CHIKV is currently causing one of the largest chikungunya fever outbreaks reported in the past 40 y [
]. The magnitude of the epidemics has surprised the population, policy makers, and public health specialists, although chikungunya is not uncommon in nearby regions of this part of the world. Our phylogenetic analyses based on partial glycoprotein E1 sequences indicate that the Indian Ocean outbreak was caused by the same strain on Réunion, Seychelles, Mayotte, Madagascar, and Mauritius islands, and show that the outbreak strain is related to East-, Central-, and South-African isolates. Although, to our knowledge, no outbreak was reported recently in East Africa, this scenario is compatible with the human population exchanges between East Africa and Comoros, where the outbreak is believed to have started. Whether the epidemic strain results from the evolution of an enzootic strain, as previously described for eastern equine encephalitis virus [
], cannot be excluded. Sequencing of additional isolates from recent outbreaks in Africa and from possible local reservoirs should define more precisely the origin of the Indian Ocean outbreak.
This study represents the first survey, to our knowledge, of intra-outbreak analysis of CHIKV nucleotide variation on a genomic scale. The availability of multiple nearly complete genome sequences allowed us to deduce the probable history of successive molecular evolutionary changes that may have occurred while the outbreak was still ongoing. This evolutionary scenario is the most likely based on the six consensus sequences obtained, although additional isolates and determination of quasispecies heterogeneity are needed to obtain a more precise picture of viral evolution during the outbreak.
Whereas E1-226A was the only genotype observed during the first period of the Réunion outbreak (March–June 2005), our data show the emergence and predominance of genotype E1-226V, which was observed from the beginning of September 2005 and experienced a spectacular rise in frequency. The appearance of E1-226V preceded by at least 3 mo the explosive epidemic peak of mid-December 2005, and the link between this aa change and the rate of transmission thus deserves further investigation. Interestingly, a mutation at residue 226 in SFV was observed to release the cholesterol dependence of the virus [
]. It is thus possible that such a mutation provides a selective advantage to the virus in mosquitoes, which are cholesterol auxotrophs. Indeed, in the SFV E1 crystal structure, position 226 is located in the ij loop, in contact with the fusion peptide, in a region of the protein that is predicted to interact with the target membrane. On the other hand, this change may be selectively neutral, and other evolutionary factors such as genetic drift or a founder effect could have favoured E1-226V by chance alone. In Sindbis virus, a change at position 226 alone was not sufficient to release cholesterol dependence [
]. Our data thus point to interesting experiments comparing the cholesterol dependence of the isolates differing at position 226 in E1.
To date, only CHIKV laboratory strains, passaged many times on mosquito or mammalian cells, had been entirely sequenced [
]. We provide for the first time nearly complete nucleotide sequences of six clinical isolates passaged in vitro only once or twice (see
). Limiting the number of passages is crucial because the infecting viral population may correspond to a quasispecies [
], i.e., a mixed viral population with genotypes co-existing in an equilibrium governed by a balance between mutation and natural selection. Repeated in vitro passages may act as a filter on this population. For example, the presence in S27 of an Arg codon instead of the opal stop codon in Indian Ocean isolates is probably explained by numerous in vitro passages of S27, as evolution of opal to Arg was observed experimentally in ONN viruses [
]. Whereas it may be advantageous for viral quasispecies to maintain the opal codon in vivo, an Arg codon may confer a selective advantage in vitro, as observed for the closely related SFV [
]. In the present study, careful inspection of the chromatograms traces identified three codons with double peaks observed both on forward and reverse traces (
). In all three cases, alternative bases corresponded to non-synonymous codons that may thus alter viral fitness. It is tempting to speculate that quasispecies diversity in vivo might facilitate the access to different body sites, such as the central nervous system. For example, selection for a subset of genotypes harbouring the changes observed in CSF isolate 06.27 may be associated with invasion of the CSF [
]. These observations underscore that the genome sequence of laboratory “reference” strains may not accurately reflect the natural situation, as the genotypic complexity of quasispecies in vivo is subject to erosion by in vitro selection. Since the Indian Ocean isolates sequenced here were subjected to in vitro selection for only a few generations, they probably correspond more closely to the in vivo genotypes than previously sequenced chikungunya strains.
