Many investigators have used viral nucleotide sequence data and phylogenetic methods to understand genetic relationships between viruses, as well as the epidemiology of viral disease. Phylogenetic studies have shown that dengue viruses can move long distances between continents (24
) as well as short distances between neighboring countries (25
). Our goal was to use a phylogenetic approach to understand recent DHF outbreaks caused by DENV-3 infections in the Indian subcontinent and Latin America.
Previous phylogenetic analysis of DENV-3 has principally relied on complete or partial sequences of the pre-M/M and E genes (13
). Our analysis used a 708-base segment, positions 437–1,145, spanning pre-M/M and a portion of the E gene, coding for 236 amino acids. This region was selected because it both conserved the original phylogenetic relationship identified by Lanciotti et al. and, in preliminary analysis with previously established sequences, captured 44% of the variable sites within DENV-3, subtype III Sri Lankan sequences. No insertion/deletion mutations and no hypervariable regions were detected in this span.
A total of 40 DENV-3 sequences, including 21 sequences available from GenBank and 19 newly determined Indian subcontinent and African sequences () were compared. Dates of isolation ranged from 1963 to 1998. With the exception of 63PuertoR, all sequences were from low-passage (<4) virus cultures. Several approaches to phylogenetic analysis, including maximum likelihood, parsimony, and distance methods, were compared. All approaches yielded identical or nearly identical topologies. Results presented here used the Tamura-Nei algorithm to calculate genetic distances and the minimum evolution method to create the trees (). The tree identifies four distinct lineages that correspond to the region of isolation, reproducing the same evolutionary relationship first described by Lanciotti, et al. (23
). Subtype I includes isolates from Southeast Asia and the South Pacific islands; subtype II consists of isolates from Thailand; subtype III is comprised of isolates from the Indian subcontinent, East Africa, and a single isolate from Samoa; and subtype IV includes Puerto Rico and Tahiti. Similarity within subtypes was high, with subtype III showing the greatest mean similarity (98.4%), followed by subtypes I, II, and IV ().
Summary of within- and between-subtype nucleotide mean similarity for dengue virus 3 isolates shown in a
All 24 Sri Lankan, Indian, and East African strains fell into subtype III (). The circulating virus genotypes within this region have remained closely related over the relatively long period of 18 years (1981–1998), indicating that countries bordering the western Indian Ocean form a geographically distinct region with regard to DENV-3 viruses. DENV-2 viruses in the regions also form a subtype with a similar geographic distribution (26
). Frequent trade between East Africa, Western Indian Ocean islands, and the Indian subcontinent may have been responsible for the movement of dengue viruses throughout the region (26,28
). Rico-Hesse, for example, demonstrated the introduction of DENV-2 to Africa from islands in the Indian Ocean (26). The earliest subtype III virus on record is an isolate from India in 1966; this virus occupies a node that is ancestral to all the subsequent Asian and African isolates (R.S. Lanciotti, pers. comm.), suggesting that the DENV-3, subtype III viruses have their origin in the Indian subcontinent and have subsequently spread out of the region.
In Sri Lanka, regular epidemics of DHF have been observed only since 1988. DENV-3 is responsible for many of the infections that progress to DHF (20
). DENV-3 isolates obtained before and after the emergence of DHF are very closely related and belong to subtype III, indicating that the emergence of DHF on the island is not due to the introduction of a new subtype from outside the region. However, within subtype III, most Sri Lankan isolates (except for 93SriLan1) from before and after the emergence of DHF segregated into two distinct clades, designated groups A and B (). Group A, with nine isolates from 1981 to 1989, consists of viruses collected up to the year epidemic DHF emerged in Sri Lanka but contains no isolates from later than 1989. Group B includes eight isolates from 1989 to 1998 but none from before 1989. Temporally, the two groups are continuous, by virtue of sharing isolates in 1989. Group A includes isolate 89SriLan2, while group B contains 89SriLan1, 89SriLan3, and 89SriLanA. However, the groups do not form a continuous lineage; they share a common ancestor only at the node for subtype III (). Group B shares ancestral nodes with isolates from India and East Africa. Because the Indian and East African isolates overlap temporally with group A (all isolates are from the 1980s), group A and group B lineages likely diverged sometime before 1981 and followed distinct evolutionary pathways.
