou and Nei, 1987).

Maximum likelihood methods, by contrast, choose the tree that best explains the evolutionary patterns of the sequence data given an explicit model of sequence evolution (Cavalli-Sforza et al., 1964; Felsenstein, 1973, 1981). Unlike the previously mentioned methods, this method allows for comparison of tree likelihoods to determine how well a competing set of trees fit a defined model (or set of models) of sequence evolution.

In these standard phylogenetic settings, the best-fit tree is determined based entirely on the available sequence data and the model of sequence evolution imposed on the data; only under ML methods are multiple sequence evolution models available. Closely related to ML method, however, is Bayesian inference, which brings together the likelihood (the probability of the data given the model parameters) and the prior (the probability of the model parameters 33

absent the data). Bayesian statistics involve a search for the optimal hypothesis that maximizes posterior probability (the probability of the hypothesis given the data) (Holder and Lewis, 2003).

To make complex hypothesis testing more tractable, Bayesian approaches generally rely on the

Markov Chain Monte Carlo (MCMC) algorithm to approximate posterior distributions. MCMC works by constructing a Markov chain and, with each step of the chain, creating perturbations in a hypothetical tree and accepting or rejecting the tree based on a Metropolis-Hastings probability (Hastings, 1970; Metropolis et al., 1953). This involves a search of the parameter space along a chain of millions of hypothetical trees and provides an estimate of the probability that a given tree bears the most correct phylogeny. The posterior probability of the tree then represents a combination of the prior probability of the tree given informed priors (based on biological knowledge, rather than the observed data) and the likelihood of the observed data given the evolutionary model. Hence, one in-built advantage of Bayesian phylogenetic methods is that they result in a (posterior) distribution of trees, rather than single representation of evolutionary history as in all other approaches.

Bayesian phylogenetic inference programs such as BEAST (Bayesian Evolutionary

Analysis by Sampling Trees, Drummond and Rambaut, 2007) utilize genetic sequence data and time-structured sampling information under a strict or relaxed evolutionary clock model to infer past population dynamics, including substitution rate, divergence times, and demographic growth, and returns a posterior set of rooted, time-structured trees. Using a Bayesian MCMC algorithm, the analysis returns the rate of nucleotide substitution and the time to the most recent common ancestor (TMRCA) at the root and at each of the nodes. The TMRCA is estimated through application of the coalescent theory, which also allows for the estimation of a variety of demographic parameters (Drummond et al., 2002). In contrast to traditional phylogenetic methods, which aim to recover evolutionary relationships among distinct ‘species’, coalescent inference considers the evolution of genes within a single species. The coalescent is based on the theory of neutral variation and genetic drift, such that any two alleles have an equal 34

probability of coalescing in each generation. The probability that any two alleles shared a common ancestor (coalesced) in the previous generation is 1/N for a haploid population, with coalescent events in successive generations occurring independently of prior generations. This process essentially reconstructs genetic drift in reverse time to determine the most recent common ancestor (MRCA) of a set of sequences – in the context of this study, a viral population. Importantly, coalescent inference can only be applied in the context of neutral evolution, when no evolutionary forces (such as natural selection) have acted to change the distribution of genetic variation (Rosenberg and Nordborg, 2002). Application of the coalescent to pathogens such as DENV is particularly valuable because it infers population history backwards in time and thus does not necessitate sampling of all individuals, but instead a randomly sampled population within the whole. From the coalescent, we can infer a host of other characteristics of viral populations, including changes in population size over time such as those occurring during epidemic spread or boom-bust periods of transmission, the timing of emergence of a novel strain, and the rate and direction of viral dispersal between geographic regions (Drummond et al., 2003, 2002; Lemey et al., 2009b; Pybus and Rambaut, 2009). As with other MCMC methods, BEAST estimates evolutionary and demographic parameters over a posterior sample of trees, thus incorporating phylogenetic and demographic uncertainty in tree estimation (Drummond et al., 2005).

Phylogeographic Methods

Important to the study of infectious disease dynamics, and central to this thesis, is an understanding of how pathogens move through and into populations. The timescale of epidemic spread matches well with that of RNA virus evolution, such that viral genomes carry accumulate informative mutations that, with accompanying metadata such as the dates and location of infection, carry important information on how viruses move through populations over 35

time (Holmes, 2008). Reconstructing evolutionary history and spatial dynamics from viral gene sequences provides key insights into the underlying evolutionary dynamics of epidemic and endemic disease spread, and thus can be used to inform prevention strategies (Rambaut et al.,

2008; Russell et al., 2008) and determine the source of and routes by which pathogens move into new populations (Lefeuvre et al., 2010; Lemey et al., 2009a; Talbi et al., 2010).

The evolution and development of methods for phylogeographic inference has followed a similar trajectory to that of the phylogenetic methods. Most models treat the geographic state of each taxon as its own discrete character (similar to that of a nucleotide or amino acid in the sequence alignment), in a manner independent of the molecular data, and utilize the evolutionary history to infer the ancestral geographic source location of a viral population at some time in the past. Under the simplest generalized parsimony approach, the reconstruction that requires the fewest changes in the geographic character state across the tree to explain the distribution of the locations at the tips of the tree is preferred (Maddison and Maddison, 2003).

Character state transitions may be weighted if there is reason to expect that some transitions may happen more frequently than others. Hypothesis testing under the parsimony model is problematic, as it offers no means of determining the probability of the estimated ancestral states (Nepokroeff et al., 2003; Nielsen, 2002). More importantly, this approach is vastly oversimplified because it disregards two important sources of error: the uncertainty associated with estimating ancestral state locations on a given tree and error in phylogenetic estimation of a single ‘best’ tree (Ronquist, 2004). This last source of error can be overcome by using a set of weighted trees rather than a single ‘best’ tree, thus incorporating topological uncertainty into the reconstruction, but still ignoring the uncertainty associated with reconstructing the geographic history. These methods are particularly unreliable when dispersal scenarios are complicated and involve frequent movement between areas – characteristics that are often associated with human pathogens.

Parametric approaches such as maximum likelihood (ML) (Mooers and Schluter, 1999;

Pagel, 1994, 1999) improve upon these models by utilizing an explicit stochastic model of evolution and branch length information to estimate the probability of ancestral state changes occurring along each branch. Given a tree topology, branch lengths, and the geographic distribution of the taxa, the ML approach determines the value of phylogeographic parameters that maximize the probability of observing the data. ML methods incorporate uncertainty in ancestral state reconstruction by evaluating all alternative reconstructions during the estimation of ancestral states (Pagel, 1994, 1999). However, again, ancestral state reconstructions are determined over a fixed tree topology with fixed branch lengths, and thus do not account for phylogenetic uncertainty (Nepokroeff et al., 2003).

As noted above, uncertainty is best incorporated by methods of Bayesian phylogeographic inference through random sampling from the posterior distribution of trees and other parameters using the MCMC algorithm. To model changes in geographic location state

(dispersal), discrete-state continuous time Markov chain models are utilized in a manner mathematically equivalent to those used to model DNA evolution (Sanmartin et al., 2008). This facilitates the estimation of migration rates between pairs of locations and estimates ancestral locations for nodes in the tree. Dominant migration routes within the data set can be inferred across the full distribution of trees by employing Bayesian variable selection within BEAST

(Lemey et al., 2009a), which also provides model averaging over uncertainty in the connectivity between locations and host populations, and allows for the imposition of priors on migration rate matrices for the testing of hypotheses related to virus dispersal.

Aims of thesis

Understanding the evolutionary and epidemiological dynamics of dengue viruses is critical to determining the underlying factors that make them so successful in invading and becoming established in new populations, and the means by which they persist following establishment. Despite considerable progress made in understanding the spatial and temporal dynamics of DENV transmission over the last decade, epidemiological data alone fail to capture the dispersal of viral lineages over space and time. While these data are integral to the implementation of control methods and health systems management during epidemics, a lack of understanding of the way the virus moves, rather than the disease, may pose a significant danger in the context of phase III vaccine trials that have recently been initiated, and the likely eventual rollout of the tetravalent vaccine. In the event that a vaccine escape occurs or a novel strain emerges from the sylvatic cycle, it is important to understand how the virus may spread in order to prevent it from occurring.

Investigation of the dispersal of DENV on local, regional, and global scales provides insight into the process by which DENV transmission periodically transitions from endemic to epidemic, as well as the scenarios in which invasion and establishment are permitted in non- endemic populations. The latter of these is particularly important if we are to predict and combat the future spread of dengue to new populations, a process that is clearly ongoing, as this thesis demonstrates. An understanding of the specific viral populations circulating in an area and locations from which they entered may also be important to epidemiological investigations of viral interactions with the host immune system, the vector, and the population- level immune landscape. More generally, the knowledge gained by this work can be applied to investigate the spread of other vector-borne pathogens and the effects of host movement on pathogen dispersal.

The combination of large-scale full genome and E gene sequencing efforts with improved computational and methodological developments in recent years provides an important and timely opportunity to investigate the phylogeography of DENV and identify aspects of DENV dynamics that may necessitate more effort, resources, and the collection of more data to fully understand. The overall aim of this thesis is, therefore, to characterize the spatial and temporal spread of DENV at various spatial scales and investigate the role of human movement in its dispersal. The four chapters that follow move in the direction of spatial spread, from local rural populations, to urban and suburban, national/regional, and finally, to the global level. The final chapter provides a synthesis of these studies and suggests topics that warrant further investigation, as well as data that would be needed to address remaining questions.

CHAPTER 2

FREQUENT IN-MIGRATION AND HIGHLY FOCAL TRANSMISSION OF DENGUE VIRUSES AMONG CHILDREN IN KAMPHAENG PHET, THAILAND

Abstract

Understanding the microevolution of dengue virus (DENV) is essential to the deployment of successful intervention strategies. We performed a phylogenetic analysis of the envelope (E) genes of DENV-1, -2, -3, and -4 isolates collected from school-based cohort and village-based cluster studies in Kamphaeng Phet, Thailand, between 2004 and 2007 to describe the spatial and temporal patterns of DENV transmission within a rural population where a future vaccine efficacy trial was planned. Our analysis revealed considerable genetic diversity within the study population, with multiple lineages within each serotype circulating for various lengths of time during the study period. These results suggest that DENV is frequently introduced into both semi-urban and rural areas in Kamphaeng Phet from other populations. In contrast, the persistence of viral lineages across sampling years was observed less frequently. Analysis of phylogenetic clustering indicated that DENV transmission was highly spatially and temporally focal and that DENV transmission generally occurred in homes rather than at school. Overall, the strength of temporal clustering suggests that seasonal bottlenecks in local DENV populations facilitate the invasion and establishment of viruses from outside of the study area, in turn reducing the extent of lineage persistence.

Introduction

Dengue is the leading cause of mosquito-borne viral disease worldwide, and dengue fever (DF) and dengue hemorrhagic fever (DHF) continue to increase in both incidence and geographic range. Recent estimates are that over 50 million DENV infections occur each year, including 500,000 hospitalizations for DHF, primarily among children (Guzman et al., 2010;

WHO, 2007). Dengue viruses (DENV) are single-stranded, positive-sense RNA viruses (family

Flaviviridae, genus Flavivirus) that are comprised of four antigenically distinct serotypes (or viruses; DENV-1, DENV-2, DENV-3, and DENV-4) that co-circulate in many endemic areas in the tropics and sub-tropics. The phenomenon of co-circulation of multiple DENV serotypes is referred to as hyperendemicity, and is believed to increase the risk of severe disease in a population (Burke et al., 1988; Kliks et al., 1988; Sangkawibha et al., 1984). Hyperendemicity has also complicated vaccine development, and as yet there is no commercially available, licensed vaccine, although a number of vaccine prospects are currently being investigated in clinical trials (Webster et al., 2009; Wilder-Smith et al., 2010).

An understanding of the spatial and temporal patterns of DENV transmission at various scales is crucial to the design and implementation of vaccine trials and future immunization efforts, and is central to determining the factors responsible for both the emergence and the persistence of DENV. Molecular epidemiological studies of global DENV populations play a key role in understanding the mechanisms of DENV transmission by elucidating critical aspects of epidemiological history, including virus population growth and decline, lineage replacement events, patterns of spatial migration, and rates of evolutionary change. The growing data base of full and partial DENV genome sequences has resulted in several detailed phylogeographic studies, but the focus of these has generally been on endemic and epidemic transmission in urban and semi-urban areas (Rabaa et al., 2010a; Raghwani et 41

al., 2011; Schreiber et al., 2009). To date, few studies have examined long-term DENV transmission dynamics in endemic rural areas, where challenges to vector control and vaccine implementation may differ from those in more densely populated regions due to environment, social factors, and public health infrastructure.

The prospect of vaccine trials and eventual large-scale vaccination to combat dengue necessitates an examination of the dynamics of endemic DENV transmission and spread at various scales over multiple years. Long-term cohort studies of DENV infection in children in rural Kamphaeng Phet, Thailand, have greatly increased our understanding of the epidemiology and evolution of DENV in an endemic rural area, and provide a unique context to assess DENV transmission dynamics in rural areas (Endy et al., 2002a, 2002b, 2011; Gibbons et al., 2007; Jarman et al., 2008; Mammen et al., 2008; Yoon et al., 2012; Zhang et al., 2006).

In this study we examined the spatial and temporal patterns of DENV transmission using sequence data from isolates obtained over several years in a longitudinal cohort undergoing school absence-based surveillance and a concurrent geographic cluster study (Mammen et al.,

2008; Yoon et al., 2012). The objectives of this study were to characterize the genetic diversity of DENV circulating within a rural population and to investigate whether DENV transmission was spatially and temporally focal within this area of stable hyperendemic transmission.

Characterizing the genetic diversity and patterns of transmission in this population is particularly relevant given that Kamphaeng Phet is a planned site for upcoming dengue vaccine trials.

Results and Discussion

Microevolution of DENV in Kamphaeng Phet over four dengue seasons

To investigate the genetic diversity and structure of DENV populations circulating in

Kamphaeng Phet (Figure 2-1) from 2004 to 2007, we sequenced and analyzed the E genes of viruses collected during school-based surveillance and geographically-based community cluster studies. At least three serotypes circulated in our study population during each year of the study, with all four serotypes detected only in 2005. This pattern generally reflects the relative proportions of DENV isolated through passive surveillance across the region during the period of sampling, although passive surveillance data from the Kamphaeng Phet Provincial Hospital detected all four serotypes circulating each year (Figure 2-1). Years in which a given serotype was not detected in our study population corresponded to instances in which that serotype was detected at less than 10% frequency during passive surveillance. This suggests that our sampling and sequencing protocol adequately captures the diversity of the viral population circulating in KPP in a given year.

400 350 300 250 2007 200 2006 150 2005 100 2004 50 0 DENV1 DENV2 DENV3 DENV4

Figure 2-1. Kamphaeng Phet Provincial Hospital data of all reported and PCR confirmed dengue cases from Muang District, Kamphaeng Phet during 2004 to 2007. Data collected from this study are included.

Only a single genotype was represented for each serotype in this data set (DENV-1, genotype I; DENV-2, Asian I genotype; DENV-3, genotype II; DENV-4, genotype I). These four genotypes have been the dominant circulating genotypes in Thailand and much of mainland

Southeast Asia for several years, and were all present among previous Thai E gene sequences from Thailand dating back to 2001 and earlier. Although other genotypes were detected in

Thailand prior to our study, only DENV-1 genotype I, DENV-2 Asian I genotype, and DENV-3 genotype II were isolated in a study performed in Kamphaeng Phet in 2001; no DENV-4 sequences were isolated during the 2001 study (Jarman et al., 2008). Thus, it appears that these were the dominant circulating genotypes in Kamphaeng Phet, at least through 2007.

Phylogenetic analysis revealed considerable viral genetic diversity within genotypes in the Kamphaeng Phet region during this period, demonstrated by the presence of phylogenetically distinct lineages within each serotype. Among the DENV-1 isolates detected, five major lineages were present, suggestive of at least five separate virus introductions prior to or during the sampling period (Figures 2-2 and 2-3). One lineage was detected briefly in 2005, but was not observed in the following years. Three lineages circulated in 2006, the year in which DENV-1 incidence greatly increased in the region (Figure 2-1). All of these lineages persisted into the 2007 season, during which one additional lineage was detected. With the exception of Lineage 4, these viral populations were most closely related to viruses isolated in

Thailand in the previous decade. Lineage 4 may have been a recent introduction from

Singapore (the most closely related geographical region on the phylogenetic tree) that invaded

Na Bo Kham sub-district and persisted within that population for at least two years. In all of the lineages except Lineage 2, viruses commonly mixed among sub-districts within a single season.

Despite the small number of sequences obtained, the DENV-2 population also showed considerable genetic diversity. Four lineages were detected between 2004 and 2007, with little mixing detected among sub-districts within a given season (Figures 2-2 and 2-4). Interestingly, however, Lineage 2 was first detected in rural Na Bo Kham in 2005, and was not isolated again 44

until in 2007 in the most cosmopolitan of our populations, Nong Pling. This suggests that in this environment viral populations may sometimes move from rural to more densely populated areas rather than the opposite, as has been reported for other locations (Rabaa et al., 2010a;

Raghwani et al., 2011).

Kon Tee Na Bo Kam Nakon Chum Nong Pling Thep Na Korn

Lineage 1

Lineage 2 DENV-1 Lineage 3 Lineage 4 Lineage 5 Lineage 1 Lineage 2 DENV-2 Lineage 3 Lineage 4 Lineage 1 DENV-3 Lineage 2

Lineage 1

DENV-4 Lineage 2 Lineage 3 Lineage 4 2004 2005 2006 2007

Figure 2-2. The timing of isolation of DENV lineages circulating in each of five participating sub-districts of Kamphaeng Phet, Thailand during 2004 – 2007. Horizontal bars indicate the time period in which specified lineages were isolated during school-based surveillance (conducted from June to November). Vertical grey bars indicate periods of the year in which cohort sampling was inactive (December to May).

Although DENV-3 exhibited the lowest incidence in the study population during this time period, one lineage was detected in 2004 and a different lineage was detected in 2005 (Figures

2-3 and 2-6). Both of these were derived from lineages that were circulating in Thailand in the previous decade, which suggests that rare lineages may persist in rural populations even when disease incidence is low.

DENV-4 was the dominant serotype detected in Kamphaeng Phet during the first two years of this study. Two lineages were present in 2004; one was represented by a single isolate detected in Na Bo Kham (Lineage 3), and the other was represented by three distinct, primarily 45

sub-district-specific clusters, showing limited mixing in a given year (Lineage 1) (Figures 2-3 and

2-7). Lineage 1 appeared to persist as the dominant DENV-4 lineage in the population throughout the study, although multiple distinct populations were present through 2007. Two additional introductions were observed in the Na Bo Kham and Nakon Chum sub-districts in

2006 (Lineage 4 and Lineage 2, respectively). Based on the sequences in this study, these introductions did not appear to result in extended transmission in the region. Interestingly, while most lineages detected appeared to have entered following long-term transmission within

Thailand, a single sequence isolated in Na Bo Kham in 2006 (Lineage 4) may have entered through the re-introduction of a Thai lineage following extended transmission elsewhere in

Southeast Asia.

It is interesting that many of recent introduction events occurred in Na Bo Kham sub- district; this is the most rural of the populations studied and more distantly positioned and isolated relative to the other study populations. A previous study performed in Kamphaeng Phet during a single, high incidence dengue season detected the greatest genetic diversity among viruses in the most densely populated areas (Jarman et al., 2008), while the current study indicates the opposite. Because most of these lineages did not appear to immediately move into the greater population of Kamphaeng Phet, it is possible that the DENV populations present in Na Bo Kham differ greatly from the rest of the area because there is less mixing with the other populations we sampled. Alternatively, Na Bo Kham may depend on a different population center through which viral variants are introduced. Differences among DENV from

Na Bo Kham and the rest of the sub-districts studied indicate that geographically distinct patterns of human movement are important processes in the structure and dynamics of DENV populations (Stoddard et al., 2009).

* Cambodia 2003 - 2008 Viet Nam 2006 - 2008

Thailand|KPP|B1337_06|KT02|4-35|M|MQ|2006 Thailand|KPP|KDS06222|KT02|4-35|I|10|2006 Thailand|KPP|KDS06258|KT02|Scho|S|10|2006 Thailand|KPP|KDS06263|KT13|Scho|S|10|2006 Thailand|KPP|KDS06283|KTxx|Scho|S|10|2006 Thailand|KPP|KDV06233|KT02|4-35|V|--|2006 Thailand|KPP|KDV06251|KT02|4-35|V|10|2006 Thailand|KPP|B0819_06|NP07|4-16|M|MQ|2006 Thailand|KPP|KDS04142|KT13|Scho|S|10|2006 Kon Tee Thailand|KPP|KDS04202|NP07|Scho|S|08|2006 Thailand|KPP|KDS04259|NP07|Scho|S|08|2006 Thailand|KPP|KDS04261|KT13|Scho|S|10|2006 Na Bo Kham Thailand|KPP|KDS04499|NP07|Scho|S|08|2006 Thailand|KPP|KDS04586|NP07|Scho|S|07|2006 Thailand|KPP|KDS04620|KT13|4-14|I|10|2006 Thailand|KPP|KDS04792|NP07|4-16|I|07|2006 Nakon Chum Thailand|KPP|KDS05170|KT03|Scho|S|10|2006 Thailand|KPP|KDS05261|KT13|Scho|S|10|2006 Thailand|KPP|KDS05365|KT03|Scho|S|10|2006 Nong Pling Thailand|KPP|KDS05368|KT02|Scho|S|10|2006 Thailand|KPP|KDS05484|NPxx|Scho|S|08|2006 Thailand|KPP|KDS05487|KT02|4-22|I|10|2006 Thep Na Korn * Thailand|KPP|KDS05558|KT02|Scho|S|10|2006 Thailand|KPP|KDS05572|KT03|Scho|S|10|2006 Thailand|KPP|KDS05706|NP07|4-28|I|08|2006 Thailand|KPP|KDS05728|KT03|Scho|S|10|2006 Thailand|KPP|KDS05769|KT03|Scho|S|10|2006 Thailand|KPP|KDS05978|KT02|Scho|S|10|2006 Thailand|KPP|KDS06006|TN04|Scho|S|04|2006 Thailand|KPP|KDS06010|NP02|4-33|I|07|2006 Thailand|KPP|KDS06043|KT03|Scho|S|10|2006 Thailand|KPP|KDS06160|KT13|Scho|S|10|2006 Thailand|KPP|KDV04056|NP07|4-07|V|0|2006 Thailand|KPP|KDV04267|KT02|4-09|V|10|2006 Thailand|KPP|KDV04539|KT02|4-09|V|10|2006 Thailand|KPP|KDV04804|KT02|4-15|V|10|2006 Lineage 1 Thailand|KPP|KDV04824|NP07|4-16|V|07|2006 Thailand|KPP|KDV05510|KT02|4-22|V|10|2006 Thailand|KPP|KDS06892|NC05|5-10|I|11|2007 Thailand|KPP|KDS07380|NC06|5-17|I|11|2007 Thailand|KPP|KDS07460|NC05|5-19|I|11|2007 Thailand|KPP|KDS07788|NC03|Scho|S|11|2007 * Thailand|KPP|KDV07392|NC06|5-17|V|11|2007 Thailand|KPP|KDV07393|NC06|5-17|V|11|2007 Thailand|KPP|KDV07396|NC06|5-17|V|11|2007 Thailand|KPP|KDV07398|NC06|5-17|V|0L|2007 Thailand|KPP|B0161_07|TN04|5-06|M|MQ|2007 Thailand|KPP|KDS06331|KT11|4-37|I|10|2006 Thailand|KPP|KDS06553|NP02|5-04|I|07|2007 Thailand|KPP|KDS06560|TN04|5-06|I|04|2007 Thailand|KPP|KDV06342|KT11|4-37|V|10|2006 Thailand|KPP|KDV06346|KT11|4-37|V|0A|2006 Thailand|KPP|KDV06404|KT11|4-37|V|10|2006 Thailand|KPP|KDS06959|NP03|Scho|S|07|2007 * Thailand|KPP|KDV08490|NC04|5-28|V|0F|2007 Thailand|KPP|KDS07060|NBxx|Scho|S|01|2007 Thailand|ThD1_0116_97|1997 Thailand|KPP|B0388_07|NB05|5-13|M|MQ|2007 Thailand|KPP|KDS08452|NC04|5-28|I|11|2007 * Thailand|KPP|KDS08532|NC04|Scho|S|11|2007 Thailand|KPP|KDV08491|NC04|5-28|V|0L|2007 Thailand|KPP|B0078_06|NC05|4-01|M|MQ|2006 Thailand|KPP|KDS03780|NC05|4-01|I|11|2006 Thailand|KPP|KDS03888|NC05|Scho|S|11|2006 Thailand|KPP|KDS04173|NC05|Scho|S|11|2006 * Thailand|KPP|KDS04175|NCxx|Scho|S|11|2006 Thailand|KPP|KDS04263|NCxx|Scho|S|11|2006 Lineage 2 Thailand|KPP|KDS04285|NC05|4-10|I|11|2006 Thailand|KPP|KDS04975|NP06|Scho|S|08|2006 Thailand|KPP|KDV03941|NC05|4-01|V|11|2006 Thailand|KPP|KDV04339|NC05|4-10|V|0F|2006 * Thailand|KPP|KDS06471|TN14|Scho|S|06|2007 Thailand|KPP|KDS06523|TN14|Scho|S|06|2007 * Thailand|KPP|KDS03779|NCxx|Scho|S|11|2006 Thailand|KPP|KDS05366|KTxx|Scho|S|10|2006 Vietnam|BID_V4087|2008 Vietnam|BID_V1839|2007 * HM181952|Cambodia|BID_V4253|2007 Vietnam|BID_V817|2006 Thailand|ThD1_0075_02|2002 Thailand|KPP|KDS02944|NP06|3-08|I|08|2005 * Thailand|KPP|KDV02921|NP06|3-05|V|--|2005 Lineage 3 Thailand|ThD1_0049_01|2001 Thailand|KPP|01A00626|2001 Thailand|KPP|01A00529|2001 Thailand|KPP|01A00463|2001 * Thailand|KPP|01A00289|2001 Thailand|KPP|01A00082|2001 GQ357689|Singapore|EHI_DED75807|2007 Thailand|KPP|ThD1_K0851_01|2001 Thailand|KPP|ThD1_K0080_01|2001 * Thailand|ThD1_0141_00|2000 Thailand|KPP|ThD1_K0107_98|1998 Thailand|ThD1_0726_99|1999

Singapore 2003 - 2008

Thailand|KPP|B1074_06|NB13|4-20|M|MQ|2006 Thailand|KPP|KDS05361|NB13|4-20|I|02|2006 * Thailand|KPP|KDS05362|NB13|Scho|S|02|2006 Thailand|KPP|KDS05498|NB17|4-23|I|03|2006 Thailand|KPP|KDS04965|TN15|4-18|I|04|2006 EU069598|Singapore|S159_03|2003 Thailand|KPP|B0341_07|NB05|5-12|M|MQ|2007 Thailand|KPP|B0344_07|NB05|5-12|M|MQ|2007 Thailand|KPP|B0348_07|NB05|5-12|M|MQ|2007 Thailand|KPP|KDS06958|NB05|5-12|I|03|2007 * Thailand|KPP|KDS06985|NB05|Scho|S|03|2007 Lineage 4 Thailand|KPP|KDS07135|NB05|5-14|I|03|2007 Thailand|KPP|KDS07826|NB05|5-23|I|03|2007 Thailand|KPP|KDV07912|NB05|5-23|V|0E|2007 Thailand|KPP|KDS08156|NB17|5-25|I|03|2007 * Thailand|KPP|KDV08195|NB17|5-25|V|03|2007 Thailand|KPP|KDV08204|NB17|5-25|V|0E|2007 EU069610|Singapore|S209_03|2003 EU069618|Singapore|T3156_04|2004

Viet Nam 2003 - 2008

Myanmar 2000 - 2002 Thailand 2001

Thailand|ThD1_0762_97|1997 Thailand|ThD1_0277_97|1997 Thailand|KPP|B0007_07|NB13|5-01|M|MQ|2007 Thailand|KPP|KDS06463|NB13|5-01|I|02|2007 * Thailand|KPP|KDS06498|NB13|5-03|I|02|2007 Thailand|KPP|KDS06549|NB13|Scho|S|02|2007 Lineage 5 Thailand|KPP|KDV06528|NB13|5-03|V|02|2007 Thailand|ThD1_0431_94|1994 Thailand|ThD1_0134_00|2000 * Thailand|ThD1_0175_02|2002 Thailand|ThD1_0002_95|1995 Thailand|ThD1_0067_99|1999 Thailand|ThD1_0289_97|1997 0.0040

Figure 2-3. ML tree of representative E gene sequences of DENV-1 genotype I. Districts of Kamphaeng Phet are designated by different colors. Bootstrap support values ≥85% are indicated by an asterisk next to the node. The tree is midpoint rooted for purposes of clarity. 47

* Viet Nam 2003 - 2009

Thailand|KPP|a12|01A00262|2001 Thailand|BID_V2310|2001 Thailand|KPP|KDV07275|NP07|5-15|V|--|2007 Thailand|KPP|KDS06887|NP03|Scho|S|07|2007 * Thailand|KPP|KDS06722|NP07|5-08|I|08|2007 Thailand|KPP|KDS07259|NP07|5-15|I|08|2007 Thailand|KPP|B0414/07|NP07|5-15|M|MQ|2007 Thailand|KPP|KDS06986|NP03|Scho|S|07|2007 Thailand|KPP|KDS07288|NP03|Scho|S|07|2007 Lineage 2 * Thailand|KPP|KDS07291|NP03|Scho|S|07|2007 Thailand|KPP|KDS07286|NP02|Scho|S|07|2007 * Thailand|KPP|KDS07564|NP02|5-21|I|07|2007 Thailand|KPP|KDV07630|NP02|5-21|V|07|2007 Thailand|KPP|KDS02933|NB02|3-06|I|02|2005 * Thailand|KPP|B0731/05|NB02|3-06|M|MQ|2005 Thailand|KPP|KDS00858|TN02|Scho|S|06|2004 Thailand|KPP|KDS00305|NCxx|Scho|S|11|2004 Thailand|KPP|KDS00725|NC06|Scho|S|11|2004 * Thailand|KPP|KDS01228|NC06|Scho|S|11|2004 Lineage 3 * Thailand|KPP|KDS01296|NC06|Scho|S|11|2004 Thailand|KPP|KDS01527|NC06|Scho|S|11|2004 Thailand|D2_Hu_Thailand_NIID20_2004|2004 Cambodia|D2_Cambodia_0708ATw|2007 Thailand|D2_Thailand_0606ATw|2006 Thailand|0408A_Tw|2004 * Cambodia 2007 - 2008 * Viet Nam 2009

Singapore|EHI.DED117008|2008 Thailand|Th_D2_0078_01|2001 Thailand|Th_D2_0981_00|2000 Thailand|KPP|a08|01A00105|2001 Thailand|BID_V2298|2001 Thailand|KPP|a08|01A00063|2001 Thailand|KPP|a08|01A00177|2001 Thailand|KPP|a08|01A00251|2001 Thailand|KPP|a08|01A00465|2001 Thailand|KPP|a08|01A00515|2001 Thailand|KPP|a09|01A00589|2001 * Thailand|KPP|a09|200105079|2001 Thailand|BID_V2293|2001 Thailand|BID_V2294|2001 Thailand|BID_V2295|2001 Thailand|BID_V2296|2001 Thailand|BID_V2297|2001 Thailand|BID_V2299|2001 Cambodia|BID_V2044|2005 * Cambodia|BID_V2045|2005 * GU131927|Cambodia|2007 * Thailand|KPP|KDS00697|NB02|2-06|I|02|2004 Lineage 4 Thailand|KPP|a02|01A00275|2001 * Thailand|BID_V2283|2001 Cambodia|BID_V2036|2003 0.0020

Figure 2-4. ML tree of representative E gene sequences of the Asian I genotype of DENV-2. Districts of Kamphaeng Phet are designated by different colors. Bootstrap support values ≥85% are indicated by an asterisk next to the node. The tree is midpoint rooted for purposes of clarity.

* Cambodia 2003 - 2008

FJ639727|Cambodia|2005 FJ639726|Cambodia|2004 FJ639724|Cambodia|2003 HM181934|Cambodia|2006 HM181933|Cambodia|2006 GU131940|Cambodia|2007 GU131911|Cambodia|2006 GU131910|Cambodia|2006 GU131909|Cambodia|2006 GU131908|Cambodia|2006 GQ868634|Cambodia|2006 GU131935|Cambodia|2007 FJ639731|Cambodia|2007 FJ639730|Cambodia|2006 Nakon Chum GU131942|Cambodia|2007 * FJ639712|Cambodia|2007 GU131915|Cambodia|2007 GU131903|Cambodia|2008 Thep Na Korn GU131944|Cambodia|2007 GU131943|Cambodia|2007 EU448440|Cambodia|0609aTwxx|2006 GQ868628|Cambodia|2005 GU131937|Cambodia|2007 * FJ639728|Cambodia|2005 GQ868629|Cambodia|2005 FJ639729|Cambodia|2006 GU131946|Cambodia|2008 GU131933|Cambodia|2006 GU131904|Cambodia|2005 GU131906|Cambodia|2003 FJ639725|Cambodia|2003 GU131916|Cambodia|2007 GU131905|Cambodia|2008

* Viet Nam 2005 - 2008

DQ518656|Vietnam|0409aTwxx|2004 EU448442|Vietnam|0310aTwxx|2003 DQ518655|Vietnam|9609aTwxx|1996 DQ518654|Vietnam|9809aTwxx|1998 EU482457|Vietnam|2006 EU482454|Vietnam|2006 AY676410|Thailand|0835_01xx|2001 AY676402|Thailand|1959_01xx|2001 AY676399|Thailand|0377_98xx|1998 AY676388|Thailand|0092_98xx|1998 AY676414|Thailand|0989_00xx|2000 AB111083|Thailand|417HuNIID|2000 AB111082|Thailand|40HuNIIDx|2000 AY676383|Thailand|0328_02xx|2002 FN429909|Malaysia|25850xxxx|2002 FN429908|Malaysia|25811xxxx|2002 DQ518664|Thailand|0211aTwxx|2002 * AY676406|Thailand|1283_98xx|1998 AY676390|Thailand|0940_98xx|1998 AY676411|Thailand|0723_99xx|1999 AY676407|Thailand|1094_01xx|2001 FN429905|Malaysia|22366xxxx|2000 AY676400|Thailand|0343_98xx|1998 GQ868627|Cambodia|2002 GQ868626|Cambodia|2001 FJ639723|Cambodia|2003 FJ639722|Cambodia|2002 * FJ639721|Cambodia|2002 FJ639720|Cambodia|2001 FJ639719|Cambodia|2000 AY676398|Thailand|0411_97xx|1997 AY676364|Thailand|0006_97xx|1997 AY676405|Thailand|1309_97xx|1997 AY676404|Thailand|1465_97xx|1997 AY676397|Thailand|0436_97xx|1997 AY676396|Thailand|0514_98xx|1998 AY676393|Thailand|0808_98xx|1998 AY676395|Thailand|0546_98xx|1998 AY145729|Thailand|D97_0144x|1997 AY676366|Thailand|0195_94xx|1994 AY145727|Thailand|D96_330xx|1997 AY912458|Thailand|0657_207x|1998 DQ518662|Taiwan|9811aTwxx|1998 DQ518661|Thailand|9709aTwxx|1997 AY676365|Thailand|0005_96xx|1996 Thailand|KPP|KDS00998|NC06|Scho|S|11|2004 DQ518663|Thailand|9807aTwxx|1998 Lineage 1 AB111084|Thailand|17HuNIIDx|1996 * AY676373|Thailand|0123_95xx|1995 AY145724|Thailand|D95_0014x|1995 AY145718|Thailand|D92_423xx|1992 AY145723|Thailand|D94_283xx|1995 AY676386|Thailand|0240_92xx|1992 AY145720|Thailand|D93_044xx|1993 Thailand|KPP|KDS03458|TN15|3-09|I|04|2005 Thailand|KPP|B0924_05|TN15|3-09|M|MQ|2005 * Thailand|KPP|B0907_05|TN15|3-09|M|MQ|2005 * Thailand|KPP|KDV03472|TN15|3-09|V|04|2005 Lineage 2 * DQ518660|Thailand|0308aTwxx|2003 AY676408|Thailand|1017_00xx|2000 AY676413|Thailand|0595_99xx|1999 AY676387|Thailand|0115_99xx|1999 AY676377|Thailand|0058_97xx|1997 AY676412|Thailand|0650_97xx|1997 * AY676403|Thailand|1687_98xx|1998 AY676391|Thailand|0903_98xx|1998

0.0020

Figure 2-5. ML tree of representative E gene sequences of DENV-3 genotype II. Districts of Kamphaeng Phet are designated by different colors. Bootstrap support values ≥85% are indicated by an asterisk next to the node. The tree is midpoint rooted for purposes of clarity.

Thailand|KPP|B0434_04|NB13|2-04|M|MQ|2004 Thailand|KPP|B0437_04|NB13|2-04|M|MQ|2004 Thailand|KPP|KDS00397|NB13|2-04|I|02|2004 Thailand|KPP|KDS00507|NB13|Scho|S|02|2004 Thailand|KPP|KDS00818|NB13|2-08|I|02|2004 * Thailand|KPP|KDS01280|NBxx|Scho|S|01|2004 Thailand|KPP|KDS01419|NB02|Scho|S|02|2004 Thailand|KPP|KDV00437|NB13|2-04|V|02|2004 Thailand|KPP|KDV00443|NB13|2-04|V|02|2004 Thailand|KPP|KDV00647|NB13|2-04|V|02|2004 Thailand|KPP|KDV00704|NB13|2-04|V|--|2004 Thailand|KPP|KDS05476|TNxx|Scho|S|05|2006 Thailand|KPP|KDS05704|TN03|Scho|S|05|2006 Thailand|KPP|KDS05727|TN03|Scho|S|05|2006 Thailand|KPP|KDS05787|TN16|Scho|S|05|2006 * Thailand|KPP|KDS05849|TN03|4-30|I|05|2006 Thailand|KPP|KDS05892|TN03|4-31|I|05|2006 Thailand|KPP|KDV05931|TN03|4-31|V|0C|2006 Kon Tee Thailand|KPP|KDV06137|TN03|4-31|V|0B|2006 Thailand|KPP|B0161_06|NP03|4-03|M|MQ|2006 Thailand|KPP|KDS03808|NP03|4-03|I|07|2006 Na Bo Kham Thailand|KPP|KDS04496|NPxx|Scho|S|07|2006 Thailand|KPP|KDV03839|NP03|4-03|V|07|2006 Thailand|KPP|KDV04109|NP03|4-03|V|0D|2006 Nakon Chum Thailand|KPP|KDS02514|NP02|Scho|S|07|2005 Thailand|KPP|KDS02585|NP02|Scho|S|07|2005 Thailand|KPP|KDS02827|NP02|Scho|S|07|2005 Nong Pling Thailand|KPP|KDS03028|NP02|Scho|S|07|2005 Thailand|KPP|KDS03106|NP03|Scho|S|07|2005 * Thailand|KPP|KDS03174|NP03|Scho|S|07|2005 Thep Na Korn Thailand|KPP|KDS03224|NP02|Scho|S|07|2005 Thailand|KPP|KDS03225|NP03|Scho|S|07|2005 Thailand|KPP|KDS03244|NP02|Scho|S|07|2005 Thailand|KPP|KDS03255|NP03|Scho|S|07|2005 Thailand|KPP|KDS03275|NP02|Scho|S|07|2005 Thailand|KPP|KDS03445|NP02|Scho|S|07|2005 Lineage 1 Thailand|KPP|KDS06599|NB15|Scho|S|03|2007 * Thailand|KPP|KDS08438|NB02|5-27|I|02|2007 Thailand|KPP|KDS03903|TN02|4-05|I|06|2006 Thailand|KPP|KDV03935|TN02|4-05|V|0J|2006 Thailand|KPP|KDV04233|TN14|4-06|V|06|2006 Thailand|KPP|B1167_06|NP03|4-26|M|MQ|2006 Thailand|KPP|KDS03489|NP07|Scho|S|08|2005 Thailand|KPP|KDS03625|NP07|Scho|S|08|2005 Thailand|KPP|KDS03653|NP07|Scho|S|07|2005 Thailand|KPP|KDS04000|NP07|4-07|I|07|2006 Thailand|KPP|KDS04087|NP07|Scho|S|08|2006 Thailand|KPP|KDS04404|NC03|Scho|S|11|2006 * Thailand|KPP|KDS05144|NP02|Scho|S|07|2006 Thailand|KPP|KDS05640|NP03|4-26|I|07|2006 Thailand|KPP|KDV04048|NP07|4-07|V|07|2006 Thailand|KPP|KDV04053|NP07|4-07|V|0K|2006 Thailand|KPP|KDV04054|NP07|4-07|V|0G|2006 * Thailand|KPP|KDV04058|NP07|4-07|V|0|2006 * Thailand|KPP|KDV04061|NP07|4-07|V|0G|2006 Thailand|KPP|KDV04062|NP07|4-07|V|08|2006 Thailand|KPP|KDV04066|NP07|4-07|V|0I|2006 Thailand|KPP|KDV04366|NP07|4-07|V|0|2006 EU448455|Cambodia|2005 Thailand|KPP|B1521_xx|NB13|2-16|M|MQ|2004 Thailand|KPP|B1523_xx|NB13|2-16|M|MQ|2004 Thailand|KPP|KDS01385|NB13|2-16|I|02|2004 Thailand|KPP|KDS01744|NB13|2-21|I|02|2004 * Thailand|KPP|KDS00265|NB02|2-01|I|02|2004 Thailand|KPP|KDS00427|NB17|Scho|S|03|2004 EU448454|Thailand|2007 Thailand|KPP|KDS06353|NPxx|Scho|S|07|2006 * EU448456|Thailand|2005 Thailand|KPP|KDS00759|NC06|Scho|S|11|2004 Thailand|KPP|KDS00779|NC06|Scho|S|11|2004 Thailand|KPP|KDS00862|KT02|2-10|I|10|2004 Thailand|KPP|KDS00996|NP06|Scho|S|08|2004 * Thailand|KPP|KDS01056|NP11|Scho|S|08|2004 Thailand|KPP|KDS01061|NP11|2-12|I|08|2004 EU448457|Thailand|2003 AY618942|Thailand|0485_01|2001 AY618947|Thailand|0438_02|2002 AY618935|Thailand|0274_00|2000 AY618945|Thailand|0352_02|2002 AY618936|Thailand|0319_00|2000 AY618941|Thailand|0440_01|2001 * AY618938|Thailand|0759_00|2000 * AY618944|Thailand|2014_01|2001 AY618939|Thailand|0761_00|2000 AY618946|Thailand|0357_02|2002 AY618948|Thailand|0501_02|2002 AY618943|Thailand|1218_01|2001 Thailand|KPP|KDS05979|NC04|4-32|I|11|2006 AY618982|Thailand|1448_98|1998 Lineage 2 AY618985|Thailand|0152_99|1999 AY618987|Thailand|0521_99|1999 Thailand|KPP|B1379_04|NB13|2-04|M|MQ|2004 AY618968|Thailand|0358_92|1992 AY618973|Thailand|0600_94|1994 Lineage 3 AY618974|Thailand|0100_95|1995 AY618977|Thailand|0229_96|1996 AY618969|Thailand|0333_93|1993 AY618980|Thailand|1142_98|1998 AB111087|Thailand|99_10_1HuNIID|1999 AY618984|Thailand|0003_99|1999 AY618975|Thailand|0485_95|1995 AY618976|Thailand|0109_96|1996 AY618967|Thailand|0261_92|1992 AY618971|Thailand|0792_93|1993 AY618965|Thailand|0348_91|1991 AY618964|Thailand|0103_91|1991 AY618966|Thailand|0557_91|1991 AY618972|Thailand|0034_94|1994 EU478408|Myanmar|2006 EU478410|Myanmar|2006 AY550909|LK|1978 AY618983|Thailand|1571_98|1998 AY786192|Vietnam|TG508|2001 AY786196|Vietnam|SG5426|2002 EU448452|Vietnam|Vietnam.0402aTw|2004 AY786188|Vietnam|AG2703|2000 AY786198|Vietnam|TN2693|1999 AY786195|Vietnam|SG5198|2002 AY786190|Vietnam|BD2017|1999 AY786193|Vietnam|TN306|2001 AY786189|Vietnam|AG583|2001 AY786191|Vietnam|TV1363|1998 AY786202|Vietnam|480Vietnam|2002 EU448450|Vietnam|2006 EU448451|Vietnam|2002 * AY786201|Vietnam|2001 * AY786194|Vietnam|VL2777|2000 AY786197|Vietnam|TN2511|1999

Thailand|KPP|KDS03914|NBxx|Scho|S|01|2006 * EU448453|Cambodia|2003 Lineage 4 0.0040

Figure 2-6. ML tree of representative E gene sequences of DENV-4 genotype I. Districts of Kamphaeng Phet are designated by different colors. Bootstrap support values ≥85% are indicated by an asterisk next to the node. The tree is midpoint rooted for purposes of clarity. 50

Spatial and temporal clustering of DENV in Kamphaeng Phet

Among village clusters initiated by dengue cases detected in the school-based cohort, nearly all of the DENV viruses sequenced from both humans and mosquitoes within a cluster during the 14-day period following the initiation of sampling exhibited identical or nearly identical

E gene sequences to the index case of their respective clusters. All but three of these sampling clusters involved viruses originating from the same lineage within a single serotype. The three exceptions are described here. First, in the case of DENV-1, one cluster of three samples included an index case and one village contact that possessed identical E genes in Lineage 2 and a second village contact that was infected with a virus in Lineage 1, which was found to be circulating simultaneously in multiple sub-districts during the 2006 and 2007 seasons. Second, a single DENV-1 virus was isolated from a cluster investigation in which the index case and all other village contacts were infected with DENV-4. This DENV-1 isolate fell within Lineage 1 and was the first virus collected within a large phylogenetic cluster of identical E gene sequences isolated in the same village for the following seven to 31 days, from elsewhere in Nong Pling sub-district on days 62 and 97 after the first isolation, and from Thep Na Korn and Kon Tee sub- districts in the following 97 and nine to 108 days, respectively, spanning 2006 to 2007. Third, among DENV-4 isolates, one cluster in Na Bo Kham in 2004 (Lineage 1) included a virus of a divergent lineage (Lineage 3). Multiple closely related viruses were isolated from both humans and mosquitoes, forming a well-supported clade with other viruses from the same sub-district in the same season. A single mosquito was found to be infected with a divergent virus, comprising the only isolate of DENV-4 Lineage 3 detected in this study. That this lineage was only detected in a mosquito and we did not detect evidence of further transmission within the study population, as well as the number of lineages represented by a single isolate in the data set, suggests that multiple DENV lineages commonly enter this population in a single season, some of which fade

out without detection, or may persist at low levels with little to no detection in the human population.

Both the sequence data presented here and the epidemiological data collected from the cohort and cluster studies (Mammen et al., 2008) suggest that DENV transmission is indeed highly spatially and temporally focal rather than occurring via simultaneous circulation and mixing of multiple DENV lineages across the region (Yoon et al., 2012). For confirmation, we performed a statistical analysis of the strength of phylogenetic clustering of viruses sampled closely in space and time. At all levels of spatial aggregation (sub-district, village, school, and sampling cluster) and for all three serotypes investigated (DENV-1, DENV-2, and DENV-4), we detected a significant relationship between phylogeny and space; i.e., there was more clustering by spatial variable than expected by chance alone (Table 2-1). Similarly, we found a strong phylogenetic clustering of viruses by year of sampling. Hence, we conclude that viral genetic diversity in this population tends to turnover on an annual basis, although lineages may occasionally persist over multiple seasons.

Table 2-1. Phylogeny-trait association tests of phylogenetic structure of geographic, temporal, and clinical traits of DENV isolated in Kamphaeng Phet, Thailand from 2004 – 2007. Traits showing significant phylogenetic structure are indicated in bold.

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nalysis included separate groups for mosquito isolates or unknown traits where applicable. 1 AI and PS scores are based on mean values representing the association between trait and phylogeny. 2 P-values are based on median values representing the association between trait and phylogeny.

Among the levels of spatial aggregation analyzed, the home sub-district and village of subjects showed stronger phylogenetic clustering (sub-district association index (AI): 0.09 to

0.31, parsimony score (PS): 0.3 to 0.51; village AI: 0.28 to 0.5, PS: 0.55 to 0.61) for all three serotypes than did the school (AI: 0.45 to 0.62, PS: 0.63 to 0.65) and sampling clusters (AI: 0.34

to 0.67, PS: 0.68 to 0.77) (Table 2-3). This result is not unexpected because epidemiological data suggest that DENV transmission in this region primarily occurs at a person’s home or the home of a friend or relative (Mammen et al., 2008; Yoon et al., 2012). Our virological data, therefore, support these earlier results and indicate that these trends continued over the four- year time period. No significant phylogenetic clustering of DENV was observed for age, sex, or clinical syndrome, except in the case of DENV-2, for which relatively weak phylogenetic structure was detected according to clinical syndrome (Table 2-1). This result was strongly influenced, however, by the existence of two clusters of unknown syndrome and non- hospitalized DF, as well as, by the small number of DENV-2 sequences isolated in this study. It does not appear to result from differences in virulence among viral populations. The use of coding scheme 1, in which subjects associated with no school and mosquitoes were coded as

NA and mosquito, respectively, produced weaker associations in the school-based analysis than when these were coded according to the school of the respective index case (data not shown). Specifically, these sequences rarely showed clustering, but instead interrupted or reduced the size of school-based clusters.

By restricting our analysis to those sequences obtained from index cases and through school-based surveillance, we were able to further assess whether highly focal sampling in the village-based cluster design was a primary factor influencing the spatial and temporal structure detected. Results of these analyses were generally similar to those obtained using the full set of sequences (Table 2-2). Significant clustering was observed at the sub-district, village, school and 100-meter radius cluster levels and within years for both DENV-1 and DENV-4. The strength of village-level clustering was reduced, however, relative to that of sub-districts and schools (Table 2-2). Further, the diversity within this subsample was comparable to that observed in the full data set, with the loss of only a single lineage (DENV-4 Lineage 3) in subsampling. These results indicate that distinct viral populations may be present in areas separated by only a few kilometers, and suggest that school-based virological surveillance 54

alone captures much of the genetic diversity of DENV within a given area. As such, school- based surveillance may be a practical and efficient indicator of DENV circulating in communities in endemic areas.

Table 2-2. Phylogeny-trait association tests of phylogenetic structure of geographic and temporal traits of DENV-1 and DENV-4 populations isolated via school-based surveillance in Kamphaeng Phet, Thailand from 2004 – 2007. Traits showing significant phylogenetic structure are indicated in bold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and PS scores are based on mean values representing the association between trait and phylogeny. 2 P-values are based on median values representing the association between trait and phylogeny.

