Genetic Evidence for Rift Valley Fever Outbreaks in Madagascar Resulting from Virus Introductions from the East African Mainland rather than Enzootic Maintenance

ABSTRACT Rift Valley fever virus (RVFV), a mosquito-borne phlebovirus, has been detected in Madagascar since 1979, with occasional outbreaks. In 2008 to 2009, a large RVFV outbreak was detected in Malagasy livestock and humans during two successive rainy seasons. To determine whether cases were due to enzootic maintenance of the virus within Madagascar or to importation from the East African mainland, nine RVFV whole genomic sequences were generated for viruses from the 1991 and 2008 Malagasy outbreaks. Bayesian coalescent analyses of available whole S, M, and L segment sequences were used to estimate the time to the most recent common ancestor for the RVFVs. The 1979 Madagascar isolate shared a common ancestor with strains on the mainland around 1972. The 1991 Madagascar isolates were in a clade distinct from that of the 1979 isolate and shared a common ancestor around 1987. Finally, the 2008 Madagascar viruses were embedded within a large clade of RVFVs from the 2006–2007 outbreak in East Africa and shared a common ancestor around 2003 to 2004. These results suggest that the most recent Madagascar outbreak was caused by a virus likely arriving in the country some time between 2003 and 2008 and that this outbreak may be an extension of the 2006–2007 East African outbreak. Clustering of the Malagasy sequences into subclades indicates that the viruses have continued to evolve during their short-term circulation within the country. These data are consistent with the notion that RVFV outbreaks in Madagascar result not from emergence from enzootic cycles within the country but from recurrent virus introductions from the East African mainland.

Rift Valley fever virus (RVFV) is a mosquito-borne, negative-strand RNA virus in the family Bunyaviridae, genus Phlebovirus (24). RVFV readily infects livestock, including sheep, goats, and cattle, and causes mortality in young and adult animals, as well as sweeping abortion storms. Although infection in humans is most often self-limiting, 1% to 2% of cases result in more-severe disease, with clinical signs ranging from hepatitis to encephalitis to hemorrhagic syndrome. Transmission can occur as a result of contact with infected livestock or directly from the bite of infected mosquitoes.
RVFV was first isolated in 1931, in Kenya (6,11), but Bayesian coalescent analyses suggest that the most recent common ancestor (MRCA) of extant RVFVs can be traced back as far as the late 1800s (3,4). The virus can be found throughout continental Africa, Madagascar, and, more recently, in the Arabian Peninsula. Epizootic/epidemic cycles appear to be pe-riodic and are often associated with climatic variations, such as increased rainfall, which can be used to generate predictive risk maps for RVFV activity in East Africa (2,7,15). Elevated rainfall results in expanding mosquito populations, which, in turn, allow for greater RVFV transmission not only within the vector populations but to other susceptible species as well. Recently Bird et al. (3) confirmed that RVFV activity also occurs during interepizootic/interepidemic periods, albeit at much lower levels.
In Madagascar, RVFV was first isolated in 1979. At that time, the virus was isolated from mosquitoes in the absence of reports of disease in humans or livestock (12,16). Eleven years later (1990 to 1991), RVFV circulation was detected in the east coast and central highland regions of the country, with reports of human cases and high abortion rates in cattle (18)(19)(20)(21). This particular outbreak was thought to have been due to the introduction of livestock from the mainland; however, because import/export records were scarce in Madagascar, this assumption could not be confirmed. Although serological evidence indicated possible RVFV exposure of slaughterhouse workers in the mid-1990s (31), no outbreaks were observed for at least the next 15 years. In 2008, Madagascar experienced its largest RVFV outbreak to date. More than 700 suspected human cases and 26 fatalities were reported during two successive rainy seasons: January through May 2008 and November 2008 through March 2009 (1). Additionally, a large number of livestock, predominantly cattle, were affected. This outbreak came on the heels of a substantial RVFV outbreak in 2006 to 2007 on the East African mainland, predominantly in Kenya, Tanzania, and Somalia (3,17,30), and occurred during the time when RVFV cases were reported from Mayotte, the Republic of the Comoros, and South Africa between 2007 and 2009 (27,28).
Although the predictive model developed for the Horn of Africa failed to show heightened risk for RVFV infection in Madagascar in 2008 (5), a nationwide cross-sectional serological survey showed extensive and widespread RVFV exposure at that time (1). Preliminary phylogenetic analyses using partial sequences showed that Malagasy viruses were similar to those found circulating during the 2006-2007 East African outbreak; however, the initial source of the 2008 Madagascar outbreak remained unclear (1). The purpose of this study, therefore, was to generate whole genomic sequences from 2008 Malagasy outbreak samples to determine whether cases most likely represented enzootic maintenance within Madagascar or epizootic/epidemic spread or importation from East Africa.