We propose two likely scenarios that led to the emergence of group B viruses in Sri Lanka. One possibility is that the group B viruses were introduced from India or East Africa into Sri Lanka because Indian and East African isolates from the mid-1980s are closely related to Sri Lankan group B viruses (). Of the two regions, India is the more likely source because of geographic proximity to Sri Lanka, although the East African viruses could be the direct ancestors of the group B viruses. Another possibility is that both groups co-circulated in Sri Lanka in the early 1980s, with group B being a minor population. Some selective force operating in the late 1980s may have shifted the balance in favor of group B viruses. In either case, group B viruses emerged in Sri Lanka because a subtype III variant already established in the greater region became more common in Sri Lanka and not because a novel virus evolved and emerged de novo on the island.
DENV-3, subtype III was detected in the Americas during DF and DHF outbreaks in Nicaragua and Panama in 1994 (11
). Subsequently, the virus has spread to many countries in Latin America, and DENV-3–associated DHF was confirmed in several countries (13
) (). To establish the relationship of recent Latin American DENV-3 isolates to each other and to the previously identified Indian subcontinent and East African subtype III isolates, we sequenced and analyzed a 966-base region spanning positions 179–1,145 on the viral genome, capturing a portion of the C gene, all of pre-M/M gene, and a portion of the E gene. This region adds 288 positions to the 5′ end of the sequences initially presented in this study.
Figure 3 Map of the spread of dengue virus 3 (DENV-3), subtype III through Latin America and the Caribbean. The introduction of DENV-3, subtype III was first reported in November 1994 in Nicaragua and Panama. This virus strain has been isolated, identified, and (more ...)
Forty-three isolates were sequenced (21 from Mexico and Central and South America, 16 from Sri Lanka, 1 from India, and 5 from East Africa) (). The D3H-87 belonging to DENV-3, subtype 1 was used as an outgroup. Years of isolation ranged from 1983 to 2001, an 18-year span. Except for the D3H-87 outgroup, all other isolates were low-passage clinical isolates. Most of the nucleotide mutations were silent: only 12 amino acid positions showed any variability and only 2 positions showed variability in more than one isolate. Consequently, the evolutionary relationships observed in this analysis likely reflect the results of genetic drift and are unlikely to have been influenced by host-specific selection events on this portion of the genome (26
All sequences included in this analysis fell within subtype III (data not shown). Several approaches to phylogenetic analysis were compared, and all approaches yielded identical or nearly identical topologies. We used the Tamura-Nei algorithm to calculate genetic distances and the minimum evolution method to create the trees (). Bootstrap values are shown at critical nodes. Despite the high overall similarity of the isolates in this analysis, geographically and temporally distinct groups formed separate lineages. Generally, two separate lineages formed within subtype III. The first consists of group A viruses isolated from 1981 to 1989 in Sri Lanka. These viruses have been associated only with DF. The second is composed of Sri Lankan group B, Indian, East African, and all of the isolates from Mexico and Central and South America.
Within group A, members are closely related, with a nucleotide mean similarity of 99.4% (). Within the expanded group B and related viruses, three distinct clades exist: a group of closely related Sri Lankan isolates from 1989 to 1998, 5 East African isolates from 1985 to 1993, and 21 isolates from 1994 to 2001 from Latin America. Isolates 84India and 93SriLan1 are less closely related to the other geographically distinct isolates in the larger second lineage.
Summary of within- and between-group nucleotide mean similarity for the dengue virus 3, subtype III virus isolates shown in a
The isolates from Latin America all emerge from a common node on the tree, suggesting a single introduction of a virus and the subsequent diversification of the virus population from the founding strain. The DENV-3, subtype III isolates from Nicaragua, Panama, and Costa Rica are closest to the Latin American group’s originating node, with the more recent isolates found farther from that node, reflecting the viral population’s ongoing evolution after the point source introduction.
The internal branch from the Latin American group shares a common node with the isolates from East Africa. The common hypothetical ancestor for Latin America and East Africa then shares a common node with the Sri Lankan group B virus isolates. Both on the phylogenetic tree and in pair-wise comparisons (), the Latin American group was more closely related to the isolates from East Africa than to the group B Sri Lankan isolates. Furthermore, the East African isolates pre-date the earliest Latin American isolates by 9 years, while the less closely related Sri Lankan group B isolates are nearly contemporaneous with the Latin American isolates. Therefore, the point source DENV-3 introduction into Latin America is most likely to have its origins in East Africa and not the Indian subcontinent ().