An important observation from our study is the strength of temporal DENV clustering by year. This suggests that seasonal bottlenecks commonly occur in this population and, hence, that only a subset of viral genetic diversity within a given year survives into the next. This may in turn result in regular reductions in population immunity and competition among viruses, thus allowing the introduction and dissemination of viruses from outside of the study area. While persistence of viral lineages and in situ evolution occur over multiple seasons in some cases, these processes are relatively limited compared to that of viral migration (i.e., importation), which appears to occur via movement of infected people, play a dominant role in generating diversity within each serotype, and contribute to dynamics in patterns of DENV transmission.

This finding is similar to that reported for Kamphaeng Phet in 2001 (Jarman et al., 2008), and further indicates that it is essential to account for the entry and re-entry of DENV lineages from outside, and potentially distant populations, in the planning and implementation of vaccine 55

programs. In the context of forthcoming vaccine trials and the potential for imperfect vaccination and complex immune interactions, these results imply that the risk of DENV transmission and severe disease within a community will not only be determined by the vaccination levels within that geographic area, but will also be affected by vaccination status and movement of people in neighboring regions in addition to numerous social and environmental factors. This suggests an as yet unquantified risk posed by heterogeneous vaccination in endemic regions, and an important topic of future study in endemic and epidemic dengue transmission settings.

Materials and Methods

Study area and selection of schools

The study protocol was approved by the Institutional Review Boards of the Thai Ministry of Public Health (MOPH), Walter Reed Army Institute of Research (WRAIR), University of

Massachusetts Medical School (UMMS), University of California at Davis (UCD) and San Diego

State University (SDSU). Written informed consent and assent was obtained from all study participants and/or their parents.

The study area and design have been described previously (Endy et al., 2011; Mammen et al., 2008; Yoon et al., 2012). The epidemiology of dengue is well characterized in this region of Thailand due to long-term cohort studies and surveillance conducted since the 1980s by the

Armed Forces Research Institute of Medical Sciences (AFRIMS), a joint research endeavor of the U.S. Armed Forces and the Royal Thai Army Medical Departments. In brief, the study was conducted in Muang District, Kamphaeng Phet, north-central Thailand, a relatively sparsely populated region of Thailand with 233,033 residents in ~63,500 housing structures located over

1962 km2. DENV transmission in this region peaks during an annual ‘dengue season’ from

June to November (Endy et al., 2002b).

Eleven primary schools were selected for participation based on the presence of dengue cases among their students in the previous five years, proximity to the AFRIMS field station

(including road access), and interest of the school administrators. Selected schools were associated with 32 villages (8445 houses). During the first half of the study, twenty of these villages (4685 houses) were selected for inclusion in the cluster study based on the density of houses, favoring those with houses in close proximity to one another (<100 m). During the second half of the study, all houses within the villages were mapped and used during cluster investigations. Unique codes were assigned for each of the 8445 houses and the associated

spatial coordinates and demographics of residents were entered into a Geographic Information

Systems (GIS) database [MapInfo (2000) version 6·0; MapInfo Corporation, Troy, New York].

Selection of school children and village clusters

Primary school children in kindergarten through grade six were followed by active school absence-based surveillance during June to November of each year. An acute blood sample was drawn from cohort subjects who reported a fever in the previous seven days or who had a measured temperature ≥38ºC. A convalescent blood sample was drawn 14 days later along with an evaluation of symptoms. Details on study design and overall results have been published previously (Mammen et al., 2008; Yoon et al., 2012). Acute blood samples underwent testing, including semi-nested reverse transcriptase polymerase chain reaction (RT-PCR) for detection of DENV RNA according to the protocol of Lanciotti et al. (Lanciotti et al., 1992) with the following modifications. Avian myeloblastosis virus (AMV RT, Promega, Madison, WI) reverse transcriptase was used in the first round RT-PCR. The concentrations of primers used in the

RT-PCR and nested reactions were reduced from 50 pmol to 12.5 pmol per reaction and the nested PCR amplification cycles were increased to 25.

Cohort subjects who were dengue PCR-positive and negative from an acute blood sample drawn within three days of illness onset served as an “index” case for a positive cluster investigation; non-dengue acutely ill subjects served as index cases for negative control clusters.

In each cluster, ten to 25 children aged six months to 15 years living within a 100-meter radius of the index case were enrolled regardless of symptomatology. These contact subjects were evaluated at days 0, 5, 10, and 15 with temperature measurement and a symptom questionnaire covering the previous five days. Blood samples were collected on days 0 and 15 and tested for dengue by RT-PCR; virus isolation was attempted in C6/36 cells from all cohort and cluster dengue PCR-positive samples. 58

Entomologic Evaluations

Female adult Ae. aegypti were collected using backpack aspirators from inside and immediately surrounding each home within the cluster (Mammen et al., 2008; Yoon et al.,

2012). Female Ae. aegypti were screened for DENV by RT-PCR using a modified protocol

(Johnson et al., 2004). In brief, pools of ten mosquitoes were made by combining 14 µl from individual mosquito suspensions (in 100 µl of RPMI containing 1% L-glutamine and 10% heat- inactivated FBS) and clarified by centrifugation at 8000 rpm at 4°C for 20 min. From positive pools, individual mosquitoes were assayed by serotype-specific PCR using 14 µl of the original suspension in 126 µl of diluent. DENV from individual mosquitoes were amplified by intrathoracic inoculation in Toxorhynchites splendens mosquitoes and/or by passaging three times in C6/36 cells.

Virus isolation and sequencing

For human DENV PCR-positive samples, virus isolation was performed in C6/36 cells and/or Toxorhynchites splendens mosquitoes as previously described (Kuno et al., 1985; Rosen and Gubler, 1974). Viral RNA was prepared for sequencing from 140 µl of human serum, mosquito suspension, or culture fluid using the QIAamp viral RNA mini kit (QIAGEN, Germany) according to the manufacturer’s instructions. RT-PCR was performed using random hexamer oligonucleotides with the SuperScript first-strand synthesis system (Invitrogen) according to the manufacturer’s instructions. Sequencing of all PCR-positive, isolation-positive samples was attempted but insufficient RNA and isolation failures resulted in some PCR positive cohort and cluster samples to be excluded in this study. The DNA fragments of the envelope gene region of 97 DENV-1, 23 DENV-2, 5 DENV-3, and 74 DENV-4 were amplified by PCR using 5 µl of cDNA in a 50 µl reaction mixture containing 0.3 mM dNTPs, 2.5 U AmpliTaq DNA polymerase

(Applied Biosystems), 1x PCR buffer, 1.5 mM MgCl2 and 15 pmol of each forward and reverse primer. PCR reactions for DENV-1, -3, and 4 were subjected to 1 cycle of 95oC for 5 min, 35 cycles of 94oC for 30 sec, 50oC for 1 min, 72oC for 2 min, and 1 cycle of 72oC for 15 min. PCR reactions for DENV-2 were subjected to the same thermal conditions as the others, except the annealing temperature was changed to 55oC. The PCR-amplified DNA fragments were purified using QIAquick PCR purification kits (QIAGEN) according to the manufacturer’s instructions.

Purified DNA fragments were used for sequencing.

Sequencing reactions were performed using the DYEnamic ET Dye Terminator sequencing kit (GE Healthcare Bio-Sciences) according to the manufacturer’s instructions.

Sequencing primers are available upon request. Sequencing products were cleaned by standard precipitation prior to sequencing in a MegaBACE 500 automated DNA sequencer (GE

Healthcare Bio-Sciences). Overlapping nucleic acid sequences were combined for analysis and edited using SEQUENCHER software (Gene Code Corporation).

Phylogenetic analysis

A total of 97 DENV-1, 23 DENV-2, 5 DENV-3, and 74 DENV-4 E gene sequences were obtained from blood samples of infected children included in the school-based cohort and from mosquitoes and infected children detected in village clusters in Kamphaeng Phet from 2004 to

2007 (Table 2-3). School-based surveillance accounted for the majority of DENV isolated in this study; 95 were not associated with cluster investigations and 47 served as cluster index cases

(Table 2-4). From 50 positive and 53 negative cluster investigations initiated during the study period, 47 index case dengue sequences were available and 19 clusters yielded at least one

DENV E gene sequence from a village contact, while eight clusters yielded sequences obtained exclusively from local mosquitoes. There were an additional four sequences obtained from three negative clusters. 60

Table 2-3. Number of DENV sequences obtained from school children and community clusters in five sub-districts of Kamphaeng Phet, Thailand from 2004 – 2007.

Number of Number of Number of Year Serotype sequences subdistricts villages 2004 DENV-1 0 0 0 DENV-2 7 3 4 DENV-3 1 1 1 DENV-4 24 5 14 2005 DENV-1 2 2 2 DENV-2 3 3 3 DENV-3 4 1 1 DENV-4 15 3 8 2006 DENV-1 59 5 16 DENV-2 2 2 2 DENV-3 0 0 0 DENV-4 33 4 10 2007 DENV-1 36 5 16 DENV-2 11 3 5 DENV-3 0 0 0 DENV-4 2 1 2

Table 2-4. Number of DENV sequences obtained via school absence-based surveillance within a childhood cohort and village cluster studies activated by illness within the cohort in Kamphaeng Phet, Thailand from 2004 – 2007.

Index case initiating Village School Serotype village cluster contact of Mosquito surveillance investigation* index case DENV-1 38 25 24 10 DENV-2 12 6 2 3 DENV-3 1 1 1 2 DENV-4 34 15 18 1 * Index cases were detected through school-based surveillance and subsequently selected for village cluster investigation. These are not included in school surveillance counts.

To place these isolates within the context of global DENV evolution, complete E gene sequences of all four serotypes with known date and country of sampling were collected from

GenBank. Sequences were manually aligned using Se-AL v2.0a11 (available from http://tree.bio.ed.ac.uk/software/), and initial trees were inferred in PAUP using a neighbor- joining algorithm using the HKY85 model of nucleotide substitution (Wilgenbusch and Swofford, 61

2003). This enabled us to obtain a provisional estimate of the pattern of genetic diversity of

DENV isolated within the study. All study isolates from each serotype were found to be of a single, predominantly Asian genotype (DENV-1, genotype I; DENV-2, Asian I genotype; DENV-

3, genotype II; DENV-4, genotype I; see Results). Consequently, global sequence data sets were sub-sampled to include only a subset of closely related sequences of Asian origin from each of the detected genotypes: 1161 DENV-1 (1485 nt), 531 DENV-2 (1485 nt), 304 DENV-3

(1479 nt), and 166 DENV-4 (1485 nt) E gene sequences.

Maximum likelihood (ML) phylogenetic trees of each genotype were then estimated using PAUP [23], utilizing the GTR+I+Γ4 model of nucleotide substitution, which was determined by ModelTest [24] to be the best-fit to the data in hand, and employing tree bisection- reconnection (TBR) branch swapping. Bootstrap resampling (1000 replicate neighbor-joining trees under the substitution model described above) was performed to assess phylogenetic support for individual nodes.

Analysis of spatial and temporal structure

To determine the extent of spatial and temporal structure of DENV within the study area,

Bayesian Maximum Clade Credibility (MCC) phylogenetic trees were inferred for the E gene sequences of DENV-1, DENV-2, and DENV-4 collected in Kamphaeng Phet during 2004-2007 using a Bayesian Markov Chain Monte Carlo (MCMC) method implemented in the BEAST package (v1.6.1) (Drummond and Rambaut, 2007). The small number of DENV-3 sequences collected during the study precluded their inclusion in this analysis. A strict molecular clock, a

TN93+Γ4 model of nucleotide substitution (determined by ModelTest (Posada and Crandall,

1998) to be the best-fit model of nucleotide substitution for the Kamphaeng Phet-specific DENV data sets) with two codon position divisions (1+2, 3), and a constant population size model were used for all analyses. Mixing under more complex evolutionary and demographic models 62

(including a Bayesian skyline plot model) was poor, and Bayes Factor tests indicated this model to be the most appropriate for each of the four serotypes independently.

We used the BaTS (Bayesian Tip-association Signiﬁcance testing) program to assess the extent of spatial and temporal structure among DENV populations in Kamphaeng

Phet from the posterior samples of trees returned by the BEAST analysis described above

(Drummond and Rambaut, 2007; Parker et al., 2008). The BaTS program outputs an

Association Index (AI) and a Parsimony Score (PS), for which 0 indicates complete population subdivision and 1 suggests random mixing (panmixis), as well as a Maximum Clade (MC) score for each character state (location, school, year, etc.) that indicates the extent of clustering for that state compared to all others. Sequences were coded and tested for clustering by; (i) home sub-district (5 sub-districts; KNT - Kon Tee, NBK - Na Bo Kham, NKC - Nakon Chum, NPL -

Nong Pling, TNK - Thep Na Korn), (ii) home village (4 to 9 per district), (iii) school (11 primary schools included in the cohort study and school-based surveillance, isolates from 12 additional schools obtained in the cluster study), (iv) age (0-4, 5-9, to 10-15 years), (v) sex, (vi) clinical syndrome of the patient (includes non-hospitalized DF, hospitalized DF, hospitalized DHFII, hospitalized DHFIII, unknown syndrome) and (vii) year of sampling. For age, sex, and clinical syndrome, mosquitoes were included as a separate group, and unknown clinical traits were recorded as NA. In the case of the school-based analysis, two coding schemes were utilized:

(1) subjects with no school listed on file (primarily children younger than school-age) and viruses isolated from mosquitoes were coded as NA and mosquito, respectively, and (2) mosquitoes and subjects associated with no school were coded with the school of the index case of the cluster in which they were identified. Results using scheme 2 are reported here. Results were generally similar, but weaker, using scheme 1.

The cluster design of the study involved focal sampling over very narrow intervals of time and space and, therefore, had the potential to bias the results of BaTS analysis. To ensure that this study design was not adversely affecting our inference of spatial and temporal patterns, 63

Bayesian MCMC trees were also inferred for DENV-1 and DENV-4 isolates obtained only from index cases and through school-based surveillance. Of the isolates obtained through school- based surveillance and not treated as index cases in the cluster study, only a single sequence from each school per year was utilized to account for potential over- or under-sampling among schools. Phylogeny-trait association tests were then performed on a subset of 37 DENV-1 sequences and 28 DENV-4 sequences, considering the home sub-district, village, school of subjects, and year of sampling. Too few sequences were available to perform this analysis on the DENV-2 data set after subsampling.

Nucleotide sequence accession numbers

DENV E-gene sequences were deposited in GenBank with the following accession numbers: DENV-1 JQ993108-JQ993204, DENV-2 JQ993205-JQ993227 DENV-3 JQ993228-

JQ993232 DENV-4 JQ993233-JQ993306.

CHAPTER 3

PHYLOGEOGRAPHY OF RECENTLY EMERGED DENV-2 IN VIET NAM

Rabaa MA, Hang VTT, Wills B, Farrar J, Simmons CP, Holmes EC (2010) Phylogeography of recently emerged DENV-2 in southern Viet Nam. PLoS Negl Trop Dis. 4(7):e766.

Abstract

Revealing the dispersal of dengue viruses (DENV) in time and space is central to understanding their epidemiology. However, the processes that shape DENV transmission patterns at the scale of local populations are not well understood, particularly the impact of such factors as human population movement and urbanization. Herein, we investigated trends in the spatial dynamics of DENV-2 transmission in the highly endemic setting of southern Viet Nam.

Through a phylogeographic analysis of 168 full-length DENV-2 genome sequences obtained from hospitalized dengue cases from 10 provinces in southern Viet Nam, we reveal substantial genetic diversity in both urban and rural areas, with multiple lineages identified in individual provinces within a single season, and indicative of frequent viral migration among communities.

Focusing on the recently introduced Asian I genotype, we observed particularly high rates of viral exchange between adjacent geographic areas, and between Ho Chi Minh City, the primary urban center of this region, and populations across southern Viet Nam. Within Ho Chi Minh

City, patterns of DENV movement appear consistent with a gravity model of virus dispersal, with viruses traveling across a gradient of population density. Overall, our analysis suggests that Ho

Chi Minh City may act as a source population for the dispersal of DENV across southern Viet

Nam, and provides further evidence that urban areas of Southeast Asia play a primary role in

DENV transmission. However, these data also indicate that more rural areas are also capable of maintaining virus populations and hence fueling DENV evolution over multiple seasons.

Introduction

Dengue viruses (DENV) are mosquito-borne RNA viruses (family Flaviviridae) that exist as four antigenically distinct viruses or serotypes (DENV-1 through DENV-4) and show complex immunological interactions within a human host and at the epidemiological scale (Gubler, 2002).

Current estimates suggest that more than half of the world’s population resides in dengue endemic areas, with 40 million symptomatic infections occurring annually, over two million of which may be severe enough to require hospitalization (Pediatric Dengue Vaccine Initiative,

2009). Although most DENV infections are asymptomatic, the virus is responsible for significant morbidity in the developing world, and places a considerable burden on health systems during periods of both endemic and epidemic transmission (DeRoeck et al., 2003).

The burden of dengue is highest in Southeast Asia, where all four DEN viruses currently circulate. Viet Nam shows consistently high levels of DENV transmission, with the incidence of dengue hemorrhagic fever (DHF) and dengue shock syndrome (DSS) ranging from 36 to

405/100,000 population per year between 1998 and 2008 (World Health Organization Western

Pacific Region Office, 2009; World Health Organization, 2009b), and the southern part of the country accounting for 85% of cases nationally (World Health Organization Western Pacific

Region Office, 2008). Most of these cases occur in children, who experience an annual exposure risk of ~10% (Bartley et al., 2002a; Thai et al., 2005). Hyperendemicity was observed in southern Viet Nam as early as the 1960s, with all four viruses discovered in mosquito specimens collected in and around Ho Chi Minh City (HCMC; formerly Saigon) just years after the first DHF epidemics swept through South Viet Nam and cities across Southeast Asia

(Russell et al., 1969). DHF epidemics were reported in children from rural villages on the

Mekong River and in urban HCMC in 1963, although a suspected DHF outbreak occurred in the

Mekong Delta region in 1960. The disease was first recognized in rural areas, and the movement of the virus was proposed to have followed human movement and commerce on the

Mekong River (Halstead et al., 1965), still a major transportation route between HCMC and the

Deltaic provinces (Lin et al., 2000). Although incidence tends to peak during the rainy season, the tropical monsoon climate and high human population densities in southern Viet Nam allow year-round transmission of DENV (Thai et al., 2005). Rapid urbanization and socioeconomic changes in recent decades may have also contributed to the establishment of what is now a relatively stable, highly endemic transmission pattern in the region.

In recent years, a number of studies have utilized gene sequence data to investigate the spatial and temporal dynamics of DENV in a restricted geographic area (Balmaseda et al.,

2010; Jarman et al., 2008) or during a single epidemic period (Schreiber et al., 2009). An understanding of the evolution and spatial spread of DENV in a larger, stably endemic region over a period of several years may provide important information on the origins of epidemic strains, reveal how and in what populations DENV are able to persist at low levels of infection when immunological and ecological conditions do not favor epidemic activity, and perhaps assist in the prediction of the transmission patterns of newly emergent DENV lineages.

To this end, the aim of this study was to reveal trends in the spatial dynamics of DENV-2 in southern Viet Nam by using full-length genome sequence data obtained from patients admitted to a tertiary referral hospital in HCMC. These data provide a unique opportunity to investigate the spatial dynamics of different lineages of DENV-2 within a geographical region characterized by high endemicity. By using recently developed phylogeographic methods we address the following key questions: (i) Does HCMC, as the primary urban center of the region, act as a source population for dengue viruses circulating in southern Viet Nam? (ii) Within

HCMC, are the highest density population areas acting as foci of viral dispersal throughout the city and the region? (iii) At this scale, does DENV dispersal adhere to a predictable population- or density-based model of transmission?

Results

Overall patterns of geographic structure

Our initial trait association (AI and PS) tests of phylogeographic structure rejected the null hypothesis of no association between sampling location and phylogeny at all of the spatial levels tested for the Asian I DENV-2 genotype in southern Viet Nam (Table 3-1). Hence, these genome sequence data possess at least some geographic structure. The use of index ratios of the observed values to those expected under panmixis (in which 0 indicates complete population subdivision and 1 suggests random mixing [panmixis]) allows the strength of the association between geography and phylogeny to be characterized further. Accordingly, although panmixis was rejected by our analysis, the AI and PS index ratios for the Asian I genotype approached 1, indicating relatively frequent virus movement between geographic areas. In contrast, no significant associations between phylogeny and geography were observed at any of the spatial levels analyzed in the American/Asian genotype (Table 3-1), and

BSSVS location reconstruction performed using American/Asian DENV-2 revealed no significant patterns of spatial diffusion, likely due to the small number of samples available

(MCC trees shown in Figures S3-4 and S3-5). We therefore focused the rest of our study on the Asian I genotype.

Table 3-1. Phylogeny-trait association tests of phylogeographic structure of DENV-2 in southern Viet Nam. !

Statistic Spatial clustering Index Ratio, observed to Observed value Expected value P-value expected (95% CI) (95% CI) (95% CI) Association Index (AI) Asian I 10 Provinces 0.60 (0.48-0.75) 5.3 (4.8-5.8) 8.9 (7.7-10) <0.01 9 Provinces, 3 HCMC 0.68 (0.59-0.81) 7.8 (7.4-8.3) 11.4 (10.3-12.5) <0.01 9 Provinces, 11 HCMC 0.69 (0.62-0.77) 9 (8.5-9.5) 13.1 (12.4-13.8) <0.01 District Groups American/Asian 6 Provinces 0.82 (0.54-1.54) 1.4 (1.2-1.7) 1.7 (1.1-2.2) 0.18 5 Provinces, 3 HCMC 0.92 (0.67-1.50) 2.3 (2-2.7) 2.5 (1.8-3) 0.31 5 Provinces, 9 HCMC 0.84 (0.66-1.13) 3.1 (2.7-3.5) 3.7 (3.1-4.1) 0.06 District Groups Parsimony Score (PS) Asian I 10 Provinces 0.75 (0.72-0.80) 40 (40-40) 53.2 (50-55.8) <0.01 9 Provinces, 3 HCMC 0.84 (0.80-0.89) 64 (64-64) 76.3 (72.1-79.7) <0.01 9 Provinces, 11 HCMC 0.92 (0.89-0.97) 85.6 (85-87) 92.8 (89.6-95.9) <0.01 District Groups American/Asian 6 Provinces 1.01 (0.90-1.00) 10 (10-10) 9.9 (9-10) 1 5 Provinces, 3 HCMC 1.01 (0.94-1.00) 16 (16-16) 15.8 (15-16) 1 5 Provinces, 9 HCMC 0.98 (0.91-1.09) 27.3 (27-28) 27.8 (25.8-29.7) 0.27 District Groups Maximum clade (MC) scores* Asian I Dong Thap (DT) NA 4 (4-4) 2 (1-3) <0.01 Tay Ninh (TN) NA 2 (2-2) 1 (1-1) <0.02 Vung Tau (VT) NA 2 (2-2) 1 (1-1) <0.02 American/Asian Ho Chi Minh City NA 11.9 (11-16) 5.5 (3.2-9) <0.03 (HCMC) * Only significant MC scores are indicated; for all other locations, p>0.05 as estimated using BaTS [31].

Spatial analysis by province

The MCC phylogeny of the Asian I genotype with BSSVS reconstructed ancestral locations (10 provinces) of the internal nodes reveals that viruses from nearly all provinces are dispersed throughout the phylogeny, most notably HCMC and DT, as well as LA and TG (Figure

1). Although some local clustering was observed, the general trend in the tree is of mixing among geographic locations, with Bayesian phylogeographic analysis estimating significant reversible diffusion pathways between AG and DT, DT and HCMC, HCMC and LA, and LA and

TG (BF>20; Table S3-1). These findings indicate viral exchange between HCMC and other provinces both adjacent to and distant from its borders, along with movement of viruses between adjacent provinces outside of HCMC (Figure 3-2). Significant clustering was also observed in three provinces: DT, TN, and VT, and may explain the overall significance of AI and

PS scores detected using BaTS (Table 3-1). 69

Figure 3-1. MCC phylogeny of the DENV- 2 Asian I genotype in southern Viet Nam (2003-2008) according to province of sampling. Tips are colored by province of sampling. Internal branches are colored based on the reconstructed ancestral state as estimated by the reversible diffusion model. Branches colored black indicate Bayesian posterior probabilities less than 0.85. Estimated support for the reconstructed ancestral state is indicated by open (>95%) and closed (>85%) diamonds.

Figure 3-2. DENV-2 Asian I genotype dispersal across provinces of southern Viet Nam. Map showing significant pathways of diffusion estimated between provinces of southern Viet Nam. Solid lines indicate diffusion pathways among provinces with a shared border. Dashed lines indicate diffusion pathways among provinces that do not have a shared border. Numbers in parentheses indicate the number of sequences included in each group.

Spatial analysis by population density within HCMC and by province

When Asian I viruses were categorized according to population density in HCMC and by province elsewhere (three HCMC groups, nine provinces), various patterns emerged. The

highly populated area of HCMC showed two strongly supported pathways of diffusion: between the Superurban and Urban districts, and between the Urban and Suburban districts (BF>30). In contrast, viral exchange between Superurban and Suburban HCMC was not supported.

Significant diffusion pathways were also observed between regions of HCMC and provinces both distant and adjacent in the Mekong Delta region, while one significant pathway was detected between two adjacent provinces west of HCMC (AG and Suburban HCMC, DT and

Urban HCMC, LA and Superurban HCMC, LA and Urban HCMC, LA and Suburban HCMC, and

LA and TG, BF>15; Table S3-2, Figures 3-3 and 3-4). These relationships largely corresponded to those detected in the provincial analysis. Sampling bias did not appear to greatly influence these results, or those by province, as similar results were obtained when overrepresented locations were subsampled (Figure S3-1).

Figure 3-3. MCC phylogeny of the DENV-2 Asian I genotype in southern Viet Nam (2003- 2008) according to population density (within HCMC) or province of sampling. Tips are colored by urban level (within HCMC) or province of sampling. Internal branches are colored based on the reconstructed ancestral state as estimated by the reversible diffusion model. Branches colored black indicate Bayesian posterior probabilities less than 0.85. Estimated support for the reconstructed ancestral state is indicated by open (>95%) and closed (>85%) diamonds.

Figure 3-4. DENV-2 Asian I genotype dispersal across a population density gradient in HCMC and southern Viet Nam. Map showing significant pathways of diffusion estimated between three regions of gradated population density within HCMC and the remaining nine provinces of southern Viet Nam. Solid lines indicate diffusion pathways among regions with a shared border. Dashed lines indicate diffusion pathways among regions that do not have a shared border. Numbers in parentheses indicate the number of sequences included in each group.

Spatial analysis by geographic proximity and population density-based district groupings within HCMC and by province

Use of finer-scale spatial classification within HCMC (11 geography- and population- based district groupings, nine provinces) revealed that samples from most geographic locations

(both within and outside of HCMC) were dispersed throughout the phylogeny of the Asian I genotype (Figure S3-2). Within HCMC, significant diffusion pathways were detected across the city (BF>15, Table S3-3), with strong mixing between areas of similar population density and relatively few diffusion pathways connecting districts in the lowest and highest population density categories, and hence suggestive of a gravity diffusion model (Figure 3-5). Additionally, seven of the eight estimated diffusion pathways with the highest support connect areas with shared borders. With respect to exchange between HCMC and the rest of southern Viet Nam, five significant diffusion pathways between four of the highest density districts of HCMC and four provinces (HCM-sup1 and LA each represented in two pathways) were inferred. Significant

diffusion was also detected between three lower-density districts and outer provinces: AG and

HCM-urb3, LA and HCM-sub1, and AG and HCM-sub2. Of these, only LA and HCM-sub1 are located directly adjacent to one another. Among the provinces outside of HCMC, significant diffusion pathways were estimated between BD and BP, BD and DN, BD and VT, BP and DN,

BP and TN, DN and TN, and DN and VT (Figure 6). No viral migration was identified among these sites in our coarse-scale analyses, and the statistical significance of these inferred diffusion pathways at the finest scale may result from the relatively small numbers of isolates from these areas and potential over-parameterization of the spatial model. Indeed, there is an inherent increase in uncertainty in these models as additional geographic groups are defined.

The loss of a significant link between LA and TG in this analysis (and between AG and DT in the previous analysis) also likely reflects the increased complexity of the phylogeographic model following the addition of spatial groups; relatively low, but significant, Bayes factors were calculated for links between these provinces in the provincial analysis (Table S1).

Figure 3-5. DENV-2 Asian I genotype dispersal among districts of Ho Chi Minh City, Viet Nam. Map indicating significant pathways of diffusion estimated between district groups within HCMC. Solid lines indicate diffusion pathways among areas with a shared border. Dashed lines indicate diffusion pathways among areas that do not have a shared border. Numbers in parentheses indicate the number of sequences included in each group.

Figure 3-6. DENV-2 Asian I genotype dispersal among districts of Ho Chi Minh City and provinces of southern Viet Nam. Map showing significant pathways of diffusion estimated between district groups within HCMC and provinces of southern Viet Nam. Solid lines indicate diffusion pathways among regions with a shared border. Dashed lines indicate diffusion pathways among regions that do not have a shared border. Numbers in parentheses indicate the number of sequences included in each group.

Importantly, the general effects of the number and scale of spatial parameters on phylogeographic model inference are currently under investigation, such that the results of this finest-scale analysis, particularly those that are inconsistent with earlier analyses, should be considered as provisional. It is also possible that these changes reflect general trends in mixing between HCMC and other communities within the region that become clearer as population characteristics of the large urban area of HCMC are taken into account in the analysis. These uncertainties not withstanding, it is interesting to note that the new connections observed at this finest spatial level are compatible with a previously undetected transmission network among the provinces east and north of HCMC, largely industrial areas that are geographically isolated from the Mekong Delta provinces showing consistent viral exchange with HCMC.

Distance and gravity model-based spatial analysis

Finally, although some movement patterns are clearly suggestive of gravity-like dynamics, and particularly within HCMC (see above), distance and gravity model-informed priors did not result in an overall better fit to the data than the default constant rate priors (as reflected in marginal log likelihoods, Table S5). Notably, gravity model-informed priors were statistically superior to distance-informed priors. This result is not entirely surprising, as physical distances alone may poorly reflect the complexity of human population movements

[38]. Additionally, both our distance and gravity model calculations were based on official government boundaries and population estimates for relatively large geographic areas, but samples (and populations) were not uniformly distributed across each of these locations, and the use of distances between centroids may not correspond to actual distances between populated areas using roads or waterways, although we would expect these to be similar.

Importantly, similar results were obtained when distances were calculated between the centroids of our viral populations instead of provinces, with slightly increased and slightly decreased likelihoods apparent when gravity model-informed and distance-informed priors were used, respectively (data not shown).

Discussion

This study documents aspects of the diffusion of DENV-2 throughout a large, highly endemic region of southern Viet Nam, and provides insight into the capacity of genome sequence data to capture potentially important trends in the spatial and temporal dynamics of

DENV. Our phylogeographic analysis suggests that the Asian I genotype of DENV-2, which likely entered southern Viet Nam in the late 1990s from elsewhere in Southeast Asia and recently displaced the American/Asian genotype as the predominant DENV-2 lineage in the region (Hang et al., 2010), has circulated consistently in the high population density region of

HCMC since at least 2000 and rapidly spread to populations across the region during the process of lineage replacement. During the period of sampling, specifically from 2003-2007, dengue incidence nearly doubled across southern Viet Nam; this increase was mostly associated with DENV-2 infection. Data suggest a high force of infection attributable to the

Asian I lineage during most of this period, which corresponds with the displacement of the

American/Asian genotype (Hang et al., 2010). The timeframe and spatial scale at which this clade replacement event was observed indicate that the Asian I lineage spread relatively rapidly into populations across the region following its introduction, and suggest that human movement likely played a significant role in the dispersal of the novel lineage. A similar process of genotype replacement appears to have occurred on different timescales in both Thailand and

Cambodia, either of which may act as a source population for southern Viet Nam (Hang et al.,

2010). Unfortunately, a detailed spatial analysis of the American/Asian genotype was not possible due to the smaller number of samples isolated during the study period.

Major urban areas of Southeast Asia have previously been proposed to play central roles in DENV epidemics, harboring the greatest viral genetic diversity and population sizes sufficiently large to allow sustained outbreaks that may subsequently spread to more rural areas, and potentially acting as harbingers of epidemic dengue activity in a given season

(Cummings et al., 2004; Gubler et al., 1979b). While the reversible nature of the diffusion pathways estimated in this study does not allow the directionality of viral movement to be determined, our results are clearly compatible with the idea that HCMC acts as a driver of viral diffusion into other locales in southern Viet Nam, either as a major source population for dengue viruses or as a mixing ground into which viruses are trafficked by the movement of migrants and visitors from rural areas into the city and are subsequently relayed out to other parts of the country through mosquito and human movements. Our analysis suggests the former of these to be more likely, as location reconstruction showed strong support for the deepest branches of the MCC tree originating in HCMC. The very large population (~6.6 million; 3155 persons/km2)

(General Statistics Office of Vietnam, 2008) of this urban area likely contains sufficiently high numbers of susceptible hosts to allow sustained year-round hyperendemic transmission, thereby providing ample opportunities for DENV to evolve within the city and its surrounding suburban districts. Additionally, the greater connectivity between HCMC and the rest of

Southeast Asia, manifest in such features as the number of airline routes, will obviously facilitate the importation of new viral lineages into this population. Finally, our study indicates that HCMC may consistently maintain multiple viral lineages of a single DENV genotype throughout the year. Indeed, it is striking that our BaTS analysis detected no large monophyletic groups (i.e.

>4.46 sequences in the provincial analysis) within HCMC at any spatial scale; lineages in which

HCMC was dominant and fell basal on the tree often included at least one virus obtained from a resident of another province, providing additional evidence for diffusion out of HCMC.

Notably, its role as the center of commerce and industry in the region makes HCMC a major acceptor of human in-migration from more rural areas, particularly to districts labeled

‘urban’ and ‘suburban’ in this analysis (General Statistical Office in Ho Chi Minh City 2005,

2006), and occasional movements of these migrants between HCMC and their home provinces may in part fuel rapid dispersal of DENV to distant provinces such as An Giang and Dong Thap, as well as to northern portions of the country. Our analysis may to some extent reflect this

movement, as significant diffusion pathways were consistently detected between these provinces and urban and suburban HCMC. It is also possible that these movements introduce new lineages from rural into urban areas, as has been suggested for the malaria parasite

(Osorio et al., 2004), thus allowing them to become established within large populations and potentially be exported to new communities across the region. Unfortunately, the relatively small numbers of sequences from many of the communities outside of HCMC preclude us from determining whether this movement from rural to urban areas plays a major role in the spatial dynamics of DENV transmission. Further, although the substantial timescales of some of these long distance migration events mean that we are unable to reject the possibility that numerous small-scale mosquito-based transmission cycles resulted in virus dispersal over long distances, the lack of closely related isolates from intermediate geographic areas suggests that the viral lineages may have traveled across these provinces rapidly enough that transmission chains were not established within the population, as would occur with human rather than mosquito movement. Greater numbers of samples from communities outside of HCMC and mosquito populations from across the region would allow us to explore this further, and to potentially confirm the presence or lack of specific viral lineages in intermediate areas. While the Asian I lineage appears to have become established within the population, ongoing sampling in southern Viet Nam would clearly allow us to capture the introduction and dispersal events of future DENV lineages, potentially on a finer spatial and temporal scale than possible here.

Despite the relatively small sample size, phylogenetic clustering of viruses, itself an indicator of potential in situ evolution, was clearly detected among viruses isolated from residents of Dong Thap, Tay Ninh, and Vung Tau. Dong Thap province in particular yielded several small monophyletic groups, as well as one well-supported clade that contained a sequence from An Giang, its neighbor to the west. This suggests that locales that are relatively geographically isolated from highly populated urban areas may experience some population subdivision similar to that observed at a smaller spatial scale in northern Thailand (Jarman et

al., 2008). The existence of a viral clade isolated exclusively from this region in late 2006 and late 2007 with an estimated divergence time of approximately two years prior to isolation (data not shown; TMRCA 95%HPD from 2005.2 to 2006.1) suggests that local transmission networks in semi-rural areas such as these western provinces may be capable of maintaining virus populations and fueling DENV evolution over multiple seasons. In addition, the detection of multiple DENV-2 lineages in Dong Thap and Long An extending through 2006 and 2007, some of which appear to have local histories dating back to previous dengue seasons, indicates that several introductions of Asian I DENV-2 have likely occurred in these provinces in recent years.

Similar findings were reported from a mixed urban-rural environment in Thailand (Jarman et al.,

2008). More generally, these findings suggest that Dong Thap and rural regions of Southeast

Asia may act as sink populations, dependent upon local seropositivity rates at a given time, with

HCMC and other major urban city centers functioning as DENV source populations for surrounding areas.

Within HCMC, the patterns of DENV movement between the three regions of varying population density are consistent with a gravity model of virus dispersal, with viruses moving down (or up) gradated population density categories even though all three regions share borders. This trend of movement across a gradient of population densities is largely upheld in the finer-scale analysis within the city, with 13 of 15 significant viral diffusion pathways in the city detected between areas of similar population density or one degree removed, and the majority of viral movement occurring between adjacent districts. Using a similar phylogeographic approach, Balmaseda et al. also observed viral movement between adjacent neighborhoods in a cohort study in Managua, Nicaragua, with some exchange occurring between more distant neighborhoods, likely attributable to transportation networks and migrant workers moving within the community (Balmaseda et al., 2010). Similar to numerous other genetic and epidemiological studies (Van Benthem et al., 2005; Jarman et al., 2008; Schreiber et al., 2009;

Thai et al., 2010b), the analysis in Nicaragua revealed marked spatial clustering in relatively

small areas (in this case, by neighborhood). Although the sampling regime undertaken here does not allow us to fully capture short transmission networks, our observation of virus dispersal over relatively short distances and between adjacent districts within HCMC highlights probable roles for local mosquito populations and small-scale human movements in the diffusion of

DENV in this highly urban area.

In sum, our study indicates that DENV moves relatively freely among human populations in southern Viet Nam, over both long and short distances, and hence suggests a major role for anthropogenic factors and urban areas as drivers of DENV dispersal in Southeast Asia.

However, the relative isolation of some areas directly adjacent to well-connected areas is not well understood, such that it is difficult to make strong conclusions on the predictability of DENV transmission dynamics within this highly endemic region; spatio-temporal variation in seropositivity may play a significant role in preventing new viral lineages from being established in some populations, and the contribution of this factor should be investigated further in future studies. Importantly, neither distance-based nor gravity models were able to explain the full complexity of the transmission dynamics, although a gravity model showed a slightly better fit to our data than did the distance-based models for both the provincial and population density- based analyses. The improved fit of the model upon the integration of population data provides further evidence that human population movement is an important factor acting on DENV dispersal in the region. Thus, it is possible that increased information on short- and long-term human population movements between rural and urban areas may provide a model with improved predictive power for estimating the future spatial spread of DENV over this area. The use of transportation information, including distances by land (such as road distance) and water travel, as well as relative costs of travel between these areas, may also increase the power of these models to determine important viral migration pathways. In the absence of this information, DENV transmission dynamics within and across cities and rural areas are difficult to predict, as many factors, including human population densities and movement, population

immunity, mosquito densities and dispersal, and the seasonality of dengue transmission intensity, as well as other unknown factors, may affect the rates of virus dispersal and establishment of new transmission networks in a locality. As these data represent the initial results of an ongoing study, the isolation of additional DENV sequences over the coming years will allow us to investigate the spatial relationships among viral lineages circulating within the country in greater detail, and may enable us to determine how specific population movement patterns such as seasonal migration and international travel affect the dispersal of viral lineages throughout the region.

Materials and Methods

Study population and data

DENV-2 genome sequences were obtained from dengue patients enrolled into a prospective clinical and virological study of dengue at the Hospital for Tropical Diseases in Ho

Chi Minh City, Viet Nam. These data and the sampling, serotyping, virus isolation, and sequencing methods used have been described elsewhere (Hang et al., 2010). Written informed consent was obtained from the patient or guardian prior to participation in the study, which was approved by the Hospital for Tropical Diseases and the Oxford University Tropical

Research Ethical Committee. Along with demographic and clinical data, the date of sampling and geographic information (Longitude, Latitude) on the location of each patient’s home were collected by research staff using a hand-held GPS device.

From 2001 to mid-2008, 187 full DENV-2 genome sequences were obtained from hospitalized dengue cases in southern Viet Nam, and their complete coding regions (10,176 nt) were manually aligned using Se-AL v2.0a11 (available from http://tree.bio.ed.ac.uk/software/).

Of these, three were initially excluded from this analysis; two sequences were missing spatial or genetic data, and one sequence was obtained from a subject who identified his or her home as being far outside of the study area (Nghe An province, >950 km north of HCMC).

As a large hospital-based study that likely included only severe cases, we would not reliably expect to capture clustering at the finest spatial and temporal scales, including direct chains of transmission. To identify such fine-scale clustering, pairwise genetic distances between the 184 remaining sequences were determined using the HyPhy package, employing the HKY85 model of nucleotide substitution with global branch length estimation parameters.

Geographic distances (WGS84 ellipsoid) between patient homes were determined using longitude and latitude coordinates with the ‘sp’ package in R (version 1.2.9) (Pebesma and

Bivand, 2005; R Development Core Team, 2009) (Figure S3-1). All sequences were categorized according to date of sampling, geographic distance and genetic distance, and potential short transmission chains were eliminated in order to prevent short-term focal transmission events from biasing the larger-scale spatial analyses. In total, 14 sequences were found to be close enough in date of sampling (<15 days) (Gubler, 1997; Vaughn et al., 1997), geographic distance (<0.8 km) (Bugher and Taylor, 1949; Harrington et al., 2005; Reiter et al.,

1995), and genetic distance (<0.0001) to other viruses in the database that they may represent direct transmission events, and were thus excluded from analysis. The resulting data set of 170 full genome DENV-2 sequences, sampled between 2001 and mid-2008, consisted of 126 viruses of the Asian I genotype, 42 of the American/Asian genotype, and two of the

Cosmopolitan genotype. Because of their small number, isolates of the Cosmopolitan genotype were assumed to be importations from outside of the study area and were removed from all analyses. The final 168 sequence data set included samples obtained from the majority of districts in HCMC (20/24 districts) and from nine additional provinces in southern Viet Nam – An

Giang (AG), Binh Duong (BD), Binh Phuoc (BP), Dong Nai (DN), Dong Thap (DT), Long An

(LA), Tay Ninh (TN), Tien Giang (TG), Vung Tau (VT) – covering an area of approximately

37,500 km2 and a population of nearly 20 million (~532 persons/km2) (General Statistics Office of Vietnam, 2008).

Phylogenetic analysis

Bayesian Maximum Clade Credibility (MCC) phylogenetic trees were inferred for the

DENV-2 genome sequences of the American/Asian and Asian I genotypes separately using a

Bayesian Markov Chain Monte Carlo (MCMC) method implemented in the BEAST package

(v1.5.2) (Drummond and Rambaut, 2007), which incorporates date of sampling information and returns rooted trees. A strict molecular clock, a GTR+Γ4 model of nucleotide substitution

(determined by Modeltest v3.7) (Posada and Crandall, 1998) with three codon positions

(substitution model, rate heterogeneity model, and base frequencies unlinked across all codon positions), and a Bayesian skyline coalescent model (five coalescent-interval groups) were used for all analyses, all of which have previously been shown to be appropriate for the analysis of

DENV (Dunham and Holmes, 2007; Hang et al., 2010; Twiddy et al., 2003). Very similar results

(no major differences in topology or coalescent times) were obtained under a relaxed

(uncorrelated lognormal) molecular clock model (results available from the authors on request).

Three independent runs of at least 150 million generations were performed, with sampling every

10,000 generations, until all parameters had reached convergence with 10% removed as burn- in. This analysis allowed us to estimate times to the most recent common ancestor (TMRCA) for key nodes on the DENV-2 phylogeny of each genotype. Nodal support is expressed as

Bayesian posterior probability values.

Analysis of spatial structure and virus dispersal patterns

To assess the overall degree of spatial admixture and geographical structure among

DENV-2 lineages in this region, we calculated values of the association index (AI) (Wang et al.,

2001) and parsimony score (PS) statistics (Slatkin and Maddison, 1989) for each genotype from the posterior samples of trees returned by BEAST using the BaTS program (Parker et al.,

2008). This method accounts for phylogenetic uncertainty in investigating phylogeny-trait correlations, with 1000 random permutations of tip locations to estimate a null distribution for each statistic. This program also allowed us to assess the level of clustering in individual locations using the monophyletic clade (MC) size statistic. The relationships among sequences were estimated on three spatial levels: (i) by province (10 spatial groups, 45 possible diffusion pathways), (ii) by population density within HCMC and by province (nine provinces and three regions within HCMC: ‘Superurban HCMC’ - population density >15,000/km2, ‘Urban HCMC’ -

population density <15,000/km2 and >2500/km2, and ‘Suburban HCMC’ - population density

<2500/km2; 66 possible diffusion pathways overall), and (iii) by geographic proximity and population density-based district groupings within HCMC and by province (11 regions within

HCMC, nine provinces; 190 possible diffusion pathways).

The strength of support for viral exchange between individual locations at each of the spatial levels was inferred using a geographically-explicit Bayesian MCMC approach implemented in BEAST (Lemey et al., 2009b), with the same coalescent models and spatial groups as described above. This method estimates a reversible diffusion rate for each potential diffusion pathway among the predefined locations while simultaneously estimating evolutionary and coalescent parameters, thereby allowing quantification of the uncertainty in ancestral state reconstructions (i.e. ancestral geographic locations). Bayesian Stochastic Search Variable

Selection (BSSVS) was used to identify the links between these locations among the posterior sets of trees that explain the most likely migration patterns among DENV-2 in southern Viet

Nam. Bayes factor (BF) tests were used to determine the statistical significance of diffusion pathways among the geographic groups. To summarize the posterior distribution of ancestral location states, nodes in the MCC trees were annotated with the modal location state for each node using TreeAnnotator, and trees were visualized using FigTree (available at http://tree.bio.ed.ac.uk/software). To account for the potential effects of sampling bias, data from highly sampled geographic areas were randomly subsampled with replacement to create smaller data sets from each geographic location, and analyses were repeated 10 times for each subsampling scheme.

Distance and gravity model-based spatial analysis

Because pathogen dispersal across geographic areas is often thought to be influenced by factors such as distance and human population density, we integrated both distance and

population-based priors into the phylogeographic analysis to mimic the effects of these factors

on the DENV population. To calculate the distances between the defined populations, centroids

for relevant geographic regions were determined using R (version 2.9.1) (Pebesma and Bivand,

2005) with the ‘sp’, ‘shapefiles’, and ‘maptools’ packages (Lewin-Koh et al., 2009; R

Development Core Team, 2009; Stabler, 2006), and utilizing a map of Viet Nam (shapefile

format) defining first and second level subnational administrative boundaries (Center for

International Earth Science Information Network (CIESIN) Columbia University, 2005).

Distances (WGS84 ellipsoidal) between centroids were subsequently estimated (R

Development Core Team, 2009). Population data for each of the provinces of Viet Nam and the

districts of HCMC in 2007 were obtained from the General Statistics Office of Viet Nam and the

Statistical Office in Ho Chi Minh City, and were used as representative population information

for all analysis (General Statistical Office in Ho Chi Minh City, 2007; General Statistics Office of

Vietnam, 2008).

To obtain prior estimates corresponding to these distance and population-based

parameters, simple gravity model calculations providing estimates of the movement of

populations (and disease dispersal) (Cij) between community i (of size Pi) and community j (of

size Pj) were calculated using the relation:

PiPj Cij = θ ρ dij

where θ is a proportionality constant and ρ adjusts the dependence of dispersal on the

€ distance (d) between the two geographic areas (Xia et al., 2004). Variables representing the

dependence of dispersal on population sizes were not utilized due to the reversibility of the

diffusion links assessed using this phylogeographic method, and could not have been reliably

estimated due to a lack of data on differential population movements in the region. Normalized

values (mean one and unit variance) for distance and gravity model calculations were then

utilized as priors to inform the rates of diffusion among geographic locations. Diffusion rate prior

distributions were fixed (F) or were sampled from multivariate Gamma prior (MGP) distributions.

Distance and gravity model-based analyses were performed at the provincial level and at the second of the spatial levels (12 groups), but were not conducted using distance or population- informed priors at the finest spatial scale (HCMC district groups and provinces, 20 groups), as these regions were not comparable in terms of distances between centroids or population sizes.

SUPPLEMENTARY FIGURES

Figure S3-1. Genetic and geographic distances among complete coding region sequences of DENV-2. Each pair is represented as a point, with the color of the point indicating the year of sampling. (a) Asian I DENV-2. Colors represent: 2003, black; 2004, blue; 2005, cyan; 2006, green; 2007, yellow, 2008, red; all years, grey. (b) American/Asian DENV-2. Colors represent: 2001, black; 2002, blue; 2003, cyan; 2004, green; 2005, yellow, 2006, red; all years, grey.

Figure S2. Results of the phylogeographic analysis on subsampled populations of DENV-2, Asian I genotype. (a) Phylogeographic analysis by province. (b) Phylogeographic analysis by urban levels within HCMC and by province outside of HCMC.

Figure S3-3. MCC phylogeny of the Asian I genotype in southern Viet Nam (2003-2008) according to district group (within HCMC) or province of sampling. Tips are colored by district group (within HCMC) or province of sampling. Internal branches are colored based on the reconstructed ancestral state as estimated by the reversible diffusion model. Branches colored black indicate Bayesian posterior probabilities less than 0.85. Estimated support for the reconstructed ancestral state is indicated by open (>95%) and closed (>85%) diamonds.

Figure S3-4. MCC phylogeny of the American/Asian genotype in southern Viet Nam (2001-2006) according to province of sampling. Tips are colored by province of sampling. Internal branches are colored based on the reconstructed ancestral state as estimated by the reversible diffusion model. Branches colored black indicate Bayesian posterior probabilities less than 0.85. Estimated support for the reconstructed ancestral state is indicated by open (>95%) and closed (>85%) diamonds.

Figure S3-5. MCC phylogeny of the American/Asian genotype in southern Viet Nam (2001-2006) according to population density (within HCMC) or province of sampling. Tips are colored by urban level (within HCMC) or province of sampling. Internal branches are colored based on the reconstructed ancestral state as estimated by the reversible diffusion model. Branches colored black indicate Bayesian posterior probabilities less than 0.85. Estimated support for the reconstructed ancestral state is indicated by open (>95%) and closed (>85%) diamonds.