The aa differences detected among the outbreak isolates may relate to biological or pathogenic characteristics of the virus. Although our viral culture results are preliminary, they clearly show phenotypic differences between the unique isolate from CSF (06.27), isolated from a neonatal encephalopathy case, and three other isolates from sera associated with either the classical form of the disease or encephalopathy. The larger foci observed in culture with 06.27 could reflect a higher replication rate of the virus and be linked to the specific aa changes identified in nsP1, nsP2, and nsP3. Single aa changes in nsP1, including a Thr/Ile change (residue 538 of Sindbis virus) [
] and a 18-nt deletion in nsP3 have previously been shown to affect neurovirulence in other alphaviruses [
]. However, in the absence of nsP1, nsP2, and nsP3 structural data, it is difficult to predict whether the specific aa changes observed in isolate 06.27 can have a structural or functional impact. We also noted that all the viral sequences determined from either the serum or the isolates from three neonatal encephalopathy cases and an adult meningo-encephalitis case had E1-226V. However, as this genotype is also observed in classical forms of the disease, one cannot conclude on a potential link of E1-226V with neuropathogenesis. Host factors also have to be considered in the occurrence of neurological forms of the disease. For example, the blood–brain crossing may be favoured by age or hypertension.
Molecular signatures of the Indian Ocean outbreak genomes were identified when they were compared to all other reported alphavirus sequences. However, these comparisons have to be considered with caution because of a potential sampling bias due to the small number of previous alphavirus sequences. Nevertheless, these features represent interesting targets for future functional studies, as well as for epidemiological follow-up. One particularly interesting feature was the E1-226V residue (see above). Another interesting molecular signature was E1–284 Asp. Although the pseudo-atomic model of the scaffold used is of modest resolution (the resolution of the crystal structure is limited—approaching 3Å—and the model results of fitting this structure into a 9Å resolution cryo-electron microscopy reconstruction), it appears that the side-chain of Asp 284 interacts with the main chain of an adjacent E1 polypeptide in the virion. Indeed, it is in a position compatible with acceptance of a hydrogen bond from a main chain amide in the contacting E1 protomer. Because the packing is very tight (see
B), it is possible that the longer glutamic acid side chain (which has an extra CH
2 group compared to Asp or Asn) may introduce a slight distortion at the contact sites, an effect that is propagated by the icosahedral T = 4 symmetry of the virion. Thus, a cooperative effect due to this change at position Asp 284 may play a role in either allowing a less efficient assembly of new particles in infected cells, or a more efficient particle disassembly process during invasion of a new cell, or a combination of both. This information can guide new site-directed mutagenesis studies, using reverse genetics, to test the effect of the Asp/Glu replacement on the virus cycle.
The magnitude and high political profile of the outbreak has underlined the critical lack of knowledge on the biology of CHIKV, contrasting with related model alphaviruses such as Sindbis, Semliki Forest, and Ross River. This situation reflects the fact that chikungunya infection, despite infecting millions of people since its discovery, has been neglected. However, chikungunya disease is clearly responsible for disabling and persistent arthralgia, although it remains unresolved whether the symptoms are due to persistence of the virus or inappropriate immune response [
]. In addition, during the Indian Ocean CHIKV outbreak, a small proportion of the patients (about 123 out of 244,000 infected) developed severe clinical signs such as neurological signs or hepatitis. Whereas neurovirulence and neuroinvasiveness are established for several alphaviruses such as eastern equine encephalitis virus and Venezuelan equine encephalitis virus, only two CHIKV strains had previously been isolated from children with clinical signs suggestive of encephalitis and meningitis [
]. Whether CHIKV Indian Ocean strains have acquired a higher neurovirulence or neuroinvasiveness certainly deserves investigation.
In the absence of efficient vaccine or antiviral therapy, vector control is at present the only way to limit chikungunya transmission. However, the broad geographic distribution of the mosquito vectors
] may allow the expansion of CHIKV to new areas, such as the European or American continents. The molecular data reported here on clinical isolates from the current epidemics should contribute to bridge the gap of knowledge concerning this human pathogen, and will help to provide, in the middle term, more specific and powerful tools to combat it.