Long-Term Transmission of Defective RNA Viruses in Humans and Aedes Mosquitoes

See allHide authors and affiliations

Science  13 Jan 2006:
Vol. 311, Issue 5758, pp. 236-238
DOI: 10.1126/science.1115030


In 2001, dengue virus type 1 (DENV-1) populations in humans and mosquitoes from Myanmar acquired a stop-codon mutation in the surface envelope (E) protein gene. Within a year, this stop-codon strain had spread to all individuals sampled. The presence of truncated E protein species within individual viral populations, along with a general relaxation in selective constraint, indicated that the stop-codon strain represents a defective lineage of DENV-1. We propose that such long-term transmission of defective RNA viruses in nature was achieved through complementation by coinfection of host cells with functional viruses.

Many RNA virus populations show high levels of genetic diversity because of the error-prone nature of their replication with RNA-dependent RNA polymerase and their large population sizes (1). However, most studies of genetic diversity in RNA viruses have considered consensus nucleotide sequences, which represent a population average of the myriad of diverse viral genomes within an infected host. The exceptions tend to be longitudinal studies of viruses generating chronic infections, such as the human immunodeficiency virus (2) and the hepatitis C virus (3). There have been few equivalent studies of RNA viruses causing acute infections, although they provide an opportunity to determine how population genetic diversity is shaped by rapidly alternating cycles of intra- and interhost evolution. Flaviviruses, such as dengue virus (DENV), constitute an ideal model system because they have distinct human and mosquito hosts and because transmission rates vary dramatically among “wet” and “dry” seasons as a result of the availability of breeding sites for the Aedes mosquito vectors. In areas where dengue is hyperendemic and three or four DENV serotypes may cocirculate, outbreaks occur in cycles of 3 to 5 years, often with changes in the relative proportions of each virus serotype sampled (4, 5).

To reveal the evolution of RNA virus populations within and among hosts, we examined DENV-1 recovered from patients at the Yangon Children's Hospital in Myanmar between August 2000 and December 2002, as well as from mosquitoes sampled at the same time (5, 6). This sampling period covered a year (2000) in which relatively few cases of dengue occurred (1229 in Yangon) and two dengue “seasons” (2001, 2002) during which the largest numbers of dengue cases on record were reported in Yangon (7383 and 5185 cases, respectively). In the 2001 season, 95% of the viruses recovered were DENV-1, and most patients had primary infections and mild disease (5). In contrast, all four DENV serotypes were recovered from patients in 2002 in the following proportions: DENV-1, 43%; DENV-2, 35%; DENV-3, 4%; and DENV-4, 18%. 2002 also was characterized by a higher incidence of patients with secondary infections and more severe disease.

We sequenced 20 clones corresponding to the viral envelope (E) glycoprotein genes of DENV-1 populations from 14 individuals—11 from sera from dengue patients and 3 from isolates from individual mosquitoes— at a single time point during their infection. Ten clones from an additional mosquito also were sequenced as described previously (7). The number of variable nucleotide sites within individual viral populations ranged from 47 to 162, the average pairwise genetic diversity within each sample was approximately 1%, and the average ratio of nonsynonymous to synonymous substitutions (dN/dS) was 0.565 (Table 1) (8). Despite such extensive diversity, the majority (mean 77%) of the mutations occurred only once within an individual, and overall consensus sequences were more conserved; three pairs of individuals had identical consensus sequences, and the total number of variable sites among consensus sequences was 76. This was reduced to only 23 if the divergent population from patient D1.Myanmar.43826/01 was excluded. Such contrasting patterns of genetic variation within and among individuals suggest that the majority of the mutations in DENV are deleterious and eventually removed by purifying selection, resulting in consensus sequences that are relatively stable in the long term and in lower dN/dS ratios than observed at the intrahost level (9, 10). Further evidence that these data represent genetic diversity before the action of widespread natural selection was that stop codons were detected in at least one clone from 13 of the 15 virus populations, mostly the result of frame shifts after single-nucleotide deletions.

Table 1.

Characteristics of DENV-1 populations. Virus populations sampled from individual mosquitoes are shown in italics. DF, dengue fever; DHF, dengue hemorrhagic fever; DSS, dengue shock syndrome; S, number of segregating sites; π, mean pairwise genetic diversity; dN/dS, mean number of nonsynonymous/synonymous substitutions per site.