Figure 4 Global spread of dengue virus 3 (DENV-3), subtype III, which has been continuously circulating in the Indian subcontinent from the 1960s to the present. The virus was first isolated from East Africa in 1985 in Mozambique and subsequently from Kenya (1991) (more ...)
Little is known about dengue activity in Africa, particularly DENV-3 (32
). DENV-3 was first detected on the African continent in 1984 to 1985 during an outbreak in Mozambique (32
). Later studies of U.S. troops in Africa and the Persian Gulf suggested that DENV-3 is endemic in those regions but largely undetected (33
). Our results show that all East African DENV-3 isolates belong to subtype III. The fact that DENV-3 was only first isolated from East Africa in 1985, whereas the viruses were present in the Indian subcontinent at least as far back as 1966 (R.S. Lanciotti, pers. comm.), suggests that DENV-3, subtype III was introduced from the Indian subcontinent into East Africa in or before 1984 (). This introduction led to the establishment of a stable East African group of DENV-3, subtype III because all the isolates from Mozambique, Kenya, and Somalia isolated from 1985 to 1993 form a distinct clade within subtype III ().
The DENV-3, subtype III viruses introduced into Latin America are most closely related to subtype III viruses in East Africa (). Although we can only speculate about the exact mode of transport of DENV-3 into Latin America, we propose that Panama, with its canal that attracts goods as well as civilians and military personnel from other parts of the world, may have been the point of introduction of subtype III into the Americas. Similarly, the introduction of DENV-2 in 1981 into Cuba may be attributable to Cuban military personnel traveling between Southeast Asia and Cuba (24
Epidemiologic and clinical studies on dengue in Indonesia in the 1970s pointed to strain differences between DENV-3 viruses contributing to transmission and disease severity (35
). Despite their overall similarity at the nucleotide level, the DENV-3, subtype III isolates examined in this study have been associated with severe or mild disease outbreaks (). Sri Lankan group A viruses were isolated during a time of little to no DHF, while group B viruses were isolated after the emergence of DHF in Sri Lanka. The emergence of DHF in Sri Lanka was not accompanied by a change in dengue transmission or the abundance of any particular serotype (21
). Implicating DENV-3 directly as the cause of DHF in Sri Lanka has been difficult because few virus isolates are available from DHF patients in Sri Lanka. However, during dengue surveillance studies in 1997, only DENV-3 was isolated from hospitalized dengue cases, whereas DENV-1, DENV-2, and DENV-3 were isolated from patients visiting outpatient clinics (22
). These observations suggest that DENV-3 is responsible for severe dengue disease in Sri Lanka. Further studies are required to better establish the relative contribution of DENV-3 to severe disease in Sri Lanka.
The current studies support a viral genetic basis for severe and mild disease outbreaks. We found that the population of DENV-3 viruses associated with DHF in Sri Lanka did not appear to be direct descendants of the group A viruses that were circulating before DHF emerged in that country. The Sri Lankan 1989–1997 isolates are more closely related to the isolates from East Africa and the isolates from the Americas than they are to the isolates from 1981 to 1989 in Sri Lanka (). All three groups of subtype III viruses (Sri Lankan group B, East African group, and Latin American group) associated with DHF are more closely related to each other than they are to the pre–DHF group A viruses from Sri Lanka (). Thus, all the viruses within subtype III are closely related (mean 98.4% identity at the nucleotide level), yet they form distinct phylogenetic groups associated with mild or severe disease.
The Sri Lankan group B viruses may be associated with severe disease unlike group A viruses because the group B viruses are inherently more virulent. Alternatively, the ability of preexisting dengue antibody to neutralize group A viruses and enhance group B viruses may account for the observed associations with severe and mild disease. In a recent study, antibodies against American DENV-1 viruses neutralized the Native American DENV-2 genotype better than the Southeast Asian DENV-2 genotype that is currently circulating in the Americas and causing DHF (37
). This study lends support to the idea that Asian DENV-2 may produce a more severe disease not because of inherent virulence properties but because persons with previous primary DENV-1 infections may enhance infection with this genotype and neutralize infections with the Native American DENV-2 genotype. Similarly, DENV-2 and -3 are the common serotypes in Sri Lanka, and persons with previous primary DENV-2 infections could neutralize the DENV-3 group A viruses better than the group B viruses. This difference may explain the unexpected emergence of DHF associated with group B. Further comparative studies with group A and B viruses are needed to understand their association with mild and severe disease, respectively.