CHAPTER 4

DENGUE VIRUS IN SUB-TROPICAL NORTHERN AND CENTRAL VIET NAM: POPULATION IMMUNITY AND CLIMATE SHAPE PATTERNS OF VIRAL INVASION AND MAINTENANCE

Abstract

Dengue is the cause of epidemic and endemic disease across the tropics and subtropics. Incidence is particularly high in much of Southeast Asia, where hyperendemic transmission plagues both urban and rural populations. However, endemicity has not been established in some areas with climates that do not support year-round viral transmission. An understanding of how dengue viruses (DENV) enter these environments and whether the viruses persist in inapparent local transmission cycles is central to understanding how DENV emerges in areas at the margins of endemic transmission. Dengue is endemic in the tropical southern Viet Nam, while increasingly large seasonal epidemics have occurred in the north over the last decade. Here, we investigate the spread of DENV-1 across Viet Nam to determine the routes by which the virus enters northern and central Viet Nam. Through phylogeographic analysis of 1,765 envelope (E) gene sequences, we identify frequent viral movement between neighboring populations and strong local clustering, suggestive of vector-mediated movement at a local scale. Long-distance migration of DENV between human population centers also occurred regularly and on short time-scales, indicating human-mediated viral invasion into northern Viet Nam. Well-connected human populations in southern Viet Nam are the primary source of DENV circulating throughout the country, while central and northern Viet Nam act as sink populations, likely due to reduced population connectivity in the central regions and the

influence of seasonal weather patterns on vector survival and DENV replication in the north.

Finally, phylogeographic modeling suggests that viral movement follows a gravity model and indicates that population immunity along with spatial connectivity may play an important role in shaping patterns of DENV transmission.

Introduction

Dengue viruses (DENV) are single-stranded, positive-sense mosquito-borne RNA viruses (family Flaviviridae), within which there is considerable genetic diversity and limited gene flow on both global and local scales (Holmes and Burch, 2000). All four serotypes or viruses (DENV-1 to DENV-4) are capable of infecting humans and result in a spectrum of clinical outcomes, ranging from asymptomatic to severe disease. The disease burden and geographic range of dengue have greatly expanded over the previous 50 years. In the 1960s, fewer than ten countries reported a total of ~15,000 dengue cases to WHO annually. Rapid urbanization and global travel have fueled the global spread and establishment of DENV populations across the tropics and sub-tropics, and recent estimates suggest that 35 million symptomatic cases now occur in 124 countries every year (Beatty et al., 2011; World Health

Organization, 2009a). There are no licensed dengue vaccines and no specific interventions beyond supportive therapy are available to treat the disease.

Endemic dengue transmission occurs in across the tropics, while sub-tropical regions may experience epidemics of varying size that subside with the shift to winter temperatures.

Little to no detectable transmission occurs in these populations during the low season, and in some areas transmission remains at very low levels for several years following an epidemic

(Cuong et al., 2011; Lin et al., 2010; Wu et al., 2010). Within Asia, phylogenetic studies of

DENV in southern China and Taiwan suggest that seasonal epidemics often originate in infected travelers returning from endemic Southeast Asia (Chen, 2011; Huang et al., 2007; Jing et al., 2012; Shu et al., 2009; Zheng et al., 2009). It is not clear whether DENV lineages can also persist across seasons in these environments. In addition, there has been no investigation of the determinants of viral introduction in human populations located at the geographic and climatic edge of endemic transmission and closely connected to endemic areas by both land and air travel. A better understanding of the roles of mosquito- and human-meditated

movement in the invasion of DENV into these environments would provide important insights into the expansion of the geographic range of dengue and its emergence in subtropical environments.

High human population densities and suitable ecological conditions have supported hyperendemic DENV transmission in southern Viet Nam since at least the early 1960s

(Halstead et al., 1965). Coupled with epidemiological and virological surveillance, these data suggest that southern Viet Nam is subject to fairly stable endemic transmission (Hang et al.,

2010; Rabaa et al., 2010b; Raghwani et al., 2011; Thai et al., 2005, 2010a). In contrast, dengue is considered to be ‘emerging’ in northern Viet Nam, where annual incidence rates show an increasing trend over the previous decade (Cuong et al., 2011). While the clinical burden of dengue across Southeast Asia generally falls on children less than 15 years of age, 85% of reported dengue cases in Hanoi occur in adults (World Health Organization Western Pacific

Region Office, 2008). This suggests that DENV transmission has not yet become endemic in northern Viet Nam. The sub-tropical climate in northern Viet Nam, which experiences hot, humid summers and cool, dry winters, relative to the consistent heat and distinct rainy and dry seasons that occur in tropical southern Viet Nam likely contributes to the apparent lack of endemic DENV transmission. In particular, winter temperatures may reduce vector feeding activity and mosquito infection rates while lengthening the extrinsic incubation period and shortening vector lifespan (Derrick and Bicks, 1958; Lambrechts et al., 2011; Scott et al., 2000;

Watts et al., 1987), thus decreasing DENV transmission such that the likelihood of DENV being maintained in the population is reduced despite low levels of immunity in the population.

Revealing how dengue viruses move between geographic localities is integral to understanding its epidemiology in both endemic and epidemic areas. While the molecular epidemiology of DENV has previously been investigated in southern Viet Nam and neighboring countries (Jarman et al., 2008; Rabaa et al., 2010b; Raghwani et al., 2011; Schreiber et al.,

2009; Wu et al., 2011), the DENV population of northern Viet Nam has not yet been described. 95

Understanding the epidemiology and evolution of this emerging pathogen in a population at the margins of transmission will provide valuable insights into the process by which DENV transitions from epidemic to endemic transmission, and may reveal general factors that influence pathogen emergence in human populations.

The aim of this study was to investigate the spread of DENV-1 across Viet Nam over the course of a decade and determine the routes by which viral populations enter northern and central Viet Nam. For this, we utilized envelope (E) gene sequence data obtained from

Vietnamese hospitals and those gathered in other studies across Southeast Asia. These data are unique in that they include the first DENV sequences available for phylogenetic analysis in northern Viet Nam. We utilized a large data set (n = 1,765 sequences) of DENV-1 isolates collected from across Southeast Asia, where Genotype I has been the dominant circulating

DENV-1 lineage since at least 1980. With these data we investigated the movement of this lineage into Viet Nam and addressed the following questions: (i) Does highly endemic southern

Viet Nam act as a source population for DENV circulating in other regions of the country? (ii) Do

DENV populations persist over multiple seasons in central and northern Viet Nam? (iii) What factors determine the patterns of dispersal of DENV lineages to new environments across Viet

Nam and within Southeast Asia?

Results

Phylogeography of DENV-1 in Southeast Asia and Viet Nam

We determined the E gene sequences of DENV-1 isolates collected from 81 dengue patients reporting to Vietnamese hospitals throughout the country and combined these with related Southeast Asian DENV-1 E gene sequences collected between 1997 and 2009. The full data set included 46 sequences from northern Viet Nam (1998-2009), 34 sequences from central Viet Nam (2004-2009), and 461 sequences from southern Viet Nam (2003-2008), and

70, 63, and 31 envelope gene sequences from Thailand (1997-2007), Cambodia (2000-2008), and Singapore (2003-2008), respectively. All viruses were collected subsequent to a clade replacement event that occurred within the DENV-1 population in Thailand (and likely across the entire region) in the mid-1990s and has been attributed to enhanced transmission capacity within the vector (Lambrechts et al., 2012). This appears to have been the primary DENV-1 lineage circulating in mainland Southeast Asia for at least a decade, although the lack of samples from Cambodia and Viet Nam in early years prevents investigation of the means by which this lineage initially spread through the region. Importantly, however, Thailand,

Cambodia, and Viet Nam all experienced the co-circulation and maintenance of multiple lineages for several years and importation of novel viruses from other countries (Figure 4-1).

(A) (C) Clade 7 China Clade 5 Myanmar RRD

Gulf Laos NCC of Tonkin Clade 2 Clade 4 Thailand SCC

CHL Cambodia

Gulf DNB Of Thailand MKD HCMC

South China Sea Andaman Sea

1985 1990 1995 2000 2005 2010

Malaysia

Singapore

(B) 35 30 Temp. 25 (°C) 20 15 10 Jan Mar May Jul Sep Nov Clade 1

Precip. 40 (cm) 20

0 Jan Mar May Jul Sep Nov Clade 6

Figure 4-1. Regional phylogeography among 705 DENV-1 genotype 1 sequences isolated in Southeast Asia from 1998 to 2009. A) Map of Southeast Asia. B) Monthly averages of climate factors in the three primary regions of Viet Nam (mean maximum and mean minimum monthly temperatures, mean precipitation). Data for the largest city in each region (Hanoi, Danang, Ho Chi Minh City) were obtained from the World Meteorological Organization (http://worldweather.wmo.int/) and are colored as in A. C) Maximum clade credibility (MCC) tree showing phylogeographic relationships among Southeast Asian DENV-1 E gene sequences. Branch colors correspond to locations indicated in the map (Singapore in purple). Closed diamonds indicate posterior probability support ≥ 0.85. Open diamonds indicate ancestral location state probability ≥ 0.85.

The inclusion of contemporaneous DENV-1 sequences from Thailand (2005-2007) in this analysis showed that at least two lineages, closely related to previous Thai isolates (1997-

2002) and not derived from viral populations present elsewhere in mainland Southeast Asia, co- circulated in Thailand. Our phylogeographic analyses provided no support for Cambodia or Viet

Nam acting as a source of recent DENV-1 lineages circulating in Thailand [mean Markov Jump counts (95% highest posterior density [HPD]): KH to TH, 0.16 (0, 1); South VN to TH, 0.04 (0,

0); Central VN to TH, 0.08 (0, 1); North VN to TH, 0.54 (0, 3)], while moderate support was provided for migration routes from Thailand to Cambodia and from Cambodia to southern Viet

Nam (Table 4-1, Supplementary Table 4-2).

Table 4-1. Viral migration patterns across the complete data set (full tree) and in Vietnamese clades only. The number of viral introductions is represented by Markov Jump counts (posterior expected number of state transitions, with 95% highest posterior density (HPD) intervals between location in Southeast Asia). Only significant Markov Jump counts are shown.

South VN Regional TH to KH SG to KH to to Central South VN to North VN Model North VN South VN VN Full tree 4.9 (2, 7) 2.3 (1, 4) 8.7 (6, 11) 14.7 (13, 17) 14.4 (11, 17) Clade 1 3.0 (2, 3) 1.9 (1, 2) ! ! ! Clade 2 2.1 (1, 3) 3.6 (2, 6) Clade 4 7.4 (6, 9) 5.0 (3, 7) ! ! ! Clade 5 2.0 (1, 2) Clade 6 2.2 (1, 3) ! ! ! ! Clade 7 !! 2.0 (1, 3) !! !! !!

Local HCM to HCM to HCM to MKD to TH to KH KH to HCM SE to RRD HCM to SE MKD to SE Model SCC RRD MKD HCM Full tree 4.3 (1, 7) !! 6.9 (3, 10) 13.6 (10, 17) 6.0 (3, 9) 8.2 (4, 10) 12.3 (5, 18) 52.5 (43, 63) 11.1 (5, 16) 33.5 (21, 46) Clade 1 2.8 (1, 4) 6.0 (2, 9) 4.5 (1, 8) 4.6 (1, 7) 22.1 (17, 27) ! ! ! ! ! Clade 2 !! !! !! !! !! !! !! 17.2 (13, 21) !! !! Clade 4 7.2 (5, 9) 4.5 (2, 6) 5.4 (2, 7) 20.1 (14, 25) ! ! ! ! ! ! Clade 5 !! !! !! 1.9 (1, 2) !! !! !! 4.5 (3, 6) !! !! Clade 6

Clade 7 !! !! !! !! !! !! !! !! !! !! !

Due to differences in sampling densities over time, conclusions on the geographic origins of some lineages cannot be made. However, lineages in which contemporaneous sequences are present in both Cambodia and Viet Nam (Clades 1, 4, 5, and 6) strongly support a Cambodian origin of Vietnamese viruses (Figure 4-1). Although frequent movement of viruses between locations was observed between 1990 and 2009, very strong clustering by country and sub-national region within Viet Nam indicates that gene flow is much higher within the defined geographic areas than between them [Regional analysis – full phylogeny: AI = 0.06 (0.05, 0.08),

PS = 0.16 (0.15, 0.17), subsampled trees: averaged AI: 0.12 (0.10, 0.15), averaged PS = 0.25 99

(0.24, 0.27); Local analysis – full phylogeny: AI = 0.19 (0.17, 0.21), PS = 0.27 (0.25, 0.28), subsampled trees: averaged AI: 0.25 (0.22, 0.30), averaged PS = 0.38 (0.35, 0.40)].

At least eight distinct lineages of DENV-1 Genotype 1 entered Vietnam between 1990 and 2007 and persisted until at least 2007-2009. While all viral diversity within the country was represented in samples from the south, viral populations in the northern and central regions were less diverse. Nearly all viruses isolated from northern and central Viet Nam clustered within the diversity of the south with the exception of viruses isolated in northern Vietnam prior to 2003 and a single divergent virus collected in the Central Highlands in 2004 (basal to Clades

5 and 7). While we find strong support for a Cambodian origin of most Vietnamese lineages,

Clades 7 and 2 instead show possible importation from Singapore and Thailand, respectively.

Clade 7 contained two examples of importation of novel lineages from Singapore into northern

Viet Nam followed by localized, short-term transmission during a single year and apparent fade- out with the onset of winter (Table 4-2), when temperatures are low and vector populations are expected to be reduced. These analyses also suggested migration of viruses from Singapore into Thailand in 2004 followed by sustained co-circulation with indigenous Thai DENV-1 through the 2007 dengue season (Figure 4-1).

Table 4-2. Estimates of the Time to the Most Recent Common Ancestor (TMRCA) and the time of the last viral isolate of clusters circulating in central and northern Viet Nam.

100

Clade Location Mean TMRCA (95% HPD) Most recent isolation date within cluster 1 Central Viet Nam 2008.9 (2008.6, 2009.0) 2009.4 North Viet Nam 1990.5 (1987.0, 1993.7) 2002.5 2 Central Viet Nam 2003.9 (2003.4, 2004.2) 2009.5 2004.4 (2003.9, 2004.7) 2004.8 North Viet Nam 2008.4 (2008.0, 2008.7) 2008.8 2008.5 (2008.1, 2008.7) 2008.8 4 2006.6 (2006.3, 2006.9) 2007.3 2008.5 (2008.0, 2008.8) 2009.5 Central Viet Nam 2008.5 (2008.1, 2008.9) 2009.3 2008.8 (2008.3, 2009.0) 2009.6 5 NA !! !! 2008.5 (2008.5, 2008.6) 2008.7 6 North Viet Nam 2009.1 (2008.8, 2009.4) 2009.9 2006.0 (2005.9, 2006.3) 2006.5 7 North Viet Nam 2008.1 (2007.7, 2008.6) 2008.9 !

A previous analysis of DENV-1 within southern Viet Nam indicated that Clade 2 became established there in 2002 (Raghwani et al., 2011). The addition of isolates from northern Viet

Nam in this study showed a considerably longer history of this lineage in the country, beginning in the late 1980s/early 1990s. Ancestral state reconstruction suggested that this lineage migrated from Thailand into northern Viet Nam, but Markov Jump counts at the basal node are quite low (Regional model: 0.18, Local model: 0.40) due to the existence of only a few sequences from viruses recovered during the early period of invasion and long branches between lineages. This was investigated further in clade-specific analyses and is discussed below.

Viral migration within Viet Nam

Of six viral lineages involved in transmission within the central and northern regions of

Viet Nam, five showed invasion and dispersal throughout the country and frequent movement between areas of interest. To investigate the spread of DENV-1 genotype 1 within Viet Nam, we estimated Markov Jump counts between locations across the full phylogeny and in subsampled data sets, and analyzed specific viral clades to investigate fine-scale spatial and

101

temporal patterns of dispersal. Using the Regional asymmetric phylogeographic migration model, we determined that southern Viet Nam was the likely source population for Vietnamese viruses in Clades 1, 4, 5, and 6 and for Clade 2 viruses isolated after 2002 (Table 4-1). No support was found for viral migration between the central and northern regions of the country

[mean Markov Jump counts across the full phylogeny (95% HPD): central VN to northern VN,

0.23 (0, 1); northern VN to central VN, 0.26 (0, 1)]. Finer scale spatial analysis using the Local model showed that DENV tended to move between neighboring areas, but also implicated Ho

Chi Minh City as the primary source population for the entire country. Significant migration was detected from this densely populated urban area into the surrounding Mekong Delta and

Southeast regions, as well as to the South Central Coast and the distant Red River Delta, which includes the capital city, Hanoi (Table 4-1, Figure 4-2). These relationships were consistent across most clades and within sub-sampled data sets (Supplementary Table 4-3), which suggested that inferred relationships are not an artifact of dense sampling in the south.

102

Clade 1 Clade 2 Clade 4

RRD

NCC

SCC

CHL

MKD HCMC

Clade 5 Clade 6 Clade 7

Singapore

Figure 4-2. Branch-specific Markov Jump counts indicating significant migration and establishment of viral lineages across Viet Nam. The number of arrows corresponds to the number of migration events resulting in successful establishment of transmission in the recipient population with branch-specific Markov Jump probability ≥ 0.85.

The southern Mekong Delta and Southeast regions also acted as secondary centers of viral diversity within the country. The Mekong Delta region is the inferred entry site of Clade 1 into Viet Nam and experienced long-term transmission of these viruses as well as sub-lineages

103

in Clades 2 and 4. However, viruses only disseminated from this region into areas with which it shares borders, namely Ho Chi Minh City and the Southeast (Table 4-1, Figure 4-2). The

Mekong Delta acted as a significant source for these populations across the full phylogeny based largely on several migration events and subsequent establishments in Ho Chi Minh City in Clade 1 (Table 4-1, Figure 4-3), as well as a number of migrations into Ho Chi Minh City in

Clades 1 and 4 and to the Southeast in Clades 1, 2 and 4 (each represented by a single isolate). Notably, no links were detected between the Mekong Delta and central or northern viral populations across the entire phylogeny. In contrast, the Southeast region did not appear to play a significant role in maintaining diversity within the south and instead generally acted as an acceptor of viruses from Ho Chi Minh City and, to a lesser extent, from the Mekong Delta.

However, analysis of the full data set revealed this area as a significant source of viral populations appearing in the Red River Delta (Table 4-1), and branch-specific Markov Jump counts suggest this area as an independent source of established DENV populations in the Red

River Delta and North Central Coast in Clades 4 and 6 (Figure 4-2).

104

Clade 6

2008 2009 2010

Clade 1

Clade 2

2004 2006 2008 2010 1990 1995 2000 2005 2010

RRD

NCC

Clade 4

SCC

CHL

SE MKD HCMC

2002 2004 2006 2008 2010

Figure 4-3. Local phylogeography in a sample of Vietnamese DENV-1 clades. Maximum clade credibility (MCC) trees show phylogeographic relationships among Vietnamese DENV-1 E gene sequences. Branch colors correspond to locations indicated in the map of Viet Nam. Closed diamonds indicate posterior probability support ≥ 0.85 on relevant nodes. Open diamonds indicate ancestral location state probability ≥ 0.85.

105

While Clades 1, 4, 5, and 6 showed similar patterns of invasion from Cambodia into southern Viet Nam and rapid dissemination across the country between 2001 and 2007, Clade

2 appeared to have entered Viet Nam in the late 1980s/early 1990s, followed by a period of extended transmission with little change in diversity prior to 2002. Ancestral state reconstruction weakly suggests that this lineage entered the country in the North (North Central

Coast) and persisted there for over a decade before reaching southern Viet Nam in 2002, likely via migration into Ho Chi Minh City. Due to heterogeneity in sampling densities and times, support for a northern origin of this lineage was limited, with root location state probabilities of

0.35 and 0.74 for the Local (North Central Coast) and Regional models (northern Viet Nam).

Clade-specific Markov Jump counts also provided low support for migration from north to south

[Regional Model, 1.0 (0, 2); Local model, 0.84 (0, 2)]. Thus, the origins of this lineage in Viet

Nam cannot be determined at present.

Invasion and maintenance of DENV-1 in northern and central Viet Nam

To determine whether distinct viral populations persisted in northern and central Viet

Nam over multiple seasons, we investigated the timing of invasion and establishment of novel viral sub-lineages (that is, viral clusters of two or more sequences originating from the same location as inferred by ancestral state reconstruction) and the potential co-circulation of distinct lineages in these areas. Viral invasion and establishment were detected in northern Viet Nam in

1990, 2004, 2008, and 2009, and in central Viet Nam in 2003, 2006, and 2008. Except in Clade

2, these data suggest that invading lineages in northern Viet Nam did not persist in the region over multiple dengue seasons (Table 4-2).

Of eight northern transmission clusters, the times to common ancestry (TMRCAs) of three clusters originating in Singapore (Clade 7, 2006 and 2008) and the Southeast (Clade 6,

2009) suggest that these lineages entered the north in January/February (Table 4-2), when 106

temperatures and precipitation in most of northern Viet Nam are at their winter lows. While we would expect winter temperatures to greatly limit transmission during this early period of the year, all three of these lineages survived into the hot, rainy season, until at least July (Clade 7,

2006) and November (Clade 7, 2008; Clade 6, 2009). The remaining northern invasion events occurred in the middle of the year via viral migration from the south (Ho Chi Minh City and the

Southeast). This period matches the timing of seasonal increases in dengue cases throughout the country and in Hanoi (Cuong et al., 2011), when viral migration and establishment are more likely due to suitable climate conditions for the vector in the north.

In contrast to the north, six central Vietnamese transmission clusters (one each in 2003,

2006, and four clusters in 2008) suggested that seasonal invasion in this region occurred during the dengue season in the second half of the year (Table 4-2), with uninterrupted chains of transmission often maintained into the next dengue season. Among these persistent lineages, one primarily central Vietnamese sub-lineage in Clade 2 became established in the Central

Highlands around 2003 and was later isolated in the South Central Coast region, where it was maintained into the 2009 dengue season. The concurrent invasion and co-circulation of multiple clades was also common in this region (two in 2006, Clades 2 and 4; five in 2008-2009, Clades

1, 2 and 4).

Testing models of DENV phylogeography

To determine how viral transmission routes within Southeast Asia were influenced by population and geographic factors, we compared the fit of the data of a variety of phylogeographic models related to human population size, distance, and DENV transmission intensities. Although slightly different results were obtained for the Local and Regional models

(Figure 4-4, Table 4-3), phylogeographic models that utilized simple distance- and population- based gravity model priors generally showed a good fit to the data relative to the equal rates 107

and single factor models (distance, population, etc.), as indicated by lower values of marginal log likelihood AICM estimates. However, the incorporation of the Relative Endemicity factor

(REf), which reflects DENV transmission intensities in both recipient and donor populations, further improved the fit of gravity models to the data under the Regional geographic scheme.

The REf alone showed a consistently good fit to the data relative to other single factor models, and the best overall fit was shown by the REf + Gravity Model (Model 6) both for the full phylogeny and randomly subsampled data sets. Importantly, the performance of the Sample

Size model (Model 7) was not significantly better than other models of viral movement in the full or subsampled data sets. This indicates that sample sizes were not the most important factor determining the patterns of transmission observed here, and hence that spatial bias in sampling did not significantly influence the results of these analyses.

108

25 Full Tree, Local model * 20 Subsample average, Local model * Full Tree, Regional model *

15 Subsample average, Regional model

10 * 5

-5

marginal log likelihood (AICM) likelihood log marginal

-10

-15

-20

-25 REf Equal Distance Population Gravity REf Sample Rates + Size Gravity Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7

Figure 4-4. Differences in relative marginal log likelihood estimates for fitting of the asymmetric phylogeographic model using Akaike’s information criterion through MCMC (AICM). Results show model fittings of Regional and Local phylogeographic model priors for the full DENV-1 genotype 1 data set and for 10 data sets including a maximum of 50 randomly subsampled sequences per location. Models with the best performance for each geographic and sampling scheme are indicated with a star.

109

Discussion

This study documents the dispersal of DENV-1 to populations across Viet Nam and provides evidence that viral populations are regularly introduced to northern Viet Nam from external populations, although endemic transmission has not been established. Strong clustering at country, regional, and local scales indicated that the viral diversity present in a given area is determined primarily by local gene flow, in which vector-mediated dispersal is likely to be an important factor. Frequent movement between neighboring locations under the Local geographic model may reflect a combination of human movement and vector dispersal in the movement of DENV between human populations in close proximity, as observed at smaller geographic scales (Balmaseda et al., 2010; Rabaa et al., 2010b; Raghwani et al., 2011).

However, many of the medium- and long-distance migration events observed here occur on a timescale of less than one year and we found no support for viral migration between the central and northern regions of Viet Nam, as expected if vector-based seasonal movement up the coast and long-term maintenance of viral populations in each locale were the primary factors responsible for the viral diversity observed in these populations. Hence, these findings suggest that local viral transmission patterns may be in part determined by the vector, but that mosquito- mediated viral migration is not driving the long-range dispersal of DENV from south to north.

Our results also show that the DENV population in sub-tropical northern Viet Nam is characterized by regular seasonal invasions of lineages from the highly endemic south, with no detectable persistence into the following dengue season. Regardless of the time of year in which invasion occurred, our phylogenetic data suggest that invading lineages experience severe seasonal bottlenecks and regular fade-out in northern Viet Nam at the end of each year, when temperatures in much of the north drop below those considered optimal for survival of the vector and efficient transmission of the virus by Aedes aegypti (Derrick and Bicks, 1958;

Lambrechts et al., 2011; Watts et al., 1987). These findings are supported by epidemiological

110

data indicating that this region supports an epidemic dynamic with little to no transmission outside of the ‘dengue season’ (Cuong et al., 2011). In addition, that three separate introductions to the Red River Delta in 2008 resulted in established transmission clusters that co-circulated during the dengue season suggests that the importation of viruses from endemic regions is largely responsible for seasonal dengue outbreaks in sub-tropical northern Viet Nam.

In contrast, although central Viet Nam commonly experiences seasonal invasions of DENV from the densely populated south, these viral populations are regularly maintained over multiple years and co-circulate with newly invading viruses. Southern Viet Nam and the central areas from which most of our isolates were obtained share similar tropical climates that are conducive to year-round maintenance of mosquito populations and efficient viral replication and transmission within the vector, and thus may explain the persistence of DENV populations within this region. Although climate is likely to be a major factor in the maintenance of DENV in these areas (Gubler et al., 2001; Patz et al., 1996; Shang et al., 2010; Shope, 1991), the importation and short-term establishment of DENV in the north during the coldest and driest months of the year indicate that these seasonal conditions do not sufficiently reduce vector populations or viral replication within the vector to fully block DENV transmission. While much of northern Viet Nam is dominated by Aedes albopictus (Higa et al., 2010), Aedes aegypti populations are prominent in Hanoi and persist at reduced levels indoors during the winter

(unpublished vector control data). These localized indoor vector populations may permit low levels of transmission when outdoor temperatures are low, and could result in future establishment of endemic transmission as dengue activity increases in the city.

Clade 2 represents one possible example of DENV persistence in northern Viet Nam.

Our analysis suggests that this lineage entered through the north and became established in the

North Central Coast, where it may have persisted for over a decade prior to its invasion and establishment in the south. Although the possibility of long-term transmission in northern Viet

Nam cannot be excluded, the lack of contemporaneous sequence data from the south and the 111

distant relationships between these basal sequences result in a lack of resolution at this early time point and thus low support for an inferred ancestral location of the lineage. Additionally, a number of other sequences sampled from early time points in the north (Clade 4) also fall at basal positions in the phylogeny, although generally within the diversity of southern Viet Nam.

Thus, the relationships among these northern sequences do not necessarily indicate their persistence. Instead, they may represent multiple importations from DENV populations in the south in the 1990s that experienced a significant bottleneck in the early 2000s, prior to the entry and rapid establishment of Clades 1, 4, 5, and 6, although the lack of any signature of intermediate diversity in the highly sampled south argues against this hypothesis.

Among the more recently sampled sequences, there are a number of interesting patterns of viral dispersal within Viet Nam. Ho Chi Minh City and the Mekong Delta experience high transmission intensities and were highly sampled relative to the other populations

(Supplementary Table 1). Previous studies indicated that Ho Chi Minh City acted as a clear source of DENV diversity in the south (Rabaa et al., 2010b; Raghwani et al., 2011). Here, we show that the role of the city as a primary source population extends across the entire country.

The Mekong Delta, in contrast, was a source of DENV for populations only within the south.

While fewer samples were available for the Southeast, this region was suggested as a significant source of viruses circulating in the north (Red River Delta and North Central Coast) using the full data set. The Southeast region maintains high population densities in areas adjacent to HCMC and is the site of increasing numbers of industrial parks and emerging regional economic centers (“No Title,” 2009). Economic incentives fuel human migration to

HCMC and Southeast, which accept large numbers of migrants (mostly young adults) from the

Mekong Delta, Central Coast regions, and the Red River Delta (“Migration and urbanization in

Vietnam: patterns, trends and differentials,” 2010). Importantly, young adults from the north may be dengue-naïve due to reduced exposure and a lack of endemic transmission in the north, and thus may be at high risk of infection and illness after arriving in the southern industrial areas 112

(Ooi et al., 2006). Travel between the economically important regions of Ho Chi Minh City, the

Southeast, and Hanoi likely creates opportunities for viruses from this area to invade the distant northern population. Indeed, the most recent common ancestor of multiple seasonal viral clades in the north, including lineages from southern Viet Nam and Singapore, is often inferred during the early months of the year. This timing corresponds with the Tet holiday during which travel across Viet Nam is increased, possibly fueling the dissemination of the virus into new areas. Dengue transmission is low nationwide during this period but persists in densely populated areas of the south. That these winter invasions were not observed in the middle of the country may speak to higher levels of cross-immunity in the population during the low dengue transmission season, to a lower connectivity of this population to other regions, or to insufficient sampling this area.

Previous studies in southern Viet Nam have suggested that viruses move through the area along somewhat predictable human migration routes based on estimated connectivity between populations (Rabaa et al., 2010b; Raghwani et al., 2011). These studies assessed

DENV movement on fine spatial scales and in a region with relatively homogeneous immune profiles, and the applicability of these migration models in areas with vastly different transmission patterns and levels of immunity was not investigated. Here, we compared a variety of epidemiological models reflecting patterns of human and mosquito movement and showed that models incorporating patterns of human migration fit the data relatively well. The simple gravity model showed the best fit to the data under the Local geographic model and suggests that the movement of DENV between areas can largely be explained by the connectivity of the human populations at this scale. Accordingly, populations may show greater connections with a distant city than with a closer, smaller population based on the draw of the city to external populations; this is reflected in the frequency of medium- and long-distance viral migration events between population centers in our phylogenies. However, the addition of even a crude factor related to population immunity and transmission intensities (REf) in the recipient 113

and donor populations improves upon the fit of the gravity model in the Regional geographic analysis. This discrepancy is not surprising given that REf estimates were based on limited data and extrapolated to all locations within a region, and thus may not reflect complex heterogeneities in immunity and transmission at finer spatial scales. The finding that the REf performs well against other factors indicates that transmission intensity in both the recipient and donor populations plays an important role in shaping the likelihood of viral invasion. The co- circulation and frequent invasion of DENV-1 lineages across all populations suggests that large susceptible populations exist across the region even in areas of high transmission intensity. In this case the role of immunity may be limited relative to the number of infections circulating in a given area over time. The REf estimation is a greatly over-simplified indicator of relative population immunity and transmission intensities, and is based on extrapolation of the mean age of infection from multiple studies that used diverse methods and surveillance data sources.

However, these data may not be directly comparable, and it is clear that this crude estimation technique and the underlying data should be refined further.

We employed multiple methods to control for sampling bias, which generally showed that heterogeneities in sampling densities were not strongly affecting the results of our migration analysis. However, the small number of sequences obtained from areas in northern and central

Viet Nam may have been insufficient to capture long-term persistence of rare lineages co- circulating with the dominant invading viruses. Additionally, differences in sampling over time hindered inference at deep locations in the tree and prevented us from determining the origins of Clade 2, a possible long-term northern DENV lineage. Given recent increases in DENV incidence in northern Viet Nam and its geographic position at the margins of endemic transmission, additional sampling in this area is clearly warranted. Greater understanding of the processes by which DENV invades sub-tropical northern Viet Nam and the potential of this area to maintain long-term autochthonous viral transmission would yield important information relevant to sub-tropical and temperate areas at risk of DENV invasion worldwide. 114

Materials and Methods

Patient population and envelope gene sequencing

Dengue viruses were recovered from suspected dengue patients presenting to hospitals across Vietnam as part of routine diagnostic serology. A total of DENV-1 viruses were collected from hospitals in north, central, and southern Viet Nam and sequenced according to the

Queensland University of Technology protocol. DENV were isolated from patient sera using

C6-36 Aedes albopictus mosquito cells (Thu et al., 2004). The E gene of these isolates were amplified by RT-PCR and sequenced as described previously (A Nuegoonpipat et al., 2004).

Sequences are catalogued in GenBank under accession numbers XXX-XXX. Twenty additional

DENV-1 viruses were collected from cases presenting to hospitals in Hanoi and sequenced according to the OUCRU Hanoi protocol. These have been assigned Genbank accession numbers HQ591537-HQ591556.

Phylogenetic analysis

The DENV-1 isolates from northern and central Viet Nam sequenced here were combined with DENV-1 envelope (E) gene sequences collected from GenBank to comprise all

DENV-1 sequences from across Asia for which the year and country of sampling were known.

Nucleotide alignments of 1765 full-length DENV-1 E gene sequences (1485 nt), including 80 isolates from northern and central Viet Nam, were manually constructed using Se-AL (Rambaut,

2002). To infer phylogenetic relationships for the complete data set of DENV-1 sequences and to identify geographic regions with phylogenetic links to northern and central Viet Nam, we utilized the maximum likelihood (ML) method available in PhyML, incorporating a GTR model of nucleotide substitution with gamma-distributed rate variation among sites and a heuristic SPR

115

branch-swapping search algorithm (Guindon and Gascuel, 2003). This initial analysis indicated that all northern and central Vietnamese DENV-1 sequences belong to a primarily Southeast

Asian subset of Genotype I, comprising viral sequences from Thailand, Cambodia, Singapore, and southern Vietnam. A second alignment was then constructed using 80 E gene sequences from northern and central Viet Nam and 625 unique E gene sequences that were isolated from the surrounding regions between 1997 and 2009 and for which the exact date of sampling was known. A phylogenetic analysis was undertaken using the Bayesian Markov Chain Monte Carlo

(MCMC) method implemented in BEAST and incorporating the date of sampling (Drummond and Rambaut, 2007). For this analysis we employed a codon-structured SDR06 model of substitution, a relaxed molecular clock as in (Raghwani et al., 2011), and a Bayesian skyline prior (BSP; 5 piecewise constant groups). The MCMC chain was run for 100 million iterations, with sub-sampling every 10,000 iterations. All parameters reached convergence, as assessed visually using Tracer (v.1.5). The initial 10% of the chain was removed as burn-in, and maximum clade credibility (MCC) trees were summarized using TreeAnnotator (v.1.6.2).

Spatial analysis

To investigate the routes of invasion of DENV-1 into northern and central Viet Nam, we categorized the geographic areas in Viet Nam using (i) a ‘Local geographic model’ – categorized by governmentally-defined region (MKD: Mekong Delta, HCM: Ho Chi Minh City,

SE: Southeast, CHL: Central Highlands, SCC: South Central Coast, NCC: North Central Coast,

RRD: Red River Delta), and by country outside of Viet Nam (KH: Cambodia, SG: Singapore,

TH: Thailand), and (ii) a ‘Regional geographic model’ – categorized by larger regions (North:

RRD and NCC, Central: CHL and SCC, South: SE, HCM and MKD; abbreviations defined below) and elsewhere by country as in the previous scheme. We inferred rates of viral migration between regions using an asymmetric model of discrete diffusion across Southeast 116

Asia and within Viet Nam (Lemey et al., 2009a). Posterior distributions of trees were estimated under a phylogenetic model using the MCMC method implemented in BEAST using BEAGLE

(Ayres et al., 2012; Drummond and Rambaut, 2007). This model incorporated the date of sampling and a relaxed molecular clock, Bayesian skyline prior, and the SRD06 codon position model as described above. The MCMC chain was run for 100 million iterations, with subsampling every 10,000 iterations, and all parameters reached convergence. The initial 10% of the chain was removed as burn-in, and Maximum Clade Credibility (MCC) trees including ancestral location-state reconstructions were summarized using TreeAnnotator (v.1.6.2). The expected number of location state transitions conditional on the location-related sequence data was determined using Markov Jump counts, summarized per branch and for the complete evolutionary history. Markov Jump counts of the expected number of geographic state transitions along branches provide a quantitative measure of gene flow between regions, representing successful viral introduction from one region to another, and are not heavily influenced by single isolate introductions (Minin and Suchard, 2008a, 2008b). Finally, parsimony score (PS) and association index (AI) tests were utilized to assess the extent of geographic structure across all trees using the Bayesian Tip-association Significance Testing

(BaTS) program (Parker et al., 2008) based on the posterior distribution of trees generated in the BEAST analysis described above.

To account for potential sampling bias, posterior distributions were also estimated as above for ten data sets that were randomly subsampled, with replacement, to include no more than 50 sequences per region for each of the ten major geographic regions in the data set

(Cambodia, Singapore, Thailand, and Viet Nam: Mekong Delta, Ho Chi Minh City, Southeast,

Central Highlands, South Central Coast, North Central Coast, Red River Delta). Subsampled data sets included 338 sequences. In addition, six major clades containing northern and central

Vietnamese isolates were identified in trees inferred from the full data set and were analyzed individually using the asymmetric discrete diffusion model to assess potential differences in the 117

spatial and temporal patterns of diffusion among them, to identify routes of invasion into northern and central Viet Nam, and to determine whether lineages were maintained over multiple years in these newly sampled populations.

Hypothesis-based spatial analysis

To investigate specific hypotheses of DENV epidemiology in Viet Nam we set rate priors to specific values to construct a series of (asymmetric) phylogeographic models that might reflect the epidemiology of DENV in Viet Nam. These models were analyzed using both the full and subsampled data sets for the Regional and Local geographic models. These models were:

(i) a geographic diffusion model that assumes equal rates of viral migration between all regions of interest (Model 1, Equal Rates), (ii) a model based on the connectedness of the populations in question using the inverse of the Euclidian distance between the largest cities in each region

(Model 2, Distance), (iii) a population size-based model, utilizing the population of the largest city from which sequences were collected in each region as representative of the influence of that region in attracting migration from other locations (Model 3, Population) (Singapore, 2011;

Thailand, 2009; “The 2009 Vietnam population and housing census: completed results,” 2010,

“Cambodia General Population Census 2008,” 2010), and (iv) a gravity model incorporating geographic distances and population size data from both the recipient and donor locations

(Model 4: Gravity Model) (Rabaa et al., 2010b; Xia et al., 2004).

We utilized two additional models that incorporated aspects of human immunity and transmission intensity, reflecting the fact that serotype-specific and short-term cross-protective immunity may impede viral invasion and establishment (Lourenco and Recker, 2010). Given that the average/median age of infection acts as an indicator of transmission intensity

(Anderson and May, 1991), we determined the ratio of the mean age of reported cases (DF and

DHF) in the recipient population to that in the donor population, and utilized this as a crude 118

indicator of the likelihood of viral invasion based on relative transmission intensities and endemicity (Anders et al., 2011; Cummings et al., 2009; Cuong et al., 2011; Egger et al., 2008;

Schmidt et al., 2011; Thai et al., 2011; Vong et al., 2010). We refer to this ratio as the Relative

Endemicity factor (REf) (Model 5, REf). Due to a lack of age-specific case data from local populations, these estimates were extrapolated to all locations within the same geographic region. In Model 6, we integrated this immunity measure as a proportionality constant in gravity model calculations (REf + Gravity Model).

Finally, we investigated the effects of sample size on phylogeographic inference using a rate matrix based on the sample size of the donor population (Model 7, Sample Size). Model priors were normalized (mean one and unit variance) and incorporated into asymmetric matrices that allow for directional rates to vary between individual location pairs. A posterior simulation- based analogue Akaike’s information criterion through MCMC (AICM) was implemented using likelihoods specific to the geographic model priors, and marginal log likelihood estimates for each model were compared to determine the best fit model to the data in hand (Baele et al.,

2012; Raftery et al., 2007).

Ethical approval

Genotyping of de-identified viruses collected in this study was undertaken under Human

Research Ethics Approval 0700000910 from the Queensland University of Technology.

119

SUPPLEMENTARY TABLES

Supplementary Table 1. Geographic distribution of DENV-1 sequences collected in Viet Nam from 1998 to 2009. Location codes are as follows: RRD, Red River Delta; NCC, North Central Coast; SCC, South Central Coast; CHL, Central Highlands; SE, Southeast; HCM, Ho Chi Minh City; MKD, Mekong Delta. QUT indicates viruses collected and sequenced using the Queensland University of Technology protocol. OUCRU Hanoi and OUCRU HCMC indicate collected and sequenced using the protocols of the respective Oxford University Clinical Research Unit.

Location Number of Sequences Year Source 1 2003 QUT 1 2004 QUT RRD 23 2008 20 OUCRU Hanoi; 3 QUT 5 2009 QUT 30 All 1 1998 QUT 1 1999 QUT 3 2002 QUT 1 2003 QUT 4 2004 QUT NCC 2 2006 QUT 1 2007 QUT 2 2008 QUT 1 2009 QUT 16 All 1 2005 QUT 5 2006 QUT 7 2007 4 QUT; 3 OUCRU HCMC SCC 4 2008 QUT 12 2009 QUT 29 All 5 2004 QUT CHL 5 All 6 2006 OUCRU HCMC 6 2007 OUCRU HCMC SE 14 2008 OUCRU HCMC 26 All 9 2003 OUCRU HCMC 1 2004 OUCRU HCMC 13 2005 OUCRU HCMC HCM 82 2006 OUCRU HCMC 87 2007 OUCRU HCMC 99 2008 OUCRU HCMC 291 All 1 2004 OUCRU HCMC 29 2006 OUCRU HCMC 86 2007 OUCRU HCMC MKD 25 2008 24 OUCRU HCMC; 1 QUT 3 2009 QUT 144 All

120

Supplementary Table 2. Entry and persistence of DENV-1 lineages in Viet Nam. Entry locations with >85% root state probability support are indicated in bold. Inferred region of Inferred local site of TMRCA - Vietnamese sequences Most recent entry into Viet Nam entry into Viet Nam (95% HPD) isolation date (state probability) (state probability) Clade 1 South (0.99) MKD (0.62) 2004.1 (2001.8, 2006.3) 2009.5 Clade 2 North (0.69) NCC (0.52) 1990.5 (1987.0, 1993.7) 2009.5 Clade 4 South (0.62) HCM (0.90) 2001.5 (2000.2, 2002.7) 2009.6 Clade 5 South (0.98) HCM (0.97) 2003.5 (2003.0, 2003.8) 2009.5 Clade 6 South (0.97) SE (0.75) 2007.7 (2007.3, 2008.0) 2009.9 North (0.96) NCC (1.0) 2006.0 (2005.9, 2006.3) 2006.5 Clade 7 North (0.99) RRD (0.97) 2008.1 (2007.7, 2008.4) 2008.9 !

Supplementary Table 3. Viral migration patterns across the complete data set (full tree) and in subsampled data sets. The number of viral introductions is represented by Markov Jump counts (posterior expected number of state transitions, with 95% highest posterior density [HPD] intervals between location in Southeast Asia). All significant Markov Jump counts are shown. Regional SG to SG to KH to South VN to South VN to North TH to KH Within South VN (Local Model Only) Model TH North VN South VN Central VN VN Full tree 4.9 (2, 7) 2.3 (1, 4) 8.7 (6, 11) 14.7 (13, 17) 14.4 (11, 17)

Subsample 50 4.5 (1.3, 2.4 (0.9, 4) 5.7 (2.1, 9) 14.0 (10.3, 13.2 (8.6, 17) (averaged) 7.4) 17)

KH to HCM to HCM to SE to HCM to HCM to MKD to MKD to ! Local Model !TH to KH ! ! ! ! ! ! ! ! ! ! HCM SCC RRD RRD SE MKD SE HCM 33.5 (21, Full tree 4.3 (1, 7) 6.9 (3, 10) 13.6 (10, 17) 6.0 (3, 9) 8.2 (4, 10) 12.3 (5, 18) 52.5 (43, 63) 11.1 (5, 16) 46)

Subsample 50 3.6 (0.1, 2.1 (0.3, 11.2 (5.6, 4.8 (0.8, 9.3 (2.2, 7.7 (0.9, 10.2 (2.5, 15.0 (8.7, 13.9 (5.1,

(averaged) 6.7) 5.0) 16.0) 8.7) 11.1) 13.7) 17.5) 21.0) 22.4) !

121

CHAPTER 5

MODERN GLOBAL DENGUE DIVERSITY: OUT OF ASIA

Abstract

The first severe dengue/dengue hemorrhagic fever (DHF) epidemics emerged in tropical

Asia in the 1950s. Since then, this area has experienced consistently high incidence of dengue fever (DF) and DHF, and all four serotypes continuously circulate in cities and some rural areas throughout the region. Here, we utilize large data sets of dengue virus (DENV) envelope (E) gene sequences for all of the four DENV serotypes and perform a phylogeographic analysis to investigate the pathways by which DENV lineages have dispersed globally over time.

Genotype-specific datasets reveal that all modern DENV genotypes originate in Asia, primarily

South and Southeast Asia, and indicate that these regions continue to harbor considerable genetic diversity and distinct viral lineages characterized by their geographic distributions. South

Asian lineages appear to be the most geographically widespread among all DENV lineages, and suggest that virus or host populations from this area may have important characteristics that allow for this global spread. Although Southeast Asia has long been considered a primary source of DENV diversity, this is the first study to indicate a role for South Asia as a major global source of DENV. The maintenance of separate ‘local’ lineages within South Asia, maritime

Southeast Asia, and mainland Southeast Asia suggest that these may have been maintained by long-term immune-mediated competition among viruses circulating within tropical Asia.

122

Introduction

Dengue viruses (DENV) are single-stranded positive-sense RNA viruses (genus

Flavivirus) that cause significant morbidity and mortality across the tropics and subtropics.

Approximately 50 million DENV infections occur annually, mostly among the 2.5 billion people currently living in dengue-endemic countries in the tropics and sub-tropics (World Health

Organization, 2009a). Dengue viruses comprise an antigenic complex of four phylogenetically and phenotypically distinct viruses (also referred to as serotypes) based on their antigenic cross-reactivities. Molecular-clock studies suggest that distinct serotypes began to diverge no more than 2000 years ago while still in a primarily non-human primate-mosquito infection cycle

(Dunham and Holmes, 2007; Twiddy et al., 2003). Although there has been some debate over the geographic origins of DENV due to the presence of human and sylvatic cycles in both Africa and Asia, the detection of all four DENV serotypes circulating in the primate population in isolated forests of Malaysia (Rudnick, 1965, 1978; Rudnick et al., 1967) suggests that early diversification of the DENV occurred in primate populations in Asia. Southeast Asia has long shown the highest apparent DENV diversity and prevalence in human populations and has been suggested as the site of origin of the DENV (Holmes, 2009; Vasilakis and Weaver, 2008).

The emergence of the four DENV serotypes into human endemic/epidemic cycles was followed by the rapid radiation and spread of genetic diversity over the last 100 to 200 years.

This recent diversity is structured in multiple genotypes within each serotype, arbitrarily defined as having no more than 6% sequence divergence (Rico-Hesse, 1990). Using E gene sequences, phylogenetic studies corroborate the existence of these genotypic clusters and suggest a geographic component to the diversity observed within serotypes (Chungue et al.,

1995; Goncalvez et al., 2002; Holmes and Twiddy, 2003; Lanciotti et al., 1994, 1997; Rico-

Hesse, 2003; Twiddy et al., 2002a; Wittke et al., 2002). However, all of these studies were conducted a decade ago, and recent additions of sequences from poorly sampled areas are

123

expected to provide a more comprehensive view of the modern global diversity and dispersal of

DENV.

Epidemiological surveillance data suggest that the dispersal of all four DENV serotypes throughout the tropics and sub-tropics is nearly complete (World Health Organization, 2009b).

However, the finding that genotype-level diversity within a country may be a primary determinant of the spatial scope, severity, and frequency of outbreaks suggests that characterization at this level might be a valuable predictive indicator for cost-effective allocation of health systems resources and control efforts (Bennett et al., 2006; Hang et al., 2010;

Lambrechts et al., 2012; OhAinle et al., 2011; Twiddy et al., 2002a, 2002b; Zhang et al., 2005).

Hyperendemic DENV transmission (of all four serotypes) has been maintained in populations throughout Southeast Asia since at least the 1950s. The reported burden of dengue is the highest here, where large DHF epidemics now occur every 2-4 years, along with year-round but highly seasonal transmission during interepidemic periods (Cummings et al.,

2004, 2009; Goh, 1995; Huy et al., 2010; Thai et al., 2010). South Asia also shows a long history of hyperendemic transmission, although few reports of DENV transmission came out of the region prior to 1990 (Carey et al., 1964; Chaturvedi et al., 1970; Myers et al., 1970) and reported incidence has generally been quite low (World Health Organization, 2009b). The endemic disease burden here appears to be much lower than that in Southeast Asia and DHF was rarely reported prior to the 1990s despite long-term hyperendemic transmission in the region. Over the last decade, DHF epidemics in this region have become more frequent and widespread (Baruah and Dhariwal, 2011; Dorji et al., 2009; Gupta et al., 2006; Lucas et al.,

2000; Malla et al., 2008).

What remain poorly understood from serotype-level incidence data alone are how viral lineages entered each population and what levels of intra-serotype diversity each of these populations may harbor. To examine this, we utilized a large data set of publicly available

DENV envelope (E) gene sequences for all four serotypes and performed standard 124

phylogeographic analyses. These analyses allowed us to reconstruct ancestral state locations across phylogenies to visualize genetic diversity and population histories within geographic regions and to identify recurrent migration pathways between locations.

125

Results and Discussion

Global sources of DENV diversity

To investigate the potential existence of global DENV source populations, we performed a regional spatial analysis over all major human DENV genotypes and determined significant directional connections between populations. The number of significant migration pathways detected between each set of geographic locations was counted over all of the genotype- specific phylogenies to investigate whether consistent links were shown across multiple phylogenies, suggestive of a recurrent migration pathway between locations (Figure 5-1).

Sink Figure 5-1. Number of

North Central Indian Maritime East genotypes for which a America America Africa Ocean SE Asia Asia South Middle South Mainland significant migration America Caribbean East Asia SE Asia Oceania pathway was detected between each of the 12 North America regions. The source South America ● ● (donor) population is indicated on the y-axis. Central America ● ● ● The sink (recipient)

Caribbean population on the x-axis. Circle size indicates the Africa ● number of genotypes in which a significant Middle East ●

ce directional migration route r Indian Ocean between two regions was

Sou detected from the 12 South Asia ● ● ● ● ● ● ● ● genotype-specific analyses. The largest circle (maritime Maritime SE Asia ● ● ● ● ● Southeast Asia to East Mainland SE Asia ● ● ● ● Asia) indicates significant migration along this route in East Asia ● ● ● seven of 12 trees, and the

Oceania smallest circle indicates a link in a single tree only. Circles are colored according to the recipient

(‘sink’) region for the purpose of clarity.

