Epidemic dynamics revealed in dengue evolution

Abstract

Dengue is an emerging tropical disease infecting tens of millions of people annually. A febrile illness with potentially severe hemorrhagic manifestations, dengue is caused by mosquito-borne viruses (DENV-1 to -4) that are maintained in endemic transmission in large urban centers of the tropics with periodic epidemic cycles at 3- to 5-year intervals. Puerto Rico (PR), a major population center in the Caribbean, has experienced increasingly severe epidemics since multiple dengue serotypes were introduced beginning in the late 1970s. We document the phylodynamics of DENV-4 between 1981 and 1998, a period of dramatic ecological expansion during which evolutionary change also occurs. The timescale of viral evolution is sufficiently short that viral transmission dynamics can be elucidated from genetic diversity data. Specifically, by combining virus sequence data with confirmed case counts in PR over these two decades, we show that the pattern of cyclic epidemics is strongly correlated with coalescent estimates of effective population size that have been estimated from sampled virus sequences using Bayesian Markov Chain Monte Carlo methods. Thus, we show that the observed epidemiologic dynamics are correlated with similar fluctuations in diversity, including severe interepidemic reductions in genetic diversity compatible with population bottlenecks that may greatly impact DENV evolutionary dynamics. Mean effective population sizes based on genetic data appear to increase prior to isolation counts, suggesting a potential bias in the latter and justifying more active surveillance of DENV activity. Our analysis explicitly integrates epidemiologic and sequence data in a joint model that could be used to further explore transmission models of infectious disease.

FIG. 1.

Endemic and epidemic periods of dengue virus in PR. (A) Number of suspected cases of DF/DHF by year reported to the CDC-Dengue Brach between 1981 and 1998. (B) Percentage of identifications of each serotype relative to the total of positive serotype identifications by tissue culture isolation per year. Numbers in parentheses indicate the numbers of DENV-4 isolations made at the CDC San Juan, Dengue Branch each year. DENV-1 (light gray, bottom), DENV-2 (white), DENV-3 (black), and DENV-4 (dark gray, top). Isolates were made on a subset of reported cases, and the two estimates of dengue population size (suspected and confirmed cases) are strongly correlated. Hyperendemic transmission is demonstrated by the occurrence of multiple cocirculating serotypes at a given time. All four serotypes of dengue exhibit seasonal peaks in transmission as well as similar interannual but alternating periodicity in transmission; with the appearance of DENV-3 in 1998, all other serotypes underwent a substantial reduction in subsequent years (data not shown).

FIG. 2.

ML phylogeny of DENV-4 in PR from 1981 to 1998, throughout a turbulent transmission history. ML phylogeny based on concatenated structural (Capsid, Membrane, and Envelope) and nonstructural genes (NS1 partial, NS2A, and NS4B), and the 3′ NTR (n = 4,016), using the GTR + I + Γ model of evolution. Support given at nodes based on 100 ML replicates implemented in RAxML Black Box webserver (Stamatakis et al. 2008). Accession numbers of NCBI sequences used in this study: AY152036–AY152363 (Bennett et al. 2003) and GU318306-GU318318 (this study, shown in bold).

FIG. 3.

Effective population size estimates in terms of effective numbers of infections per month based on viral genetic diversity and coalescent patterns, superimposed on the number of DENV-4 isolates by month (A) directly and (B) with a 7-month upward adjustment of N_e (e.g., ahead in time).