Virus populationDate isolated (mm/yyyy)Day of symptomsInfection/diseaseNo. clonesNo. clones with stop codonSUnique mutations (%)πdN/dS
D1.Myanmar.Mos059/01 06/2001 20 0 83 58 (70) 0.010 0.523
D1.Myanmar.Mos194/01 06/2001 20 10 162 141 (87) 0.015 0.690
D1.Myanmar.Mos206/01 06/2001 10 6 79 13 (16) 0.023 0.121
D1.Myanmar.Mos305/01 06/2001 20 10 85 69 (81) 0.009 0.855
D1.Myanmar.37045/00 08/2000 4 Primary/DHFII 20 0 54 52 (96) 0.004 0.464
D1.Myanmar.38518/01 05/2001 5 Primary/? 20 0 143 115 (80) 0.013 0.265
D1.Myanmar.40530/01 07/2001 2 Primary/DHFI 20 1 128 106 (83) 0.012 0.233
D1.Myanmar.40906/01 07/2001 5 Primary/DF 20 0 70 44 (63) 0.008 0.558
D1.Myanmar.43549/01 09/2001 2 Primary/DHFII 20 0 51 46 (90) 0.004 0.846
D1.Myanmar.43826/01 10/2001 2 Primary/DHFI 20 0 54 53 (98) 0.004 0.635
D1.Myanmar.47185/02 07/2002 4 Primary/DSS 20 9 87 67 (77) 0.009 0.631
D1.Myanmar.47662/02 07/2002 2 Primary/DF 20 1 78 65 (83) 0.007 0.345
D1.Myanmar.48572/02 08/2002 2 Primary/DF 20 11 47 34 (72) 0.006 1.107
D1.Myanmar.49440/02 09/2002 4 Primary/DF 20 11 89 70 (79) 0.009 0.624
D1.Myanmar.50457/02 12/2002 3 Secondary/DF 20 9 112 86 (77) 0.012 0.582
Mean 88 68 (77) 0.001 0.565

However, the most striking feature of these data was the high frequency of a stop codon at amino acid E 248 (nucleotide 742), caused by a C → U transition. This stop-codon mutation was found in 68 of the 290 sequences (23%), representing 9 of the 15 individuals (6 human, 3 mosquito). It was first detected in three of four virus populations from mosquitoes collected in June 2001 as the outbreak in that year began, and then in one of the five virus populations recovered from patients during the 2001 dengue season. However, all five virus populations sampled in 2002 contained viruses with the stop codon at E 248. Moreover, in three samples—D1.Myanmar.Mos206/01, D1.Myanmar.48572/02, and D1.Myanmar. 49440/02—the stop codon was the most common form.

To determine whether this stop-codon mutation arose independently in each patient, such that it represents a hot spot for mutation or recombination, or was transmitted among patients, we conducted a phylogenetic analysis on all 290 DENV-1 sequences (Fig. 1) (11). This revealed that the stop-codon sequences formed a single phylogenetic group that was strongly associated with seven other amino acid changes at E 242, 269, 316, 363, 373, 417, and 458, and no evidence of recombination. Hence, the stop-codon variant represents a distinct lineage of DENV-1 that has been transmitted among humans and mosquitoes for at least 18 months. To determine the age of each amino acid mutation associated with the stop-codon lineage, we mapped their occurrence onto the phylogenetic tree using parsimony (12). The eight amino acid changes that defined the stop-codon lineage fell on the main trunk of the tree, with most synonymous mutations specific to individual viruses. The first amino acid changes to appear (those that fall deepest in the tree) were the stop-codon mutation at E 248 and the only change 5′ to it, at E 242 (Fig. 1).The remaining six amino acid changes were acquired sequentially from the stop codon toward the carboxy terminal of the E protein; that is, in the order E 269, E 316, E 363 and E 373 together, E 417, and finally E 458. Our phylogenetic analysis also revealed the existence of a second distinct viral lineage, defined by seven amino acid changes. This comprised 25 sequences from four individuals, including all those from patient D1.Myanmar.43826/01, which explains why this individual has such a divergent consensus sequence.

Fig. 1.

Maximum-likelihood phylogenetic tree depicting the evolutionary relationships among 290 DENV-1 E gene sequences sampled from 15 patients and mosquitoes in Myanmar from 2000 to 2002. The stop codon (shaded) and the second-variant lineages are indicated. The phylogenetic locations of the mutations associated with the stop-codon lineage are also indicated, all of which result in amino acid changes. 1 = E 242 (Thr → Ser); 2 = E 248 (Gln → Stop); 3 = E 269 (Glu → Gly); 4 = E 316 (Gln → Arg); 5 = E 363 (Lys → Asn); 6 = E 373 (Ser → Phe); 7 = E 417 (Asp → Gly); 8 = E 458 (Gly → Glu). All horizontal branch lengths are drawn to a scale of nucleotide substitutions per site.