126

First, these data indicate that the Western Hemisphere is somewhat isolated, with each region seeding viruses into other regions of the Americas; the region as a whole accepts viral populations from across tropical Asia. This is well known based on previous phylogenetic studies (Allicock et al., 2012; Carrington et al., 2005; Foster et al., 2004) and is not surprising given the relative geographic isolation of the Americas to other endemic regions. Less well known, however, are the specific routes by which DENV enters the region. Here, multiple significant links are detected between Asia and Central America, and from Asia (multiple from

South Asia) into the Caribbean. These findings suggest that viruses initially introduced to the

Americas through Central America and the Caribbean have origins in Asia, with the latter of primarily South Asian origin. The DENV populations in North and South America, however, show no significant connections with regions outside of the Americas, and thus appear to be seeded exclusively by populations in the American tropics.

On a global level, these data suggest that all global DENV populations included in this study, except those in North and South America, at some point harbored viral populations that were likely introduced directly from South Asia, with multiple trees showing migration pathways from South Asia to Africa, the Caribbean, the Middle East, the Indian Ocean Islands, East Asia, and Oceania, as well as both regions of Southeast Asia, which are suggested here as secondary global DENV sources. South Asia has frequently introduced viruses to maritime

Southeast Asia, and showed the highest number of migration events of all those shown into

Africa, the Middle East, and the Indian Ocean Islands. Maritime and mainland Southeast Asia, by contrast, rarely acted as sources of viral populations to the west, but were the most frequently detected sources of viruses for the opposite region of Southeast Asia, as well as for

East Asia and Oceania. Migration from South Asia into Southeast Asia is somewhat surprising given that both regions of Southeast Asia consistently report much higher DF/DHF incidence than all countries in South Asia (World Health Organization, 2009b) and are better sampled over 127

time. Additionally, relatively few genotype introductions were detected from these regions into

South Asia.

In the example of DENV-2, which includes the American-Asian, Asian I, American, and

Cosmopolitan genotypes (named for their distributions in early phylogenetic studies, along with the Asian II genotype not included in this study), some examples of these patterns can be observed. Figure 5-2 shows the evolutionary and spatial relationships within the American and

Cosmopolitan genotypes. The American genotype, named for its primary distribution in the

Americas prior to the recent sequencing of historical virus samples, was first detected in

Trinidad in 1953 and was not detected again until 1969, appearing in Puerto Rico, after which it moved into Central and South America (Figure 5-2a). It circulated in these areas until at least

1995 and 1996, respectively. Until recently, the origins of this genotype were unknown. Its isolation in Trinidad in 1953 coincided with a DENV-2 epidemic in the Caribbean, of which this lineage was presumably the cause. Dengue activity died down in the Americas soon after due to mosquito control efforts by the Pan American Health Organization to eradicate yellow fever from 1947 to 1963 (Gubler, 1997), but the DENV-2 American lineage was isolated again within a few years following the discontinuation of the program and spread into the late 1990s. Using historical sequences that have recently come out of India and the islands of Oceania, ancestral state reconstruction suggests that viruses of the American genotype that circulated in the

Americas after the 1960s likely had re-entered from South Asia, where they circulated continuously from 1950 to at least 2001. Ancestral state reconstruction at the root of the tree shows significantly strong posterior support (>85%) for origins of this genotype in South Asia

(specifically, India).

Although the Trinidad 1953 sequence was the first to be sampled from this lineage, ancestral state reconstruction places the root of the phylogeny in South Asia based on diversity across the posterior set of trees. The most recent common ancestor (MRCA) of the American genotype is inferred to have existed in 1923 (1910-1938). This corresponds well to historical 128

traffic between Trinidad and India. Thousands of Indian citizens were brought over to the

Caribbean by the British as servants from 1845 to 1917 (Williams, 2010). After the abolition of these practices, a wave of immigrants followed the same path to Trinidad over at least the next several years. It appears likely that the movement of people between India and Trinidad introduced the lineage isolated in Trinidad in 1953 and caused a limited outbreak over two years

(Downs, 1964) rather than this being a long-term autochthonously circulating lineage. The cessation of vector control in 1963 probably allowed a re-introduction and (re-)estabiishment of this lineage from India into the region.

The Cosmopolitan genotype also circulated in the same location (India) in South Asia from the late 1960s until at least 2007 (Figure 5-2b). The Cosmopolitan genotype was named as such for its nearly global distribution; by the early 1990s, it had been detected across Asia, and in Oceania, Africa, Central America, the Middle East, and the Indian Ocean Islands.

Interestingly, the genotype is comprised of two primary clades, clearly based on geography.

Considerable geographic dispersal is shown within each of these clades, suggesting that the lineage may be particularly fit to invade new environments (Twiddy et al., 2002a) or possibly that the human populations of these regions are particularly good at dispersing the virus globally. Given the patterns of dispersal shown in Figure 5-1 and across all trees (data not shown), either hypothesis may be supported and further study is necessary. It is worth noting that the American and Cosmopolitan genotypes circulated simultaneously for ~20 years in

South Asia, with the most recent American genotype virus detected in 2001 and the

Cosmopolitan in 2007. This may indicate a genotype replacement among DENV-2 lineages in the region, although more sampling is needed to confirm this.

129

Figure 5-2. MCC phylogenies of the DENV-2 American and Cosmopolitan genotypes according to region of sampling. Tips are colored by location of sampling. Internal branches are colored according to the reconstructed ancestral state location estimated using the asymmetric migration model. Nodes of interest are indicated with an asterisk if posterior support > 0.80. a) Caribbean * * Central America DENV-2 South America American genotype * * * * South Asia

Indian Ocean Islands

Africa * Middle East * * * Oceania * East Asia

* * Maritime SE Asia

1930 1940 1950 1960 1970 1980 1990 2000 Mainland SE Asia

DENV-2 * * Cosmopolitan genotype *

* * * * * * *

* * * * * * * * * * * * * * * * * * * * * * * * * * * 1970 1980 1990 2000 2010

130

Figure 5-3 shows the evolutionary and spatial relationships within the American-Asian and Asian I genotypes. The Asian-American genotype shows a clear source in mainland

Southeast Asia, with limited migration into maritime Southeast Asia and significant migration into East Asia in the mid-1980s (Figure 5-3a). A single introduction of the Asian-American genotype from this region into the Caribbean appears to have seeded a widespread epidemic in the region. The entry of this genotype to the Americas in the 1980s has been associated with the first DHF outbreaks in the region, and it has been proposed that the invasion and dispersal of the American-Asian genotype around 1980 replaced this lineage in the Americas due to a higher transmission fitness conferred by more efficient infection and dissemination of the virus in the Aedes aegypti mosquito vector relative to the American genotype (Armstrong and Rico-

Hesse, 2001; Lorono-Pino et al., 2004).

The Asian I genotype has circulated in mainland Southeast Asia since at least the mid-

1950s, but was detected at relatively low levels until an upsurge in the late 1990s/early 2000s.

This lineage was detected in Thailand as early as 1958 but did not become established at significant levels in other countries until 2003, when it swept into both Cambodia and Viet Nam, causing widespread epidemics and, within only four years, apparently complete replacement of the American-Asian genotype across Southeast Asia. In vivo studies in mosquitoes and humans attributed this replacement event to a higher viremia occurring in human infections with

DENV-2 Asian I infections relative to the American-Asian genotype (Hang et al., 2010).

Together, phylogeographic analysis of the DENV-2 genotypes illustrates a number of aspects common among all four serotypes (Supplementary Figures 5-1 to 5-8 show genotypes of DENV-1, -3, and -4). First, that mainland Southeast Asia, maritime Southeast Asia, and

South Asia all apparently harbor significant DENV diversity and act as global sources for essentially all of the current global DENV lineages circulating today. Within each serotype, phylogeographic analysis suggests that each of the genotypes analyzed here have origins in

Asia (Table 5-1). 131

North America

Caribbean

Central America Figure 5-3. MCC phylogenies of the DENV- a) * South America 2 American-Asian and Asian I genotypes DENV-2 * according to region of American-Asian * South Asia sampling. Tips are colored genotype Caribbean by location of sampling. Central America Indian Ocean Islands South America Internal branches are South Asia colored according to the * Oceania reconstructed ancestral * Africa Island SE Asia state location estimated using the asymmetric * * migration model. Nodes of Middle East * interest are indicated with an asterisk if posterior * Oceania support > 0.80. *

* East Asia * * 1940 1950 1960 1970 1980 1990 2000 * * * Island SE Asia * * 1980 1985 1990 1995 2000 2005 Mainland SE Asia b) DENV-2 Caribbean Asian I genotype Central America

South America * * South Asia South Asia Indian Ocean Islands Africa * Oceania Middle East Oceania

East Asia East Asia Island SE Asia * Mainland SE Asia Maritime SE Asia

* Mainland SE Asia

* * *

* * 1960 1970 1980 1990 2000 2010 1970 1980 1990 2000 2010

132

Table 5-1. Inferred ancestral locations and times to the most recent common ancestor (TMRCA) of each of the major DENV genotypes. Suggested Serotype Genotype Source source Other regions involved Mean TMRCA** country mainland SE Asia, maritime SE Asia, Middle 1939 I East Asia* Japan* East, Oceania, South Asia (1936, 1944) maritime The Caribbean, East Asia, Indian Ocean, 1962 IV mainland SE Asia, Oceania DENV-1 Southeast Asia Philippines (1956, 1969) Africa, Caribbean, Central America, East Asia, Indian Ocean, mainland SE Asia, 1951 V South Asia India maritime SE Asia, Middle East, North (1947, 1955) America, Oceania, South America Caribbean, Central America, maritime SE 1923 American South Asia India Asia, Oceania, South America (1910, 1938) American- mainland Caribbean, Central America, East Asia, 1977 Thailand Asian Southeast Asia maritime SE Asia, South America (1976, 1978) DENV-2 mainland Central America, East Asia, maritime SE 1957 Asian I Thailand Southeast Asia Asia, Oceania, South Asia (1956, 1958) Cosmo- maritime Africa, Central America, East Asia, Indian 1965 Indonesia Ocean, mainland SE Asia, Middle East, politan Southeast Asia Oceania, South Asia (1963, 1967) maritime 1968 I Indonesia East Asia, mainland SE Asia, Oceania Southeast Asia (1966, 1971) mainland 1971 II Thailand East Asia, maritime SE Asia, South Asia DENV-3 Southeast Asia (1969, 1972) Africa, Caribbean, Central America, East Asia, Indian Ocean, mainland SE Asia, 1979 III South Asia Sri Lanka maritime SE Asia, Middle East, Oceania, (1978, 1981) South America maritime The East Asia, mainland SE Asia, Oceania, 1942 I Southeast Asia Philippines South Asia (1933, 1949) DENV-4 maritime Caribbean, Central America, East Asia, 1968 II Indonesia Southeast Asia mainland SE Asia, Oceania, South America (1966, 1971) !

* Although Japan is the inferred origin of DENV-1 genotype I, this is based on the sequencing of a single virus isolated in 1943 during World War II, the earliest sequence in the entire data set, during an outbreak caused by imported cases in Nagasaki. As Japan is not a site of endemic transmission, the outbreak likely resulted from an infected soldier returning from somewhere in Southeast Asia. Three closely related viruses were isolated in Hawaii in 1944 and 1945. Hawaii does experience some authocthonous transmission (Effler et al., 2005; Imrie et al., 2010), but was also a primary site used for the movement of troops and supplies between the mainland United States and the Pacific Theater during and after the War (U.S. Commander Navy Installations Command, 2012). The inferred origin of all the current diversity in DENV-1 genotype 1 arises in mainland Southeast Asia (likely Thailand) in 1968. The presence of DENV-1 in Thailand was determined by serological study as early as 1956 (Halstead, 1966). ** Mean TMRCA estimates are generally close, but some fall outside of the bounds of estimates from previous studies (generally more recently in time than inferred by other studies) (Araujo et al., 2008; Twiddy et al., 2003; Villabona-Arenas and Zanotto, 2011). These must be confirmed in secondary analyses, with testing of both strict and relaxed clocks. However, this should not significantly affect inferred spatial relationships.

Interestingly, in the cases of DENV-1 and DENV-3, each of the three primary modern lineages appears to originate in one of these three locations, with little and mostly modern or

133

transient circulation of the other two genotypes in the other two locations (Figure 5-4). Within

DENV-2 and DENV-4, results are not dissimilar. DENV-2 phylogenies suggest minimal mixing of viral populations between the three geographic regions, with each region inferred as a source population for one genotype (although support for a maritime Southeast Asian origin of the

Cosmopolitan lineage is low); mainland Southeast Asia harbored two primary viral populations in recent history, among which a recent genotype-level replacement has been documented

(Hang et al., 2010), and South Asia harbored two viral populations, among which a possible genotype-level replacement has taken place. DENV-4 diversity is comprised of only two primary genotypes. Both are inferred to be of maritime Southeast Asian origin, with genotype II invading Oceania and the Americas and genotype I invading South Asia and mainland

Southeast Asia, becoming established in the mid-1960s and early 1970s, respectively.

Following these migration events, both South Asia and mainland Southeast Asia have maintained local DENV-4 populations, again with little mixing between the three primary South and Southeast Asian populations.

Frequent seeding of viruses between South and Southeast Asia suggests that viruses from these areas are regularly exchanged (Figure 5-4). However, these importations rarely result in establishment of viral populations in the recipient region. The overall lack of mixing among these viral populations is somewhat surprising given their geographic proximity and the high incidence rates reported around the region (at least in the case of Southeast Asia), which should provide ample opportunities for viral invasion to occur. However, if incidence is high in the recipient population and is dominated by local viruses, immune-mediated competition will limit the number of susceptibles available for infection with a novel lineage upon invasion, and the likelihood of viral establishment will be greatly decreased. Here, the continued dominance of distinct ‘local’ viral populations in South Asia, maritime Southeast Asia, and mainland

134

Southeast Asia suggest that such a phenomenon occurs here. Importantly, this would suggest that dengue is grossly underreported in South Asia, as the disease burden here has generally been reported to be very low.

Figure 5-4. Timeline of emergence and circulation of DENV genotypes in South, maritime Southeast, and mainland Southeast Asia.

DENV-1 DENV-2 American Genotype I

American-Asian Genotype IV

Asian I

Genotype V Cosmopolitan

1945 1955 1965 1975 1985 1995 2005 1945 1955 1965 1975 1985 1995 2005 DENV-3 DENV-4

Genotype I Genotype I

Genotype II

Genotype II Genotype III

1945 1955 1965 1975 1985 1995 2005 1945 1955 1965 1975 1985 1995 2005 South Asia Maritime Southeast Asia Mainland Southeast Asia

To investigate factors that may dictate the global dispersal of DENV, priors were estimated to reflect two primary hypotheses, distance-based spread and disease burden-based spread. These were tested and the fit of each model was assessed using AICM marginal likelihood estimates. Generally, distance-based models showed a slightly better fit to the data than the equal rates model when limited migration was shown between areas separated by long distances (Figure 5-5). As expected, the distance-based model shows a poor fit to the data when genotypes were particularly ‘cosmopolitan’, which include DENV-1 genotype V, DENV-2

135

Cosmopolitan genotype, and DENV-3 genotype III. The DENV-2 American genotype also shows a poor fit to the data. Each of these involves a major South Asian lineage that acts as a source for global populations and suggest that the distance between South Asia and recipient populations is not a major barrier to DENV gene flow. In contrast, distance-based models showed a generally good fit for mainland Southeast Asian lineages, which tended to be much less much more localized.

For the dual burden model, priors were estimated using the ratio of average DF/DHF incidence reported in the DengueNet database (World Health Organization, 2009b) for the source population from 1955 to 2005 to that reported from the recipient population during the same period (this included many zero reports, which were not considered the calculation of averages). This ratio provides a crude approximation of differential transmission intensities between each pair of locations. We would expect these data to reflect: i.) the dependence of exportation from the source population on transmission intensity at the source and ii.) the dependence of importation and establishment in the recipient population on the immunity in this population. Again, the fit of the model to the data was generally good for primarily Southeast

Asian lineages, while a poor fit was shown for South Asian lineages. There is a simple explanation for this discrepancy. South Asia, even since the start of large DF/DHF epidemics in the last two decades, has reported extremely low incidence to the WHO database. However, the high levels of diversity observed in South Asia, even within a single genotype, indicate that this area has experienced considerable and consistent transmission since at least the 1950s, when the first epidemics emerged across South and Southeast Asia, and suggest that there has been gross underreporting from South Asia over the last 50 years. However, the finding that the dual burden model fit well to data dominated by Southeast Asia suggests that this crude indicator is generally adequate when appropriate disease burden data are available.

136

Figure 5-5. Differences in relative marginal log likelihood estimates for fitting of the asymmetric phylogeographic model using Akaike’s information criterion through MCMC (AICM). Results show model fittings of distance- and disease burden-based phylogeographic model priors for each genotype relative to the equal rates model. Higher values indicate a better of the model prior to the data at hand, such that for DENV-1 genotype 1, the dual burden model shows the best fit relative to the equal rates and distance-based models. The inferred source of the lineage is indicated below the x-axis.

20 Distance Dual Burden

5 AICM

! 0

-5 Asia South Asia South Asia South maritime SE Asia maritimeSE Asia maritimeSE Asia maritimeSE Asia maritimeSE Asia maritimeSE mainland SE Asia mainlandSE Asia mainlandSE Asia mainlandSE

-10

-15 East Asia/mainland SE Asia Asia/mainlandSE East I IV V Am. Am.-As. Asian I Cosmo I II III I II DENV-1 DENV-2 DENV-3 DENV-4

Notably, a marked difference in migration routes connecting these regions as source populations for regions outside of tropical Asia was observed in the phylogenies and can be observed by comparing Figures 5-2b and 5-3b, as well as in Figure 5-1. Mainland Southeast

Asia is rarely detected a source of DENV for areas outside of Asia, while South Asia and maritime Southeast Asia regularly exchange viruses globally. This pattern is shown within multiple lineages (see Supplementary Figures 5-1 to 5-8) and suggests that lineages circulating in mainland Southeast Asia may differ from those circulating elsewhere in tropical Asia. I

137

hypothesize that these differences in global spread may not be due to viral factors, but instead to characteristics of the populations in which they circulate. Populations in South Asia and maritime Southeast Asia have historically been more mobile and likely to migrate or travel to new regions than those in mainland Southeast Asia due to differences in economics, language

(English is not a standard language spoken anywhere in mainland Southeast Asia; South Asia and maritime Southeast Asia have large English-speaking populations), and education. To test this, I will need to approach analyses considering both viral factors and host population factors in separate analyses. An initial analysis of selection pressures at the genotype level would be used to determine whether different genotypes have been subject to different selection pressures, specifically the more ‘cosmopolitan’ genotypes versus ‘local’ genotypes as in Twiddy et al. (2002b). To test the human mobility hypothesis, data on human mobility would need to be collected or an indicator of this (such as GDP, language, etc.) used to investigate the role of human mobility in the spread of these lineages.

138

Conclusions

The data presented here represent preliminary analyses to identify general trends in

DENV dispersal and diversity over time. These data indicate a clear, previously undescribed delineation between viral lineages indigenous to mainland Southeast Asia, maritime Southeast

Asia, and South Asia, and suggest that these three regions are the source of all modern global

DENV diversity. Additionally, the somewhat consistent finding that each of these populations maintains a distinct genotype or lineage within each serotype suggests that the diversity shown in these regions has been generated over a long history of transmission and immune interaction at the regional level. This analysis has generated hypotheses related to the means by which diversity is dispersed from tropical Asia, and these will be tested in the near future. The finding that South Asia harbors viral diversity comparable to that observed in the two populations of

Southeast Asia suggests that similar processes may have generated this diversity, and the role of immune-mediated interaction in this process must be explored in the future.

139

Materials and Methods

Sequence data and phylogenetic analysis

All available full- or nearly full-length DENV envelope (E) gene sequences (DENV-1, -2, -

3, -4) with both the year and location of sampling available were downloaded from GenBank

(2163, 2071, 1170, 605 sequences, respectively). Nucleotide alignments were created for each serotype and manually aligned using Se-Al (v2.0a11 Carbon, http://tree.bio.ed.ac.uk/software/seal).

Likely recombinants and redundant sequences originating in the same country and year were identified and removed from all data sets. To infer phylogenetic relationships for complete serotype-specific data sets and identify the primary genotypes within each, we utilized the maximum likelihood (ML) method available in RAxML (Stamatakis, 2006), incorporating a

GTRCAT model of nucleotide substitution with gamma-distributed rate variation among sites with 500 bootstrap replicates to assess the robustness of individual nodes on the tree. Using the resultant trees, human DENV genotype-specific alignments were created if the genotype included taxa from multiple countries and did not show obvious evidence of multiple laboratory errors, as in the case of the DENV-2 Asian II genotype. Genotypes included in the analysis were: DENV-1 genotypes I, IV, V (Goncalvez et al., 2002); DENV-2 American, American-Asian,

Asian I, Cosmopolitan genotypes (Twiddy et al., 2002a); DENV-3 genotypes I, II, III (Araujo et al., 2008); DENV-4 genotypes I, II (Villabona-Arenas and Zanotto, 2011). Excluded genotypes include: DENV-1 genotype II, which includes only three sequences from Thailand in the late

1950s/early 1960s, genotype III consists of sylvatic sequences; DENV-2 Asian II, which includes multiple sequences of questionable provenance; DENV-3 genotype IV includes only five taxa, genotype V includes only 13 taxa; DENV-4 genotype III includes five sequences from Thailand).

Phylogenetic analyses were undertaken on each data set using the Bayesian Markov Chain

140

Monte Carlo (MCMC) method implemented in BEAST and incorporating the date of sampling

(Drummond and Rambaut, 2007). For this analysis we employed a codon-structured SDR06 model of substitution, a relaxed molecular clock, and an extended Bayesian skyline prior

(EBSP). The MCMC chain was run for up to 200 million iterations, with sub-sampling every

20,000 iterations. Parameters reached convergence in most analyses, but some of the larger datasets have not yet converged (due to time constraints), as assessed visually using Tracer

(v.1.5). The initial 10% of the chain was removed as burn-in, and maximum clade credibility

(MCC) trees were summarized using TreeAnnotator (v.1.6.2).

Spatial analysis

To investigate the global migration routes of DENV and identify possible global sources of diversity, we categorized sequence data for each of the genotypes separately according to the region of the world from which they were sampled, including: Africa, the Caribbean, Central

America, East Asia, the Indian Ocean Islands, the Middle East, North America, Oceania, South

America, South Asia, and two categories for Southeast Asia, mainland Southeast Asia

(Cambodia, Laos, Myanmar, Thailand, Vietnam) and maritime Southeast Asia (Brunei, East

Timor, Indonesia, Malaysia, the Philippines, Singapore). The Southeast Asian countries were separated based on a previously described partial phylogenetic division in the distribution of prevailing DENV lineages between the two regions (Holmes, 2009) and a visual confirmation of this phenomenon using ML trees. Due to inherent biases in the collection of sequence data as well as computational constraints on estimation using extremely large data sets, we randomly subsampled all genotype-level sequence alignments such that they included no more than 20 sequences per region per year (Table 5-2).

141

Table 5-2. Description of DENV sequences used in this study.

Number of Number of Range of years Region countries taxa sampled sampled Africa 14 11 1968 - 2010 Caribbean 503 23 1953 - 2010 Central America 248 8 1980 - 2010 East Asia 178 3 1943 - 2010 Indian Ocean Islands 15 4 1977 - 2008 mainland SE Asia 873 5 1958 - 2011 maritime SE Asia 483 5 1956 - 2011 Middle East 21 2 1992 - 2010 North America 5 1 2009 - 2010 Oceania 102 16 1944 - 2010 South America 535 9 1981 - 2011 South Asia 136 5 1956 - 2009 !

We then inferred rates of viral migration between locations using an asymmetric model of discrete diffusion (Lemey et al., 2009). This method estimates an asymmetric diffusion rate for each potential migration pathway among the predefined locations while simultaneously estimating nucleotide substitution rates, divergence times, and demographic histories, thereby allowing quantification of the uncertainty in ancestral state reconstructions (i.e. ancestral geographic locations). Posterior distributions of trees were estimated under a phylogenetic model using the MCMC method implemented in BEAST using BEAGLE (Ayres et al., 2012;

Drummond and Rambaut, 2007). This model incorporated the date of sampling and a relaxed molecular clock, extended Bayesian skyline prior, and the SRD06 codon position model as described above, with Bayesian Stochastic Search Variable Selection (BSSVS) used to identify migration routes with the highest probability between these locations among the posterior sets of trees. The MCMC chain was run for 200 million iterations, with sub-sampling every 20,000

142

iterations, and parameters reached convergence in most analyses (with the exception of the

DENV-2 American-Asian genotype). The initial 10% of the chain was removed as burn-in, and

Maximum Clade Credibility (MCC) trees including modal ancestral location-state reconstructions were summarized using TreeAnnotator (v.1.6.2). Bayes factor (BF) tests were used to determine the statistical significance of diffusion pathways among the geographic locations, with

BF ≥ 6 indicating a significant migration route connecting two locations. BF estimates are inferred over the entire phylogeny and thus represent significant migration pathways inferred over the posterior distribution of trees rather than individual migration events, many of which may occur within a single inferred phylogeny (genotype). In cases where the model is nearing overparameterization or parameters have not reached convergence, the model occasionally estimates high Bayes Factor values that do not appear well supported in the MCC trees. Thus, we applied a secondary filtering method to determine whether these links were well supported.

This consisted of an averaging of the frequency of specified geographic location changes across the posterior set of trees for each genotype and subsequent removal of BF-based migration routes with an estimated frequency < 0.85, which is indicative of less than one significant migration event between two locations across the inferred trees. All possible migration pathways were subjected to this secondary filtering method, and only pathways with

BF ≥ 6 and an average frequency > 0.85 across the posterior set of trees are reported.

Hypothesis-based spatial analysis

To investigate specific hypotheses related to the processes that are expected to play a role in the global migration of DENV, we set rate priors to specific values to construct a series of

(asymmetric) phylogeographic models that might reflect regional differences in epidemiology

143

and human movement. These models were:

i.) Equal rates – assumes equal rates of viral migration between all regions of interest,

ii.) Distance – assumes that viral exchange is dependent upon the distance between

populations, using the inverse of the log distance between the centroids of each region

as an indicator of the likelihood of viral exchange, such that rate priors between two

regions are equal in both directions,

iii.) Dual burden – assumes that disease burden at both the donor and recipient locations

determine the likelihood of exporting virus, using a simple ratio of the average

incidence in the donor population versus the average incidence in the recipient

population (incidence averaged over all countries contributing sequences within a

region and all years of reporting in the WHO DengueNet data base (World Health

Organization, 2009b); due to seemingly infrequent reporting from some regions,

reports of zero incidence were not counted among averages).

Model priors were normalized (mean one and unit variance) and incorporated into asymmetric matrices that allow for directional rates to vary between individual location pairs. A posterior simulation-based analogue Akaike’s information criterion through MCMC (AICM) was implemented using likelihoods specific to the geographic model priors, and marginal log likelihood estimates for each model were compared to determine the best fit model to the data in hand (Baele et al., 2012; Raftery et al., 2007).

144

Supplementary Figures

Supplementary Figure 5-1. MCC phylogeny of DENV-1 genotype I according to region of sampling. Tips are colored by location of sampling. Internal branches are colored according to the reconstructed ancestral state location estimated using the asymmetric migration model.

0.09.648 0.005.93 0.8 1 0.5 0.46 0.08 0.54 0.43 0.240.11 0.18 0..2034 0.96 00..4565 00.8.068 1 1 1 0.27 0.85 1 0.11 1 0.69 0.090.09.843 0.9 1 0.11 0.21 0.92 0.37 1 0.007.74 0.13 0.57 0.79 00.2.113 0.89 0.103.99 0.19 0.907.17 0.53 1 0.29 0.3 0.1 0.36 1 1 0.29 0.91 0.86 1 0.71 0.02 0.28 0.002.03 0.72 0.99 1 0.43 0 1 0.160.13 1 0.03 1 0.11 1 0.040.62 1 DENV-1 1 0.903.54 1 0.44 0.17 1 0.21 1 0.73 0.23 0.25 0.18 0.104.103.08 0.17 0.19 South Asia Genotype I 1 0.901.19 0.47 0.805.93 0.14 0.34 1 1 0.13 1 0.24 0.06 0.94 0 Middle East 0 0 0.06 0.37 1 0.07 1 0.29 0 0.02 1 0.97 0.62 0.74 0.3549 East Asia 1 0.78 0.430.56 1 0.69 0.90.77 1 1 0.53 1 0.57 0.10.91 Island SE Asia 0.45 0.7 0.99 0.99 0.97 0.5 0.91 0.404.29 0.09 0.61 0.04 1 0.01 0.84 0.49 0 0.33 1 Mainland SE Asia 0.03 0.33 1 0.25 1 1 0.02 0.83 0.2 1 0.02.013 0.61 0.95 0.04.252 0 0.97 0.99 1 1 0.89 0.30.82 0.95 0.99 0.46 1 0.22 1 0.05 1 0.7 1 1 0.33 0.29 1 0 0.007.19 0.09.724 0.48 0.97 1 0.03.21 1 0.65 0.95 0.99 1 0.32 0 1 1 1 1 1 0.11 0.39 1 0.05 0.65 0.35 0.03 0.94 0.09 0.99 0.39 1 1 0.104.58 1 0.33 0.98 0.8 0.98 0.001.28 1 0 0.18 0.05 0.03 0.04 0.11 00 0.54 0.43 1 0.02 0.32 0.36 1 1 0.22 1 1 0.99 1 1 0.96 1 0.97 1 0.99 0.970.48 0.89 0.96 0.94 1 0.67 1 1 0.57 0.92 1 1 1 1 0.26 0.15 1 0.9 0.005.07 0 0.33 0.030.98 0.13 1 1 0.00.705 1 1 0.99 0.48 0.08 0 1 00.0.017 0.18 1 1 0.18 0.62 0.29 0.99 0.11 0.74 1 0.910.26 1 0.81 0..046 1 0.13 0.2 0.42 0.96 0.07 0.33 1 0.6 0.91 0.75 0.91 0.99 0.98 0.91 0.37 1 1 0.91 0.19 1 0.930.2 1

1940.0 1950.0 1960.0 1970.0 1980.0 1990.0 2000.0 2010.0 2020.0

145

Supplementary Figure 5-2. MCC phylogeny of DENV-1 genotype IV according to region of sampling. Tips are colored by location of sampling. Internal branches are colored according to the reconstructed ancestral state location estimated using the asymmetric migration model.

0.27 0.22 0.35

0.8 0.25 0.15 Caribbean 1 0.99 0.71 DENV-1 0.16 Indian Ocean Islands 1

0.51 0.23 0.63 Oceania Genotype IV 1 0.09

1 East Asia 0.13 1 0.25 0.18 0.68

1 Island SE Asia 0.43 0.5 1 0.42 0.72 0.98 Mainland SE Asia 1 1

0.39 0.98 0.07 0.24 1 0.19 1

0.98 1

0.43 0.68 0.31 1 0.99 1

1 0.7

1 0.13 0.63 0.12 0.14 0.14 1 0.63 1 0.88

0.99 1 0.86

0.23 1 0.85 0.99 0.45 0.9 1 1 0.56 1 1 1 1 0.85 0.3

0.04 0.81 1 1 0.1 0.52 0.1 0.04 0.33 1 0.99 0.53

0.76 1 1 1

1960.0 1970.0 1980.0 1990.0 2000.0 2010.0 2020.0

Supplementary Figure 5-3. MCC phylogeny of DENV-1 genotype V according to region of sampling. Tips are colored by location of sampling. Internal branches are colored according to the reconstructed ancestral state location estimated using the asymmetric migration model.

0.0.311 0.00.601 0.41 0.11 0.67 0.73 0.68 0.03.819 0.65 1 1 0.1 0.96 1 North America 0.29 0.1 0.97 0.05 0.04.323 0.97 1 1 0.301.99 Caribbean 0.94 0.34 0.35 1 0.23 1 0.53 1 0.05 0.57 1 0.03 1 Central America 0.340.31 0.02 1 0.35 0.46 1 1 0 0.98 0.05 0.74 0.03 0.28 South America 0.01 0.16 0 0.54 1 0.94 0.97 0.54 0.09 00.0.044 1 0.107.05 South Asia DENV-1 0.89 0.03 1 0.01 0.14 0.00.101 0.83 0.33 0.09 0.39 00.1.129 Indian Ocean Islands 1 0.92 Genotype V 1 0.95 1 0.56 0.29 1 0.08 0.17 1 0.18 0.99 Africa 0.06 0.12 0.86

00.0.172 1 0.2 0.05 0.15 0.01 0.64 Middle East 0.01 1 1 0.01 0.07 0.02 0.88 0.06 0.13 0.93 1 1 0.53 Oceania 0.37 1 0.0.3 3 0.79 1 0.11 0.61 0.95 0.96 0.1 0.97 0.3 0.17 0.64 East Asia 0.99 0.91 0.85 1 0.05 0.32 0.0.505 0.007.02 0.19 1 0.91 1 Island SE Asia 0.61 1 1 1 0.88 1 0.07 1 1 0.19 1 0.08 1 1 Mainland SE Asia 1 0.87 1 0.96 0.98 0.04 1 0.95 0.4 0.42 0.76 1 1 1 1 1 1 0.84 0.78 0.18 0.74 1 0.73 0.48 0.24 1 1 0.58 0.48 0.02 1 1 0.86 0.01 0.17 0.001.08 0.88 0 0.06 1 0.99 0.009.97 0.201.05 0.99 1 0.59 0.98 0.98 0.43 0.13 0.63 1 1 0.49 1 0.22 0.75 0.84 0.89 0.83 1 1 0.99 1 0.17 0.51 0.13 0.12 0.98 1 1 0.29 0.09 0.43 0.08 0.19 0.06 0.33 1 0.28 1 1 1 1 0.08 0.101.14 0.45 0.74 1 0.98 1 0.97 0.06 1 1 1 0.99 1 0.31 0.94 1 0.99 0.84 1 0.93 0.92 1 1 1 1 0.81 0.3 1 1 0.49 1 1 0.9 1 0.52 0.9 0.19 1 0.96 0.15 0.98 1 0.6 1 1 0.99 1

1950.0 1960.0 1970.0 1980.0 1990.0 2000.0 2010.0 2020.0

146

Supplementary Figure 5-4. MCC phylogeny of DENV-3 genotype I according to region of sampling. Tips are colored by location of sampling. Internal branches are colored according to the reconstructed ancestral state location estimated using the asymmetric migration model.

0.62 0.07 0.74 0.38 1 0.32 Oceania 1 0.73 1 0.15 0.4 East Asia 0.18 1 0.91 1

0.24 Island SE Asia 0.16 0.18 1 0.41 0.99 0.8 1 Mainland SE Asia 0.25 0.17 0.18 0.2 0.84 1 1 DENV-3 1

0.89 1 0.21 0.15 Genotype I 0.27 0.52 1 0.73 1

1 1 0.28 0.65 0.14 1 0.71 0.18

0.67 1 1 1 0.53 0.95 0.24 0.44 0.29

0.23 1 0.52 0.97 0.96 1 0.09 0.51 1

0.14

0.96 1 1 0.46 0.77 1 0.25

0.15 0.17 0.47 1 1 1 0.94 0.88 1 0.43 0.99 1 1 1

1 0.69 0.99 0.99 0.96 0.77 0.9

0.49 1 0.34 0.29 1 0.29 0.99 0.55 0.86 0.87

0 1 0.02 0.03 0.07

0.48 0.14 0.05 1 1 1 0.21 0.48 1970.0 1980.0 1990.0 2000.0 2010.0

Supplementary Figure 5-5. MCC phylogeny of DENV-3 genotype II according to region of sampling. Tips are colored by location of sampling. Internal branches are colored according to the reconstructed ancestral state location estimated using the asymmetric migration model.

1 1 0.89

DENV-3 0.94

0.83 Genotype II 1 South Asia

East Asia 0.95 1 0.94

Island SE Asia 1 1 1

Mainland SE Asia

1 1

0.96 1

1 0.88

1 1

0.94

0.84 1 1 0.8 1

0.89 1 1 1 1 1

0.96 0.95 1

0.82 1 0.94 0.81 1

1970.0 1975.0 1980.0 1985.0 1990.0 1995.0 2000.0 2005.0 2010.0

147

Supplementary Figure 5-6. MCC phylogeny of DENV-3 genotype III according to region of sampling. Tips are colored by location of sampling. Internal branches are colored according to the reconstructed ancestral state location estimated using the asymmetric migration model.

0 0.01 0.03 0 0 0 0.06 00.02 0 0.92 0.109.42 0.04 Caribbean 0.08 0 0.01.06 0.11 0.03 0.66 0.01 0.13 0.1 0.99 0.05 0.05 0.12 0.27 0.12 1 0.970.28 0.93 0.87 0.9 0.97 Central America 0.99 1 0.97 0.14 1 DENV-3 0.6 0.08 0.020.09 0.99 1 0.98 0.05 0.15 0.24 0.13 0.99 0.67 0.04 1 1 South America 1 0.98 0.01 1 0.35 Genotype III 0.01 1 0.39 1 1 0.76 1 0.1 0.14 0.79 0.73 0.05 1 0.99 0.79 South Asia 0.08 0.12 0.99 0.03 0.03 0.43 0.08 0.67 0.33 1 0.97 0.94 1 1 0.99 0.51 0.86 1 1 Indian Ocean Islands 0.990.4 0.0.217 0.908.14 0.96 0.33 0.160.96 0.53 0.909.16 0.99 0.1 0.003..60073 0.98 0.3 0.36 Africa 0.990.4 0.5 0.79 0.63 1 1 0.16 0.95 0.96 0.99 0.38 0 0.99 0.82 0 0.02 0.00.207 1 1 0.001.02 Middle East 0.79 0.91 1 0.68 0.01 0.05 1 0.97 0.88 0.030.00.409 0.12 1 0.94 0.07 0.92 0.64 Oceania 0.73 1 1 0.59 0.93 0.61 0.17 0.99 1 0.08 0.22 0.69 0.95 1 0.1 0.36 0.26 1 0.84 0.98 East Asia 0.97 1 0.01 0.15 0.56 0 0 0 0 0.96 0 00.02 0 0.94 1 0.99 0 0.46 1 Island SE Asia 1 0.85 0.82 0.73 0 0.010.80.036 0.7 1 0.06 0.22 0.46 0.77 0.16 1 0.660.48 0.52 1 Mainland SE Asia 0.06 00..0038 1 1 0.303.07 0.72 1 0.45 0.02 1 0 0.97 0.88 0.46 1 0.02 0.78 0.27 0 0 0 0.01 0.53 0.85 0 0.02 1 1 1 0.23 1 0.55 1 0.99 1 1 0.03 0.07 0.97 0.03 0.07 0.1 0.22 0.96 0 0.4 0.99 0.95 0.98 0.96 0 1 0.34 0.34 0 0.03 0 0 0 0.05 0.07 0 0.1 0.98 0 0 0 0.53 1 0.08 0.03 1 0.34 1 0.29 0 0 0.01 0.25 0.25 0.08 0.31 0.01.11 0.99 0.97 1 0.12 0.99 0.09 0.86 1 0.2 0.84 0.02.806 0.17 0.29 0.98 0.402.09 0.06 0.46 0.07 1 0.28 0.1 0.48 0.42 0.95 0.18 1 1 1 0.39 1 0.6 0.67 0.91 1 1 0.12 1 0.99 0.52 1 0.16 0.46 1 1 1 0.09 0.1 00.0.022 0.97 1 0.04 1 0.38 1 1 0.470.14 0.19 1 0.74 0.99 0.97 0.93 0.08 1 0.87 1 0.68 0 0.01 0.03 0 0.09 0.99 0 0.99 0 0.90.180.31 0.02 0.09 1 0.29 1 0.9 0.98 0.03 0.92 0.29 0.68 0.97 0.53 0.58 1 1 0.180.02 1 0.38 1 1 1 1 1 0.99 0.93 0.89 0.65 0.16 1 0.5 1 0.9

1970.0 1980.0 1990.0 2000.0 2010.0 2020.0

Supplementary Figure 5-7. MCC phylogeny of DENV-4 genotype I according to region of sampling. Tips are colored by location of sampling. Internal branches are colored according to the reconstructed ancestral state location estimated using the asymmetric migration model.

0.86 0.3 1 0.93 0.91 1 0.73

0.22 0.02 0.85 0.81 0.08 0.04 0.05 1 0.82 1 1 1 0.07 0.98 1 DENV-4 0.28 0.34 0.12 1 0.88 0.92 South Asia Genotype I 0.9 0.99 0.21 0.21 1 1 East Asia 0.25 1

0.69

0.92 Island SE Asia 1

0.94 0.53 0.46 1 Mainland SE Asia 0.98 0.24 0.45 1 1 1 1 0.11 1 0.92 0.01 1 0.07

0.06 0.12 0.78 0.24 0.25 0.54 1 0.15 1 1 0.98 0.86 1 1 0.68 1 1 0.87

1 0.99

0.93 0.99 0.89

1 1 0.99 1 0.34 1 1 0.72 0.49 0.23 0.41 1

1 1 0.74 1 0.95 1 0.42 1

0.93

0.63 1 0.97 1 1 1 1 1 1 0.64

0.96 0.93 1 0.25 1 1 1940.0 1950.0 1960.0 1970.0 1980.0 1990.0 2000.0 2010.0

148

Supplementary Figure 5-8. MCC phylogeny of DENV-4 genotype II according to region of sampling. Tips are colored by location of sampling. Internal branches are colored according to the reconstructed ancestral state location estimated using the asymmetric migration model.

0.13 0.01 0.03 0.89 0.11 0.06 0.16 0.12 0.74 1 0.03 0.94 0.96 1 0.02.116 1 0.61 1 1 1 1 0.95 1 0.29 1 0.74 0.080.98 0.03 0.99 0.26 0.17 1 0.08 Caribbean 0.99 0.47 0.909.99 0.98 0.94 1 1 DENV-4 0.86 1 0.82 0.990.96 0.93 0.91 Central America 1 0.202.25 0.1 1 0.18 0.94 0.97 1 Genotype II 0.15 0.01 1 0.04 0.16 0.08 South America 0.87 1 1 0.02 0.21 1 0.07 1 0.93 1 0.21 0.00.602 Oceania 0.11 0.98 0.08 0.27 0.97 1 1 0.00.815 0.48 0.17 0.08 0.96 0.14 East Asia 0.8 0.17 0.35 0.91 1 0.36 1 0.24 1 1 1 0.99 0.16 Island SE Asia 0.92 0.16 1 0.11 1 00.2.086 0.420.83 1 0.37 0.22 1 1 0.12 Mainland SE Asia 0.93 0.61 0.99 0.45 0.26 0.99 0.96 0.61 0.03 0.920.21 0.02 0.1 0.01.15 0.33 1 0.98 0.91 0.74 0.54 0.05 0.71 0.99 0.02 1 0.21 0.23 0.22 0.5 0.97 1 1 1 0.903.42 1 0.94 1 0.24 1 0.31 1 0.26 0.120.99 0.01 0.05 0.31 1 0.02 1 0.01 0.09 1 0.41 1 0.89 00..0016 0.94 0.01 0.87 1 0.780.3 1 1 0.7 0.56 1 1 1 0.3 0.17 0.05 0.03 0.97 1 0.75 0.03 1 0.89 0.79 1 0.91 0.12 1 1 0.37 0.09 0.01 0.23 0.203.11 0.99 0 1 0.980.39 0.0.106 0.02 0.02 0.14 0.42 1 0.24 0 0.01 0.82 0 0.06 0.15 0.16 1 0.71 1 0.38

0.2 0.65 1 1 1 0.58 1 1 1 1 1 1 0.43 1 0.58 0.39 0.990.76 00.2.166 1 0.1 0.99 0.92 0.11 0.21 1 0.07 0.97 0.02 0.1 1 1 1 0.99 0.05 0.93 1 1 1 0.32 1 0.01 0.14 0 0.97 0.97 0 0 0.07 0.99 0.03 0.96 1 00.00.105 1 1 1 1 1 0.909.37 1 0.46 1 1 1 0.99 1 0.58 1 0.85 0.35 1 0.98

1960.0 1970.0 1980.0 1990.0 2000.0 2010.0 2020.0

149

CHAPTER 6

DISCUSSION

This thesis addresses several of the central questions in the evolution and epidemiology of the dengue viruses through the analysis of large, spatially-related whole genome and envelope gene sequence data sets to investigate viral dynamics at multiple spatial and temporal scales, from short-term intra-country transmission to long-term regional dispersals. These analyses reveal a complex pattern of virus evolution and dispersal, with Asia acting as a source population for DENV diversity at the global level and urban areas within Asia regularly acting to seed the surrounding rural areas, and one another, with new lineages. These hypotheses have been proposed previously (Cummings et al., 2004; Gubler and Clark, 1995; Holmes, 2009), but have not been investigated using robust phylogenetic and phylogeographic methods and intensively sampled data from all areas of relevance. The recent development of such methods now allows for DENV dispersal and invasion dynamics to be quantified at multiple scales.

Collectively, the four phylogenetic analyses contained herein provide detailed descriptions of the spatio-temporal dynamics of DENV evolution across a diverse set of populations and spatial scales. Through intensive sampling of viruses in both endemic and epidemic transmission environments, we detect substantial genetic diversity characterized by strong spatial subdivision in most settings, and determine a role for climate in local DENV dynamics as well as a primary role for human-mediated dispersal of DENV over various spatial and temporal scales: from a rural site of hyperendemic transmission over four years in Thailand

(Chapter 2), to a highly urban setting and surrounding rural areas during the five-year invasion and establishment of a novel genotype in hyperendemic southern Viet Nam (Chapter 3), to endemic southeast Asia and subtropical northern Viet Nam, where dengue appears to be

150

emerging (Chapter 4), and finally, to over 50 years of epidemic and endemic transmission sampled globally (Chapter 5).

DENV dynamics in the hyperendemic rural population in Thailand (Chapter 2) and epidemic setting of urban northern Viet Nam (Chapter 4) share surprising commonalities.

Rather than evolving in situ from the previous season, the genetic diversity observed among

DENV circulating in these locations generally appears to be imported from other populations on an annual basis. These locations share a similar climate dynamic, with lower temperatures and precipitation in the winter relative to the hot, wet summer months in which dengue is particularly active, but more importantly, these lower temperatures are expected to limit viral replication in the vector such that transmission efficiency is greatly reduced (Lambrechts et al., 2011; Watts et al., 1987). In northern Viet Nam, lower temperatures appear to result in severe population bottlenecks such that no viral populations survive from year to year [although this was only assessed for one serotype here as well as another unreported serotype (DENV-2) for which the dynamic is identical], while populations in rural Thailand experience limited in situ evolution over multiple seasons within the study area. Generally, we would expect higher density populations to be more likely to maintain virus populations due to the availability of greater numbers of susceptibles throughout the year and a shorter distance required to travel between them.

However, unlike northern Viet Nam, the site in rural Thailand has experienced hyperendemic transmission for decades and, thus, population-level immune dynamics here may better allow for the maintenance of low levels of transmission throughout the year via higher incidence of severe dengue disease through secondary infection, which confers higher viremia and therefore increases transmission fitness. Notably, however, we find that viral populations do invade and become established in northern Viet Nam during the winter months, which suggests that although a climate-mediated population bottleneck may exist, the vector population is

151

sufficiently high to maintain some transmission, thus incidence and diversity may need only reach a threshold before these bottlenecks no longer result in extinction of the northern

Vietnamese lineages.

Further evidence that climate plays a role in local viral establishment and evolution comes from the south of Viet Nam, where hyperendemic transmission of local viral lineages is maintained in urban settings for several years at a time (Chapter 3 and 4). Here, transmission remains highly seasonal, though significant transmission is detected outside of the dengue season and temperatures remain high enough that vector populations and DENV replication efficiency are not expected to vary substantially across seasons. Invasion and establishment here appear to occur very rapidly when the immune landscape is permissive (Chapter 4) or when introduced viral populations carry a selective advantage (Chapter 3). The urban center of the region, Ho Chi Minh City, maintains virtually all of the diversity shown across the country, and frequently seeds virus to other parts of the country, from rural populations surrounding the city to urban populations in the north. Notably, it appears that rural populations here are also capable of sustained viral transmission for several seasons at a time. This finding is particularly important, as it implies that temperature may play a greater role in dictating viral persistence in a given area than host population density. However, the role of the immune landscape in viral persistence here is not well understood and should be further investigated through virological and sero-epidemiological studies.

A key result of this thesis is the finding that viral migration at local and regional scales appears to be determined in large part by patterns of human movement in and out of urban areas and, at the regional and global scales, by the burden of disease in both the donor and recipient populations (although human movement was not specifically assessed in the global analysis, anecdotal evidence suggests that it also plays an important role). While significant

152

population subdivision characterizes all of the data sets used in these analyses, we also observe significant and rapid viral dispersal out of population centers such as Ho Chi Minh City and Singapore into locales with lower viral diversity and transmission intensity; these key factors appear to play a role in determining whether viral populations become established in the host population into which they are introduced. Certainly, there is significant human movement between Ho Chi Minh City and Singapore, given that they are a shorter distance from one another, act as primary economic centers, and share greater numbers of flights than do the northern Vietnamese city of Hanoi and Singapore (some flight data were collected for the analysis conducted in Chapter 4 but were insufficient). However, the current permissive immune landscape of Hanoi may allow for more frequent viral establishment than in the south due to a lack of competition for susceptible hosts. Whether these establishments will eventually develop into local DENV populations remains an open question, and it is here, perched on the edge of emergence, that resources should be directed if we are to understand this process and the factors that determine its success or failure.

The immune landscape and viral diversity present in the population also appear to play significant roles in determining the global dynamics of DENV dispersal. In the analyses conducted in Chapter 5, mainland Southeast Asia, maritime Southeast Asia, and South Asia exhibit fairly strong spatial subdivision, harboring largely independent viral lineages within each serotype with only occasional mixing amongst geographic areas. The relative lack of mixing between these areas given their geographic proximity and high incidence (at least in the case of

Southeast Asia), as well as the overall frequency of viral seeding events between these areas and subsequent lack of evidence for establishment in many of these cases, suggests that immune-mediated competition may to a large extent dictate the modern geographic distributions of DENV in hyperendemic Asian populations. Additionally, the high incidence of dengue

153

disease (and likely infection) in these regions relative to most other populations makes it more likely that they will seed viruses into new populations if susceptible or currently infected hosts travel between these areas and populations in which immune competition is not a dominating factor or indigenous viral fitness is low relative to the invading virus.