There are two possible explanations for the long-term transmission of the stop-codon lineage of DENV-1: (i) that “read-through” of a stop codon, as occurs in some alphaviruses (13), has allowed the production of a full-length and functional E protein, or (ii) that it represents a defective strain of DENV-1 that is transmitted through the population by means of a complementation mechanism. To test these hypotheses, we determined the number of E protein species within viral samples containing the stop-codon variant (Fig. 2) (14). Mosquito cells infected with the prototype strain of DENV-1 (HAW), which has a functional codon at E 248, contained a single species of E protein with a molecular weight of 54 kD, which is approximately that predicted from its complete amino acid sequence. In contrast, cells infected with viruses from D1.Myanmar.48572/02, characterized by a mixture of wild-type and stop codons at E 248, contained two forms of E protein. One resembled the full-length E, whereas a smaller band had a molecular weight of 27 kD, that predicted for the amino-terminal region of the E protein up to a stop codon at E 248. The stop-codon mutation therefore generates a prematurely truncated and hence nonviable E protein.

Fig. 2.

Indirect ELISA with a monoclonal antibody (M17) to DENV-1 E protein on Western blots of lysates of Aedes albopictus cells (C6-36) infected with the prototype strain DENV-1 HAW or a Myanmar population of DENV-1, which contains a stop codon at E248.

Additional evidence that the stop-codon lineage comprised defective viruses was supplied by an analysis of selection pressures (8) and substitution patterns. Most notably, there was about a twofold increase in dN/dS in the region 3′ to the stop codon in the stop-codon lineage, and the dN/dS value observed, 1.020, was exactly that expected under strictly neutral evolution (Table 2). A similar increase in dN/dS was observed 5′ to the stop codon in the stop-codon lineage. This suggests that selection pressures in this region have also been relaxed, even though protein translation seems to proceed normally. There was no evidence for positive selection in either gene region (15). Further, all eight amino acid changes that characterized the stop-codon lineage were nonconservative and occurred at sites that were invariant among all available DENV-1 E gene sequences sampled from localities other than Myanmar (n = 155), which suggests that they have a major effect on fitness. Similarly, these mutations were not predicted to have any effect on RNA secondary structure, either on the positive or the negative strands, and frequent stop codons were observed in alternate reading frames. Hence, the stop-codon lineage is associated with the stepwise accumulation of amino acid changes by mutation-drift that would most likely be deleterious in a fully functional virus. However, the underlying cause of the sequential ordering of substitutions in the stop-codon lineage, chance or otherwise, is unknown.

Table 2.

Numbers of nonsynonymous and synonymous substitutions per site (dN/dS) for different regions and lineages of DENV-1. Because of the large size of the wild-type DENV-1 lineage, this analysis was conducted on a random sample of 66 sequences from 14 individuals.

Lineage (n)5′ to stop (247 amino acids)3′ to stop (247 amino acids)
Wild type (66) 0.234 0.494
Stop codon (68) 0.586 1.020

These results have important implications for the evolutionary dynamics of RNA viruses. First, although the consensus nucleotide sequence from each infected individual is similar, a number of distinct lineages have been stably transmitted among human and mosquito populations. Hence, the consensus sequence paints a misleading picture of population genetic diversity, which may have a complex genealogical history. More striking was the observation that an evolutionary lineage that is associated with defective viruses due to the acquisition of a stop codon has been transmitted for 18 months among humans and mosquitoes and through periods of relatively high and low DENV prevalence. The most likely mechanism for such sustained transmission is complementation, in which defective genomes are rescued through the parasitism of functional proteins from wild-type viruses. Although there is growing evidence for the importance of complementation in RNA viruses (1619), that defective viruses can be retained through so many cycles of transmission in nature is unexpected. Indeed, DENV-1 recovered from patients in New Caledonia in February 2003 and in Singapore in August 2004 also contained the stop codon at E248, indicating that this defective strain has entered transmission cycles thousands of kilometers apart. Such long-term complementation not only implies that multiple infections of the same cell are commonplace (18) but that long-term population sizes are large enough to prevent severe bottlenecks at transmission. That at least three lineages of DENV-1 have been transmitted among humans and mosquitoes in Myanmar also argues against the occurrence of widespread population bottlenecks.

It seems counterintuitive that mutations that generate intragenic stop codons or that disrupt the structure or function of the E protein (for example, the disruption of a disulfide bridge by a Cys-Arg change at E 121) should reach such high frequency within viral populations. It is theoretically possible that the transmission of defective lineages is selectively advantageous for co-infecting wild-type viruses by ensuring the presence of viruses differentially adapted to human or mosquito cells, decoying host immune responses, or allowing the production of extra capsid and membrane proteins. However, we observed a general relaxation in selective constraint in the stop-codon lineage, and in most evolutionary models complementation has an adverse effect on fitness because deleterious mutations are allowed to persist in the population (20). In this context, it is noteworthy that the increased frequency of the stop-codon strain was concomitant with a major reduction in DENV-1 prevalence in Myanmar. Whether this was a cause, a consequence, or coincident with these larger scale epidemiological patterns is unknown, although “hyperparasites,” as perhaps represented by the stop-codon strain of DENV-1, are predicted to have important effects on pathogen transmission and virulence (21).

References and Notes

View Abstract

Navigate This Article