Importantly, the finding that these three regions maintain the greatest viral diversity and also tend to harbor distinct DENV genotypes with little mixing between the regions strongly supports the idea that the genotypes emerged separately in Asia as a result of vicariance, rather than through a selectively-driven process (Holmes, 2009). The diversity maintained in

Southeast Asia is fairly well-studied and appears to regularly undergo a process of lineage birth and death based on changing serotype frequencies and population immunity, and in some cases, on differences in viral fitness. However, a primary role for South Asia as a global source population for DENV has not been previously investigated due to a dearth of sequence data and a general assumption that DENV activity was low in the region following a lack of reporting between early epidemics in the 1950s and recent reports of epidemic and endemic activity starting in the 1990s. The sequencing of historical DENV samples from this region fills a number of gaps in our understanding of how the DENV genotypes evolved and eventually emerged in populations across the world, while highlighting a number of additional features that question our current understanding of DENV dynamics. Generally, the most cosmopolitan lineages within DENV-1, -2, and -3 either originate in or have moved through South Asia as they expanded globally. However, reports of disease burden over the last 50 years in South Asia do not suggest the levels of transmission and diversity that are shown by the sequence data included here. If, after a brief period of DHF/severe dengue outbreaks in the 1950s, these types of epidemics were rare in the population until an upsurge in the 1990s, it is imperative that we determine how and why these virus populations circulated for so long in the population without

154

causing the severe disease observed across Southeast Asia during the same period.

Additionally, the widespread distributions of viruses that originate in this region suggest that there may be some important aspects of the DENV or human populations of South Asia that allow for the extensive dispersal and establishment of local viruses in new host populations.

Clearly, further investigation is needed, including follow-up studies to Chapter 5 of this thesis, to investigate a specific role for human movement over time in determining global DENV dispersal and to determine whether viral factors such as transmission fitness or adaptation to a wider range of vector species aid in the spatial spread of these lineages.

Together, the analyses presented here indicate that viral dynamics and dispersal must be considered in the context of the host and vector populations, environments, and immune landscapes in which they circulate, but show that viral gene sequence data are an incredibly powerful and unmatched means by which to infer viral dynamics and understand the spatio- temporal spread of a pathogen rather than of a disease. The use of large viral sequence data sets collected with matched epidemiological data obtained over a wide distribution of geographic locations during epidemics, endemic activity, and periods of emergence would be particularly valuable in determining how viral dynamics and immune dynamics interact at the individual and population levels.

Questions on the geographic location and populations in which DENV emerged and the processes by which the four serotypes diverged remain unanswered. While it may not be possible to answer these questions with certainty, additional sampling of human, primate, and mosquito viruses in poorly sampled areas that may experience primate cycles and/or epizootic transmission of DENV including populations across central Africa, rural Malaysia and Indonesia, and tropical South Asia will likely provide great insight into how DENV initially moved into the human population, as well as the likelihood that new DENV lineages may once again emerge

155

out of the primate cycle. The possibility of this occurring has serious implications for the use of a tetravalent vaccine that confers limited immunity to divergent lineages, and should be a focus of research efforts for both biologists and vaccine scientists as clinical and community trials of

DENV vaccine move in the direction of implementation on a large scale.

156

REFERENCES

A Nuegoonpipat, A., Berlioz-Arthaud, A., Chow, V., Endy, T., Lowry, K., Mai le, Q., Ninh, T.U., Pyke, A., Reid, M., Reynes, J.M., Su Yun, S.T., Thu, H.M., Wong, S.S., Holmes, E.C., Aaskov, J., 2004. Sustained transmission of dengue virus type 1 in the Pacific due to repeated introductions of different Asian strains. Virology. 329(2), 505-512.

Aaskov, J., Buzacott, K., Field, E., Lowry, K., Berlioz-Arthaud, A., Holmes, E.C., 2007. Multiple recombinant dengue type 1 viruses in an isolate from a dengue patient. J Gen Virol. 88(Pt 12), 3334-3340.

AbuBakar, S., Wong, P.F., Chan, Y.F., 2002. Emergence of dengue virus type 4 genotype IIA in Malaysia. J Gen Virol. 83(Pt 10), 2437-2442.

Acosta, E.G., Castilla, V., Damonte, E.B., 2008. Functional entry of dengue virus into Aedes albopictus mosquito cells is dependent on clathrin-mediated endocytosis. J Gen Virol. 89(Pt 2), 474-484.

Adams, B., Boots, M., 2006. Modelling the relationship between antibody-dependent enhancement and immunological distance with application to dengue. J Theor Biol. 242(2), 337-346.

Adams, B., Holmes, E.C., Zhang, C., Mammen Jr., M.P., Nimmannitya, S., Kalayanarooj, S., Boots, M., 2006. Cross-protective immunity can account for the alternating epidemic pattern of dengue virus serotypes circulating in Bangkok. Proc Natl Acad Sci U S A. 103(38), 14234-14239.

Adams, B., Kapan, D.D., 2009. Man bites mosquito: understanding the contribution of human movement to vector-borne disease dynamics. PLoS One. 4(8), e6763.

Allicock, O.M., Lemey, P., Tatem, A.J., Pybus, O.G., Bennett, S.N., Mueller, B.A., Suchard, M.A., Foster, J.E., Rambaut, A., Carrington, C.V., 2012. Phylogeography and Population Dynamics of Dengue Viruses in the Americas. Mol Biol Evol.

Amberg, S.M., Nestorowicz, A., McCourt, D.W., Rice, C.M., 1994. NS2B-3 proteinase-mediated processing in the yellow fever virus structural region: in vitro and in vivo studies. Journal of virology. 68(6), 3794-802.

Anders, K.L., Nguyet, N.M., Chau, N.V., Hung, N.T., Thuy, T.T., Lien le, B., Farrar, J., Wills, B., Hien, T.T., Simmons, C.P., 2011. Epidemiological factors associated with dengue shock syndrome and mortality in hospitalized dengue patients in Ho Chi Minh City, Vietnam. Am J Trop Med Hyg. 84(1), 127-134.

Anderson, J.R., Rico-Hesse, R., 2006. Aedes aegypti vectorial capacity is determined by the infecting genotype of dengue virus. Am J Trop Med Hyg. 75(5), 886-892.

Anderson, R., 2003. Manipulation of cell surface macromolecules by flaviviruses. Adv Virus Res. 59, 229- 74.

Anderson, R.M., May, R.M., 1991. Infectious diseases of humans: dynamics and control. Oxford University Press, New York.

157

Araujo, J.M., Nogueira, R.M., Schatzmayr, H.G., Zanotto, P.M., Bello, G., 2008. Phylogeography and evolutionary history of dengue virus type 3. Infect Genet Evol. 9(4), 716-725.

Armstrong, P.M., Rico-Hesse, R., 2001. Differential susceptibility of Aedes aegypti to infection by the American and Southeast Asian genotypes of dengue type 2 virus. Vector Borne Zoonotic Dis. 1(2), 159-168.

Ashour, J., Laurent-Rolle, M., Shi, P.Y., Garcia-Sastre, A., 2009. NS5 of dengue virus mediates STAT2 binding and degradation. J Virol. 83(11), 5408-5418.

Ashour, J., Morrison, J., Laurent-Rolle, M., Belicha-Villanueva, A., Plumlee, C.R., Bernal-Rubio, D., Williams, K.L., Harris, E., Fernandez-Sesma, A., Schindler, C., Garcia-Sastre, A., 2010. Mouse STAT2 restricts early dengue virus replication. Cell Host Microbe. 8(5), 410-421.

Ayres, D.L., Darling, A., Zwickl, D.J., Beerli, P., Holder, M.T., Lewis, P.O., Huelsenbeck, J.P., Ronquist, F., Swofford, D.L., Cummings, M.P., Rambaut, A., Suchard, M.A., 2012. BEAGLE: An Application Programming Interface and High-Performance Computing Library for Statistical Phylogenetics. Systematic Biology. 61(1), 170-173.

Azeredo, E.L., De Oliveira-Pinto, L.M., Zagne, S.M., Cerqueira, D.I., Nogueira, R.M., Kubelka, C.F., 2006. NK cells, displaying early activation, cytotoxicity and adhesion molecules, are associated with mild dengue disease. Clin Exp Immunol. 143(2), 345-356.

Baele, G., Lemey, P., Bedford, T., Rambaut, A., Suchard, M.A., Alekseyenko, A.V., 2012. Improving the accuracy of demographic and molecular clock model comparison while accomodating phylogenetic uncertainty . Mol Biol Evol.

Balmaseda, A., Hammond, S.N., Tellez, Y., Imhoff, L., Rodriguez, Y., Saborio, S.I., Mercado, J.C., Perez, L., Videa, E., Almanza, E., Kuan, G., Reyes, M., Saenz, L., Amador, J.J., Harris, E., 2006. High seroprevalence of antibodies against dengue virus in a prospective study of schoolchildren in Managua, Nicaragua. Trop Med Int Health. 11(6), 935-942.

Balmaseda, A., Standish, K., Mercado, J.C., Matute, J.C., Tellez, Y., Saborio, S., Hammond, S.N., Nunez, A., Aviles, W., Henn, M.R., Holmes, E.C., Gordon, A., Coloma, J., Kuan, G., Harris, E., 2010. Trends in patterns of dengue transmission over 4 years in a pediatric cohort study in Nicaragua. J Infect Dis. 201(1), 5-14.

Barbosa da Silva, J., Siqueira Jr., J.B., Coelho, G.E., Vilarinhos, P.T., Pimenta Jr., F.G., 2002. Dengue in Brazil: current situation and prevention and control activities. Epidemiol Bull. 23(1), 3-6.

Bartelma, G., Padmanabhan, R., 2002. Expression, purification, and characterization of the RNA 5’- triphosphatase activity of dengue virus type 2 nonstructural protein 3. Virology. 299(1), 122-132.

Bartley, L.M., Carabin, H., Vinh Chau, N., Ho, V., Luxemburger, C., Hien, T.T., Garnett, G.P., Farrar, J., 2002a. Assessment of the factors associated with flavivirus seroprevalence in a population in Southern Vietnam. Epidemiol Infect. 128(2), 213-220.

Bartley, L.M., Donnelly, C.A., Garnett, G.P., 2002b. The seasonal pattern of dengue in endemic areas: mathematical models of mechanisms. Trans R Soc Trop Med Hyg. 96(4), 387-397.

158

Baruah, K., Dhariwal, A.C., 2011. Epidemiology of dengue, its prevention and control in India. J Indian Med Assoc. 109(2), 82-86.

Beatty, M.E., Beutels, P., Meltzer, M.I., Shepard, D.S., Hombach, J., Hutubessy, R., Dessis, D., Coudeville, L., Dervaux, B., Wichmann, O., Margolis, H.S., Kuritsky, J.N., 2011. Health economics of dengue: a systematic literature review and expert panel’s assessment. Am J Trop Med Hyg. 84(3), 473-488.

Beaty, B.J., Bishop, D.H., 1988. Bunyavirus-vector interactions. Virus research. 10(4), 289-301.

Benarroch, D., Selisko, B., Locatelli, G.A., Maga, G., Romette, J.L., Canard, B., 2004. The RNA helicase, nucleotide 5’-triphosphatase, and RNA 5'-triphosphatase activities of Dengue virus protein NS3 are Mg2+-dependent and require a functional Walker B motif in the helicase catalytic core. Virology. 328(2), 208-218.

Bennett, K.E., Beaty, B.J., Black, W.C. th, 2005. Selection of D2S3, an Aedes aegypti (Diptera: Culicidae) strain with high oral susceptibility to Dengue 2 virus and D2MEB, a strain with a midgut barrier to Dengue 2 escape. J Med Entomol. 42(2), 110-119.

Bennett, S.N., Holmes, E.C., Chirivella, M., Rodriguez, D.M., Beltran, M., Vorndam, V., Gubler, D.J., McMillan, W.O., 2006. Molecular evolution of dengue 2 virus in Puerto Rico: positive selection in the viral envelope accompanies clade reintroduction. J Gen Virol. 87(Pt 4), 885-893.

Van Benthem, B.H., Vanwambeke, S.O., Khantikul, N., Burghoorn-Maas, C., Panart, K., Oskam, L., Lambin, E.F., Somboon, P., 2005. Spatial patterns of and risk factors for seropositivity for dengue infection. Am J Trop Med Hyg. 72(2), 201-208.

Black, W.C. th, Bennett, K.E., Gorrochotegui-Escalante, N., Barillas-Mury, C.V., Fernandez-Salas, I., de Lourdes Munoz, M., Farfan-Ale, J.A., Olson, K.E., Beaty, B.J., 2002. Flavivirus susceptibility in Aedes aegypti. Arch Med Res. 33(4), 379-388.

Brown, A.W., 1977. Yellow fever, dengue and dengue hemorrhagic fever, in: Howe, G.M. (Ed.), A World Geography of Human Diseases. Academic Press, New York, pp. 271-317.

Bugher, J.C., Taylor, M., 1949. Radiophosphorus and Radiostrontium in Mosquitoes. Preliminary Report. Science (New York, N.Y.). 110(2849), 146-7.

Burke, D.S., Nisalak, A., Johnson, D.E., Scott, R.M., 1988. A prospective study of dengue infections in Bangkok. Am J Trop Med Hyg. 38(1), 172-180. de la C Sierra, B., Kouri, G., Guzman, M.G., 2007. Race: a risk factor for dengue hemorrhagic fever. Arch Virol. 152(3), 533-542.

Cardosa, J., Ooi, M.H., Tio, P.H., Perera, D., Holmes, E.C., Bibi, K., Abdul Manap, Z., 2009. Dengue virus serotype 2 from a sylvatic lineage isolated from a patient with dengue hemorrhagic fever. PLoS Negl Trop Dis. 3(4), e423.

Cardosa, M.J., Wang, S.M., Sum, M.S., Tio, P.H., 2002. Antibodies against prM protein distinguish between previous infection with dengue and Japanese encephalitis viruses. BMC Microbiol. 2, 9.

159

Carrington, C.V., Foster, J.E., Pybus, O.G., Bennett, S.N., Holmes, E.C., 2005. Invasion and maintenance of dengue virus type 2 and type 4 in the Americas. J Virol. 79(23), 14680-14687.

Casals, J., Brown, L.V., 1954. Hemagglutination with arthropod-borne viruses. J Exp Med. 99(5), 429- 449.

Cavalli-Sforza, L.L., Barrai, I., Edwards, A.W., 1964. Analysis of human evolution under random genetic drift. Cold Spring Harbor symposia on quantitative biology. 29, 9-20.

Center for International Earth Science Information Network (CIESIN) Columbia University, 2005. Poverty Mapping Project: Small Area Estimates of Poverty and Inequality.

Chambers, T.J., Hahn, C.S., Galler, R., Rice, C.M., 1990. Flavivirus genome organization, expression, and replication. Annual review of microbiology. 44, 649-88.

Chan, V.F., 1987. Virological and epidemiological studies of Dengue haemorrhagic fever in the Philippines. Southeast Asian J Trop Med Public Health. 18(3), 275-277.

Chareonsirisuthigul, T., Kalayanarooj, S., Ubol, S., 2007. Dengue virus (DENV) antibody-dependent enhancement of infection upregulates the production of anti-inflammatory cytokines, but suppresses anti-DENV free radical and pro-inflammatory cytokine production, in THP-1 cells. J Gen Virol. 88(Pt 2), 365-375.

Chareonsook, O., Foy, H.M., Teeraratkul, A., Silarug, N., 1999. Changing epidemiology of dengue hemorrhagic fever in Thailand. Epidemiol Infect. 122(1), 161-166.

Chaturvedi, U.C., Kapoor, A.K., Mathur, A., Chandra, D., Khan, A.M., Mehrotra, R.M., 1970. A clinical and epidemiological study of an epidemic of febrile illness with haemorrhagic manifestations which occurred at Kanpur, India, in 1968. Bull World Health Organ. 43(2), 281-287.

Chau, T.N., Hieu, N.T., Anders, K.L., Wolbers, M., Lien le, B., Hieu, L.T., Hien, T.T., Hung, N.T., Farrar, J., Whitehead, S., Simmons, C.P., 2009. Dengue virus infections and maternal antibody decay in a prospective birth cohort study of Vietnamese infants. J Infect Dis. 200(12), 1893-1900.

Chen, M.C., Lin, C.F., Lei, H.Y., Lin, S.C., Liu, H.S., Yeh, T.M., Anderson, R., Lin, Y.S., 2009. Deletion of the C-terminal region of dengue virus nonstructural protein 1 (NS1) abolishes anti-NS1-mediated platelet dysfunction and bleeding tendency. J Immunol. 183(3), 1797-1803.

Chen, S., 2011. The origin of dengue viruses caused the DF outbreak in Guangdong province, China, in 2006. Infect Genet Evol. 11(5), 1183-1187.

Chen, S.P., Yu, M., Jiang, T., Deng, Y.Q., Qin, C.F., Han, J.F., Qin, E.D., 2008a. Identification of a recombinant dengue virus type 1 with 3 recombination regions in natural populations in Guangdong province, China. Arch Virol. 153(6), 1175-1179.

Chen, S.T., Lin, Y.L., Huang, M.T., Wu, M.F., Cheng, S.C., Lei, H.Y., Lee, C.K., Chiou, T.W., Wong, C.H., Hsieh, S.L., 2008b. CLEC5A is critical for dengue-virus-induced lethal disease. Nature. 453(7195), 672-676.

Chen, Y., Maguire, T., Hileman, R.E., Fromm, J.R., Esko, J.D., Linhardt, R.J., Marks, R.M., 1997. Dengue 160

virus infectivity depends on envelope protein binding to target cell heparan sulfate. Nat Med. 3(8), 866-871.

Chen, Y.C., Wang, S.Y., King, C.C., 1999. Bacterial lipopolysaccharide inhibits dengue virus infection of primary human monocytes/macrophages by blockade of virus entry via a CD14-dependent mechanism. J Virol. 73(4), 2650-2657.

Chew, A., Leng, G.A., Yuen, H., Teik, K.O., Kiat, L.Y., Hong, L.C., Wells, R., 1961. A haemorrhagic fever in Singapore. Lancet. 1(7172), 307-10.

Chowell, G., Cazelles, B., Broutin, H., Munayco, C.V., 2011. The influence of geographic and climate factors on the timing of dengue epidemics in Peru, 1994-2008. BMC Infect Dis. 11, 164.

Chu, J.J.H., Ng, M.L., 2004. Infectious entry of West Nile virus occurs through a clathrin-mediated endocytic pathway. Journal of virology. 78(19), 10543-55.

Chungue, E., Burucoa, C., Boutin, J.P., Philippon, G., Laudon, F., Plichart, R., Barbazan, P., Cardines, R., Roux, J., 1992. Dengue 1 epidemic in French Polynesia, 1988-1989: surveillance and clinical, epidemiological, virological and serological findings in 1752 documented clinical cases. Trans R Soc Trop Med Hyg. 86(2), 193-197.

Chungue, E., Cassar, O., Drouet, M.T., Guzman, M.G., Laille, M., Rosen, L., Deubel, V., 1995. Molecular epidemiology of dengue-1 and dengue-4 viruses. J Gen Virol. 76 ( Pt 7), 1877-1884.

Clemens, M.J., Williams, B.R., 1978. Inhibition of cell-free protein synthesis by pppA2’p5'A2'p5'A: a novel oligonucleotide synthesized by interferon-treated L cell extracts. Cell. 13(3), 565-72.

Clyde, K., Kyle, J.L., Harris, E., 2006. Recent advances in deciphering viral and host determinants of dengue virus replication and pathogenesis. J Virol. 80(23), 11418-11431.

Coffey, L.L., Vasilakis, N., Brault, A.C., Powers, A.M., Tripet, F., Weaver, S.C., 2008. Arbovirus evolution in vivo is constrained by host alternation. Proceedings of the National Academy of Sciences of the United States of America. 105(19), 6970-5.

Cologna, R., Armstrong, P.M., Rico-Hesse, R., 2005. Selection for virulent dengue viruses occurs in humans and mosquitoes. J Virol. 79(2), 853-859.

Cologna, R., Rico-Hesse, R., 2003. American genotype structures decrease dengue virus output from human monocytes and dendritic cells. J Virol. 77(7), 3929-3938.

Colon-Gonzalez, F.J., Lake, I.R., Bentham, G., 2011. Climate variability and dengue fever in warm and humid Mexico. Am J Trop Med Hyg. 84(5), 757-763.

Condon, R., Taleo, G., Stewart, T., Sweeney, T., Kiedrzynski, T., 2000. Dengue surveillance in the Pacific Islands. Pac Health Dialog. 7(2), 122-126.

Costa, S.M., Freire, M.S., Alves, A.M., 2006. DNA vaccine against the non-structural 1 protein (NS1) of dengue 2 virus. Vaccine. 24(21), 4562-4564.

Craig, S., Thu, H.M., Lowry, K., Wang, X.F., Holmes, E.C., Aaskov, J., 2003. Diverse dengue type 2 virus 161

populations contain recombinant and both parental viruses in a single mosquito host. J Virol. 77(7), 4463-4467.

Cummings, D.A., Iamsirithaworn, S., Lessler, J.T., McDermott, A., Prasanthong, R., Nisalak, A., Jarman, R.G., Burke, D.S., Gibbons, R.V., 2009. The impact of the demographic transition on dengue in Thailand: insights from a statistical analysis and mathematical modeling. PLoS Med. 6(9), e1000139.

Cummings, D.A., Irizarry, R.A., Huang, N.E., Endy, T.P., Nisalak, A., Ungchusak, K., Burke, D.S., 2004. Travelling waves in the occurrence of dengue haemorrhagic fever in Thailand. Nature. 427(6972), 344-347.

Cummings, D.A., Schwartz, I.B., Billings, L., Shaw, L.B., Burke, D.S., 2005. Dynamic effects of antibody- dependent enhancement on the fitness of viruses. Proc Natl Acad Sci U S A. 102(42), 15259-15264. da Cunha, R.V., Dias, M., Nogueira, R.M., Chagas, N., Miagostovich, M.P., Schatzmayr, H.G., 1995. Secondary dengue infection in schoolchildren in a dengue endemic area in the state of Rio de Janeiro, Brazil. Rev Inst Med Trop Sao Paulo. 37(6), 517-521.

Cuong, H.Q., Hien, N.T., Duong, T.N., Phong, T.V., Cam, N.N., Farrar, J., Nam, V.S., Thai, K.T., Horby, P., 2011. Quantifying the emergence of dengue in Hanoi, Vietnam: 1998-2009. PLoS Negl Trop Dis. 5(9), e1322.

Das, P., Shapiro, D.S., 2002. Infectious disease surveillance update. Lancet Infect Dis. 2(5), 267.

DeRoeck, D., Deen, J., Clemens, J.D., 2003. Policymakers’ views on dengue fever/dengue haemorrhagic fever and the need for dengue vaccines in four southeast Asian countries. Vaccine. 22(1), 121-129.

Deen, J.L., Harris, E., Wills, B., Balmaseda, A., Hammond, S.N., Rocha, C., Dung, N.M., Hung, N.T., Hien, T.T., Farrar, J.J., 2006. The WHO dengue classification and case definitions: time for a reassessment. Lancet. 368(9530), 170-173.

Dejnirattisai, W., Jumnainsong, A., Onsirisakul, N., Fitton, P., Vasanawathana, S., Limpitikul, W., Puttikhunt, C., Edwards, C., Duangchinda, T., Supasa, S., Chawansuntati, K., Malasit, P., Mongkolsapaya, J., Screaton, G., 2010. Cross-reacting antibodies enhance dengue virus infection in humans. Science. 328(5979), 745-748.

Derrick, E.H., Bicks, V.A., 1958. The limiting temperature for the transmission of dengue. Australas Ann Med. 7(2), 102-107.

Diallo, M., Ba, Y., Faye, O., Soumare, M.L., Dia, I., Sall, A.A., 2008. Vector competence of Aedes aegypti populations from Senegal for sylvatic and epidemic dengue 2 virus isolated in West Africa. Trans R Soc Trop Med Hyg. 102(5), 493-498.

Diallo, M., Sall, A.A., Moncayo, A.C., Ba, Y., Fernandez, Z., Ortiz, D., Coffey, L.L., Mathiot, C., Tesh, R.B., Weaver, S.C., 2005. Potential role of sylvatic and domestic African mosquito species in dengue emergence. Am J Trop Med Hyg. 73(2), 445-449.

Diamond, M.S., Roberts, T.G., Edgil, D., Lu, B., Ernst, J., Harris, E., 2000. Modulation of Dengue virus infection in human cells by alpha, beta, and gamma interferons. J Virol. 74(11), 4957-4966.

162

Diaz, F.J., Black, W.C. th, Farfan-Ale, J.A., Lorono-Pino, M.A., Olson, K.E., Beaty, B.J., 2006. Dengue virus circulation and evolution in Mexico: a phylogenetic perspective. Arch Med Res. 37(6), 760-773.

Dorji, T., Yoon, I.K., Holmes, E.C., Wangchuk, S., Tobgay, T., Nisalak, A., Chinnawirotpisan, P., Sangkachantaranon, K., Gibbons, R.V., Jarman, R.G., 2009. Diversity and origin of dengue virus serotypes 1, 2, and 3, Bhutan. Emerg Infect Dis. 15(10), 1630-1632.

Drake, J.W., Holland, J.J., 1999. Mutation rates among RNA viruses. Proceedings of the National Academy of Sciences of the United States of America. 96(24), 13910-3.

Drummond, A., Pybus, O.G., Rambaut, A., 2003. Inference of viral evolutionary rates from molecular sequences. Adv Parasitol. 54, 331-358.

Drummond, A.J., Nicholls, G.K., Rodrigo, A.G., Solomon, W., 2002. Estimating mutation parameters, population history and genealogy simultaneously from temporally spaced sequence data. Genetics. 161(3), 1307-20.

Drummond, A.J., Rambaut, A., 2007. BEAST: Bayesian evolutionary analysis by sampling trees. BMC Evol Biol. 7, 214.

Drummond, A.J., Rambaut, A., Shapiro, B., Pybus, O.G., 2005. Bayesian coalescent inference of past population dynamics from molecular sequences. Molecular biology and evolution. 22(5), 1185-92.

Dunham, E.J., Holmes, E.C., 2007. Inferring the timescale of dengue virus evolution under realistic models of DNA substitution. J Mol Evol. 64(6), 656-661.

Durand, J.P., Vallee, L., de Pina, J.J., Tolou, H., 2000. Isolation of a dengue type 1 virus from a soldier in West Africa (Cote d’Ivoire). Emerg Infect Dis. 6(1), 83-84.

Egger, J.R., Ooi, E.E., Kelly, D.W., Woolhouse, M.E., Davies, C.R., Coleman, P.G., 2008. Reconstructing historical changes in the force of infection of dengue fever in Singapore: implications for surveillance and control. Bull World Health Organ. 86(3), 187-196.

Egloff, M.P., Decroly, E., Malet, H., Selisko, B., Benarroch, D., Ferron, F., Canard, B., 2007. Structural and functional analysis of methylation and 5’-RNA sequence requirements of short capped RNAs by the methyltransferase domain of dengue virus NS5. J Mol Biol. 372(3), 723-736.

Endy, T.P., Anderson, K.B., Nisalak, A., Yoon, I.K., Green, S., Rothman, A.L., Thomas, S.J., Jarman, R.G., Libraty, D.H., Gibbons, R.V., 2011. Determinants of inapparent and symptomatic dengue infection in a prospective study of primary school children in Kamphaeng Phet, Thailand. PLoS Negl Trop Dis. 5(3), e975.

Endy, T.P., Chunsuttiwat, S., Nisalak, A., Libraty, D.H., Green, S., Rothman, A.L., Vaughn, D.W., Ennis, F.A., 2002a. Epidemiology of inapparent and symptomatic acute dengue virus infection: a prospective study of primary school children in Kamphaeng Phet, Thailand. Am J Epidemiol. 156(1), 40-51.

Endy, T.P., Nisalak, A., Chunsuttitwat, S., Vaughn, D.W., Green, S., Ennis, F.A., Rothman, A.L., Libraty, D.H., 2004. Relationship of preexisting dengue virus (DV) neutralizing antibody levels to viremia and severity of disease in a prospective cohort study of DV infection in Thailand. J Infect Dis. 189(6),

163

990-1000.

Endy, T.P., Nisalak, A., Chunsuttiwat, S., Libraty, D.H., Green, S., Rothman, A.L., Vaughn, D.W., Ennis, F.A., 2002b. Spatial and temporal circulation of dengue virus serotypes: a prospective study of primary school children in Kamphaeng Phet, Thailand. Am J Epidemiol. 156(1), 52-59.

Eram, S., Setyabudi, Y., Sadono, T.I., Sutrisno, D.S., Gubler, D.J., Sulianti Saroso, J., 1979. Epidemic dengue hemorrhagic fever in rural Indonesia. II. Clinical studies. Am J Trop Med Hyg. 28(4), 711- 716.

Falgout, B., Pethel, M., Zhang, Y.M., Lai, C.J., 1991. Both nonstructural proteins NS2B and NS3 are required for the proteolytic processing of dengue virus nonstructural proteins. J Virol. 65(5), 2467- 2475.

Felsenstein, J., 1973. Maximum-likelihood estimation of evolutionary trees from continuous characters. American journal of human genetics. 25(5), 471-92.

Felsenstein, J., 1981. Evolutionary trees from DNA sequences: a maximum likelihood approach. Journal of molecular evolution. 17(6), 368-76.

Felsenstein, J., 1982. How can we infer geography and history from gene frequencies? Journal of theoretical biology. 96(1), 9-20.

Ferguson, N., Anderson, R., Gupta, S., 1999. The effect of antibody-dependent enhancement on the transmission dynamics and persistence of multiple-strain pathogens. Proc Natl Acad Sci U S A. 96(2), 790-794.

Fernandez-Garcia, M.D., Mazzon, M., Jacobs, M., Amara, A., 2009. Pathogenesis of flavivirus infections: using and abusing the host cell. Cell Host Microbe. 5(4), 318-328.

Figueiredo, M.A., Rodrigues, L.C., Barreto, M.L., Lima, J.W., Costa, M.C., Morato, V., Blanton, R., Vasconcelos, P.F., Nunes, M.R., Teixeira, M.G., 2010. Allergies and diabetes as risk factors for dengue hemorrhagic fever: results of a case control study. PLoS Negl Trop Dis. 4(6), e699.

Focks, D.A., Haile, D.G., Daniels, E., Mount, G.A., 1993. Dynamic life table model for Aedes aegypti (diptera: Culicidae): simulation results and validation. J Med Entomol. 30(6), 1018-1028.

Foster, J.E., Bennett, S.N., Carrington, C.V., Vaughan, H., McMillan, W.O., 2004. Phylogeography and molecular evolution of dengue 2 in the Caribbean basin, 1981-2000. Virology. 324(1), 48-59.

Foster, J.E., Bennett, S.N., Vaughan, H., Vorndam, V., McMillan, W.O., Carrington, C.V., 2003. Molecular evolution and phylogeny of dengue type 4 virus in the Caribbean. Virology. 306(1), 126-134.

Franco, L., Di Caro, A., Carletti, F., Vapalahti, O., Renaudat, C., Zeller, H., Tenorio, A., 2010. Recent expansion of dengue virus serotype 3 in West Africa. Euro Surveill. 15(7).

General Statistical Office in Ho Chi Minh City, 2007. Population and population density in 2007 by districts.

General Statistical Office in Ho Chi Minh City 2005, 2006. Statistical yearbook of Ho Chi Minh City. Ho 164

Chi Minh City, Viet Nam.

General Statistics Office of Vietnam, 2008. Population and population density by province, 2008.

Gibbons, R.V., Kalanarooj, S., Jarman, R.G., Nisalak, A., Vaughn, D.W., Endy, T.P., Mammen Jr., M.P., Srikiatkhachorn, A., 2007. Analysis of repeat hospital admissions for dengue to estimate the frequency of third or fourth dengue infections resulting in admissions and dengue hemorrhagic fever, and serotype sequences. Am J Trop Med Hyg. 77(5), 910-913.

Glaziou, P., Chungue, E., Gestas, P., Soulignac, O., Couter, J.P., Plichart, R., Roux, J.F., Poli, L., 1992. Dengue fever and dengue shock syndrome in French Polynesia. Southeast Asian J Trop Med Public Health. 23(3), 531-532.

Godding, C.C., 1890. An Account of an Obscure Outbreak of Dengue Occurring on Board H.M.S. Agamemnon while Stationed at Zanzibar between November, 1888, and September, 1889. Br Med J. 1(1520), 352-354.

Goh, K.T., 1995. Changing epidemiology of dengue in Singapore. Lancet. 346(8982), 1098.

Gollins, S.W., Porterfield, J.S., 1985. Flavivirus infection enhancement in macrophages: an electron microscopic study of viral cellular entry. The Journal of general virology. 66 ( Pt 9), 1969-82.

Gollins, S.W., Porterfield, J.S., 1986. The uncoating and infectivity of the flavivirus West Nile on interaction with cells: effects of pH and ammonium chloride. The Journal of general virology. 67 ( Pt 9), 1941-50.

Goncalvez, A.P., Engle, R.E., St Claire, M., Purcell, R.H., Lai, C.J., 2007. Monoclonal antibody-mediated enhancement of dengue virus infection in vitro and in vivo and strategies for prevention. Proc Natl Acad Sci U S A. 104(22), 9422-9427.

Goncalvez, A.P., Escalante, A.A., Pujol, F.H., Ludert, J.E., Tovar, D., Salas, R.A., Liprandi, F., 2002. Diversity and evolution of the envelope gene of dengue virus type 1. Virology. 303(1), 110-119.

Gorbalenya, A.E., Donchenko, A.P., Koonin, E.V., Blinov, V.M., 1989. N-terminal domains of putative helicases of flavi- and pestiviruses may be serine proteases. Nucleic acids research. 17(10), 3889- 97.

Gordon Smith, C.E., 1956a. The history of dengue in tropical Asia and its probable relationship to the mosquito Aedes aegypti. J Trop Med Hyg. 59(10), 243-251.

Gordon Smith, C.E., 1956b. A localized outbreak of dengue fever in Kuala Lumpur: epidemiological and clinical aspects. Med J Malaya. 10(4), 289-303.

Gould, E.A., Gallian, P., De Lamballerie, X., Charrel, R.N., 2010. First cases of autochthonous dengue fever and chikungunya fever in France: from bad dream to reality! Clin Microbiol Infect. 16(12), 1702-1704.

Gould, E.A., Solomon, T., 2008. Pathogenic flaviviruses. Lancet. 371(9611), 500-509.

Graham, R.R., Juffrie, M., Tan, R., Hayes, C.G., Laksono, I., Ma’roef, C., Erlin, Sutaryo, Porter, K.R., 165

Halstead, S.B., 1999. A prospective seroepidemiologic study on dengue in children four to nine years of age in Yogyakarta, Indonesia I. studies in 1995-1996. Am J Trop Med Hyg. 61(3), 412-419.

Grenfell, B.T., Bjørnstad, O.N., Kappey, J., 2001. Travelling waves and spatial hierarchies in measles epidemics. Nature. 414(6865), 716-23.

Gromowski, G.D., Barrett, A.D., 2007. Characterization of an antigenic site that contains a dominant, type-specific neutralization determinant on the envelope protein domain III (ED3) of dengue 2 virus. Virology. 366(2), 349-360.

Gubler, D.J., 1997. Dengue and dengue hemorrhagic fever: its history and resurgence as a global public health problem, in: Gubler, D.J., Kuno, G. (Eds.), Dengue and Dengue Hemorrhagic Fever. CAB International, New York, pp. 1-22.

Gubler, D.J., 2002. Epidemic dengue/dengue hemorrhagic fever as a public health, social and economic problem in the 21st century. Trends Microbiol. 10(2), 100-103.

Gubler, D.J., Nalim, S., Tan, R., Saipan, H., Sulianti Saroso, J., 1979a. Variation in susceptibility to oral infection with dengue viruses among geographic strains of Aedes aegypti. Am J Trop Med Hyg. 28(6), 1045-1052.

Gubler, D.J., Reiter, P., Ebi, K.L., Yap, W., Nasci, R., Patz, J.A., 2001. Climate variability and change in the United States: potential impacts on vector- and rodent-borne diseases. Environ Health Perspect. 109 Suppl , 223-233.

Gubler, D.J., Rosen, L., 1976. Variation among geographic strains of Aedes albopictus in susceptibility to infection with dengue viruses. Am J Trop Med Hyg. 25(2), 318-325.

Gubler, D.J., Rosen, L., 1977. Quantitative aspects of replication of dengue viruses in Aedes albopictus (Diptera: Culicidae) after oral and parenteral infection. J Med Entomol. 13(4-5), 469-472.

Gubler, D.J., Suharyono, W., Sumarmo, Wulur, H., Jahja, E., Sulianti Saroso, J., 1979b. Virological surveillance for dengue haemorrhagic fever in Indonesia using the mosquito inoculation technique. Bull World Health Organ. 57(6), 931-936.

Guindon, S., Gascuel, O., 2003. A simple, fast, and accurate algorithm to estimate large phylogenies by maximum likelihood. Systematic Biology. 52(5), 696-704.

Gupta, E., Dar, L., Kapoor, G., Broor, S., 2006. The changing epidemiology of dengue in Delhi, India. Virol J. 3, 92.

Guzman, M.G., Halstead, S.B., Artsob, H., Buchy, P., Farrar, J., Gubler, D.J., Hunsperger, E., Kroeger, A., Margolis, H.S., Martinez, E., Nathan, M.B., Pelegrino, J.L., Simmons, C., Yoksan, S., Peeling, R.W., 2010. Dengue: a continuing global threat. Nat Rev Microbiol. 8(12 Suppl), S7-16.

Guzman, M.G., Kouri, G., 2003. Dengue and dengue hemorrhagic fever in the Americas: lessons and challenges. J Clin Virol. 27(1), 1-13.

Guzman, M.G., Kouri, G., Bravo, J., Soler, M., Martinez, E., 1991. Sequential infection as risk factor for dengue hemorrhagic fever/dengue shock syndrome (DHF/DSS) during the 1981 dengue 166

hemorrhagic Cuban epidemic. Mem Inst Oswaldo Cruz. 86(3), 367.

Guzman, M.G., Kouri, G., Martinez, E., Bravo, J., Riveron, R., Soler, M., Vazquez, S., Morier, L., 1987. Clinical and serologic study of Cuban children with dengue hemorrhagic fever/dengue shock syndrome (DHF/DSS). Bull Pan Am Health Organ. 21(3), 270-279.

Guzman, M.G., Kouri, G., Valdes, L., Bravo, J., Vazquez, S., Halstead, S.B., 2002. Enhanced severity of secondary dengue-2 infections: death rates in 1981 and 1997 Cuban outbreaks. Rev Panam Salud Publica. 11(4), 223-227.

Halstead, S.B., 1965. Dengue and hemorrhagic fevers of Southeast Asia. Yale J Biol Med. 37(6), 434- 454.

Halstead, S.B., 1979. In vivo enhancement of dengue virus infection in rhesus monkeys by passively transferred antibody. J Infect Dis. 140(4), 527-533.

Halstead, S.B., 1980. Dengue haemorrhagic fever--a public health problem and a field for research. Bull World Health Organ. 58(1), 1-21.

Halstead, S.B., 1982. Immune enhancement of viral infection. Prog Allergy. 31, 301-364.

Halstead, S.B., 1992. The 20th century dengue pandemic: need for surveillance and research. World Health Stat Q. 45(2-3), 292-298.

Halstead, S.B., 2003. Neutralization and antibody-dependent enhancement of dengue viruses. Adv Virus Res. 60, 421-467.

Halstead, S.B., Chow, J.S., Marchette, N.J., 1973. Immunological enhancement of dengue virus replication. Nat New Biol. 243(122), 24-26.

Halstead, S.B., Nimmannitya, S., Cohen, S.N., 1970. Observations related to pathogenesis of dengue hemorrhagic fever. IV. Relation of disease severity to antibody response and virus recovered. Yale J Biol Med. 42(5), 311-328.

Halstead, S.B., Nimmannitya, S., Yamarat, C., Russell, P.K., 1967. Hemorrhagic fever in Thailand; recent knowledge regarding etiology. Jpn J Med Sci Biol. 20 Suppl, 96-103.

Halstead, S.B., Voulgaropoulos, E.M., Tien, N.H., Udomsakdi, S., 1965. Dengue hemorrhagic fever in South Vietnam: report of the 1963 outbreak. Am J Trop Med Hyg. 14(5), 819-830.

Halstead, S.B., Yamarat, C., 1965. Recent Epidemics of Hemorrhagic Fever in Thailand. Observations Related to Pathogenesis of a “New” Dengue Disease. Am J Public Health Nations Health. 55, 1386- 1395.

Hammon, W.M., 1973. Dengue hemorrhagic fever--do we know its cause? Am J Trop Med Hyg. 22(1), 82-91.

Hammon, W.M., Rudnick, A., Sather, G.E., 1960. Viruses associated with epidemic hemorrhagic fevers of the Philippines and Thailand. Science. 131(3407), 1102-1103.

167

Hammond, S.N., Balmaseda, A., Perez, L., Tellez, Y., Saborio, S.I., Mercado, J.C., Videa, E., Rodriguez, Y., Perez, M.A., Cuadra, R., Solano, S., Rocha, J., Idiaquez, W., Gonzalez, A., Harris, E., 2005. Differences in dengue severity in infants, children, and adults in a 3-year hospital-based study in Nicaragua. Am J Trop Med Hyg. 73(6), 1063-1070.

Hang, V.T.T., Holmes, E.C., Duong, V., Nguyen, T.Q., Tran, T.H., Quail, M., Churcher, C., Parkhill, J., Cardosa, J., Farrar, J., Wills, B., Lennon, N.J., Birren, B.W., Buchy, P., Henn, M.R., Simmons, C.P., 2010. Emergence of the Asian 1 genotype of dengue virus serotype 2 in viet nam: in vivo fitness advantage and lineage replacement in South-East Asia. PLoS Negl Trop Dis. 4(7), e757.

Hanley, K.A., Nelson, J.T., Schirtzinger, E.E., Whitehead, S.S., Hanson, C.T., 2008. Superior infectivity for mosquito vectors contributes to competitive displacement among strains of dengue virus. BMC Ecol. 8, 1.

Hansen, J.A., 1990. Our race to the future. J Am Mosq Control Assoc. 6(3), 361-365.

Harrington, L.C., Scott, T.W., Lerdthusnee, K., Coleman, R.C., Costero, A., Clark, G.G., Jones, J.J., Kitthawee, S., Kittayapong, P., Sithiprasasna, R., Edman, J.D., 2005. Dispersal of the dengue vector Aedes aegypti within and between rural communities. Am J Trop Med Hyg. 72(2), 209-220.

Hastings, W.K., 1970. Monte Carlo sampling methods using Markov Chains and their applications. Biometrika. 57, 97-109.

Hayes, C.G., Phillips, I.A., Callahan, J.D., Griebenow, W.F., Hyams, K.C., Wu, S.J., Watts, D.M., 1996. The epidemiology of dengue virus infection among urban, jungle, and rural populations in the Amazon region of Peru. Am J Trop Med Hyg. 55(4), 459-463.

Henchal, E.A., Henchal, L.S., Schlesinger, J.J., 1988. Synergistic interactions of anti-NS1 monoclonal antibodies protect passively immunized mice from lethal challenge with dengue 2 virus. J Gen Virol. 69 ( Pt 8), 2101-2107.

Higa, Y., Yen, N.T., Kawada, H., Son, T.H., Hoa, N.T., Takagi, M., 2010. Geographic distribution of Aedes aegypti and Aedes albopictus collected from used tires in Vietnam. J Am Mosq Control Assoc. 26(1), 1-9.

Hii, Y.L., Rocklov, J., Ng, N., Tang, C.S., Pang, F.Y., Sauerborn, R., 2009. Climate variability and increase in intensity and magnitude of dengue incidence in Singapore. Glob Health Action. 2.

Hirshman, J.H., 1976. Communicable disease in the South Pacific Islands, 1. Med J Aust. 2(20), 758-760.

Ho, L.J., Hung, L.F., Weng, C.Y., Wu, W.L., Chou, P., Lin, Y.L., Chang, D.M., Tai, T.Y., Lai, J.H., 2005. Dengue virus type 2 antagonizes IFN-alpha but not IFN-gamma antiviral effect via down-regulating Tyk2-STAT signaling in the human dendritic cell. J Immunol. 174(12), 8163-8172.

Holder, M., Lewis, P.O., 2003. Phylogeny estimation: traditional and Bayesian approaches. Nature reviews. Genetics. 4(4), 275-84.

Holmes, E.C., 2003. Patterns of intra- and interhost nonsynonymous variation reveal strong purifying selection in dengue virus. J Virol. 77(20), 11296-11298.

168

Holmes, E.C., 2008. Evolutionary history and phylogeography of human viruses. Annual review of microbiology. 62, 307-28.

Holmes, E.C., 2009. The evolution and emergence of RNA viruses. Oxford University Press, Oxford, United Kingdom.

Holmes, E.C., Burch, S.S., 2000. The causes and consequences of genetic variation in dengue virus. Trends Microbiol. 8(2), 74-77.

Holmes, E.C., Tio, P.H., Perera, D., Muhi, J., Cardosa, J., 2009. Importation and co-circulation of multiple serotypes of dengue virus in Sarawak, Malaysia. Virus Res. 143(1), 1-5.

Holmes, E.C., Twiddy, S.S., 2003. The origin, emergence and evolutionary genetics of dengue virus. Infect Genet Evol. 3(1), 19-28.

Hotta, S., 1952. Experimental studies on dengue. I. Isolation, identification and modification of the virus. J Infect Dis. 90(1), 1-9.

Huang, J.H., Liao, T.L., Chang, S.F., Su, C.L., Chien, L.J., Kuo, Y.C., Yang, C.F., Lin, C.C., Shu, P.Y., 2007. Laboratory-based dengue surveillance in Taiwan, 2005: a molecular epidemiologic study. Am J Trop Med Hyg. 77(5), 903-909.

Huang, K.J., Yang, Y.C., Lin, Y.S., Huang, J.H., Liu, H.S., Yeh, T.M., Chen, S.H., Liu, C.C., Lei, H.Y., 2006. The dual-specific binding of dengue virus and target cells for the antibody-dependent enhancement of dengue virus infection. J Immunol. 176(5), 2825-2832.

Hurtado-Diaz, M., Riojas-Rodriguez, H., Rothenberg, S.J., Gomez-Dantes, H., Cifuentes, E., 2007. Short communication: impact of climate variability on the incidence of dengue in Mexico. Trop Med Int Health. 12(11), 1327-1337.

Huy, R., Buchy, P., Conan, A., Ngan, C., Ong, S., Ali, R., Duong, V., Yit, S., Ung, S., Te, V., Chroeung, N., Pheaktra, N.C., Uok, V., Vong, S., 2010. National dengue surveillance in Cambodia 1980-2008: epidemiological and virological trends and the impact of vector control. Bull World Health Organ. 88(9), 650-657.

Jarman, R.G., Holmes, E.C., Rodpradit, P., Klungthong, C., Gibbons, R.V., Nisalak, A., Rothman, A.L., Libraty, D.H., Ennis, F.A., Mammen Jr., M.P., Endy, T.P., 2008. Microevolution of Dengue viruses circulating among primary school children in Kamphaeng Phet, Thailand. J Virol. 82(11), 5494-5500.

Jenkins, G.M., Pagel, M., Gould, E.A., de A Zanotto, P.M., Holmes, E.C., 2001. Evolution of base composition and codon usage bias in the genus Flavivirus. Journal of molecular evolution. 52(4), 383-90.

Jessie, K., Fong, M.Y., Devi, S., Lam, S.K., Wong, K.T., 2004. Localization of dengue virus in naturally infected human tissues, by immunohistochemistry and in situ hybridization. J Infect Dis. 189(8), 1411-1418.

Jetten, T.H., Focks, D.A., 1997. Potential changes in the distribution of dengue transmission under climate warming. Am J Trop Med Hyg. 57(3), 285-297.

169

Jindadamrongwech, S., Thepparit, C., Smith, D.R., 2004. Identification of GRP 78 (BiP) as a liver cell expressed receptor element for dengue virus serotype 2. Arch Virol. 149(5), 915-927.

Jing, Q.L., Yang, Z.C., Luo, L., Xiao, X.C., Di, B., He, P., Fu, C.X., Wang, M., Lu, J.H., 2012. Emergence of dengue virus 4 genotype II in Guangzhou, China, 2010: Survey and molecular epidemiology of one community outbreak. BMC Infect Dis. 12(1), 87.

Johnson, A.J., Roehrig, J.T., 1999. New mouse model for dengue virus vaccine testing. J Virol. 73(1), 783-786.

Johnson, B.W., Chambers, T.V., Crabtree, M.B., Guirakhoo, F., Monath, T.P., Miller, B.R., 2004. Analysis of the replication kinetics of the ChimeriVax-DEN 1, 2, 3, 4 tetravalent virus mixture in Aedes aegypti by real-time reverse transcriptase-polymerase chain reaction. Am J Trop Med Hyg. 70(1), 89-97.

Johnson, K.M., Halstead, S.B., Cohen, S.N., 1967. Hemorrhagic fevers of Southeast Asia and South America: a comparative appraisal. Prog Med Virol. 9, 105-158.

Kay, B.H., Fanning, I.D., Mottram, P., 1989. Rearing temperature influences flavivirus vector competence of mosquitoes. Med Vet Entomol. 3(4), 415-422.

Kliks, S.C., Nimmanitya, S., Nisalak, A., Burke, D.S., 1988. Evidence that maternal dengue antibodies are important in the development of dengue hemorrhagic fever in infants. Am J Trop Med Hyg. 38(2), 411-419.

Kouri, G.P., Guzman, M.G., Bravo, J.R., 1987. Why dengue haemorrhagic fever in Cuba? 2. An integral analysis. Trans R Soc Trop Med Hyg. 81(5), 821-823.

Kramer, L.D., Ebel, G.D., 2003. Dynamics of flavivirus infection in mosquitoes. Advances in virus research. 60, 187-232.

Krishnan, M.N., Sukumaran, B., Pal, U., Agaisse, H., Murray, J.L., Hodge, T.W., Fikrig, E., 2007. Rab 5 is required for the cellular entry of dengue and West Nile viruses. J Virol. 81(9), 4881-4885.

Kuhn, R.J., Zhang, W., Rossmann, M.G., Pletnev, S.V., Corver, J., Lenches, E., Jones, C.T., Mukhopadhyay, S., Chipman, P.R., Strauss, E.G., Baker, T.S., Strauss, J.H., 2002. Structure of dengue virus: implications for flavivirus organization, maturation, and fusion. Cell. 108(5), 717-725.

Kuno, G., 2009. Emergence of the severe syndrome and mortality associated with dengue and dengue- like illness: historical records (1890 to 1950) and their compatibility with current hypotheses on the shift of disease manifestation. Clin Microbiol Rev. 22(2), 186-201, Table of Contents.

Kuno, G., Chang, G.J., Tsuchiya, K.R., Karabatsos, N., Cropp, C.B., 1998. Phylogeny of the genus Flavivirus. Journal of virology. 72(1), 73-83.

Kuno, G., Gubler, D.J., Santiago de Weil, N.S., 1985. Antigen capture ELISA for the identification of dengue viruses. J Virol Methods. 12(1-2), 93-103.

Kurane, I., Brinton, M.A., Samson, A.L., Ennis, F.A., 1991. Dengue virus-specific, human CD4+ CD8- cytotoxic T-cell clones: multiple patterns of virus cross-reactivity recognized by NS3-specific T-cell

170

clones. J Virol. 65(4), 1823-1828.

Kurosu, T., Chaichana, P., Yamate, M., Anantapreecha, S., Ikuta, K., 2007. Secreted complement regulatory protein clusterin interacts with dengue virus nonstructural protein 1. Biochem Biophys Res Commun. 362(4), 1051-1056.

Lai, C.Y., Tsai, W.Y., Lin, S.R., Kao, C.L., Hu, H.P., King, C.C., Wu, H.C., Chang, G.J., Wang, W.K., 2008. Antibodies to envelope glycoprotein of dengue virus during the natural course of infection are predominantly cross-reactive and recognize epitopes containing highly conserved residues at the fusion loop of domain II. J Virol. 82(13), 6631-6643.

Lai, M.M., 1992. RNA recombination in animal and plant viruses. Microbiological reviews. 56(1), 61-79.

Lambrechts, L., Fansiri, T., Pongsiri, A., Thaisomboonsuk, B., Klungthong, C., Richardson, J.H., Ponlawat, A., Jarman, R.G., Scott, T.W., 2012. Dengue-1 virus clade replacement in Thailand associated with enhanced mosquito transmission. J Virol. 86(3), 1853-1861.

Lambrechts, L., Paaijmans, K.P., Fansiri, T., Carrington, L.B., Kramer, L.D., Thomas, M.B., Scott, T.W., 2011. Impact of daily temperature fluctuations on dengue virus transmission by Aedes aegypti. Proc Natl Acad Sci U S A. 108(18), 7460-7465.

Lanciotti, R.S., Calisher, C.H., Gubler, D.J., Chang, G.J., Vorndam, A.V., 1992. Rapid detection and typing of dengue viruses from clinical samples by using reverse transcriptase-polymerase chain reaction. J Clin Microbiol. 30(3), 545-551.

Lanciotti, R.S., Gubler, D.J., Trent, D.W., 1997. Molecular evolution and phylogeny of dengue-4 viruses. J Gen Virol. 78 ( Pt 9), 2279-2284.

Lanciotti, R.S., Lewis, J.G., Gubler, D.J., Trent, D.W., 1994. Molecular evolution and epidemiology of dengue-3 viruses. J Gen Virol. 75 ( Pt 1), 65-75.

Lee, J.M., Crooks, A.J., Stephenson, J.R., 1989. The synthesis and maturation of a non-structural extracellular antigen from tick-borne encephalitis virus and its relationship to the intracellular NS1 protein. The Journal of general virology. 70 ( Pt 2), 335-43.

Lefeuvre, P., Martin, D.P., Harkins, G., Lemey, P., Gray, A.J.A., Meredith, S., Lakay, F., Monjane, A., Lett, J.-M., Varsani, A., Heydarnejad, J., 2010. The spread of tomato yellow leaf curl virus from the Middle East to the world. PLoS pathogens. 6(10), e1001164.

Leitmeyer, K.C., Vaughn, D.W., Watts, D.M., Salas, R., Villalobos, I., de, C., Ramos, C., Rico-Hesse, R., 1999. Dengue virus structural differences that correlate with pathogenesis. J Virol. 73(6), 4738- 4747.

Lemey, P., Rambaut, A., Drummond, A.J., Suchard, M.A., 2009a. Bayesian phylogeography finds its roots. PLoS Comput Biol. 5(9), e1000520.

Lemey, P., Rambaut, A., Drummond, A.J., Suchard, M.A., 2009b. Bayesian phylogeography finds its roots. PLoS computational biology. 5(9), e1000520.

Lewin-Koh, N.J., Bivand, R., Pebesma, E.J., Archer, E., Baddeley, A., 2009. maptools: tools for reading 171

and handling spatial objects. R package version 0.7-26.

Lewis, J.A., Chang, G.J., Lanciotti, R.S., Kinney, R.M., Mayer, L.W., Trent, D.W., 1993. Phylogenetic relationships of dengue-2 viruses. Virology. 197(1), 216-224.

Li, H., Clum, S., You, S., Ebner, K.E., Padmanabhan, R., 1999. The serine protease and RNA-stimulated nucleoside triphosphatase and RNA helicase functional domains of dengue virus type 2 NS3 converge within a region of 20 amino acids. J Virol. 73(4), 3108-3116.

Li, L., Lok, S.M., Yu, I.M., Zhang, Y., Kuhn, R.J., Chen, J., Rossmann, M.G., 2008. The flavivirus precursor membrane-envelope protein complex: structure and maturation. Science. 319(5871), 1830-1834.

Libraty, D.H., Young, P.R., Pickering, D., Endy, T.P., Kalayanarooj, S., Green, S., Vaughn, D.W., Nisalak, A., Ennis, F.A., Rothman, A.L., 2002. High circulating levels of the dengue virus nonstructural protein NS1 early in dengue illness correlate with the development of dengue hemorrhagic fever. J Infect Dis. 186(8), 1165-1168.

Limonta, D., Torres, G., Capo, V., Guzman, M.G., 2008. Apoptosis, vascular leakage and increased risk of severe dengue in a type 2 diabetes mellitus patient. Diab Vasc Dis Res. 5(3), 213-214.

Lin, C.C., Huang, Y.H., Shu, P.Y., Wu, H.S., Lin, Y.S., Yeh, T.M., Liu, H.S., Liu, C.C., Lei, H.Y., 2010. Characteristic of dengue disease in Taiwan: 2002-2007. Am J Trop Med Hyg. 82(4), 731-739.

Lin, F.Y., Vo, A.H., Phan, V.B., Nguyen, T.T., Bryla, D., Tran, C.T., Ha, B.K., Dang, D.T., Robbins, J.B., 2000. The epidemiology of typhoid fever in the Dong Thap Province, Mekong Delta region of Vietnam. The American journal of tropical medicine and hygiene. 62(5), 644-8.

Lin, S.R., Hsieh, S.C., Yueh, Y.Y., Lin, T.H., Chao, D.Y., Chen, W.J., King, C.C., Wang, W.K., 2004. Study of sequence variation of dengue type 3 virus in naturally infected mosquitoes and human hosts: implications for transmission and evolution. J Virol. 78(22), 12717-12721.

Lindenbach, B.D., Rice, C.M., 1999. Genetic interaction of flavivirus nonstructural proteins NS1 and NS4A as a determinant of replicase function. Journal of virology. 73(6), 4611-21.

Lindenbach, B.D., Thiel, H.-J., Rice, C.M., 2007. Flaviviridae: the viruses and their replication, in: Knipe, D.M., Howley, P.M. (Eds.), Fields Virology. Lippincott-Raven, Philadelphia, pp. 1101-1152.

Lisova, O., Hardy, F., Petit, V., Bedouelle, H., 2007. Mapping to completeness and transplantation of a group-specific, discontinuous, neutralizing epitope in the envelope protein of dengue virus. J Gen Virol. 88(Pt 9), 2387-2397.

Lobigs, M., 1993. Flavivirus premembrane protein cleavage and spike heterodimer secretion require the function of the viral proteinase NS3. Proceedings of the National Academy of Sciences of the United States of America. 90(13), 6218-22.

Loke, H., Bethell, D.B., Phuong, C.X., Dung, M., Schneider, J., White, N.J., Day, N.P., Farrar, J., Hill, A.V., 2001. Strong HLA class I--restricted T cell responses in dengue hemorrhagic fever: a double- edged sword? J Infect Dis. 184(11), 1369-1373.

172

Lourenco, J., Recker, M., 2010. Viral and epidemiological determinants of the invasion dynamics of novel dengue genotypes. PLoS Negl Trop Dis. 4(11), e894.

Lozach, P.Y., Burleigh, L., Staropoli, I., Navarro-Sanchez, E., Harriague, J., Virelizier, J.L., Rey, F.A., Despres, P., Arenzana-Seisdedos, F., Amara, A., 2005. Dendritic cell-specific intercellular adhesion molecule 3-grabbing non-integrin (DC-SIGN)-mediated enhancement of dengue virus infection is independent of DC-SIGN internalization signals. J Biol Chem. 280(25), 23698-23708.

Lu, L., Lin, H., Tian, L., Yang, W., Sun, J., Liu, Q., 2009. Time series analysis of dengue fever and weather in Guangzhou, China. BMC Public Health. 9, 395.

Lucas, G.N., Amerasinghe, A., Sriranganathan, S., 2000. Dengue haemorrhagic fever in Sri Lanka. Indian J Pediatr. 67(7), 503-504.

Mackenzie, J.M., Jones, M.K., Young, P.R., 1996a. Immunolocalization of the dengue virus nonstructural glycoprotein NS1 suggests a role in viral RNA replication. Virology. 220(1), 232-240.

Mackenzie, J.M., Khromykh, A.A., Jones, M.K., Westaway, E.G., 1998. Subcellular localization and some biochemical properties of the flavivirus Kunjin nonstructural proteins NS2A and NS4A. Virology. 245(2), 203-215.

Mackenzie, J.S., la Brooy, J.T., Hueston, L., Cunningham, A.L., 1996b. Dengue in Australia. J Med Microbiol. 45(3), 159-161.

Maddison, D.R., Maddison, W.P., 2003. MacClade 4: analysis of phylogeny and character evolution. Sunderland: Sinauer Associates.

Malla, S., Thakur, G.D., Shrestha, S.K., Banjeree, M.K., Thapa, L.B., Gongal, G., Ghimire, P., Upadhyay, B.P., Gautam, P., Khanal, S., Nisaluk, A., Jarman, R.G., Gibbons, R.V., 2008. Identification of all dengue serotypes in Nepal. Emerg Infect Dis. 14(10), 1669-1670.

Mammen, M.P., Pimgate, C., Koenraadt, C.J., Rothman, A.L., Aldstadt, J., Nisalak, A., Jarman, R.G., Jones, J.W., Srikiatkhachorn, A., Ypil-Butac, C.A., Getis, A., Thammapalo, S., Morrison, A.C., Libraty, D.H., Green, S., Scott, T.W., 2008. Spatial and temporal clustering of dengue virus transmission in Thai villages. PLoS Med. 5(11), e205.

Marchette, N.J., Halstead, S.B., Falkler Jr., W.A., Stenhouse, A., Nash, D., 1973. Studies on the pathogenesis of dengue infection in monkeys. 3. Sequential distribution of virus in primary and heterologous infections. J Infect Dis. 128(1), 23-30.

Mathew, A., Rothman, A.L., 2008. Understanding the contribution of cellular immunity to dengue disease pathogenesis. Immunol Rev. 225, 300-313.

Meade, M., 1976. A new disease in Southeast Asia: man’s creation of dengue haemorrhagic fever. Pacific Viewpoint. 17, 133-146.

Medin, C.L., Fitzgerald, K.A., Rothman, A.L., 2005. Dengue virus nonstructural protein NS5 induces interleukin-8 transcription and secretion. J Virol. 79(17), 11053-11061.

Mercado-Curiel, R.F., Esquinca-Aviles, H.A., Tovar, R., Diaz-Badillo, A., Camacho-Nuez, M., Munoz Mde, 173

L., 2006. The four serotypes of dengue recognize the same putative receptors in Aedes aegypti midgut and Ae. albopictus cells. BMC Microbiol. 6, 85.

Messer, W.B., Vitarana, U.T., Sivananthan, K., Elvtigala, J., Preethimala, L.D., Ramesh, R., Withana, N., Gubler, D.J., De Silva, A.M., 2002. Epidemiology of dengue in Sri Lanka before and after the emergence of epidemic dengue hemorrhagic fever. Am J Trop Med Hyg. 66(6), 765-773.

Metropolis, N., Rosenbluth, A.W., Rosenbluth, M.N., Teller, A.H., Teller, E., 1953. Equation of state calculations by fast computing machines. J Chem Phys. 21, 1087-1092.

Miller, S., Kastner, S., Krijnse-Locker, J., Buhler, S., Bartenschlager, R., 2007. The non-structural protein 4A of dengue virus is an integral membrane protein inducing membrane alterations in a 2K- regulated manner. J Biol Chem. 282(12), 8873-8882.

Minin, V.N., Suchard, M.A., 2008a. Counting labeled transitions in continuous-time Markov models of evolution. J Math Biol. 56(3), 391-412.

Minin, V.N., Suchard, M.A., 2008b. Fast, accurate and simulation-free stochastic mapping. Philosophical Transactions of the Royal Society B-Biological Sciences. 363(1512), 3985-3995.

Modis, Y., Ogata, S., Clements, D., Harrison, S.C., 2005. Variable surface epitopes in the crystal structure of dengue virus type 3 envelope glycoprotein. J Virol. 79(2), 1223-1231.

Monath, T.P., 1994a. Dengue: the risk to developed and developing countries. Proc Natl Acad Sci U S A. 91(7), 2395-2400.

Monath, T.P., 1994b. Yellow fever and dengue - the interactions of virus, vector, and host in the re- emergence of epidemic disease. Seminars in Virology. 5(133-145).

Mooers, A.O., Schluter, D., 1999. Reconstructing ancestor states with maximum likelihood: support for one- and two-rate models. Syst Biol. 48, 623-633.

Moya, A., Holmes, E.C., González-Candelas, F., 2004. The population genetics and evolutionary epidemiology of RNA viruses. Nature reviews. Microbiology. 2(4), 279-88.

Mukhopadhyay, S., Kuhn, R.J., Rossmann, M.G., 2005. A structural perspective of the flavivirus life cycle. Nat Rev Microbiol. 3(1), 13-22.

Munoz-Jordan, J.L., Sanchez-Burgos, G.G., Laurent-Rolle, M., Garcia-Sastre, A., 2003. Inhibition of interferon signaling by dengue virus. Proc Natl Acad Sci U S A. 100(24), 14333-14338.

Murray, J.M., Aaskov, J.G., Wright, P.J., 1993. Processing of the dengue virus type 2 proteins prM and C- prM. J Gen Virol. 74 ( Pt 2), 175-182.

Muylaert, I.R., Chambers, T.J., Galler, R., Rice, C.M., 1996. Mutagenesis of the N-linked glycosylation sites of the yellow fever virus NS1 protein: effects on virus replication and mouse neurovirulence. Virology. 222(1), 159-68.

Nagao, Y., Svasti, P., Tawatsin, A., Thavara, U., 2008. Geographical structure of dengue transmission

174

and its determinants in Thailand. Epidemiol Infect. 136(6), 843-851.

Nakhapakorn, K., Tripathi, N.K., 2005. An information value based analysis of physical and climatic factors affecting dengue fever and dengue haemorrhagic fever incidence. Int J Health Geogr. 4, 13.

Nayak, V., Dessau, M., Kucera, K., Anthony, K., Ledizet, M., Modis, Y., 2009. Crystal structure of dengue virus type 1 envelope protein in the postfusion conformation and its implications for membrane fusion. J Virol. 83(9), 4338-4344.

Nepokroeff, M., Sytsma, K.J., Wagner, W.L., Zimmer, E.A., 2003. Reconstructing ancestral patterns of colonization and dispersal in the Hawaiian understory tree genus Psychotria (Rubiaceae): a comparison of parsimony and likelihood approaches. Systematic biology. 52(6), 820-38.

Nielsen, R., 2002. Mapping mutations on phylogenies. Systematic biology. 51(5), 729-39.

Nisalak, A., Endy, T.P., Nimmannitya, S., Kalayanarooj, S., Thisayakorn, U., Scott, R.M., Burke, D.S., Hoke, C.H., Innis, B.L., Vaughn, D.W., 2003. Serotype-specific dengue virus circulation and dengue disease in Bangkok, Thailand from 1973 to 1999. Am J Trop Med Hyg. 68(2), 191-202.

Nomaguchi, M., Ackermann, M., Yon, C., You, S., Padmanabhan, R., 2003. De novo synthesis of negative-strand RNA by Dengue virus RNA-dependent RNA polymerase in vitro: nucleotide, primer, and template parameters. J Virol. 77(16), 8831-8842.

OhAinle, M., Balmaseda, A., Macalalad, A.R., Tellez, Y., Zody, M.C., Saborio, S., Nunez, A., Lennon, N.J., Birren, B.W., Gordon, A., Henn, M.R., Harris, E., 2011. Dynamics of dengue disease severity determined by the interplay between viral genetics and serotype-specific immunity. Sci Transl Med. 3(114), 114ra128.

Ooi, E.E., Goh, K.T., Gubler, D.J., 2006. Dengue prevention and 35 years of vector control in Singapore. Emerg Infect Dis. 12(6), 887-893.

Osorio, L., Todd, J., Bradley, D.J., 2004. Travel histories as risk factors in the analysis of urban malaria in Colombia. The American journal of tropical medicine and hygiene. 71(4), 380-6.

Pagel, M.A., 1994. Detecting correlated evolution on phylogenies: a general method for the comparative analysis of discrete characters. Proceedings of the Royal Society B: Biological Sciences. 255, 37- 45.

Pagel, M.A., 1999. The maximum likelihood approach to reconstructing ancestral character states on phylogenies. Syst Biol. 48, 612-622.

Pant, C.P., Jatanasen, S., Yasuno, M., 1973. Prevalence of Aedes aegypti and Aedes albopictus and observations on the ecology of dengue haemorrhagic fever in several areas of Thailand. Southeast Asian J Trop Med Public Health. 4(1), 113-121.

Parker, J., Rambaut, A., Pybus, O.G., 2008. Correlating viral phenotypes with phylogeny: Accounting for phylogenetic uncertainty. Infection Genetics and Evolution. 8(3), 239-246.

Patz, J.A., Epstein, P.R., Burke, T.A., Balbus, J.M., 1996. Global climate change and emerging infectious

175

diseases. JAMA. 275(3), 217-223.

Patz, J.A., Martens, W.J., Focks, D.A., Jetten, T.H., 1998. Dengue fever epidemic potential as projected by general circulation models of global climate change. Environ Health Perspect. 106(3), 147-153.

Pavlovic, J., Zürcher, T., Haller, O., Staeheli, P., 1990. Resistance to influenza virus and vesicular stomatitis virus conferred by expression of human MxA protein. Journal of virology. 64(7), 3370-5.

Pebesma, E.J., Bivand, R.S., 2005. sp:classes and methods for spatial data. R package version 0.9-44.

Pediatric Dengue Vaccine Initiative, 2009. Global burden of dengue.

Pokidysheva, E., Zhang, Y., Battisti, A.J., Bator-Kelly, C.M., Chipman, P.R., Xiao, C., Gregorio, G.G., Hendrickson, W.A., Kuhn, R.J., Rossmann, M.G., 2006. Cryo-EM reconstruction of dengue virus in complex with the carbohydrate recognition domain of DC-SIGN. Cell. 124(3), 485-493.

Polizel, J.R., Bueno, D., Visentainer, J.E., Sell, A.M., Borelli, S.D., Tsuneto, L.T., Dalalio, M.M., Coimbra, M.T., Moliterno, R.A., 2004. Association of human leukocyte antigen DQ1 and dengue fever in a white Southern Brazilian population. Mem Inst Oswaldo Cruz. 99(6), 559-562.

Porter, K.R., Beckett, C.G., Kosasih, H., Tan, R.I., Alisjahbana, B., Rudiman, P.I., Widjaja, S., Listiyaningsih, E., Ma’Roef, C.N., McArdle, J.L., Parwati, I., Sudjana, P., Jusuf, H., Yuwono, D., Wuryadi, S., 2005. Epidemiology of dengue and dengue hemorrhagic fever in a cohort of adults living in Bandung, West Java, Indonesia. Am J Trop Med Hyg. 72(1), 60-66.

Posada, D., Crandall, K.A., 1998. MODELTEST: testing the model of DNA substitution. Bioinformatics. 14(9), 817-818.

Pybus, O.G., Rambaut, A., 2009. Evolutionary analysis of the dynamics of viral infectious disease. Nature reviews. Genetics. 10(8), 540-50.

Quintos, F.N., Lim, L.E., Juliano, L., Reyes, A., Lacson, P., 1954. Hemorrhagic fever observed among children in the Philippines. Philippin J Pediatr. 3, 1-19.

R Development Core Team, 2009. R: A language and environment for statistical computing. Vienna, Austria.

Rabaa, M.A., Ty Hang, V.T., Wills, B., Farrar, J., Simmons, C.P., Holmes, E.C., 2010a. Phylogeography of recently emerged DENV-2 in southern Viet Nam. PLoS neglected tropical diseases. Public Library of Science, 4(7), e766.

Rabaa, M.A., Ty Hang, V.T., Wills, B., Farrar, J., Simmons, C.P., Holmes, E.C., 2010b. Phylogeography of Recently Emerged DENV-2 in Southern Viet Nam. PLoS neglected tropical diseases. 4(7), e766.

Radke, E.G., Gregory, C.J., Kintziger, K.W., Sauber-Schatz, E.K., Hunsperger, E.A., Gallagher, G.R., Barber, J.M., Biggerstaff, B.J., Stanek, D.R., Tomashek, K.M., Blackmore, C.G., 2012. Dengue outbreak in Key West, Florida, USA, 2009. Emerg Infect Dis. 18(1), 135-137.

Raftery, A.E., Newton, M.A., Statgopan, J.M., Krivitsky, P., 2007. Estimating the Integrated Likelihood via Posterior Simulation Using the Harmonic Mean Identity (with Discussion), in: Bernardo, J.M. (Ed.), 176

Bayesian Statistics 8. Oxford University Press, New York.

Raghwani, J., Rambaut, A., Holmes, E.C., Hang, V.T., Hien, T.T., Farrar, J., Wills, B., Lennon, N.J., Birren, B.W., Henn, M.R., Simmons, C.P., 2011. Endemic dengue associated with the co-circulation of multiple viral lineages and localized density-dependent transmission. PLoS Pathog. 7(6), e1002064.

Rajamanonmani, R., Nkenfou, C., Clancy, P., Yau, Y.H., Shochat, S.G., Sukupolvi-Petty, S., Schul, W., Diamond, M.S., Vasudevan, S.G., Lescar, J., 2009. On a mouse monoclonal antibody that neutralizes all four dengue virus serotypes. J Gen Virol. 90(Pt 4), 799-809.

Ramakrishnan, S.P., Gelfand, H.M., Bose, P.N., Sehgal, P.N., Mukherjee, R.N., 1964. The epidemic of acute hemorrhagic fever, Calcutta, 1963: epidemiological inquiry. The Indian journal of medical research. 52, 633-50.

Rambaut, A., 2002. Sequence alignment editor.

Rambaut, A., Pybus, O.G., Nelson, M.I., Viboud, C., Taubenberger, J.K., Holmes, E.C., 2008. The genomic and epidemiological dynamics of human influenza A virus. Nature. 453(7195), 615-9.

Rasul, C.H., Ahasan, H.A., Rasid, A.K., Khan, M.R., 2002. Epidemiological factors of dengue hemorrhagic fever in Bangladesh. Indian Pediatr. 39(4), 369-372.

Recker, M., Blyuss, K.B., Simmons, C.P., Hien, T.T., Wills, B., Farrar, J., Gupta, S., 2009. Immunological serotype interactions and their effect on the epidemiological pattern of dengue. Proc Biol Sci. 276(1667), 2541-2548.

Reiter, P., 1996. [Dengue in the Americas]. Bull Soc Pathol Exot. 89(2), 95-6; discussion 97.

Reiter, P., Amador, M.A., Anderson, R.A., Clark, G.G., 1995. Short report: dispersal of Aedes aegypti in an urban area after blood feeding as demonstrated by rubidium-marked eggs. Am J Trop Med Hyg. 52(2), 177-179.

Reyes-Del Valle, J., Chavez-Salinas, S., Medina, F., Del Angel, R.M., 2005. Heat shock protein 90 and heat shock protein 70 are components of dengue virus receptor complex in human cells. J Virol. 79(8), 4557-4567.

Rico-Hesse, R., 1990. Molecular evolution and distribution of dengue viruses type 1 and 2 in nature. Virology. 174(2), 479-493.

Rico-Hesse, R., 2003. Microevolution and virulence of dengue viruses. Adv Virus Res. 59, 315-341.

Rico-Hesse, R., Harrison, L.M., Salas, R.A., Tovar, D., Nisalak, A., Ramos, C., Boshell, J., de Mesa, M.T., Nogueira, R.M., da Rosa, A.T., 1997. Origins of dengue type 2 viruses associated with increased pathogenicity in the Americas. Virology. 230(2), 244-251.

Rodenhuis-Zybert, I.A., van der Schaar, H.M., da Silva Voorham, J.M., van der Ende-Metselaar, H., Lei, H.Y., Wilschut, J., Smit, J.M., 2010a. Immature dengue virus: a veiled pathogen? PLoS Pathog. 6(1), e1000718.

177

Rodenhuis-Zybert, I.A., Wilschut, J., Smit, J.M., 2010b. Dengue virus life cycle: viral and host factors modulating infectivity. Cell Mol Life Sci. 67(16), 2773-2786.

Rodriguez-Barraquer, I., Cordeiro, M.T., Braga, C., de Souza, W.V., Marques, E.T., Cummings, D.A.T., 2011. From Re-Emergence to Hyperendemicity: The Natural History of the Dengue Epidemic in Brazil. PLoS Neglected Tropical Diseases. 5(1), e935.

Rogers, D.J., Wilson, A.J., Hay, S.I., Graham, A.J., 2006. The global distribution of yellow fever and dengue. Adv Parasitol. 62, 181-220.

Ronquist, F., 2004. Bayesian inference of character evolution. Trends in ecology & evolution. 19(9), 475- 81.

Rosa-Freitas, M.G., Schreiber, K.V., Tsouris, P., Weimann, E.T., Luitgards-Moura, J.F., 2006. Associations between dengue and combinations of weather factors in a city in the Brazilian Amazon. Rev Panam Salud Publica. 20(4), 256-267.

Rosen, L., 1967. A recent outbreak of dengue in French Polynesia. Jpn J Med Sci Biol. 20 Suppl, 67-69.

Rosen, L., Gubler, D., 1974. The use of mosquitoes to detect and propagate dengue viruses. Am J Trop Med Hyg. 23(6), 1153-1160.

Rosenberg, N.A., Nordborg, M., 2002. Genealogical trees, coalescent theory and the analysis of genetic polymorphisms. Nature reviews. Genetics. 3(5), 380-90.

Rudnick, A., 1965. Studies of the ecology of dengue in Malaysia: a preliminary report. J Med Entomol. 2(2), 203-208.

Rudnick, A., 1978. Ecology of dengue virus. Asian J. Infect. Dis. 2, 156-160.

Rudnick, A., Marchette, N.J., Garcia, R., 1967. Possible jungle dengue--recent studies and hypotheses. Jpn J Med Sci Biol. 20 Suppl, 69-74.

Russell, C.A., Jones, T.C., Barr, I.G., Cox, N.J., Garten, R.J., Gregory, V., Gust, I.D., Hampson, A.W., Hay, A.J., Hurt, A.C., de Jong, J.C., Kelso, A., Klimov, A.I., Kageyama, T., Komadina, N., Lapedes, A.S., Lin, Y.P., Mosterin, A., Obuchi, M., Odagiri, T., Osterhaus, A.D.M.E., Rimmelzwaan, G.F., Shaw, M.W., Skepner, E., Stohr, K., Tashiro, M., Fouchier, R.A.M., Smith, D.J., 2008. The global circulation of seasonal influenza A (H3N2) viruses. Science (New York, N.Y.). 320(5874), 340-6.

Russell, P.K., Van Quy, D., Nisalak, A., Simasathien, P., Yuill, T.M., Gould, D.J., 1969. Mosquito vectors of dengue viruses in South Vietnam. Am J Trop Med Hyg. 18(3), 455-459.

Sabin, A.B., 1950. The dengue group of viruses and its family relationships. Bacteriol Rev. 14(3), 225- 232.

Sabin, A.B., 1952. Research on dengue during World War II. Am J Trop Med Hyg. 1(1), 30-50.

Saitou, N., Nei, M., 1987. The neighbor-joining method: a new method for reconstructing phylogenetic trees. Molecular biology and evolution. 4(4), 406-25.

178

Salje, H., Lessler, J., Endy, T.P., Curriero, F.C., Gibbons, R.V., Nisalak, A., Nimmannitya, S., Kalayanarooj, S., Jarman, R.G., Thomas, S.J., Burke, D.S., Cummings, D.A.T., 2012. Revealing the microscale spatial signature of dengue transmission and immunity in an urban population. Proceedings of the National Academy of Sciences. 109(24), 9535-8.

San Martin, J.L., Brathwaite, O., Zambrano, B., Solorzano, J.O., Bouckenooghe, A., Dayan, G.H., Guzman, M.G., 2010. The epidemiology of dengue in the americas over the last three decades: a worrisome reality. Am J Trop Med Hyg. 82(1), 128-135.

Sangiambut, S., Keelapang, P., Aaskov, J., Puttikhunt, C., Kasinrerk, W., Malasit, P., Sittisombut, N., 2008. Multiple regions in dengue virus capsid protein contribute to nuclear localization during virus infection. J Gen Virol. 89(Pt 5), 1254-1264.

Sangkawibha, N., Rojanasuphot, S., Ahandrik, S., Viriyapongse, S., Jatanasen, S., Salitul, V., Phanthumachinda, B., Halstead, S.B., 1984. Risk factors in dengue shock syndrome: a prospective epidemiologic study in Rayong, Thailand. I. The 1980 outbreak. Am J Epidemiol. 120(5), 653-669.

Sanmartin, I., van der Mark, P., Ronquist, F., 2008. Inferring dispersal: a Bayesian approach to phylogeny-based island biogeography, with special refernce to the Canary Islands. Journal of biogeography. 35, 428-449. van der Schaar, H.M., Rust, M.J., Chen, C., van der Ende-Metselaar, H., Wilschut, J., Zhuang, X., Smit, J.M., 2008. Dissecting the cell entry pathway of dengue virus by single-particle tracking in living cells. PLoS Pathog. 4(12), e1000244. van der Schaar, H.M., Wilschut, J.C., Smit, J.M., 2009. Role of antibodies in controlling dengue virus infection. Immunobiology. 214(7), 613-629.

Schlesinger, J.J., Brandriss, M.W., Putnak, J.R., Walsh, E.E., 1990. Cell surface expression of yellow fever virus non-structural glycoprotein NS1: consequences of interaction with antibody. The Journal of general virology. 71 ( Pt 3), 593-9.

Schmidt, W.P., Suzuki, M., Thiem, V.D., White, R.G., Tsuzuki, A., Yoshida, L.M., Yanai, H., Haque, U., Tho le, H., Anh, D.D., Ariyoshi, K., 2011. Population density, water supply, and the risk of dengue fever in Vietnam: cohort study and spatial analysis. PLoS Med. 8(8), e1001082.

Schreiber, M.J., Holmes, E.C., Ong, S.H., Soh, H.S., Liu, W., Tanner, L., Aw, P.P., Tan, H.C., Ng, L.C., Leo, Y.S., Low, J.G., Ong, A., Ooi, E.E., Vasudevan, S.G., Hibberd, M.L., 2009. Genomic epidemiology of a dengue virus epidemic in urban Singapore. J Virol. 83(9), 4163-4173.

Scott, T.W., Amerasinghe, P.H., Morrison, A.C., Lorenz, L.H., Clark, G.G., Strickman, D., Kittayapong, P., Edman, J.D., 2000. Longitudinal studies of Aedes aegypti (Diptera: Culicidae) in Thailand and Puerto Rico: blood feeding frequency. J Med Entomol. 37(1), 89-101.

Se-Thoe, S.Y., Ng, M.M., Ling, A.E., 1999. Retrospective study of Western blot profiles in immune sera of natural dengue virus infections. J Med Virol. 57(3), 322-330.

Seligman, S.J., 2008. Constancy and diversity in the flavivirus fusion peptide. Virol J. 5, 27.

Seyler, T., Grandesso, F., Le Strat, Y., Tarantola, A., Depoortere, E., 2009. Assessing the risk of

179

importing dengue and chikungunya viruses to the European Union. Epidemics. 1(3), 175-184.

Shafee, N., AbuBakar, S., 2003. Dengue virus type 2 NS3 protease and NS2B-NS3 protease precursor induce apoptosis. J Gen Virol. 84(Pt 8), 2191-2195.

Shang, C.S., Fang, C.T., Liu, C.M., Wen, T.H., Tsai, K.H., King, C.C., 2010. The role of imported cases and favorable meteorological conditions in the onset of dengue epidemics. PLoS Negl Trop Dis. 4(8), e775.

Shekhar, K.C., Huat, O.L., 1992. Epidemiology of dengue/dengue hemorrhagic fever in Malaysia--a retrospective epidemiological study. 1973-1987. Part II: Dengue fever (DF). Asia Pac J Public Health. 6(3), 15-25.

Shope, R., 1991. Global climate change and infectious diseases. Environ Health Perspect. 96, 171-174.

Shresta, S., Kyle, J.L., Beatty, P.R., Harris, E., 2004a. Early activation of natural killer and B cells in response to primary dengue virus infection in A/J mice. Virology. 319(2), 262-273.

Shresta, S., Kyle, J.L., Snider, H.M., Basavapatna, M., Beatty, P.R., Harris, E., 2004b. Interferon- dependent immunity is essential for resistance to primary dengue virus infection in mice, whereas T- and B-cell-dependent immunity are less critical. J Virol. 78(6), 2701-2710.

Shresta, S., Sharar, K.L., Prigozhin, D.M., Snider, H.M., Beatty, P.R., Harris, E., 2005. Critical roles for both STAT1-dependent and STAT1-independent pathways in the control of primary dengue virus infection in mice. J Immunol. 175(6), 3946-3954.

Shu, P.Y., Chen, L.K., Chang, S.F., Yueh, Y.Y., Chow, L., Chien, L.J., Chin, C., Lin, T.H., Huang, J.H., 2000. Dengue NS1-specific antibody responses: isotype distribution and serotyping in patients with Dengue fever and Dengue hemorrhagic fever. J Med Virol. 62(2), 224-232.

Shu, P.Y., Su, C.L., Liao, T.L., Yang, C.F., Chang, S.F., Lin, C.C., Chang, M.C., Hu, H.C., Huang, J.H., 2009. Molecular characterization of dengue viruses imported into Taiwan during 2003-2007: geographic distribution and genotype shift. Am J Trop Med Hyg. 80(6), 1039-1046.

Simmons, C.P., Dong, T., Chau, N.V., Dung, N.T., Chau, T.N., Thao le, T.T., Hien, T.T., Rowland-Jones, S., Farrar, J., 2005. Early T-cell responses to dengue virus epitopes in Vietnamese adults with secondary dengue virus infections. J Virol. 79(9), 5665-5675.

Singapore, D. of S., 2011. Statistics: population. Department of Statistics Singapore.

Sittisombut, N., Sistayanarain, A., Cardosa, M.J., Salminen, M., Damrongdachakul, S., Kalayanarooj, S., Rojanasuphot, S., Supawadee, J., Maneekarn, N., 1997. Possible occurrence of a genetic bottleneck in dengue serotype 2 viruses between the 1980 and 1987 epidemic seasons in Bangkok, Thailand. Am J Trop Med Hyg. 57(1), 100-108.

Slatkin, M., Maddison, W.P., 1989. A cladistic measure of gene flow inferred from the phylogenies of alleles. Genetics. 123(3), 603-13.

Stabler, B., 2006. shapefiles: Read and write ESRI shapefiles. R package version 0.6.

180

Stephenson, J.R., 2005. Understanding dengue pathogenesis: implications for vaccine design. Bull World Health Organ. 83(4), 308-314.

Stoddard, S.T., Morrison, A.C., Vazquez-Prokopec, G.M., Paz Soldan, V., Kochel, T.J., Kitron, U., Elder, J.P., Scott, T.W., 2009. The role of human movement in the transmission of vector-borne pathogens. PLoS Negl Trop Dis. 3(7), e481.

Strauss, J.H., Strauss, E.G., 1988. Evolution of RNA viruses. Annual review of microbiology. 42, 657-83.

Sugrue, R.J., Cui, T., Xu, Q., Fu, J., Chan, Y.C., 1997. The production of recombinant dengue virus E protein using Escherichia coli and Pichia pastoris. J Virol Methods. 69(1-2), 159-169.

Sweet, B.H., Sabin, A.B., 1954. Properties and antigenic relationships of hemagglutinins associated with the dengue viruses. J Immunol. 73(5), 363-373.

Tabachnick, W.J., Wallis, G.P., Aitken, T.H., Miller, B.R., Amato, G.D., Lorenz, L., Powell, J.R., Beaty, B.J., 1985. Oral infection of Aedes aegypti with yellow fever virus: geographic variation and genetic considerations. The American journal of tropical medicine and hygiene. 34(6), 1219-24.

Talbi, C., Lemey, P., Suchard, M.A., Abdelatif, E., Elharrak, M., Nourlil, J., Jalal, N., Faouzi, A., Echevarría, J.E., Vazquez Morón, S., Rambaut, A., Campiz, N., Tatem, A.J., Holmes, E.C., Bourhy, H., 2010. Phylodynamics and human-mediated dispersal of a zoonotic virus. PLoS pathogens. 6(10), e1001166.

Tassaneetrithep, B., Burgess, T.H., Granelli-Piperno, A., Trumpfheller, C., Finke, J., Sun, W., Eller, M.A., Pattanapanyasat, K., Sarasombath, S., Birx, D.L., Steinman, R.M., Schlesinger, S., Marovich, M.A., 2003. DC-SIGN (CD209) mediates dengue virus infection of human dendritic cells. J Exp Med. 197(7), 823-829.

Thai, K.T., Binh, T.Q., Giao, P.T., Phuong, H.L., Hung le, Q., Van Nam, N., Nga, T.T., Groen, J., Nagelkerke, N., de Vries, P.J., 2005. Seroprevalence of dengue antibodies, annual incidence and risk factors among children in southern Vietnam. Trop Med Int Health. 10(4), 379-386.

Thai, K.T., Cazelles, B., Nguyen, N.V., Vo, L.T., Boni, M.F., Farrar, J., Simmons, C.P., van Doorn, H.R., de Vries, P.J., 2010a. Dengue dynamics in Binh Thuan province, southern Vietnam: periodicity, synchronicity and climate variability. PLoS Negl Trop Dis. 4(7), e747.

Thai, K.T., Henn, M.R., Zody, M.C., Tricou, V., Nguyet, N.M., Charlebois, P., Lennon, N.J., Green, L., de Vries, P.J., Hien, T.T., Farrar, J., van Doorn, H.R., de Jong, M.D., Birren, B.W., Holmes, E.C., Simmons, C.P., 2012. High-resolution analysis of intrahost genetic diversity in dengue virus serotype 1 infection identifies mixed infections. J Virol. 86(2), 835-843.

Thai, K.T., Nagelkerke, N., Phuong, H.L., Nga, T.T., Giao, P.T., Hung, L.Q., Binh, T.Q., Nam, N.V., De Vries, P.J., 2010b. Geographical heterogeneity of dengue transmission in two villages in southern Vietnam. Epidemiol Infect. 138(4), 585-591.

Thai, K.T., Nishiura, H., Hoang, P.L., Tran, N.T., Phan, G.T., Le, H.Q., Tran, B.Q., Nguyen, N.V., de Vries, P.J., 2011. Age-specificity of clinical dengue during primary and secondary infections. PLoS Negl Trop Dis. 5(6), e1180.

181

Thailand, N.S.O., 2009. The 2008 key statistics of Thailand. National Statistical Office Thailand.

Thein, S., Aung, M.M., Shwe, T.N., Aye, M., Zaw, A., Aye, K., Aye, K.M., Aaskov, J., 1997. Risk factors in dengue shock syndrome. Am J Trop Med Hyg. 56(5), 566-572.

Thepparit, C., Smith, D.R., 2004. Serotype-specific entry of dengue virus into liver cells: identification of the 37-kilodalton/67-kilodalton high-affinity laminin receptor as a dengue virus serotype 1 receptor. J Virol. 78(22), 12647-12656.

Thu, H.M., Lowry, K., Myint, T.T., Shwe, T.N., Han, A.M., Khin, K.K., Thant, K.Z., Thein, S., Aaskov, J., 2004. Myanmar dengue outbreak associated with displacement of serotypes 2, 3, and 4 by dengue 1. Emerg Infect Dis. 10(4), 593-597.

Tolou, H.J., Couissinier-Paris, P., Durand, J.P., Mercier, V., de Pina, J.J., de Micco, P., Billoir, F., Charrel, R.N., de Lamballerie, X., 2001. Evidence for recombination in natural populations of dengue virus type 1 based on the analysis of complete genome sequences. J Gen Virol. 82(Pt 6), 1283-1290.

Trent, D.W., 1977. Antigenic characterization of flavivirus structural proteins separated by isoelectric focusing. Journal of virology. 22(3), 608-18.

Trent, D.W., Harvey, C.L., Qureshi, A., LeStourgeon, D., 1976. Solid-phase radioimmunoassay for antibodies to flavivirus structural and nonstructural proteins. Infect Immun. 13(5), 1325-1333.

Trinchieri, G., Sher, A., 2007. Cooperation of Toll-like receptor signals in innate immune defence. Nature reviews. Immunology. 7(3), 179-90.

Twiddy, S.S., Farrar, J.J., Vinh Chau, N., Wills, B., Gould, E.A., Gritsun, T., Lloyd, G., Holmes, E.C., 2002a. Phylogenetic relationships and differential selection pressures among genotypes of dengue- 2 virus. Virology. 298(1), 63-72.

Twiddy, S.S., Holmes, E.C., 2003. The extent of homologous recombination in members of the genus Flavivirus. J Gen Virol. 84(Pt 2), 429-440.

Twiddy, S.S., Holmes, E.C., Rambaut, A., 2003. Inferring the rate and time-scale of dengue virus evolution. Mol Biol Evol. 20(1), 122-129.

Twiddy, S.S., Woelk, C.H., Holmes, E.C., 2002b. Phylogenetic evidence for adaptive evolution of dengue viruses in nature. J Gen Virol. 83(Pt 7), 1679-1689.

Umareddy, I., Chao, A., Sampath, A., Gu, F., Vasudevan, S.G., 2006. Dengue virus NS4B interacts with NS3 and dissociates it from single-stranded RNA. J Gen Virol. 87(Pt 9), 2605-2614.

Umareddy, I., Tang, K.F., Vasudevan, S.G., Devi, S., Hibberd, M.L., Gu, F., 2008. Dengue virus regulates type I interferon signalling in a strain-dependent manner in human cell lines. J Gen Virol. 89(Pt 12), 3052-3062.

Uzcategui, N.Y., Camacho, D., Comach, G., Cuello de Uzcategui, R., Holmes, E.C., Gould, E.A., 2001. Molecular epidemiology of dengue type 2 virus in Venezuela: evidence for in situ virus evolution and recombination. J Gen Virol. 82(Pt 12), 2945-2953.

182

Vasilakis, N., Shell, E.J., Fokam, E.B., Mason, P.W., Hanley, K.A., Estes, D.M., Weaver, S.C., 2007. Potential of ancestral sylvatic dengue-2 viruses to re-emerge. Virology. 358(2), 402-412.

Vasilakis, N., Tesh, R.B., Weaver, S.C., 2008. Sylvatic dengue virus type 2 activity in humans, Nigeria, 1966. Emerg Infect Dis. 14(3), 502-504.

Vasilakis, N., Weaver, S.C., 2008. The history and evolution of human dengue emergence. Adv Virus Res. 72, 1-76.

Vaughn, D.W., Green, S., Kalayanarooj, S., Innis, B.L., Nimmannitya, S., Suntayakorn, S., Endy, T.P., Raengsakulrach, B., Rothman, A.L., Ennis, F.A., Nisalak, A., 2000. Dengue viremia titer, antibody response pattern, and virus serotype correlate with disease severity. J Infect Dis. 181(1), 2-9.

Vaughn, D.W., Green, S., Kalayanarooj, S., Innis, B.L., Nimmannitya, S., Suntayakorn, S., Rothman, A.L., Ennis, F.A., Nisalak, A., 1997. Dengue in the early febrile phase: viremia and antibody responses. J Infect Dis. 176(2), 322-330.

Vazeille-Falcoz, M., Failloux, A.B., Mousson, L., Elissa, N., Rodhain, F., 1999. [Oral receptivity of Aedes aegypti formosus from Franceville (Gabon, central Africa) for type 2 dengue virus]. Bull Soc Pathol Exot. 92(5), 341-342.

Vong, S., Khieu, V., Glass, O., Ly, S., Duong, V., Huy, R., Ngan, C., Wichmann, O., Letson, G.W., Margolis, H.S., Buchy, P., 2010. Dengue incidence in urban and rural Cambodia: results from population-based active fever surveillance, 2006-2008. PLoS Negl Trop Dis. 4(11), e903.

WHO, 2007. Scientific working group report on dengue. World Health Organization, Geneva, Switzerland.

Wang, E., Ni, H., Xu, R., Barrett, A.D., Watowich, S.J., Gubler, D.J., Weaver, S.C., 2000. Evolutionary relationships of endemic/epidemic and sylvatic dengue viruses. J Virol. 74(7), 3227-3234.

Wang, T.H., Donaldson, Y.K., Brettle, R.P., Bell, J.E., Simmonds, P., 2001. Identification of shared populations of human immunodeficiency virus type 1 infecting microglia and tissue macrophages outside the central nervous system. Journal of virology. 75(23), 11686-99.

Wang, W.K., Chao, D.Y., Kao, C.L., Wu, H.C., Liu, Y.C., Li, C.M., Lin, S.C., Ho, S.T., Huang, J.H., King, C.C., 2003. High levels of plasma dengue viral load during defervescence in patients with dengue hemorrhagic fever: implications for pathogenesis. Virology. 305(2), 330-338.

Watts, D.M., Burke, D.S., Harrison, B.A., Whitmire, R.E., Nisalak, A., 1987. Effect of temperature on the vector efficiency of Aedes aegypti for dengue 2 virus. Am J Trop Med Hyg. 36(1), 143-152.

Weaver, S.C., Vasilakis, N., 2009. Molecular evolution of dengue viruses: contributions of phylogenetics to understanding the history and epidemiology of the preeminent arboviral disease. Infect Genet Evol. 9(4), 523-540.

Webster, D.P., Farrar, J., Rowland-Jones, S., 2009. Progress towards a dengue vaccine. Lancet Infect Dis. 9(11), 678-687.

Wilder-Smith, A., Ooi, E.E., Vasudevan, S.G., Gubler, D.J., 2010. Update on dengue: epidemiology, virus

183

evolution, antiviral drugs, and vaccine development. Curr Infect Dis Rep. 12(3), 157-164.

Wilgenbusch, J.C., Swofford, D., 2003. Inferring evolutionary trees with PAUP*. Curr Protoc Bioinformatics. Chapter 6, Unit 6 4.

Wittke, V., Robb, T.E., Thu, H.M., Nisalak, A., Nimmannitya, S., Kalayanrooj, S., Vaughn, D.W., Endy, T.P., Holmes, E.C., Aaskov, J.G., 2002. Extinction and rapid emergence of strains of dengue 3 virus during an interepidemic period. Virology. 301(1), 148-156.

Woelk, C.H., Holmes, E.C., 2002. Reduced positive selection in vector-borne RNA viruses. Molecular biology and evolution. 19(12), 2333-6.

Wolfe, N.D., Kilbourn, A.M., Karesh, W.B., Rahman, H.A., Bosi, E.J., Cropp, B.C., Andau, M., Spielman, A., Gubler, D.J., 2001. Sylvatic transmission of arboviruses among Bornean orangutans. Am J Trop Med Hyg. 64(5-6), 310-316.

World Health Organization, 2009a. Dengue: Guidelines for treatment, prevention and control. Geneva.

World Health Organization, 2009b. DengueNet database.

World Health Organization Western Pacific Region Office, 2008. Report on the Asia-Pacific dengue program managers meeting.

World Health Organization Western Pacific Region Office, 2009. Annual dengue data in the Western Pacific Region.

Wu, J.Y., Lun, Z.R., James, A.A., Chen, X.G., 2010. Dengue Fever in mainland China. Am J Trop Med Hyg. 83(3), 664-671.

Wu, P.C., Guo, H.R., Lung, S.C., Lin, C.Y., Su, H.J., 2007. Weather as an effective predictor for occurrence of dengue fever in Taiwan. Acta Trop. 103(1), 50-57.

Wu, P.C., Lay, J.G., Guo, H.R., Lin, C.Y., Lung, S.C., Su, H.J., 2009. Higher temperature and urbanization affect the spatial patterns of dengue fever transmission in subtropical Taiwan. Sci Total Environ. 407(7), 2224-2233.

Wu, W., Bai, Z., Zhou, H., Tu, Z., Fang, M., Tang, B., Liu, J., Liu, L., Chen, W., 2011. Molecular epidemiology of dengue viruses in southern China from 1978 to 2006. Virol J. 8, 322.

Xi, Z., Ramirez, J.L., Dimopoulos, G., 2008. The Aedes aegypti toll pathway controls dengue virus infection. PLoS Pathog. 4(7), e1000098.

Xia, Y.C., Bjornstad, O.N., Grenfell, B.T., 2004. Measles metapopulation dynamics: A gravity model for epidemiological coupling and dynamics. American Naturalist. 164(2), 267-281.

Yap, T.L., Xu, T., Chen, Y.L., Malet, H., Egloff, M.P., Canard, B., Vasudevan, S.G., Lescar, J., 2007. Crystal structure of the dengue virus RNA-dependent RNA polymerase catalytic domain at 1.85- angstrom resolution. J Virol. 81(9), 4753-4765.

Yauch, L.E., Zellweger, R.M., Kotturi, M.F., Qutubuddin, A., Sidney, J., Peters, B., Prestwood, T.R., Sette, 184

A., Shresta, S., 2009. A protective role for dengue virus-specific CD8+ T cells. J Immunol. 182(8), 4865-4873.

Yazi Mendoza, M., Salas-Benito, J.S., Lanz-Mendoza, H., Hernandez-Martinez, S., del Angel, R.M., 2002. A putative receptor for dengue virus in mosquito tissues: localization of a 45-kDa glycoprotein. Am J Trop Med Hyg. 67(1), 76-84.

Yoon, I.-K., Rothman, A.L., Tannitisupawong, D., Srikiatkhachorn, A., Jarman, R.G., Aldstadt, J., Nisalak, A., Mammen, M.P., Thammapalo, S., Green, S., Libraty, D.H., Gibbons, R.V., Getis, A., Endy, T., Jones, J.W., Koenraadt, C.J.M., Morrison, A.C., Fansiri, T., Pimgate, C., Scott, T.W., 2012. Under- Recognized Mildly Symptomatic Viremic Dengue Virus Infections in Rural Thai Schools and Villages. The Journal of infectious diseases.

Yu, I.M., Zhang, W., Holdaway, H.A., Li, L., Kostyuchenko, V.A., Chipman, P.R., Kuhn, R.J., Rossmann, M.G., Chen, J., 2008. Structure of the immature dengue virus at low pH primes proteolytic maturation. Science. 319(5871), 1834-1837.

Zanotto, P.M., Gould, E.A., Gao, G.F., Harvey, P.H., Holmes, E.C., 1996. Population dynamics of flaviviruses revealed by molecular phylogenies. Proc Natl Acad Sci U S A. 93(2), 548-553.

Zeng, L., Kurane, I., Okamoto, Y., Ennis, F.A., Brinton, M.A., 1996. Identification of amino acids involved in recognition by dengue virus NS3-specific, HLA-DR15-restricted cytotoxic CD4+ T-cell clones. J Virol. 70(5), 3108-3117.

Zhang, C., Mammen Jr., M.P., Chinnawirotpisan, P., Klungthong, C., Rodpradit, P., Monkongdee, P., Nimmannitya, S., Kalayanarooj, S., Holmes, E.C., 2005. Clade replacements in dengue virus serotypes 1 and 3 are associated with changing serotype prevalence. J Virol. 79(24), 15123-15130.

Zhang, C., Mammen Jr., M.P., Chinnawirotpisan, P., Klungthong, C., Rodpradit, P., Nisalak, A., Vaughn, D.W., Nimmannitya, S., Kalayanarooj, S., Holmes, E.C., 2006. Structure and age of genetic diversity of dengue virus type 2 in Thailand. J Gen Virol. 87(Pt 4), 873-883.

Zhang, Y., Bi, P., Hiller, J.E., 2008. Climate change and the transmission of vector-borne diseases: a review. Asia Pac J Public Health. 20(1), 64-76.

Zhang, Y., Corver, J., Chipman, P.R., Zhang, W., Pletnev, S.V., Sedlak, D., Baker, T.S., Strauss, J.H., Kuhn, R.J., Rossmann, M.G., 2003. Structures of immature flavivirus particles. EMBO J. 22(11), 2604-2613.

Zheng, K., Zhou, H.Q., Yan, J., Ke, C.W., Maeda, A., Maeda, J., Takashima, I., Kurane, I., Ma, H., Xie, X.M., 2009. Molecular characterization of the E gene of dengue virus type 1 isolated in Guangdong province, China, in 2006. Epidemiol Infect. 137(1), 73-78.

Zivna, I., Green, S., Vaughn, D.W., Kalayanarooj, S., Stephens, H.A., Chandanayingyong, D., Nisalak, A., Ennis, F.A., Rothman, A.L., 2002. T cell responses to an HLA-B*07-restricted epitope on the dengue NS3 protein correlate with disease severity. J Immunol. 168(11), 5959-5965.

Zybert, I.A., van der Ende-Metselaar, H., Wilschut, J., Smit, J.M., 2008. Functional importance of dengue virus maturation: infectious properties of immature virions. J Gen Virol. 89(Pt 12), 3047-3051.

185

2009. Dengue in Africa: emergence of DENV-3, Cote d’Ivoire, 2008. Wkly Epidemiol Rec. 84(11-12), 85- 88.

2009. List of Established Industrial Parks. Vietnam Investment Development Group.

2010. Migration and urbanization in Vietnam: patterns, trends and differentials. Ministry of Planning and Investment, General Statistics Office.

2010. The 2009 Vietnam population and housing census: completed results. General Statistics Office of Viet Nam.

2010. Cambodia General Population Census 2008. Latin American and Caribbean Demographic Center, National Institute of Statistics Cambodia.

186

APPENDIX A

The emergence of rotavirus G12 and the prevalence of enteric viruses in hospitalized pediatric diarrheal patients in southern Vietnam

The Emergence of Rotavirus G12 and the Prevalence of Enteric Viruses in Hospitalized Pediatric Diarrheal Patients in Southern Vietnam

Phan Vu Tra My , Maia A. Rabaa , Ha Vinh , Edward C. Holmes , Nguyen Van Minh Hoang , Nguyen Thanh Vinh , Le Thi Phuong , Nguyen Thi Tham , Phan Van Be Bay , James I. Campbell , Jeremy Farrar , and Stephen Baker * Hospital for Tropical Diseases, Wellcome Trust Major Overseas Programme, Oxford University Clinical Research Unit, Ho Chi Minh City, Vietnam; Center for Infectious Disease Dynamics, Department of Biology, Pennsylvania State University, University Park, Pennsylvania; Hospital for Tropical Diseases, Ho Chi Minh City, Vietnam; Fogarty International Center, National Institutes of Health, Bethesda, Maryland; Dong Thap Provincial Hospital, Dong Thap, Vietnam; Centre for Tropical Diseases, University of Oxford, Oxford, United Kingdom

Abstract. Diarrhea is a major cause of childhood morbidity and mortality in developing countries, and the majority of infections are of viral etiology. We aimed to compare the etiological prevalence of the major enteric viruses in an urban and a rural setting in southern Vietnam. We simultaneously screened fecal specimens from 362 children in Ho Chi Minh City and Dong Thap province that were hospitalized with acute diarrhea over a 1-month-long period for four viral gastrointestinal pathogens. Rotavirus was the most common pathogen identified, but there was a differential prevalence of rotavirus and norovirus between the urban and rural locations. Furthermore, rotavirus genotyping and phylogenetic analysis again differentiated the genotypes by the sampling location. Our data show a disproportional distribution of enteric viral pathogens in urban and rural locations, and we provide evidence of continual importation of new rotavirus strains into southern Vietnam and report the emergence of rotavirus genotype G12.

INTRODUCTION Transmission and distribution of rotavirus A is complex and influenced by multiple social, demographic, and environ- Diarrhea is the second most common cause of childhood mental factors. 15, 16 Vietnam is a typical industrializing country mortality worldwide, estimated to be responsible for 1.76 mil- where the agents of infectious diseases are changing rapidly. 17 lion deaths annually between 2000 and 2003 and 1.87 million Existing data on viral gastrointestinal pathogens are avail- 1–3 deaths in children under the age of 5 years in 2004. It is able from sentinel surveillance in Ho Chi Minh City (HCMC) young children and infants in unindustrialized and industri- between 1998 and 2007.18– 24 However, because detection and alizing countries that are disproportionately affected, and in identification of enteric pathogens are not performed rou- such locations, diarrheal infections are often severe, frequently tinely in hospitals, little is known about the prevalence of viral 4 requiring hospitalization. The etiological agents of diarrhea diarrheal pathogens and the strains that circulate in different are numerous, including multiple viral, bacterial, and parasitic geographic and demographic locations in Vietnam. We aimed pathogens. However, it is viruses that are responsible for the to investigate the distribution of norovirus, enteric adenovirus, vast burden of morbidity and mortality, causing up to 40% of astrovirus, and rotavirus genotypes causing diarrhea in hospi- 5 all severe diarrhea cases in developing countries. The most talized children in distinct urban and rural locations in south- prevalent viruses causing endemic childhood gastroenteritis ern Vietnam. are rotavirus, norovirus, enteric adenovirus, and astrovirus. 6 Rotavirus is the dominant agent of viral diarrhea and is the suspected etiological agent in 39% and 45% of all hospital MATERIALS AND METHODS admissions related to diarrhea globally and in Asia, respectively. 7, 8 In Vietnam, rotavirus is estimated to be responsible Study sites and population. Verbal informed consent was for between 44% and 67.4% of all childhood diarrheal infec- obtained from the parents or legal guardians of minors enrolled tions requiring hospitalization.9– 11 in this study. This work was approved by the institutional Rotavirus has a genome comprised of 11 segments of double- ethical review boards of the Hospital for Tropical Diseases, stranded RNA (dsRNA) encoding six structural proteins and HCMC and Dong Thap Provincial Hospital, Dong Thap. six non-structural proteins.12 It is genetic heterogeneity in Patient recruitment was performed over one calendar month two of the structural regions that encode the viral capsid from November 1, 2008 to November 30, 2008 at two hospitals, proteins, VP7 and VP4, that permit the differentiation of pediatric ward B at the Hospital for Tropical Diseases (HTD) individual rotavirus strains.12 Sequencing of the VP7 region in HCMC and pediatric infections ward at Dong Thap Provin- (glycoprotein) defines the G type, and VP4 region sequencing cial Hospital (DTPH) in Dong Thap (DT) province. DTPH is (protease-sensitive protein) determines the P type. To date, 154 km away from HCMC and located within the Mekong 19 G and 28 P genotypes have been described, 10 and 11 of Delta region of southern Vietnam; it is a rural location, with a which, respectively, have been isolated from humans. 12, 13 Strain lower population density than HCMC. We enrolled all pediatric G1P[8] is the most frequent rotavirus A GP combination iso- (under the age of 15 years) patients who had been hospitalized lated from symptomatic humans worldwide.14 at HTD or DTPH because of acute watery diarrhea (defined as three or more loose stools within a 24-hour period) without any additional underlying complications, such as febrile convulsions, extensive dehydration, or stool-containing blood * Address correspondence to Stephen Baker, Enteric Infections Group, Hospital for Tropical Diseases, Wellcome Trust Major Overseas or mucus. The age of each patient was recorded, and a stool Programme, Oxford University Clinical Research Unit, 190 Ben Ham specimen from each patient was collected in a sterile container Tu, District 5, Ho Chi Minh City, Vietnam. E-mail: [email protected] on the day of admission and was stored at −20°C. 768 VIRAL ETIOLOGY OF ACUTE DIARRHEA AND G12 ROTAVIRUS 769

RNA extraction and virus detection. Total viral RNA was RESULTS extracted from 10% (in phosphate-buffered saline [PBS]) fecal specimens using the QIAamp viral RNA Mini kit (QIAGEN, Enteric virus prevalence. A total of 362 children from the Hilden, Germany) according to the manufacturer’s recommen- two locations (252 from HCMC and 110 from DT) with acute dations. RNA preparations were converted to complemen tary gastroenteritis were concurrently enrolled in the 1-month- DNA (cDNA) by reverse transcription (RT), and an aliquot of long study. We were able to detect rotavirus A, norovirus, RNA of each sample was stored at −80°C until required. For adenovirus, or astrovirus using EIA in 195 samples (53.9%), RT, extracted RNA was reverse-transcribed by SuperScript and eight children had more than one viral pathogen in Reverse Transcriptase III and RNase Inhibitor (Invitrogen) their stool (Table 1). In the 252 samples originating from combined with a random hexamer (Roche Diagnostics, United HCMC, 75 (29.8%) were positive for group A rotavirus, and Kingdom) according to the manufacturer’s instructions. The 34 (13.5%) were positive for norovirus ( Table 1 ). In the DT resulting cDNA were stored at −80°C. samples, 72 (65.5%) were positive for group A rotavirus, and All stool samples were screened for rotavirus A, norovi- 4 (3.6%) were positive for norovirus. Enteric adenovirus and rus (genogroups I and II), enteric adenovirus, and astrovirus astrovirus collectively comprised only a small proportion using IDEIATM direct antigen detection kits according to the of all the acute diarrheal cases in HCMC (2.4% and 2.8%, manufacturer’s instructions (Oxoid; Thermo Fisher Scientific, respectively); this finding was also the case in DT (2.7% and United Kingdom). Rotavirus outer capsid genes (VP7 and 1.8%, respectively). VP4) detection was performed by RT-polymerase chain reac- The age distribution of all the hospitalized patients with diar- tion (PCR) in stool specimens that were positive for EIA rhea was similar in both locations, with a mean of 15.8 months rotavirus A. Briefly, viral cDNA was subjected to RT-PCR (median = 13 months, range = 2–96 months) in HCMC to amplify the VP7 and VP4 genes using primers and ampli- and a mean of 15.3 months (median = 10 months, range = fication conditions as previously described. 25 Amplification of 1.5–156 months) in DT. The preponderance of patients were VP7 and VP4 regions was performed individually, and PCR less than 24 months of age (90.6% in HCMC and 94.9% reactions were predicted to generate amplicons of 881 and 663 in DT), and the most common 6-month age group was bp, respectively. PCR amplicons were visualized on 2% aga- between 7 and 12 months. This peak age group was comprised rose gels under ultraviolet (UV) light after staining with 3% of 43.6% (51/117) of the positive samples from HCMC and ethidium bromide. 48.7% (38/78) of the positive samples from DT. We found that Sequencing, genotype determination, and phylogenetic the prevalence of viral positive samples in cases from HCMC analysis. PCR amplicons from successful VP7 and VP4 PCR was significantly lower in those children younger than 6 months amplifications were DNA sequenced using the amplification and older than 24 months of age compared with those chil- < χ2 primers. PCR amplicons were purified using the QIAquick dren in the intermediate age group ( P 0.001, test). In DT, PCR purification kit (QIAGEN, Germany), and DNA con- the prevalence of viral diarrhea was lower in patients older centrations were determined using a NanoDrop ND-1000 spec- than 18 months of age compared with children aged between < χ2 trophotometer (Thermo Fisher Scientific, United Kingdom). 0 and 18 months (P 0.001, test). Furthermore, the pro- Direct DNA sequencing was performed using the BigDye Ter- portion of rotavirus infections in patients under the age of minator Cycle Sequencing kit (Applied Biosystems) according 6 months from DT was significantly higher than the cor- χ2 to the manufacturer’s recommendations. All DNA sequences responding age group from HCMC ( P = 0.0289, test). were generated using an ABI Prism 3130xl Genetic Analyzer Conversely, in the group consisting of children aged from (Applied Biosystems), and the resulting DNA sequences were 19 to 24 months, the proportion of rotavirus infections was sig- χ2 assembled using DNA Baser Sequence Assembler v3.0.17 nificantly greater in HCMC than DT (P = 0.009457, test). (Heracle Biosoft, Pitesti, Romania). Distribution of rotavirus genotypes. All stool samples were The resulting VP4 and VP7 sequences (Genbank acces- subjected to RT-PCR for the VP4 and VP7 regions of rotavirus sion numbers: VP4, FR820957–FR821065; VP7, FR822209– A. None of the samples that were negative for rotavirus using FR822321) were compared with other corresponding genotype EIA were positive by PCR, and 133 of 157 samples of the sequences using BLASTn (https://blast.ncbi.nlm.nih.gov/ positive EIA samples gave a PCR amplicon for one or both of Blast.cgi) for genotype determination. Coding sequences the loci (PCR compared with EIA: sensitivity, 90%; specificity, were manually aligned using Se-AL v2.0a11 (http://tree.bio 100%). All positive VP4 and VP7 PCR amplicons were .ed.ac.uk/software/ ), and additional sequences were trimmed to correspond with our sequences to maximize sequence Table 1 homology. Maximum likelihood trees for each of the geno- Enzyme immunoassay detection of four viral pathogens in stool sam- 26 types were inferred using RAxML v 7.0.4 employing the gen- ples from children with acute diarrhea in Ho Chi Minh City and eral time-reversible model of nucleotide substitution with a Dong Thap γ-distribution of among-site rate variation (GTR+Γ), which Number of positive samples Number of positive samples was determined using jModelTest.27 One thousand boot- Viral assay in Ho Chi Minh City (%) * in Dong Thap (%) † strap replicates were used as implemented in a rapid boot- Group A rotavirus ‡ 75 (29.8) 72 (65.5) strap algorithm available in RAxML. The resulting trees Norovirus 34 (13.5) 4 (3.6) Adenovirus 6 (2.4) 3 (2.7) were visualized in FigTree v1.3.1 ( http://tree.bio.ed.ac.uk/soft Astrovirus 7 (2.8) 2 (1.8) ware/figtree/ ), and genetic distances were estimated using Multiple positive 5 (4.3) 3 (2.7) HyPhy v2.0 (http://www .datam0nk3y.org/hyphy/doku.php ).28 Total viruses detected 122 81 All statistical analyses were performed in R version 2.9.0 (The Total positive samples 117 (46.4) 78 (73.6) R foundation for Statistical Computing); P values < 0.05 were * N = 252. † N = 110. considered statistically significant. ‡ Equivocal results were found in two samples from Ho Chi Minh City. 770 TRA MY AND OTHERS

Figure 1. The distribution of rotavirus G and P types detected in the stools of children with diarrhea in Ho Chi Minh City and Dong Thap. The graph shows the distribution of rotavirus G types 1, 2, 3, 4, and 12 and P types P[4], P[6], and P[8] in 118 and 109 VP7 and VP4 PCR amplification- positive samples, respectively. The graph is subdivided into G (N = 60) and P types (N = 56) from DT (upper, grey) and G (N = 58) and P types ( N = 53) from HCMC (lower, black).

DNA-sequenced; 118 of 133 amplicons produced a sequence (uncorrected) = 0.0774] and could be differentiated into three consistent with the VP4 region, and 109 of 133 amplicons distinct lineages (Figure 2). We could identify a significant produced a sequence consistent with the VP7 region. We phylogenetic association between the DT and HCMC G1 found that genotype G1 was the dominant circulating G type, sequences, signifying the circulation of closely related rotavirus comprising 82 of 118 (69.5%) of all typeable G types ( Figure 1 ). strains in these two locations. Comparing these G1 data with The novel emerging genotype, G12, was the second most previous rotavirus sequence data originating in Vietnam, common G type, representing 16.1% (19/118) of all G types. one lineage showed a close phylogenetic relationship to Other less common G genotypes, including G2, G3, and G4, rotaviruses also isolated in HCMC, indicating local persistence were also detected in small proportions in the remainder of of this particular lineage ( Figure 2 ).29 The distribution of the the positive samples (Figure 1). Genotype P[8] was the most P[8]-type sequences was comparable with the distribution prevalent typeable P genotype, accounting for 88.1% (96/109) observed throughout the G1 sequences, exhibiting extensive of all P amplicons, followed by P[4] (10.1%, 11/109) and P[6] genetic diversity (maximum genetic distance [uncorrected] = (1.8%, 2/109). 0.159) ( Figure 3 ). The P[8] sequences could be differentiated The most common GP combination was G1P[8], compris- into four phylogenetically distinct clusters that seem to be ing 78.5% (73/93) of all samples that were positive for both closely related to sequences from other regions around Asia amplification targets. Other globally diffuse GP genotypes, ( Figure 3 ). Sequences from the G12 lineage, which were including G2P[4] and G3P[8], were also detected but again, in primarily detected in samples from DT, were closely related a limited number of samples (10.8%, 10/93; 2.2%, 2/93, respec- to Thai and Indian sequences ( Figure 4 ). However, less overall tively) (Figure 1). The overall distribution of G and P types diversity was detected among the G12 genotype sequences differed substantially between the two locations. Genotypes (maximum genetic distance [uncorrected] = 0.0130), with only G2 and G3 were present in several rotavirus positive samples one major lineage identified ( Figure 4 ), signifying potential in HCMC, but neither was identified in positive amplifica- recent introduction of this variant. tions from DT. Genotype G12 was identified in both locations but was more prevalent in DT than in HCMC, representing DISCUSSION 28.3% and 3.4% of all G types in these locations, respectively ( P = 0.0016, χ 2 test). Furthermore, P[4] was identified in 20.8% Enteric viruses are a predominant cause of acute childhood of the P-type samples in HCMC but was not detected in DT, gastroenteritis in Vietnam. 9, 11, 20, 21, 30– 33 Our study was designed and P[6] was detected only in samples originating from DT. to examine the prevalence and distribution of four enteric Phylogenetic analysis of rotavirus sequences. We performed viruses causing hospitalization in children attending two phylogenetic analyses on the G1, G12, and P[8] sequences pro- defined healthcare centers in one rural and one urban location duced from the VP7 and VP4 amplifications, comparing them in southern Vietnam. We found a substantial proportion of with additional global sequences available in public data- diarrhea to be caused by viral pathogens and an overall domi- bases. The G1 rotavirus sequences from DT and HCMC exhib- nance of group A rotavirus. Our data suggest that, although this ited extensive genetic diversity [maximum genetic distance study represents a 1-month long snapshot of acute childhood VIRAL ETIOLOGY OF ACUTE DIARRHEA AND G12 ROTAVIRUS 771

Figure 2. Phylogenetic tree of 81 rotavirus G1 sequences from Ho Chi Minh City and Dong Thap combined with representative global rotavirus G1 sequences. Maximum likelihood phylogenetic tree was constructed from G1 sequences from the amplification and sequencing of the VP7 gene. Sequences generated from this study are indicated in black. HCMC indicates samples from Ho Chi Minh City, and DT indicates samples from Dong Thap. Vietnamese isolates from previous studies are highlighted in grey. The tree is midpoint-rooted, with all horizontal branch lengths drawn to the scale of a nucleotide substitution per site. Bootstrap values > 85% are indicated by asterisks, and triangles represent compressed regions of the tree. 772 TRA MY AND OTHERS

Figure 3. Phylogenetic tree of 96 rotavirus P[8] sequences from Ho Chi Minh City and Dong Thap combined with representative global rotavirus P[8] sequences. Maximum likelihood phylogenetic tree [VP4 gene] constructed from P[8] sequences and representative global sequences of rotavirus P[8] type. Tree rooting, bootstrap values, branch lengths, and font correspond to those factors presented in Figure 2 . VIRAL ETIOLOGY OF ACUTE DIARRHEA AND G12 ROTAVIRUS 773

Figure 4. Phylogenetic tree of 19 rotavirus G12 sequences from Ho Chi Minh City and Dong Thap combined with representative global rotavirus G12 sequences. Maximum likelihood phylogenetic tree [VP7 gene] constructed from G12 sequences and representative global sequences of rotavirus G12. Tree rooting, bootstrap values, branch lengths, and font correspond to those factors presented in Figure 2 . diarrhea, viral pathogens are predominant etiological agents sistent with multiple introductions of the G1 genotype to the of acute childhood gastroenteritis in this location, a view that region. Furthermore, only one of three G1 lineages showed a concurs with previous studies.9, 11, 20, 30 We additionally show a relationship to rotavirus sequences from HCMC in 2002–2005, variable distribution of enteric viruses in children hospitalized again supporting on-going strain introduction with limited in with acute watery diarrhea in distinct urban and rural loca- situ evolution.29 Of all typeable P strains, P[8] predominated, tions in southern Vietnam, suggesting the circulation of differ- with a comparable prevalence with North America, Australia, ent pathogens and corresponding differential infection risks. and Europe but higher prevalence than the prevalence previ- Our findings show that group A rotaviruses were the pre- ously reported in Vietnam. 9– 11, 14, 20 Again, phylogenetic analysis dominant viral cause of diarrheal disease in children in both indicates that the extensive heterogeneity observed within the sampling locations. Within the group A rotavirus-positive P[8] sequences is likely caused by multiple strain introduction samples, we identified extensive genetic diversity despite the rather than clonal expansion. limited temporal distribution and relatively small number of We identified a differential distribution of viral diarrheal samples. Such dramatic diversity within the G1 group is con- pathogens between the urban and rural locations. We note a 774 TRA MY AND OTHERS significantly higher prevalence of rotavirus genotype G12 in major enteric viral pathogens and rotavirus genotypes caus- the rural location compared with the urban location. This study ing childhood acute gastroenteritis in two distinct locations is the first to report a rotavirus genotype G12 in Vietnam, and in southern Vietnam. We highlight the need for longitudinal since the primary detection in the Philippines in 1987, it has research of enteric viruses in this location and continued mon- become increasingly prevalent worldwide. 34 Our high detec- itoring of circulating rotavirus strains for effective prevention tion rate (28.3% of all G types in DT) of this variant highlights and vaccination strategies. the capacity of this genotype to spread and become fixed in a local population, and it has direct implications for rotavi- Received June 10, 2011. Accepted for publication June 24, 2011. rus vaccination, because protective immunity of two available Acknowledgments: We are thankful to Dr. Carl Kirkwood (Royal rotavirus vaccines (RotaRix and RotaTeq) against G12 geno- Children’s Hospital, Victoria, Australia) and Winifred Dove (University type is currently undetermined. However, a high evolutionary of Liverpool, Liverpool, United Kingdom) for their technical rate of the VP7 gene (1.66 × 10−3 substitutions/site per year) support. suggests that the introduction of either vaccine may impose Financial support: This work is supported by The Wellcome Trust, a selective pressure on circulating strains and accelerate evo- Euston Road, London, United Kingdom. S.B. is supported by an Oak Foundation Fellowship through Oxford University. M.A.R. is lutionary rates, facilitating the emergence and rapid spread of supported by a National Science Foundation Graduate Research 35 variants. Such factors emphasize the need for on-going modi- Fellowship. fication and development of rotavirus vaccines and continued Authors’ addresses: Phan Vu Tra My, Nguyen Van Minh Hoang, Nguyen surveillance for genotype circulation. Thanh Vinh, James I. Campbell, Jeremy Farrar, and Stephen Baker, Secondary data supporting a differential distribution of Enteric Infections Group, Hospital for Tropical Diseases, Wellcome enteric viruses between urban and rural settings is the distribu- Trust Major Overseas Programme, Oxford University Clinical tion of norovirus in the two locations. Norovirus was the second Research Unit, Ho Chi Minh City, Vietnam, E-mails: mypvt@oucru .org , [email protected] , [email protected] , [email protected] , most commonly identified virus in the patients’ stool samples, [email protected] , and [email protected] . Maia A. Rabaa, Department 36 which is similar to global data. From the positive stool sam- of Biology, Center for Infectious Disease Dynamics, Pennsylvania ples in HCMC, norovirus constituted 29% of positive sample State University, University Park, PA, E-mail: maia.rabaa@gmail compared with only 5.1% in DT. We suggest different epide- .com. Ha Vinh, Pediatric Ward B, Hospital for Tropical Diseases, Ho miological risk factors related to this organism in these loca- Chi Minh City, Vietnam, E-mail: [email protected] . Edward C. Holmes, Fogarty International Center, National Institute of Health, Bethesda, tions, which necessitates additional investigation. Norovirus is MD, E-mail: [email protected] . Le Thi Phuong, Nguyen Thi Tham, and highly contagious and is related to isolated outbreaks in devel- Phan Van Be Bay, Infectious Ward, Dong Thap Provincial Hospital, oped countries. 37 HCMC is more densely populated and has Dong Thap, Vietnam. undergone a greater level of development with respect to the surrounding province, such as DT. Transmission and the cor- REFERENCES responding exposure to this particular pathogen are likely to 1. Bryce J , Boschi-Pinto C , Shibuya K , Black RE , 2005 . WHO esti- follow such a developmental change. In parallel to rotavirus mates of the causes of death in children . Lancet 365: infections, the majority of patients in which we detected noro- 1147 – 1152 . virus in the stool samples was in the 7–12 months age group. 2. Hogberg U , 2005 . The World Health Report 2005: “make every This age group is the key age group for children with diarrheal mother and child count”—including Africans . Scand J Public infections and poses the greatest number of epidemiological Health 33: 409 – 411 . 3. Boschi-Pinto C , Velebit L , Shibuya K , 2008 . Estimating child mor- questions. Decreased rates of all viral pathogens detected in tality due to diarrhoea in developing countries. Bull World the stool samples outside this key age group may be related to Health Organ 86: 710 – 717 . risk, exposure, healthcare-seeking behavior, maternal immu- 4. Parashar UD , Gibson CJ , Bresse JS , Glass RI , 2006 . Rotavirus and nity, and natural postinfection immunity. Additional studies on severe childhood diarrhea . Emerg Infect Dis 12: 304 – 306 . 5. Davidson G , Barnes G , Bass D , Cohen M , Fasano A , Fontaine O , the effect maternal antibody or immunity from previous expo- Guandalini S , 2002 . Infectious diarrhea in children: Working sure to norovirus should help guide vaccine development and Group Report of the First World Congress of Pediatric usage within this age group. Gastroenterology, Hepatology, and Nutrition . J Pediatr Gastro- Our study design does have several caveats; first, although enterol Nutr 35 (Suppl 2): S143 – S150 . the age distribution of enrolled patients was similar in both 6. Procop GW , 2001 . Gastrointestinal infections . Infect Dis Clin North Am 15: 1073 – 1108 . locations, such sampling was open to selection bias. However, 7. Nelson EA , Bresee JS , Parashar UD , Widdowson MA , Glass RI , we aimed to understand the distribution viral diarrheal patho- 2008 . Rotavirus epidemiology: the Asian Rotavirus Surveillance gen and not determine risk factors; therefore, a limited period Network . Vaccine 26: 3192 – 3196 . of sampling may bias the distribution of the various agents. 8. Widdowson MA , Steele D , Vojdani J , Wecker J , Parashar U , 2009 . Global rotavirus surveillance: determining the need and Second, it is known that 5% of control healthy patients (defined measuring the impact of rotavirus vaccines. J Infect Dis 200 as inpatients for non-infectious causes without diarrhea in the (Suppl 1): S1 – S8 . previous 2 weeks) between the ages of 3 months and 5 years 9. Van Man N , Luan le T , Trach DD , Thanh NT , Van Tu P , Long NT , can carry viral pathogens in their stool specimens, and addi- Anh DD , Fischer TK , Ivanoff B , Gentsch JR , Glass RI , 2005 . tional sampling is necessary to understand the transmission Epidemiological profile and burden of rotavirus diarrhea in 32 Vietnam: 5 years of sentinel hospital surveillance, 1998–2003 . dynamics and asymptomatic carriage of enteric viruses. We J Infect Dis 192 (Suppl 1): S127 – S132 . did find some dual viral infections (4.1%), but we were unable 10. Nguyen VM , Nguyen VT , Huynh PL , Dang DT , Nguyen TH , Phan to elucidate the impact in pathogenesis and their clinical sig- VT , Nguyen TL , Le TL , Ivanoff B , Gentsch JR , Glass RI , 2001 . nificance, which also may be an issue when other bacterial and The epidemiology and disease burden of rotavirus in Vietnam: sentinel surveillance at 6 hospitals . J Infect Dis 183: 1707 – 1712 . parasitic agents are considered. 11. Nguyen TA , Yagyu F , Okame M , Phan TG , Trinh QD , Yan H , In conclusion, this report is the first report of rotavirus G12 Hoang KT , Cao AT , Le Hoang P , Okitsu S , Ushijima H , 2007 . in Vietnam, and we show a differential distribution of the Diversity of viruses associated with acute gastroenteritis in VIRAL ETIOLOGY OF ACUTE DIARRHEA AND G12 ROTAVIRUS 775

children hospitalized with diarrhea in Ho Chi Minh City, 24. Nishio O , Matsui K , Lan DT , Ushijima H , Isomura S , 2000 . Vietnam . J Med Virol 79: 582 – 590 . Rotavirus infection among infants with diarrhea in Vietnam. 12. Greenberg HB , Estes MK , 2009 . Rotaviruses: from pathogenesis Pediatr Int 42: 422 – 424 . to vaccination . Gastroenterology 136: 1939 – 1951 . 25. Gomara MI , Cubitt D , Desselberger U , Gray J , 2001 . Amino acid 13. Franco MA , Greenberg HB , 2009 . Rotaviruses . Richman DD , substitution within the VP7 protein of G2 rotavirus strains Whitley RJ , Hayden FG , eds. Clinical Virology . Washington, associated with failure to serotype . J Clin Microbiol 39: DC : ASM Press , 797 – 816 . 3796 – 3798 . 14. Santos N , Hoshino Y , 2005 . Global distribution of rotavirus sero- 26. Stamatakis A , 2006 . RAxML-VI-HPC: maximum likelihood-based types/genotypes and its implication for the development and phylogenetic analyses with thousands of taxa and mixed models. implementation of an effective rotavirus vaccine . Rev Med Virol Bioinformatics 22: 2688 – 2690 . 15: 29 – 56 . 27. Posada D , 2009 . Selection of models of DNA evolution with 15. Pitzer VE , Viboud C , Simonsen L , Steiner C , Panozzo CA , Alonso jModelTest . Methods Mol Biol 537: 93 – 112 . WJ , Miller MA , Glass RI , Glasser JW , Parashar UD , Grenfell BT , 28. Pond SL , Frost SD , Muse SV , 2005 . HyPhy: hypothesis testing using 2009 . Demographic variability, vaccination, and the spatiotem- phylogenies . Bioinformatics 21: 676 – 679 . poral dynamics of rotavirus epidemics. Science 325: 290 – 294 . 29. Trinh QD , Nguyen TA , Phan TG , Khamrin P , Yan H , Hoang PL , 16. Atchison C , Iturriza-Gomara M , Tam C , Lopman B , 2010 . Maneekarn N , Li Y , Yagyu F , Okitsu S , Ushijima H , 2007 . Spatiotemporal dynamics of rotavirus disease in Europe: can cli- Sequence analysis of the VP7 gene of human rotavirus G1 iso- mate or demographic variability explain the patterns observed. lated in Japan, China, Thailand, and Vietnam in the context of Pediatr Infect Dis J 29: 566 – 568 . changing distribution of rotavirus G-types. J Med Virol 79: 17. Vinh H , Nhu NT , Nga TV , Duy PT , Campbell JI , Hoang NV , Boni 1009 – 1016 . MF , My PV , Parry C , Nga TT , Van Minh P , Thuy CT , Diep TS , 30. Nguyen TA , Hoang L , Pham le D , Hoang KT , Mizuguchi M , Phuong le T , Chinh MT , Loan HT , Tham NT , Lanh MN , Mong Okitsu S , Ushijima H , 2008 . Identification of human astrovi- BL , Anh VT , Bay PV , Chau NV , Farrar J , Baker S , 2009 . A chang- rus infections among children with acute gastroenteritis in the ing picture of shigellosis in southern Vietnam: shifting species Southern Part of Vietnam during 2005–2006 . J Med Virol 80: dominance, antimicrobial susceptibility and clinical presenta- 298 – 305 . tion . BMC Infect Dis 9: 204 . 31. Li L , Phan TG , Nguyen TA , Kim KS , Seo JK , Shimizu H , Suzuki E , 18. Landaeta ME , Dove W , Vinh H , Cunliffe NA , Campbell J , Parry Okitsu S , Ushijima H , 2005 . Molecular epidemiology of adeno- CM , Farrar JJ , Hart CA , 2003 . Characterization of rotaviruses virus infection among pediatric population with diarrhea in causing diarrhoea in Vietnamese children . Ann Trop Med Asia . Microbiol Immunol 49: 121 – 128 . Parasitol 97: 53 – 59 . 32. Bodhidatta L , Lan NT , Hien BT , Lai NV , Srijan A , Serichantalergs 19. Nguyen TV , Le Van P , Le Huy C , Weintraub A , 2004 . Diarrhea O , Fukuda CD , Cam PD , Mason CJ , 2007 . Rotavirus disease in caused by rotavirus in children less than 5 years of age in Hanoi, young children from Hanoi, Vietnam. Pediatr Infect Dis J 26: Vietnam . J Clin Microbiol 42: 5745 – 5750 . 325 – 328 . 20. Doan LT , Okitsu S , Nishio O , Pham DT , Nguyen DH , Ushijima H , 33. Nguyen TA , Hoang LP , Pham LD , Hoang KT , Okitsu S , Mizuguchi 2003 . Epidemiological features of rotavirus infection among M , Ushijima H , 2008 . Use of sequence analysis of the VP4 gene hospitalized children with gastroenteristis in Ho Chi Minh City, to classify recent Vietnamese rotavirus isolates . Clin Microbiol Vietnam . J Med Virol 69: 588 – 594 . Infect 14: 235 – 241 . 21. Anh DD , Thiem VD , Fischer TK , Canh DG , Minh TT , Tho le H , 34. Taniguchi K , Urasawa T , Kobayashi N , Gorziglia M , Urasawa S , Van Man N , Luan le T , Kilgore P , von Seidlein L , Glass RI , 2006 . 1990 . Nucleotide sequence of VP4 and VP7 genes of human rota- The burden of rotavirus diarrhea in Khanh Hoa Province, viruses with subgroup I specificity and long RNA pattern: impli- Vietnam: baseline assessment for a rotavirus vaccine trial. cation for new G serotype specificity. J Virol 64: 5640 – 5644 . Pediatr Infect Dis J 25: 37 – 40 . 35. Matthijnssens J , Heylen E , Zeller M , Rahman M , Lemey P , Van 22. Vu Nguyen T , Le Van P , Le Huy C , Nguyen Gia K , Weintraub A , Ranst M , 2010 . Phylodynamic analyses of rotavirus genotypes 2006 . Etiology and epidemiology of diarrhea in children in G9 and G12 underscore their potential for swift global spread. Hanoi, Vietnam . Int J Infect Dis 10: 298 – 308 . Mol Biol Evol 27: 2431 – 2436 . 23. Ngo TC , Nguyen BM , Dang DA , Nguyen HT , Nguyen TT , Tran 36. Patel MM , Widdowson MA , Glass RI , Akazawa K , Vinje J , Parashar VN , Vu TT , Ogino M , Alam MM , Nakagomi T , Nakagomi O , UD , 2008 . Systematic literature review of role of noroviruses Yamashiro T , 2009 . Molecular epidemiology of rotavirus diar- in sporadic gastroenteritis. Emerg Infect Dis 14: 1224 – 1231 . rhoea among children in Haiphong, Vietnam: the emergence of 37. Glass RI, Parashar UD , Estes MK , 2009 . Norovirus gastroenteritis . G3 rotavirus . Vaccine 27 (Suppl 5): F75 – F80 . N Engl J Med 361: 1776 – 1785 .

APPENDIX B

The evolutionary consequences of blood-stage vaccination on the rodent malaria Plasmodium chabaudi

196 The Evolutionary Consequences of Blood-Stage Vaccination on the Rodent Malaria Plasmodium chabaudi

Victoria C. Barclay1*, Derek Sim1, Brian H. K. Chan1, Lucas A. Nell1, Maia A. Rabaa1, Andrew S. Bell1, Robin F. Anders2, Andrew F. Read1,3,4 1 Center for Infectious Disease Dynamics, Department of Biology, The Pennsylvania State University, University Park, Pennsylvania, United States of America, 2 Department of Biochemistry, La Trobe University, Melbourne, Australia, 3 Department of Entomology, The Pennsylvania State University, University Park, Pennsylvania, United States of America, 4 Fogarty International Center, National Institutes of Health, Bethesda, Maryland, United States of America

Abstract Malaria vaccine developers are concerned that antigenic escape will erode vaccine efficacy. Evolutionary theorists have raised the possibility that some types of vaccine could also create conditions favoring the evolution of more virulent pathogens. Such evolution would put unvaccinated people at greater risk of severe disease. Here we test the impact of vaccination with a single highly purified antigen on the malaria parasite Plasmodium chabaudi evolving in laboratory mice. The antigen we used, AMA-1, is a component of several candidate malaria vaccines currently in various stages of trials in humans. We first found that a more virulent clone was less readily controlled by AMA-1-induced immunity than its less virulent progenitor. Replicated parasites were then serially passaged through control or AMA-1 vaccinated mice and evaluated after 10 and 21 rounds of selection. We found no evidence of evolution at the ama-1 locus. Instead, virulence evolved; AMA-1-selected parasites induced greater anemia in naı¨ve mice than both control and ancestral parasites. Our data suggest that recombinant blood stage malaria vaccines can drive the evolution of more virulent malaria parasites.

Citation: Barclay VC, Sim D, Chan BHK, Nell LA, Rabaa MA, et al. (2012) The Evolutionary Consequences of Blood-Stage Vaccination on the Rodent Malaria Plasmodium chabaudi. PLoS Biol 10(7): e1001368. doi:10.1371/journal.pbio.1001368 Academic Editor: David S. Schneider, Stanford University, United States of America Received November 25, 2011; Accepted June 19, 2012; Published July 31, 2012 Copyright: ß 2012 Barclay 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: Wellcome Trust UK (www.wellcome.ac.uk) (VCB, AFR) Penn State Start Up Fund (AFR). 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. Abbreviations: AMA-1, apical membrane antigen-1; RBCs, red blood cells *E-mail:[email protected]

Introduction towards inducing variant-independent immunity against these targets [7,21–27]. Evolution is a significant challenge to malaria control. Malaria Epitope evolution is not the only type of evolution that can parasites have repeatedly evolved resistance to frontline drugs occur in response to vaccination. Immunization can also promote [1,2], and mosquitoes have evolved resistance to all classes of the emergence of variants at loci other than those targeted by approved insecticides [3,4]. Here we report experimental studies vaccine-induced immunity [28]. Of particular interest are investigating how malaria parasites might evolve in response to the virulence determinants because, in theory, immunization can ‘‘natural’’ selection imposed by a blood stage malaria vaccine. under some circumstances promote the emergence and spread of There is currently no licensed malaria vaccine, but a number of strains causing more severe disease (morbidity and mortality) [28– candidates are in human trials [5–9], and a vaccine targeting the 37]. The idea that vaccines could prompt the evolution of more pre-erythrocytic stages of Plasmodium falciparum has provided partial virulent pathogens is controversial, but it has been described as protection to young children in a large phase 3 trial in Africa [10]. one of the key unexpected insights to arise from the nascent field of There are two ways parasites could evolve in vaccinated evolutionary medicine [38]. Several veterinary vaccines have populations. Vaccine developers have traditionally been con- failed in the face of more virulent strains, apparently in the cerned with epitope evolution (antigenic escape) [5,8,9,11,12]. absence of epitope evolution [39–43]. This is where pre-existing or de novo variants of target antigens Vaccination could favor virulent malaria parasites in two ways. emerge and spread because they enable parasites to evade vaccine- First, if the primary force preventing the evolution of more virulent induced immunity. Epitope evolution in response to vaccination strains is that they kill their hosts and therefore truncate their occurs in a range of infectious agents, including hepatitis B virus infectious periods, keeping hosts alive with vaccination will allow [13,14], Bordetella pertussis [15–18], and Streptococcus pneumoniae more virulent strains to circulate [28–37,44]. Second, immunity [19,20]. Epitope evolution has been of particular concern for those might be less effective against virulent strains [36]. For instance, a developing blood stage malaria vaccines because target antigens given antibody titer or a proliferating immune response might are often highly polymorphic, presumably because of natural better control slower replicating strains than more aggressive immune selection. Considerable ingenuity is currently going strains [45]. Virulence factors that reduce the efficacy of primed

PLoS Biology | www.plosbiology.org 1 July 2012 | Volume 10 | Issue 7 | e1001368 Blood-Stage Malaria Vaccine Evolutionary Effects

Author Summary epitope evolution in response to vaccination, but virulence increased. Vaccination can drive the evolution of pathogens. Most obviously, molecules targeted by vaccine-induced immu- Results nity can change. Such evolution makes vaccines less effective. A different possibility is that more virulent Our experimental evolution studies consisted of two serial pathogens are favored in vaccinated hosts. In that case, passage experiments, denoted A and B, and four separate vaccination would create pathogens that cause more harm ‘‘evaluation’’ experiments to determine the virulence of the to unvaccinated individuals. To test this idea, we studied a passaged lines, denoted experiments 1 to 4 (Table S1). rodent malaria parasite in laboratory mice immunized with a component of malaria vaccines currently in human trials. Serial Passage Generates Virulent Parasites That Are Less We found that a more virulent parasite clone was less well controlled by vaccine-induced immunity than was its less Well Controlled by AMA-1 Vaccination virulent ancestor. We then passaged parasites through Before beginning experimental evolution in vaccinated animals, sham- or vaccinated mice to study how the parasites we wanted to test whether AMA-1 vaccine-induced immunity might evolve after multiple rounds of infection of mouse would be less effective against virulent parasites. In order to generate hosts. The parasite molecule targeted by the vaccine did virulent parasites, we serially passaged a single clonal lineage of P. c. not change during this process. Instead, the parasites adami (clone DK) through 30 successive naı¨ve mice (‘‘serial passage became more virulent if they evolved in vaccinated hosts. A’’). We then tested the performance and virulence of these virulent Our data suggest that some vaccines can drive the parasites and their less virulent ancestral precursors in sham- and evolution of more virulent parasites. AMA-1-vaccinated mice (‘‘evaluation experiment 1’’). As expected, serial passage produced parasites that were more virulent in naı¨ve mice than were the ancestral parasites immune responses might also have a selective advantage in (Figure 1A–B; anemia F1,6 = 6.5, p = 0.04). Vaccination with vaccinated hosts [46]. recombinant AMA-1 reduced anemia (Figure 1A–B). It also Epitope evolution and virulence evolution are not necessarily suppressed parasite densities (Figure 1C–D). Importantly, vac- mutually exclusive (some antigens can be virulence determinants), cine-induced immunity was disproportionately effective at con- but they will have different consequences for public and animal taining the avirulent (ancestral) parasites, even though they health. Epitope evolution will erode vaccine efficacy but need not shared complete sequence identity at ama-1 with the more lead to more severe disease in unvaccinated individuals. Virulence virulent (derived) parasites (Figure 1C–D; total parasite density6 evolution on the other hand would both erode vaccine efficacy and vaccination: F1,12 = 5.4, p = 0.03). This suggests that AMA-1 cause more severe disease outcomes in unvaccinated individuals vaccination has the potential to selectively favor more virulent P. [28,35,36]. Note that virulence evolution will not occur for chabaudi parasites. Serial passage did not affect the nucleotide vaccines that induce sterilizing immunity: evolution can proceed sequence of ama-1 (Figure S1). only where vaccines are leaky so that wild-type pathogens can transmit from vaccinated hosts. Because natural immunity against Serial Passage through Vaccinated Mice Caused malaria is neither life-long nor sterilizing [47,48], it seems likely that malaria vaccines will be leaky. Enhanced Virulence, Not Target Site Evolution To investigate the consequences of blood stage malaria To test the evolutionary impact of vaccination with AMA-1, we vaccination for epitope and virulence evolution, we performed contemporaneously passaged P. c. adami DK parasites every week serial passage experiments with the rodent malaria Plasmodium for 20 wk through either sham-vaccinated mice or through mice chabaudi in laboratory mice immunized with a candidate blood vaccinated with recombinant AMA-1 (‘‘serial passage B’’). We stage vaccine. In this system, virulence, which we measure as refer to the parasite lines evolved under these contrasting weight loss and particularly anemia, is positively related to conditions as C-lines and V-lines, respectively. We set out to transmission and competitive ability [35,36]. Anaemia is due to evolve five independent replicate lines of each type, but direct red cell destruction by parasites and bystander killing by particularly in vaccinated groups, lineage loss occurred when host responses [35,36,49]. As with many pathogens [50,51], serial parasites failed to reach high enough densities to allow onward passage of P. chabaudi creates more virulent parasites [52]. Serial syringe passage. Failure to achieve transmissible densities in passage through mice immunized with live parasites augments this vaccinated hosts is likely to be an important evolutionary force. effect [30], consistent with the idea that parasites evolving in When lines were lost, sub-lines were derived from surviving lines. vaccinated populations could become more virulent. However, The full evolutionary history of the lines is shown in Figure S2. most probably, actual blood stage vaccines will consist of Throughout the 20 passages, parasite densities on the day of recombinant antigens [53–69]. Here we specifically test the passage were lower in AMA-1 vaccinated mice (Figure S3). evolutionary impact of vaccination with Apical Membrane However, the densities of those V-lines increased steadily over the Antigen-1 (AMA-1), a component of at least 10 vaccines in successive passages, presumably because of parasite adaptation to human trials [6,66–68]. Antibodies elicited by this antigen are vaccine-induced immunity. believed to confer protection by inhibiting the invasion of To test whether parasite virulence had evolved during the merozoites into red blood cells (RBCs) [55,65,69]. In nature, the passages, we evaluated the virulence of the parasite lines in naı¨ve ama-1 gene is highly polymorphic, and this antigenic diversity is mice at two time points during the evolution of the lines: once after 10 thought likely to compromise vaccine efficacy in the long term rounds of serial passage (‘‘evaluation experiment 2’’) and again after [7,70–72]. By immunizing with a highly defined single recombi- 21 rounds (‘‘evaluation experiment 3’’). In that latter experiment, we nant blood stage antigen, we could specifically determine whether also assayed the virulence of the ancestral parasites (passage 0). We antibodies raised against AMA-1 select for parasites with altered used naı¨ve mice in these experiments because the hypothesis under ama-1 sequence (epitope evolution) and/or for parasites that cause test is that evolution through AMA-1 vaccinated mice will produce more severe disease (virulence evolution). We found no evidence of parasites that do more harm to unvaccinated hosts.

PLoS Biology | www.plosbiology.org 2 July 2012 | Volume 10 | Issue 7 | e1001368 Blood-Stage Malaria Vaccine Evolutionary Effects

Figure 1. Virulence and densities of P. c. adami parasites that had undergone 30 passages in naı¨ve mice (derived) with their progenitors (ancestral) (‘‘evaluation experiment 1’’). Curves (panels A and C) represent the kinetics in up to four mice (mean 6 1 s.e.m.) that were sham-vaccinated (no symbols) or AMA-1 vaccinated (filled circles) and infected with derived (red) or ancestral parasites (blue). Interaction plots (panels B and D) show minimum parasite densities and red cell densities in sham- or AMA-1-vaccinated mice infected with ancestral parasites (blue lines) or derived parasites (red lines). Derived parasites induced more anaemia and achieved higher parasite densities than ancestral parasites during infection of naı¨ve mice (A–D; anemia F1,6 = 6.5, p = 0.04, parasites F1,6 = 22.3, p = 0.003) and AMA-1 vaccination was disproportionately less effective at containing the derived parasites (C–D; total parasite density6vaccination: F1,12 = 5.4, p = 0.03). doi:10.1371/journal.pbio.1001368.g001

Parasites passaged through AMA-1 vaccinated mice (V-lines) 20 passages, no changes in ama-1 nucleotide sequence were became more virulent than parasites passaged through sham- detected in any of the lines (Figure S1). Thus, over the course of vaccinated mice (C-lines) (Figures 2 and 3). This difference had the experiment, parasites evolved in AMA-1 immunized mice already arisen by the 10th passage and was still apparent after 21 became more virulent to naı¨ve animals, and there was no evidence passages. Thus, in naı¨ve mice, V-line parasites from both the 10th of nucleotide evolution at the ama-1 target sequence. and 21st passage ‘‘generations’’ caused more anemia than their The virulence differences apparent at the 10th round of selection comparator C-lines (Figure 2A–B; Figure 3A–B; F1,28 = 8.4, were associated with differences in parasite densities (Figure 2C– p = 0.007, and F1,27 = 6.2, p = 0.02, respectively). The V-lines also D). V-line parasites produced more parasites in total (Figure 2D; induced more anemia than the parasites from which they were F1,28 = 11.5, p = 0.002), and had higher densities on the day of derived (passage 21 versus passage 0: F1,22 = 8.2, p = 0.008). After serial passage (F1,28 = 4.3, p = 0.04) than did C-line parasites. This

PLoS Biology | www.plosbiology.org 3 July 2012 | Volume 10 | Issue 7 | e1001368 Blood-Stage Malaria Vaccine Evolutionary Effects

Figure 2. Virulence and densities in naı¨ve mice of parasites that had previously been serially passaged 10 times through mice that were sham-vaccinated or AMA-1 vaccinated (‘‘evaluation experiment 2’’). Curves (A and C) show the kinetics of five C-lines (blue) and five V-lines (red) each assayed in up to three mice. Points on the scatterplots (B and D) are individual mice infected with C-lines (filled blue circles) or V- lines (filled red triangles). Horizontal black lines indicate mean values. V-lines induced more anemia (A–B; F1,28 = 8.4, p = 0.007) and reached higher total parasite densities than their comparator C-lines (C–D; F1,28 = 11.5, p = 0.002). doi:10.1371/journal.pbio.1001368.g002 is consistent with the hypothesis that selection by AMA-1 passage 21 in AMA-1 vaccinated and sham-vaccinated mice vaccination results in faster growing parasites, and that was why (‘‘evaluation experiment 4’’). This allowed us to ask whether V- vaccine-evolved lines were more virulent. However, vaccine- lines and C-lines were better adapted to the immune environment in adapted parasites from 21 passages, while still more virulent, did which they evolved. Note that the half of this experiment conducted not achieve higher densities than C-line parasites (Figure 3D; V- in sham-vaccinated mice closely replicates our previous evaluation of lines versus C-lines: F1,27 = 1.6, p = 0.2), even though they did the virulence of the lines in naı¨ve mice (‘‘evaluation experiment 3’’). achieve higher densities than ancestral parasites (Figure 3D; Again, we found that the V-lines were more virulent than the C- passage 21 versus passage 0: F1,22 = 12.3, p = 0.002). lines in control mice (Figure 4A–B; anemia F1,38 = 4.0, p = 0.05). We performed another evaluation experiment, this time to This virulence difference was also apparent in vaccinated mice compare the virulence and performance of V- and C-lines from (Figure 4A–B; anemia F1,38 = 4.0, p = 0.05). The magnitude of the

PLoS Biology | www.plosbiology.org 4 July 2012 | Volume 10 | Issue 7 | e1001368 Blood-Stage Malaria Vaccine Evolutionary Effects

Figure 3. Virulence and densities in naı¨ve mice of parasites that had previously been serially passaged 21 times through mice that were sham-vaccinated or AMA-1 vaccinated, together with the progenitor parasites (ancestral) (‘‘evaluation experiment 3’’). Curves (A and C) show the kinetics of five C-lines (blue) and five V-lines (red) each assayed in up to three mice. Black curve is the mean of nine mice infected with the ancestral lineage. Points on the scatterplots (B and D) are individual mice infected with ancestral parasites (filled black diamonds), C-lines (filled blue circles), or V-lines (filled red triangles). Horizontal black lines indicate mean values. V-line parasites caused more anemia than the C-lines and ancestral parasites (A–B; F1,27 = 6.2, p = 0.02 and F1,22 = 8.2, p = 0.008, respectively). The V-lines also reached higher total parasite densities than the ancestral parasites (C–D; F1,22 = 12.3, p = 0.002), but the C-lines and V-lines did not differ from each other (C–D; F1,27 = 1.6, p = 0.2). doi:10.1371/journal.pbio.1001368.g003

virulence difference was unaltered by the vaccine status of the host sham-vaccinated hosts (Figure 4C–D: F1,38 = 1.9, p = 0.1), just as (Figure 4A–B; anemia, parasite6vaccination: F1,76 = 1.0, p = 0.3). they did in naı¨ve mice in evaluation experiment 3 (Figure 3). The Thus, vaccine-line parasites were more virulent in both sham- and V-lines did achieve higher densities in AMA-1-vaccinated hosts AMA-1-vaccinated hosts. (Figure 4C–D; F1,38 = 3.9, p = 0.05), as expected if indeed the V- If parasites had become adapted to the immune environment in lines were better adapted to vaccinated hosts, but this difference which they evolved, we would expect V-lines to perform best in was itself not significantly different from that observed in sham- AMA-1-vaccinated hosts and C-lines to do better than V-lines in vaccinated hosts (Figure 4C–D; parasite6vaccination: F1,76 = 2.8, sham-vaccinated hosts. In fact, C- and V-lines did equally well in p = 0.09).

PLoS Biology | www.plosbiology.org 5 July 2012 | Volume 10 | Issue 7 | e1001368 Blood-Stage Malaria Vaccine Evolutionary Effects

Figure 4. Virulence and densities of parasites that had been serially passaged 21 times in sham-vaccinated and AMA-1-vaccinated mice when assayed in sham-vaccinated or AMA-1-vaccinated mice (‘‘evaluation experiment 4’’). Curves (A and C) represents the kinetics (mean 61 s.e.m.) of five C-lines (blue) and five V-lines (red) when assayed in sham-vaccinated (no symbol) or AMA-1-vaccinated (filled circles) mice. The interaction plots show the minimum RBC (B) and total asexual parasite densities (D) reached during infection of sham- or AMA-1-vaccinated mice with C-lines (blue line) or V-lines (red line). During infection of sham- and AMA-1-vaccinated mice, V-lines induced more anemia than C-lines (A–B; F1,38 = 4.0, p = 0.05 and F1,38 = 4.0, p = 0.05, respectively), but the magnitude was not significant (A–B; anemia, parasite6vaccination: F1,76 = 1.0, p = 0.3). V-lines and C-lines performed equally well in sham-vaccinated hosts (C–D: F1,76 = 1.0, p = 0.3), and although V-lines achieved higher densities in AMA- 1-vaccinated hosts (C–D; F1,38 = 3.9, p = 0.05), the difference was not significant (Figure 4E–F; parasite6vaccination: F1,38 = 1.9, p = 0.1). doi:10.1371/journal.pbio.1001368.g004

Discussion experiments in mice to test whether the candidate malaria blood- stage vaccine AMA-1 creates within-host conditions that selec- The main evolutionary concern of malaria vaccine developers is tively favor the emergence of more virulent parasite variants. In that antigenic escape will erode vaccine efficacy [7,21–27]. three separate phenotyping experiments, we found that parasites Evolutionary biologists have raised a different concern, suggesting selected by passage through AMA-1-vaccinated mice caused more that some vaccines may drive the evolution of more virulent severe disease, removing 20% more RBCs in unvaccinated hosts pathogen variants [28–37]. Virulence evolution would put than did the parasites evolved in unvaccinated mice (Figures 2–4, unvaccinated individuals at risk of more severe disease should panels A and B). Importantly, vaccination did not select for they become infected. In this study we used serial passage antigenic escape at the ama-1 locus (Figure S1). Our data highlight

PLoS Biology | www.plosbiology.org 6 July 2012 | Volume 10 | Issue 7 | e1001368 Blood-Stage Malaria Vaccine Evolutionary Effects the importance of considering the evolutionary repercussions of promote virulence or whether the effects might be less than blood-stage vaccines. These vaccines evidently have the capacity additive. It could be argued that semi-immune individuals will to cause changes at pathogen loci other than target antigens, already naturally be imposing selection for greater virulence in the including those responsible for disease severity. field, and the effects of vaccination will be no worse. However, the In our experiments, all parasites were from the same clonal aim of vaccination programs is to increase the number of immune lineage, so variants differing in virulence must have been people in a population, and if that is achieved, a greater generated either by mutational processes or by switching of proportion of the parasite population will be evolving in immune expression among members of multigene families. Presumably the hosts. more virulent variants had a relative fitness advantage during the Second, our data show that virulence rises with serial passage, as process of serial passage and this was disproportionately larger in it does in many systems [51]. In nature, something must counter vaccinated hosts. Consistent with this, AMA-1-induced immunity within-host selection for virulence (or all pathogens would be controlled ancestral avirulent parasites more effectively than it extremely virulent). It has been hypothesized that syringe passage, controlled virulent descendant parasites (Figure 1). Virulent clones which by-passes natural transmission, eliminates this counter- out-compete less virulent clones in mixed infections [73,74]. This selection against excessive virulence that arises through host death competitive advantage could be associated with more aggressive [51]. This must be true in the limit, but the virulence increases we extraction of resources (e.g., RBCs) during infection or better observed here as a consequence of immunity are likely to be far performance in immune-mediated competition [49,75–78]. We from this limit because mouse death played no role in the selection expect that comparative expression or genomic analyses of our process in our serial passages (Figure S2). In the P. chabaudi-mouse different parasite lines will open up research programs that could model, more virulent infections are more infectious to mosquitoes shed light on the virulence determinants favored by AMA-1- [35,36], and serial passage enhances virulence and transmission induced immunity. stage production [30,52]. Virulence differences generated by Our experiments highlight the importance of considering all experimental evolution using protocols identical to ours, but using types of evolution during malaria vaccine studies. To date, reports whole-parasite immunized mice rather than a recombinant on parasite evolution in response to candidate vaccines in both antigen, were not eliminated by mosquito transmission [30,84]. human and animal trials have focused on antigenic polymorphism. If within- and between-host selection on virulence are somehow But in the few cases where virulence correlates are also available, it antagonistic, an important question is how they play out in the is impossible to disentangle the effects of antigenic polymorphism field now, and how vaccination might affect that. Our data show from virulence. For instance, in a human field trial in Papua New that the within-host selection for virulence is strengthened by Guinea with the P. falciparum ‘‘Combination B’’ blood-stage vaccine-induced immunity. vaccine, which contained recombinant 3D7 MSP-2, the vaccine Third, our experimental design involved passaging parasites was less effective against parasites of the FC27 MSP-2 genotype. every 7 d. We chose that timing because that is after a period of This was interpreted as reflecting a strain-specific protective rapid parasite population expansion (selection) but before naı¨ve response [56,79] but could have also been because the FC27 MSP- mice begin mounting a strong acquired response against malaria 2 genotypes were more virulent [80]. [49,85–90]. This meant that, in contrast to parasites in our Results from an AMA-1 vaccine trial in non-human primates vaccinated mice, our control-selected lines were under only are also consistent with the possibility that more virulent P. modest antibody-mediated immunity. Without further experimen- falciparum strains are harder to control [81]. Aotus monkeys were tation, it is unclear whether onward transmission on any other vaccinated with AMA-1 derived from the P. falciparum 3D7 strain days would lead to more or less potent selection on virulent and then challenged with one of two heterologous strains, FVO or variants. Later passage could select for parasite variants that are FCH/4. AMA-1 vaccination afforded less protection against the even more resilient against the mounting immune response; earlier FVO strain [82]. This could have been because of greater epitope passage may relax selection against competitively less able variants. dis-similarity between the AMA-1 of FV0 and 3D7 strain [81] or How that would play out in terms of transmission to mosquitoes because of the greater virulence of FVO parasites [82]. summed over the whole infectious period remains to be determined. Caveats Our data show that immunization with a recombinant malaria Conclusions vaccine can create ecological conditions that favor parasites that Our data demonstrate that immunity induced by a recombinant cause greater disease severity in unvaccinated individuals. But we antigen that is a candidate for human malaria vaccines can are a long way from being able to assess the likelihood of this increase the potency of within-host selection for more virulent occurring in human malaria populations, were a malaria vaccine malaria parasites. In contrast, we found no evolution of the to go into widespread use. Most obviously, generalizing from parasite locus controlling production of the target antigen. This animal models is notoriously difficult in malaria (reviewed in this does not exclude antigenic polymorphism as a challenge for context by [76,83]), so extreme caution is warranted. But in vaccine efficacy, nor does it mean that virulence evolution is addition to this generic issue, many potentially important inevitable in populations immunized with a leaky (non-sterilizing) considerations remain to be evaluated. Some of these are the vaccine. But it does argue that a range of evolutionary trajectories following. are possible in response to vaccination [36,44], and that epitope First, in human populations there will be variation in levels of evolution is not the only evolution that can occur. We suggest that immunity due to prior infection. Whether existing natural investigation of the impact on blood stage parasite densities and immunity will act to enhance or suppress vaccine-imposed transmission should be a standard component of all Phase 3 selection for more virulent parasite variants remains to be malaria vaccine trials [10], and that whole genome analyses of determined. In mice, live parasite-induced immunity [30] and parasites that survive and are transmitted from individuals in AMA-1-induced immunity (this study) both promote the evolution vaccinated and control arms in clinical trials should be a priority. of virulence. Further experiments are needed to determine Until there is a better understanding of the selection processes set whether both occurring together in the same host would further up by imperfect vaccination, there is no reason to think that

PLoS Biology | www.plosbiology.org 7 July 2012 | Volume 10 | Issue 7 | e1001368 Blood-Stage Malaria Vaccine Evolutionary Effects vaccine-driven evolution will occur only in genes encoding target Box 1. Evaluating Evolutionary Risk antigens. Evaluating the medium term effects of widespread vaccination (evolutionary risk) is a substantial challenge, not least Our experimental data demonstrate that widespread use because evolutionary change is likely to occur long after clinical of a malaria vaccine could create parasites that cause more trials have concluded (Box 1). More generally, there is little reason severe disease in unvaccinated individuals. However, it is to think the vaccine-driven virulence evolution we have seen will not currently possible to evaluate the likelihood of such be limited to malaria parasites. Analysis of virulence evolution in evolution. This is for a variety of reasons. range of infectious diseases for which leaky vaccines are in First, evolutionary trajectories in natural populations are widespread use would be of substantial interest. always extremely difficult to predict from laboratory studies. Our experiments began with a single clone and Material and Methods relied entirely on mutational variation that arose during our experiments. The genetic variation in virulence and Ethics Statement epitopes present in natural malaria populations are likely This study was carried out in strict accordance with the different. Malaria virulence is probably controlled by many genes, so that mutational variation in virulence may arise recommendations in the Guide for the Care and Use of more frequently than escape variation in epitopes that are Laboratory Animals of the National Institutes of Health. The typically encoded by small genetic regions. That might be protocol was approved by the Animal Care and Use Committee of why we failed to see epitope evolution in our experiments the Pennsylvania State University (Permit Number: 27452). but easily detected virulence evolution. If so, it is possible that attempts to broaden the response to AMA-1 with Parasites and Hosts multivalent vaccines [26,27,98] might, by reducing the We used the DK clone of P. chabaudi adami, which was originally range of escape options available to the parasite, make collected from thicket rats (Thamnomys rutilans) in the Congo virulence evolution more likely. Brazzaville [91–93], and subsequently cloned by limiting dilution. Even if we knew a lot about population-level genetic Laboratory genotypes are stored as stable isolates in liquid variation in virulence and epitopes, predicting evolutionary nitrogen with subscript codes used to identify their position in trajectories—and in particular evolutionary timescales— clonal history [52]. Mice in our experiments were female C57Bl/ requires additional knowledge about genetic covariation 6, at least 6–8 wk old. Parasite densities were estimated from day 4 with fitness. Virulence-transmission relationships, for ex- from samples of tail blood using Giemsa-stained thin smears and ample, are well understood in our mouse model [35,36]. red blood cell density was estimated from day 0 by flow cytometry There is circumstantial evidence that similar relationships (Beckman Coulter), or by genotype-specific real-time quantitative exist in P. falciparum, but the issue is far from settled and indeed may never be [35]. Additionally, we know very little real-time PCR (qPCR) assays as described previously [74]. For about the strength of selection that will be imposed by amplification of the DK genotype, we used the forward primer candidate malaria vaccines. Clearly vaccine coverage will previously used to amplify AS/AJ genotypes [74] and the DK be an important determinant, but so too will the strength genotpe-specific reverse primer 59 GATTGTAGAGAAGTA- of vaccine-induced within host selection, which has yet to GAAAATACA GATACAACTAA 39. be estimated in people. In our view, a profitable way forward is whole transcrip- Vaccination tome comparisons of parasites that appear in people in All mice were in one of the following three immune classes: vaccine and control arms of vaccine trials. And before naı¨ve (never vaccinated with the adjuvant or the AMA-1 antigen), novel vaccines go into widespread use, it should be a high sham-vaccinated (which were immunized with adjuvant alone), or priority to collect random samples of parasites from the vaccinated (which were immunized with AMA-1 antigen plus pre-vaccine era and then to regularly collect random adjuvant). We use that terminology consistently throughout. samples perhaps every 5 years after that. Whole tran- Immunization protocols were similar to those described by scriptome analyses of longitudinal parasite samples have Anders and others [53,94,95]. Briefly, vaccination was with the the potential to detect vaccine-driven evolution of ectodomian of the AMA-1 protein derived from P. c. adami virulence determinants. genotype DK [53]. AMA-1 was emulsified with Montanide ISA 720 adjuvant (Seppic). Each mouse was injected intra-peritoneally with a total of 10 mg of protein on two occasions with a 4-wk The second serial passage (‘‘B’’) was the experimental evolution interval. Sham-vaccinated mice were injected with Montanide phase of our study (Figure S2). This was aimed at comparing the ISA720 plus PBS. During serial passage, and during the evaluation evolutionary consequences of passaging parasites through two experiments, mice were infected with parasites 14 d after the contrasting selection treatments: sham- and AMA-1-vaccinated second immunization. mice. We used sham-vaccinated mice so as to ensure that any evolved differences could be attributed to AMA-1 antigen, and not Serial Passages the adjuvant. We initially aimed to derive five independent We conducted two separate serial passage experiments (denoted parasite lines per selection treatment. At the start (generation 1), A and B). All passages involved the syringe transfer of 0.1 ml of five mice that had been previously immunized with the AMA-1 diluted blood containing 56105 parasites between mice every 7 d. vaccine (V- lines) or a sham vaccine (C-lines) were infected with P. We first used serial passage simply to derive a more virulent c. adami genotype DK247 (generation 0) (Figure S2). Parasites from parasite lineage from the ancestral DK (‘‘serial passage experiment each one of the five mice at generation 1 were then used to infect A’’). This allowed us to test whether AMA-1-induced immunity at least two mice at generation 2 (forming a total of 10 sublines per controlled the derived (virulent) line less successfully than the treatment). Duplicate infections helped reduce the possibility of ancestral (less virulent) line. P. c. adami genotype DK294 was losing lines during the selection phase. Thus, from generation 2 to derived via serial passage of ancestral P. c. adami genotype DK122 21, parasites from each mouse within a selection treatment were after a total of 30 passages though immunologically naı¨ve mice. used to infect a fresh mouse in the next generation. Some lines

PLoS Biology | www.plosbiology.org 8 July 2012 | Volume 10 | Issue 7 | e1001368 Blood-Stage Malaria Vaccine Evolutionary Effects were lost (notably where AMA-1 vaccination induced a strongly Statistical Analysis protective anti-parasitic response) (Figure S2). When lines were All analyses were conducted in R 2.10.1 [96]. All parasite lost, blood from a mouse in another line within that treatment density data were log transformed to meet normality assumptions group was used to infect at least two other mice in the next of the models. For the analysis of evaluation experiments 2–4, generation. This protocol ensured that at each generation 10 mice which determined the consequences of evolution through sham- were infected with parasites within each selection treatment. A and AMA-1-vaccinated hosts (serial passage B), differences among total of 410 mice were used during this experimental evolution sub-line variances (C-lines and V-lines) were first analyzed using phase. mixed effect linear models with sub-line as a random effect [97]. In all experiments there were no sub-line variances with selection Virulence Phenotyping treatments so we only report the between-selection effects. For Virulence and clone performance were assessed in four separate completeness, we report the more conservative analysis, based only ‘‘evaluation’’ experiments conducted after the serial passages. In on line means, in Table S2. all cases frozen lines (P. c. adami-infected erythrocytes (IRBC)) were first introduced into naı¨ve donor mice and then into naı¨ve or Supporting Information sham-immunized experimental mice. Naı¨ve donors are used Figure S1 Nucleotide sequence of P. chabaudi ama-1. Consensus P. because exact doses to initiate experiment infections cannot be c. adami ama-1 nucleotide sequence between the derived virulent obtained from frozen stock. Note that this single passage in naı¨ve parasites from used in ‘‘evaluation’’ experiment 1, the V-lines and mice would, if it does anything, act to narrow the virulence C-lines used in ‘‘evaluation’’ experiments 3 and 4 (21 serial differences observed in our experiments. Experimental mice were 6 passages), and the ancestral lineages from which all lines were intra-peritoneally injected with 1610 IRBCs. derived and all compared to the published P. c. adami DK ama-1 Evaluation experiment 1 compared the performance of (genebank accession number U49745). There was 100% ama-1 parasites derived from serial passage A with their pre-passage sequence identity among and between all of the derived lines and progenitors in vaccinated and naı¨ve hosts (Table S1). Two mice with their ancestral lineages and to the published genebank died (one control immunized and one AMA-1 immunized both sequence (shaded in grey). The outer forward and inner reverse infected with derived parasites). These were included in the primers used for amplification and sequencing are highlighted in calculation of daily densities until death as death always occurred bold and the inner forward and outer reverse primers are shown in after the peak of infection (days 17 and 15, respectively). lowercase lettering. All traces were examined by eye for multiple Three further evaluation experiments were used to compare peaks, and none were observed. If parasites with base-pair changes the virulence and parasites dynamics of the C-lines and V-lines were present in sequenced samples, they must have been there at from serial passage B (Table S1): evaluation experiment 2, frequencies less than about 20%. parasites from passage 10 in naı¨ve mice; evaluation experiment 3, (TIF) parasites from passage 21 in naı¨ve mice; and evaluation experiment 4, parasites from passage 21 in sham- and AMA-1- Figure S2 Experimental evolution (serial passage B) in sham- and vaccinated mice. In these three evaluation experiments, we AMA-1-vaccinated animals. Schematic genealogy illustrating compared five surviving C-lines with five surviving V-lines, with passage history of the C-lines and V-lines from the ancestral each line used to infect three mice. The lines used and their lineage. Nodes represent mice. To start, five mice that had been history are as shown in Figure S2. In evaluation experiment 3, previously immunized with the AMA-1 vaccine or a sham vaccine were infected with P. c. adami genotype DK (passage 1) to initiate nine naı¨ve mice were also infected with the ancestral lineage (P. 247 the V-lines and C-lines, respectively. Parasites from each one of the chabaudi genotype DK ). During evaluation experiment 2, one 247 five mice at passage 1 were then used to infect at least two mice at mouse infected with C-line parasites died on day five and was passage 2 (forming a total of 10 sublines per treatment). From thus excluded from all analyses passage 2 to 21 parasites from each mouse within a selection treatment were used to infect a fresh mouse in the next passage. DNA Sequencing and Sequence Analysis Where parasite lines were lost (filled red circles) blood from a mouse To test selected parasites for epitope evolution, ama-1 nucleotide in another line within that treatment group was used to infect at sequences of the ancestral and derived parasites from experiment least two other mice in the next generation. Lines were lost when one and the ancestral, C- and V-line parasites from experiments 3 parasite densities were below transmissible frequencies on day 7 PI and 4 (passage 21 parasites) were established using a series of either because of vaccine-induced immunity (V-lines) or errors in overlapping oligonucleotide primers designed by reference to the dose delivered to mice (C-lines). Diamonds represent parasite lines published sequences of P. c. adami DK [94,95]. Parasite DNA was used in the different evaluation experiments. extracted as previously described [74]. AMA-1 was amplified as (TIFF) two gene fragments: Outer Forward 59 CTTGGGTAATTGT- TCCGA 39 and Inner Reverse 59 GCACTTCTAACCCTTTG- Figure S3 Parasite densities of each mouse during serial passage GT 39; Inner Forward 59 GGGTCCAAGATATTGTAG 39 and B in sham- and AMA-1- vaccinated animals. Each data point Outer Reverse 59 GGGTTTCGTCTTTTCTAC 39. PCR was represents the log parasite density of each mouse in the C-lines performed using Nova Taq (Novagen), with the thermocycle (blue circles) or V-lines (red triangles) from passage 1 to 21. Solid black lines represent the log linear regression change in parasite profile; 95uC for 12 min, then 95uC for 1 min, 57uC for 1 min, density per selection treatment over time. and 72uC for 1 min (630 cycles) ending at 72uC for 10 min. Amplified DNA was visualized on a 1% agarose gel and positive (TIF) amplifications were cleaned with QIAquick Gel extraction kit Table S1 Description of evaluation experiments 1 to 4. V and C, (Qiagen) and sequenced in both directions with the same primers V-or C-lines. Numbers indicate subline used. DK122, DK247, and that were used for amplification. Sequencing was performed by DK294, DK ancestral genotypes with subscript codes used to Penn State DNA sequencing core facility and sequences were identify their position in clonal history. aligned and analyzed using ClustalW. (DOC)

PLoS Biology | www.plosbiology.org 9 July 2012 | Volume 10 | Issue 7 | e1001368 Blood-Stage Malaria Vaccine Evolutionary Effects

Table S2 Most conservative statistical analysis of evaluation Fogarty International Center, National Institutes of Health, particularly J. experiments 2 to 4. Lloyd-Smith, for stimulating discussion. (DOC) Author Contributions Acknowledgments The author(s) have made the following declarations about their contributions: Conceived and designed the experiments: VCB AFR. We thank M. Mackinnon, M. Ferrari, S. Baigent, current members of the Performed the experiments: VCB DS BHKC LAN ASB. Analyzed the Read group, and members of the RAPIDD program of the Science & data: VCB AFR. Contributed reagents/materials/analysis tools: RFA Technology Directorate, Department of Homeland Security, and the AFR. Wrote the paper: VCB RFA AFR. Figure S2: VCB, MAR.

References 1. Hyde JE (2005) Drug-resistant malaria. Trends Parasitol 21: 494–498. 28. Gandon S, Mackinnon MJ, Nee S, Read AF (2001) Imperfect vaccines and the 2. Enserink M (2008) Signs of drug resistance rattle experts, trigger bold plan. evolution of pathogen virulence. Nature 414: 751–756. Science 322: 1776. 29. Gandon S, Day T (2003) Understanding and managing pathogen evolution: a 3. Yewhalaw D, Wassie F, Steurbaut W, Spanoghe P, Van Bortel W, et al. (2011) way forward. Trends Microbiol 11: 206–207. Multiple insecticide resistance: an impediment to insecticide-based malaria vector 30. Mackinnon M, Read AF (2004) Immunity promotes virulence evolution in a control program. PLoS One 6: e16066. doi:10.1371/journal.pone.0016066 malaria model. PLoS Biol 2: 1–7. doi:10.1371/journal.pbio.0020230 4. Ranson H, N’Guessan R, Lines J, Moiroux N, Nkuni Z, et al. (2011) Pyrethroid 31. Mackinnon M, Read AF (2004) Virulence in malaria: an evolutionary viewpoint. resistance in African anopheline mosquitoes: what are the implications for Philos Trans R Soc Lond B Biol Sci 359: 965–986. malaria control? Trends Parasitol 27: 91–98. 32. Andre´ JB, Gandon S (2006) Vaccination, within-host dynamics, and virulence 5. Crompton PD, Pierce SK, Miller LH (2010) Advances and challenges in malaria evolution. Evolution 60: 13–23. vaccine development. J Clin Invest 120: 4168–4178. 33. Day T, Galvani A, Struchiner C, Gumel A (2008) The evolutionary 6. Remarque EJ, Faber BW, Kocken CHM, Thomas AW (2008) Apical membrane consequences of vaccination. Vaccine 26: C1–C3. antigen 1: a malaria vaccine candidate in review. Trends Parasitol 24: 74–84. 34. Gandon S, Day T (2008) Evidence of parasite evolution after vaccination. 7. Anders RF (2011) The case for a subunit vaccine against malaria. Trends Vaccine 26: C4–C7. Parasitol 27: 330–334. 35. Mackinnon MJ, Gandon S, Read AF (2008) Virulence evolution in response to 8. Hviid L (2010) The role of Plasmodium falciparum variant surface antigens in vaccination: the case of malaria. Vaccine 26: C42–C52. protective immunity and vaccine development. Hum Vaccin 6: 84–89. 36. Read AF, Mackinnon MJ (2008) Pathogen evolution in a vaccinated world. In: 9. Greenwood BM, Targett GAT (2011) Malaria vaccines and the new malaria Stearns SC, Koella, JC editors Evolution in health and disease 2nd Ed. Oxford agenda. Clin Micro Infect 17: 1600–1607. University Press pp. 139–152. 10. Agnandji ST, Lell B, Soulanoudjingar SS, Fernandes JF, Abossolo BP, et al. 37. Williams PD, Day T (2008) Epidemiological and evolutionary consequences of (2011) First results of phase 3 trial of RTS,S/AS01 malaria vaccine in African targeted vaccination. Mol Ecol 17: 485–499. children. 38. Stearns SC, Nesse RM, Govindaraju DR, Ellison PT (2010) Evolutionary 11. Rappuoli R, Aderem A (2011) A 2020 vision for vaccines against HIV, perspectives on health and medicine. Proc Natl Acad Sci U S A 107: 1691–1695. tuberculosis and malaria. Nature 473: 463–469. 39. van den Berg TP, Eterradossi N, Toqui D, Meulemans G (2000) Infectious 12. Good MF (2001) Towards a blood-stage vaccine for malaria: are we following all bursal disease (Gumboro disease). Rev Sci Technol 19: 509–543. the leads? Nat Rev Immunol 1: 117–125. 40. Witter RL (2001) Protective efficacy of Marek’s disease vaccines. Curr Top 13. Zuckerman AJ (2000) Effects of hepatitis B virus mutants on efficacy of Microbiol Immunol: 57–90. vaccination. Lancet 356: 1382–1384. 41. Davidson F, Nair L (2004) Marek’s disease an evolving problem. London, 14. Huang M-L, Liao W-L, Ho M-S (2007) HBV serological markers of vaccinated Elsevier Academic Press. children in remote areas of Taiwan: emphasis on factors contributing to vaccine 42. Radford AD, Dawson S, Coyne KP, Porter CJ, Gaskell RM (2006) The failure. Vaccine 25: 6326–6333. challenge for the next generation of feline calicivirus vaccines. Vet Microbiol 15. Mooi FR, van Oirschot H, Heuvelman K, van der Heide HGJ, Gaastra W, et al. 117: 14–18. (1998) Polymorphism in the Bordetella pertussis virulence factors P.69/pertactin 43. Gimeno IM (2008) Marek’s disease vaccines: a solution for today but a worry for and pertussis toxin in The Netherlands: temporal trends and evidence for tomorrow? Vaccine 26: C31–C41. vaccine-driven evolution. Infect Immun 66: 670–675. 44. Mackinnon MJ, Marsh K (2010) The selection landscape of malaria parasites. 16. van Loo IHM, van der Heide HGJ, Nagelkerke NJD, Verhoef J, Mooi FR Science 328: 866–871. (1999) Temporal trends in the population structure of Bordetella pertussis during 1949–1996 in a highly vaccinated population. J Infect Dis 179: 915–923. 45. Antia R, Levin BR, May RM (1994) Within-host population dynamics and the 17. Kallonen T, He Q (2009) Bordetella pertussis strain variation and evolution evolution and maintenance of microparasite virulence. Am Nat 144: 457–472. postvaccination. Expert Rev Vaccines 8: 863–875. 46. Mooi FR, van Loo IHM, van Gent M, He Q, Heuvelman CJ, et al. (2009) 18. King AJ, van Gorkom T, van der Heide HG, Advani A, van der Lee S (2010) Bordetella pertussis strains with increased toxin production associated with Pertussis Changes in the genomic content of circulating Bordetella pertussis strains isolated resurgence. Emerg Infect Dis 15: 1206–1213. from the Netherlands, Sweden, Japan and Australia: adaptive evolution or drift? 47. Kwiatkowski D, Marsh K (1997) Development of a malaria vaccine. The Lancet BMC Genomics 11: 64. 350: 1696–1701. 19. Huang SS, Platt R, Rifas-Shiman SL, Pelton SI, Goldmann D, et al. (2005) Post- 48. Boyd MF, Kitchen SF (1936) On the efficiency of the homologous properties of PCV7 changes in colonizing pneumococcal serotypes in 16 Massachusetts acquired immunity to Plasmodium vivax. Am J Trop Med 16: 447–457. communities, 2001 and 2004. Pediatrics 116: e408–e413. 49. Metcalf CJE, Graham AL, Huijben S, Barclay VC, Long GH, et al. (2011) 20. Croucher NJ, Harris SR, Fraser C, Quail MA, Burton J, et al. (2011) Rapid Partitioning regulatory mechanisms of within-host malaria dynamics using the pneumococcal evolution in response to clinical interventions. Science 331: 430– effective propagation number. Science 333: 984–988. 434. 50. Ni Y, Kemp MC (1992) Strain-specific selection of genome segments in avian 21. Jeffery GM (1966) Epidemiological significance of repeated infections with reovirus coinfections. J Gen Virol 73: 3107–3113. homologous and heterologous strains and species of Plasmodium. WHO Bulletin 51. Ebert D (1998) Experimental evolution of parasites. Science 282: 1432–1436. 35: 873–882. 52. Mackinnon MJ, Read AF (1999) Selection for high and low virulence in the 22. Mendis KN, David PH, Carter R (1991) Anigenic polymorphism in malaria: is it malaria parasite Plasmodium chabaudi. Proc R Soc Lond B Biol Sci 266: 741–748. an important mechanism for immune evasion? Immunol Today 12: A34–A37. 53. Anders RF, Crewther PE, Edwards S, Margetts M, Matthew MLSM, et al. 23. Cortes A, Mellombo M, Masciantonio R, Murphy VJ, Reeder JC, et al. (2005) (1998) Immunisation with recombinant AMA-1 protects mice against infection Allele specificity of naturally acquired antibody responses against Plasmodium with Plasmodium chabaudi. Vaccine 16: 240–247. falciparum apical membrane antigen 1. Infect Immun 73: 422–430. 54. Genton B, Al-Yaman F, Anders R, Saul A, Brown G, et al. (2000) Safety and 24. Cheesman S, Raza A, Carter R (2006) Mixed strain infections and strain-specific immunogenicity of a three-component blood-stage malaria vaccine in adults protective immunity in the rodent malaria parasite Plasmodium chabaudi chabaudi in living in an endemic area of Papua New Guinea. Vaccine 18: 2504–2511. mice. Infect Immun 74: 2996–3001. 55. Hodder AN, Crewther PE, Anders RF (2001) Specificity of the protective 25. Barclay VC, Chan BHK, Anders RF, Read AF (2008) Mixed allele malaria antibody response to apical membrane antigen 1. Infect Immun 69: 3286–3294. vaccines: host protection and within-host selection. Vaccine 26: 6099–6107. 56. Genton B, Betuela I, Felger I, Al-Yaman F, Anders RF, et al. (2002) A 26. Dutta S, Lee SY, Batchelor AH, Lanar DE (2007) Structural basis of antigenic recombinant blood-stage malaria vaccine reduces Plasmodium falciparum density escape of a malaria vaccine candidate. Proc Natl Acad Sci U S A 104: 12488– and exerts selective pressure on parasite populations in a phase 1–2b trial in 12493. Papua New Guinea. J Infect Dis 185: 820–827. 27. Kusi KA, Faber BW, Thomas AW, Remarque EJ (2009) Humoral immune 57. Burns JJM, Flaherty PR, Romero MM, Weidanz WP (2003) Immunization response to mixed PfAMA1 alleles; multivalent PfAMA1 vaccines induce broad against Plasmodium chabaudi malaria using combined formulations of apical specificity. PLoS One 4: e8110. doi:10.1371/journal.pone.0008110 membrane antigen-1 and merozoite surface protein-1. Vaccine 21: 1843–1852.

PLoS Biology | www.plosbiology.org 10 July 2012 | Volume 10 | Issue 7 | e1001368 Blood-Stage Malaria Vaccine Evolutionary Effects

58. Pan W, Huang D, Zhang Q, Qu L, Zhang D, et al. (2004) Fusion of two malaria 77. Barclay VC, Ra˚bergL,ChanBHK,BrownS,GrayD,etal.(2008)CD4+T cells vaccine candidate antigens enhances product yield, immunogenicity, and do not mediate within-host competition between genetically diverse malaria antibody-mediated inhibition of parasite growth in vitro. J Immunol 172: parasites. Proc R Soc Lond B Biol Sci 275: 1171–1179. 6167–6174. 78. Mideo N, Day T, Read AF (2008) Modelling malaria pathogenesis. Cell 59. Polhemus ME, Magill AJ, Cummings JF, Kester KE, Ockenhouse CF, et al. Microbiol 10: 1947–1955. (2007) Phase I dose escalation safety and immunogenicity trial of Plasmodium 79. Cortes A, Mellombo M, Benet A, Lorry K, Rare L, et al. (2004) Plasmodium falciparum apical membrane protein (AMA-1) FMP2.1, adjuvanted with AS02A, falciparum: distribution of msp2 genotypes among symptomatic and asymptom- in malaria-naive adults at the Walter Reed Army Institute of Research. Vaccine atic individuals from the Wosera region of Papua New Guinea. Exp Parasitol 25: 4203–4212. 106: 22–29. 60. Dicko A, Sagara I, Ellis RD, Miura K, Guindo O, et al. (2008) Phase 1 study of a 80. Engelbrecht F, Felger I, Genton B, Alpers M, Beck HP (1995) Plasmodium combination AMA1 blood stage malaria vaccine in Malian children. PLoS One falciparum: malaria morbidity is associated with specific merozoite surface antigen 3: e1563. doi:10.1371/journal.pone.0001563 2 genotypes. Exp Parasitol 81: 90–96. 61. Hu J, Chen Z, Gu J, Wan M, Shen Q, et al. (2008) Safety and immunogenicity 81. Dutta S, Sullivan JS, Grady KK, Haynes JD, Komisar J, et al. (2009) High of a malaria vaccine, Plasmodium falciparum AMA-1/MSP-1 chimeric protein antibody titer against apical membrane antigen-1 is required to protect against formulated in montanide ISA 720 in healthy adults. PLoS One 3: e1952. malaria in the Aotus model. PLoS One 4: e8138. doi:10.1371/journal.- pone.0008138 doi:10.1371/journal.pone.0001952 82. Collins WE, Galland GG, Sullivan JS, Morris CL (1994) Selection of different 62. Thera MA, Doumbo OK, Coulibaly D, Diallo DA, Kone AK, et al. (2008) strains of Plasmodium falciparum for testing blood-stage vaccines in Aotus nancymai Safety and immunogenicity of an AMA-1 malaria vaccine in Malian adults: monkeys. Am J Trop Med Hyg 51: 224–232. results of a phase 1 randomized controlled trial. PLoS One 3: e1465. 83. Wargo AR, Huijben S, de Roode JC, Shepherd J, Read AF (2007) Competitive doi:10.1371/journal.pone.0001465 release and facilitation of drug-resistant parasites after therapeutic chemotherapy 63. Sagara I, Dicko A, Ellis RD, Fay MP, Diawara SI, et al. (2009) A randomized in a rodent malaria model. Proc Natl Acad Sci USA 104: 19914–19919. controlled phase 2 trial of the blood stage AMA1-C1/Alhydrogel malaria 84. Mackinnon MJ, Bell A, Read AF (2005) The effects of mosquito transmission vaccine in children in Mali. Vaccine 27: 3090–3098. and population bottlenecking on virulence, multiplication rate and rosetting in 64. Spring MD, Cummings JF, Ockenhouse CF, Dutta S, Reidler R, et al. (2009) rodent malaria. Int J Parasitol 35: 145–153. Phase 1/2a study of the malaria vaccine candidate apical membrane antigen-1 85. Druilhe P, Khusmith S (1987) Epidemiological correlation between levels of (AMA-1) administered in adjuvant system AS01B or AS02A. PLoS One 4: antibodies promoting merozoite phagocytosis of Plasmodium falciparum and e5254. doi:10.1371/journal.pone.0005254 malaria-immune status. Infect Immun 55: 888–891. 65. Xue X, Ding F, Zhang Q, Pan X, Qu L, et al. (2010) Stability and potency of the 86. Langhorne J, Gillard S, Simon B, Slade S, Eichmann K (1989) Frequencies of Plasmodium falciparum MSP1-19/AMA-1(III) chimeric vaccine candidate with CD4+ T cells reactive with Plasmodium chabaudi chabaudi: distinct response kinetics Montanide ISA720 adjuvant. Vaccine 28: 3152–3158. for cells with Th1 and Th2 characteristics during infection. Int Immunol 1: 416– 66. Anders RF, Adda CG, Foley M, Norton RS (2010) Recombinant protein 424. vaccines against the asexual blood-stages of Plasmodium falciparum. Hum Vaccin 6: 87. Stevenson MM, Huang DY, Podoba JE, Nowotarski ME (1992) Macrophage 39–53. activation during Plasmodium chabaudi AS infection in resistant C57BL/6 and 67. Pierce MA, Ellis RD, Martin LB, Malkin E, Tierney E, et al. (2010) Phase 1 susceptible A/J mice. Infect Immun 60: 1193–1201. safety and immunogenicity trial of the Plasmodium falciparum blood-stage malaria 88. von der Weld T, Langhorne J (1993) Altered response of CD4+ T cell subsets to vaccine AMA1-C1/ISA 720 in Australian adults. Vaccine 28: 2236–2242. Plasmodium chabaudi chabaudi in B cell-deficient mice. Int Immunol 5: 1343–1348. 68. Thera MA, Doumbo OK, Coulibaly D, Laurens MB, Ouattara A, et al. (2011) 89. Waki S (1994) Antibody-dependent neutrophil-mediated parasite killing in non- A field trial to assess a blood-stage malaria vaccine. N Engl J Med 365: 1004– lethal rodent malaria. Parasite Immunol 16: 587–591. 1013. 90. Mota MM, Brown KN, Holder AA, Jarra W (1998) Acute Plasmodium chabaudi 69. Kocken CHM, Withers-Martinez C, Dubbeld MA, van der Wel A, Hackett F, et chabaudi malaria infection induces antibodies which bind to the surfaces of al. (2002) High-level expression of the malaria blood-stage vaccine candidate parasitized erythrocytes and promote their phagocytosis by macrophages In Plasmodium falciparum apical membrane antigen 1 and induction of antibodies that vitro. Infect Immun 66: 4080–4086. inhibit erythrocyte invasion. Infect Immun 70: 4471–4476. 91. Landau I (1965) Description of Plasmodium chabaudi. Parasite of African rodents. 70. Marshall VM, Zhang L, Anders RF, Coppel RL (1996) Diversity of the vaccine Comptes Rendus Hebdomadaires des Seances de l Academie des Sciences 260: candidate AMA-1 of Plasmodium falciparum. Mol Bio Para 77: 109–113. 3758–3761. 92. Carter R, Walliker D (1976) Malaria parasites of rodents of the Congo 71. Gunasekera AM, Wickramarachchi T, Neafsey DE, Ganguli I, Perera L, et al. (Brazzaville). Ann Parasitol Hum Comp 51: 637–646. (2007) Genetic diversity and selection at the Plasmodium vivax Apical Membrane 93. Beale GH, Walliker D, Carter R (1978) Genetics. In: Killick-Kendrick R & Antigen-1 (PvAMA-1) locus in a Sri Lankan population. Mol Biol Evol 24: 939– Peters W editors. Rodent malaria. London Academic Press. pp. 213–45. 947. 94. Crewther PE, Matthew ML, Flegg RH, Anders RF (1996) Protective immune 72. Rajesh V, Singamsetti VK, Vidya S, Gowrishankar M, Elamaran M, et al. responses to apical membrane antigen 1 of Plasmodium chabaudi involve (2008) Plasmodium falciparum: genetic polymorphism in apical membrane antigen- recognition of strain-specific epitopes. Infect Immun 64: 3310–3317. 1 gene from Indian isolates. Exp Parasitol 119: 144–151. 95. Marshall VM, Peterson MG, Lew AM, Kemp DJ (1989) Structure of the apical 73. de Roode JC, Pansini R, Cheesman SJ, Helinski MEH, Huijben S, et al. (2005) membrane antigen I (AMA-1) of Plasmodium chabaudi. Mol Bio Para 37: 281–283. Virulence and competitive ability in genetically diverse malaria infections. Proc 96. R Development Core Team (2009) R: A language and environment for Natl Acad Sci U S A 102: 7624–7628. statistical computing. R Foundation for Statistical Computing V, Austria. ISBN 74. Bell AS, Roode JC, Sim D, Read AF (2006) Within-host competition in 3-900051-07-0, http://www.R-project.org genetically diverse malaria infections: parasite virulence and competitive success. 97. Pinheiro JC, Bates DM (2000) Fitting linear mixed-effects models. In: Statistics Evolution 60: 1358–1371. and computing. Mixed-effects models in S and S-plus. Springer Verlag New 75. Read AF, Taylor LH (2001) The ecology of genetically diverse infections. York, LLC. Science 292: 1099–1102. 98. Duan J, Mu J, Thera MA, Joy D, Kosakovsky Pond SL, et al. (2008) Population 76. Ra˚berg L, de Roode JC, Bell AS, Stamou P, Gray D, et al. (2006) The role of structure of the genes encoding the polymorphic Plasmodium falciparum apical immune-mediated apparent competition in genetically diverse malaria infec- membrane antigen 1: implications for vaccine design. Proc Natl Acad Sci U S A tions. Am Nat 168: 41–53. 105: 7857–7862.

PLoS Biology | www.plosbiology.org 11 July 2012 | Volume 10 | Issue 7 | e1001368 Figure S1.

Figure S2.

Figure S3.

Table S1.

No. Evaluation exp Immune status Parasites Subline mice

1 Sham-vaccine Ancestral Ancestral (DK122) 4

Sham-vaccine Derived Derived (DK294) 4

AMA-1 vaccine Ancestral Ancestral (DK122) 4

AMA-1 vaccine Derived Derived (DK294) 4 2 Naïve Passage 10 C2 3 Naïve C4 3 Naïve C6 3 Naïve C8 3 Naïve C10 3 Naïve Passage 10 V1 3 Naïve V3 3 Naïve V4 3 Naïve V5 3 Naïve V10 3

3 Naïve Passage 21 Ancestral (DK247) 9 Naïve C1 3 Naïve C3 3 Naïve C5 3 Naïve C7 3 Naïve C10 3 Naïve Passage 21 V2 3 Naïve V3 3 Naïve V6 3 Naïve V8 3 Naïve V9 3 4 Sham-vaccine Passage 21 C1 3 Sham-vaccine C3 3 Sham-vaccine C5 3 Sham-vaccine C7 3 Sham-vaccine C10 3 AMA-1 vaccine C1 3 AMA-1 vaccine C3 3 AMA-1 vaccine C5 3 AMA-1 vaccine C7 3 AMA-1 vaccine C10 3 Sham-vaccine V2 3 Sham-vaccine V3 3 Sham-vaccine V6 3 Sham-vaccine V8 3 Sham-vaccine V9 3 AMA-1 vaccine V2 3 AMA-1 vaccine V3 3 AMA-1 vaccine V6 3 AMA-1 vaccine V8 3 AMA-1 vaccine V9 3

Table S2.

Evaluation Exp. Linear mixed effects test (line as random effect) Result

2 Anaemia: V-lines versus C-lines in naïve mice F1,8= 8.3, p=0.02

Total parasites: V-lines verusus C-lines in naïve mice F1,8=11.5, p=0.009

Parasites on day 7: V-lines verusus C-lines in naïve mice F1,8= 3.6, p= 0.08

3 Anaemia: V-lines versus C-lines in naïve mice F1,10=6.2, p=0.03

Anaemia: V-lines versus ancestral parasites in naïve mice F1,10=9.7, p=0.01

Total parasites: V-lines verusus C-lines in naïve mice F1,10= 1.1, p=0.3

Total parasites: V-lines verusus ancestral parasites in naïve mice F1,10= 7.3, p=0.02

4 Anaemia: V-lines versus C-lines in sham-vaccinated mice F1,8= 4.0, p=0.08

Anaemia: V-lines versus C-lines in AMA-1 vaccinated mice F1,8= 2.2, p=0.1

Anaemia interaction = Line*Vaccination F1,16= 0.5, p=0.4

Total parasites: V-lines versus C-lines in sham-vaccinated mice F1,8= 2.5, p=0.1

Toatl parasites: V-lines versus C-lines in AMA-1 vaccinated mice F1,8= 4.0, p=0.07 Parasite interaction = Line*Vaccination F1,16= 2.6, p=0.1

Maia A. Rabaa Curriculum vitae Address

• Department of Biology, The Pennsylvania State University, University Park, Pennsylvania, USA, 16802 Tel: 330.904.1568; E-mail: [email protected]

Academic History

• 2012: Ph.D., Biology, The Pennsylvania State University, University Park, Pennsylvania, USA • 2006: M.H.S., International Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, USA • 2001: B.S., Biology, Walsh University, North Canton, Ohio, USA

Non-Academic History

• 2007-2008: Research assistant, Fogarty International Center, National Institutes of Health, Bethesda, Maryland, USA • 2005-2006: Research assistant, Department of International Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, USA • 2004-2009: Scientific editor/Writer, Harrisco Research Institute, Seoul, Republic of Korea • 2002-2004: University admissions counselor, Hanyoung Foreign Language High School Overseas Study Program, Seoul, Republic of Korea

Teaching Experience • • 2011: Teaching assistant, Department of Biology, The Pennsylvania State University, University Park, Pennsylvania • 2005: Teaching assistant, Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland • 2002-2003: Teacher of English and science, Hanyoung Foreign Language High School, Seoul, Republic of Korea

Selected Honors and Awards

• 2009-2012: Graduate Research Fellowship, National Science Foundation • 2008-2011: Braddock Scholarship, The Pennsylvania State University • 2008: Paul and Harriet Campbell Distinguished Graduate Fellowship, The Pennsylvania State University

Publications

1. Rabaa MA, Klungthong C, Yoon I-K, Holmes EC, Chinnawirotpisan P, Thaisomboonsuk B, Srikiatkhachorn A, et al. (2012) Frequent in-migration and highly focal transmission of dengue viruses among children in Kamphaeng Phet, Thailand. PLoS Negl Trop Dis. (Accepted) 2. Barclay VC, Sim D, Chan BHK, Nell L, Rabaa MA, Bell AS, Anders RF, Read AF. (2012) The evolutionary consequences of blood-stage vaccination on the rodent malaria Plasmodium chabaudi. PLoS Biol. 10: e1001368. 3. Tra My PV, Rabaa MA, Vinh H, Holmes EC, Hoang NV, Vinh NT, Phuong le T, Tham NT, Bay PV, Campbell JI, Farrar J, Baker S. (2011) The emergence of rotavirus G12 and the prevalence of enteric viruses in hospitalized pediatric diarrheal patients in southern Vietnam. Am J Trop Med Hyg. 85:768- 75. 4. Rabaa MA, Hang VTT, Wills B, Farrar J, Simmons CP, Holmes EC. (2010) Phylogeography of recently emerged DENV-2 in southern Viet Nam. PLoS Negl Trop Dis. 4:e766. 5. Miller MA, Viboud C, Olson DR, Grais RF, Rabaa MA, Simonsen L. (2008) Measuring years of life lost for prioritization of influenza pandemic vaccine: who should be vaccinated first? J Infect Dis. 198:305-11. 6. Miller MA, Sentz J, Rabaa MA, Mintz ED. (2008) Global epidemiology of infections due to Shigella, Salmonella serotype Typhi, and enterotoxigenic Escherichia coli. Epidemiol and Infect. 136:433-5.