MATERIALS AND METHODS
Three clinical specimens (liver or serum) and four RVFV isolates from the 2008 outbreak and two RVFVs isolated from the 1991 outbreak were sent from the Institut Pasteur in Madagascar to the Centers for Disease Control and Prevention (CDC) in Atlanta, GA, for genetic characterization (Table 1). Whole genomic sequences of approximately 11.9 kb were generated for each of the nine samples.
RNA isolation, RT-PCR, and complete genome sequencing. Total RNA was obtained directly from clinical specimens or after virus isolation on Vero E6 cells, using the ABI 6100 nucleic acid PrepStation (Applied Biosystems, Foster City, CA) and associated reagents, as described previously (3,4). Reverse transcriptase PCR (RT-PCR) was performed to amplify the three single-stranded genomic RNA segments (small [S], medium [M], and large [L]) using the SuperScript III one-step RT-PCR system with Platinum Taq HiFi (Invitrogen, San Diego, CA). The primers and thermal profiles were as described by Bird et al. (3,4), with annealing temperature adjustments (46 to 56°C) for some clinical samples. RT-PCR products were purified using ExoSAP-IT (USB Corporation, Cleveland, OH), and sequences were generated on an ABI 3730XL automated DNA sequencer using BigDye chemistry (version 3.1; Applied Biosystems, Foster City, CA). A total of 21, 52, and 50 sequencing primers were used for the S, M, and L segments, respectively; additional strain-specific primers were employed for difficult-to-sequence regions. For each genome, approximately 140 to 150 reads were obtained, with an average 6-fold redundancy at each nucleotide position.
Phylogenetic and coalescent analyses. In addition to the nine Madagascar samples described here, all available RVFV whole S, M, and L segment sequences (with dates of collection) in GenBank were used for the phylogenetic and coalescent analyses. A total of 120 S segment sequences, 77 M segment sequences, and 74 L segment sequences were examined (see Table S1 in the supplemental material). Multiple sequence alignments were generated in SeaView (13) using the MAFFT function (14), and nucleotide substitution models were selected based on Akaike's information criterion (AIC) in ModelTest, version 3.7 (22). The general time-reversible model with gamma distribution (GTRϩG) was employed for all three segments.
Bayesian coalescent analyses were performed in the BEAST (version 1.4.8) and Tracer (version 1.4.1) software packages (10). For each segment, preliminary analyses (10,000,000 generations each) were conducted to determine the most appropriate molecular clock (strict versus relaxed uncorrelated lognormal versus relaxed uncorrelated exponential) and demographic (constant versus Bayesian skyline population size) models. The relaxed uncorrelated exponential clock (9) and Bayesian skyline population size models were chosen for the S, M, and L segments on the basis of an analysis of marginal likelihoods (29). Final runs consisted of 100,000,000 to 175,000,000 generations to ensure effective sample sizes (ESSs) of at least 200. Maximum clade credibility trees were summarized with TreeAnnotator and were depicted using FigTree (10).
Nucleotide sequence accession numbers. The S, M, and L sequences of the 9 RVFVs sequenced in this study have been deposited in GenBank under the accession numbers given in Table 1.

Phylogeny.
Tree topologies based on all of the available RVFV S, M, and L sequences are shown in Fig. S1, S2, and S3 in the supplemental material, respectively. For the sake of clarity, abbreviated trees, including relevant S (   (Fig. 1 to 3). All RVFVs detected since the early 1980s in East Africa, Madagascar, and the Middle East are contained in a single clade and share a MRCA in the 1960s (S segment) (Fig. 1) or 1970s (M and L segments) (Fig. 2 and 3). The MRCA of the clade containing the 1991 Madagascar RVFVs occurred between 1986 and 1988 ( Fig. 1 to 3 (1). The combination of the genetic and serological data would be consistent with the view that in Madagascar, large outbreaks of RVF are associated with viruses introduced from the mainland, but the virus activity likely ramps up over the preceding year or two and likely persists at decreasing levels for a year or two following the large out-break. Unfortunately, disease surveillance is insufficiently sensitive to detect these pre-and postoutbreak periods of virus activity. Increased surveillance and genetic characterization of interepizootic virus isolates would shed light on this issue. RVFV activity during an interepizootic/interepidemic period is not surprising, especially in the years immediately preceding (3) or following (31) an outbreak. Even when the virus fails to become established in the long term, it is likely that there is a temporary enzootic state with short-term circulation for a couple of years surrounding a RVFV outbreak.
Similar patterns of disease spread have been observed in other areas. For instance, in Mayotte, RVFV cases in humans and cattle in 2007 and 2008 also appeared to be an expansion of the East African outbreak (27) (8,25,26). These cases provide additional illustrations of the potential for infectious agents to be imported from coastal countries in Africa to islands located off the African continent and highlight the risk of subsequent establishment and circulation in a new environment. This is especially important given that replication-competent vectors of several viruses can already be found in these areas (12,23,27). While infected mosquitoes have certainly contributed to the widespread dispersal of RVFV, the role of mosquitoes in the importation of RVFV into Madagascar is unknown. Given the 400 km or more of open water between Madagascar and the African mainland, it is highly improbable, though not impossible, that wind-borne mosquitoes could be the origin of the imported virus. More likely, virus-infected mosquitoes could have been unwittingly transported in air or sea shipments to Madagascar. However, the most likely route remains importation of virus-infected cattle or other livestock. In nearby Comoros, the outbreak of East Coast fever in 2003 and 2004 was attributed to the importation of cattle from Tanzania (8). Only limited information is available regarding the cattle trade to and from Madagascar, with some recent records documenting exportation of ruminants to the Comoros archipelago (1) and importation of live cattle from New Zealand (United Nations Commodity Trade Statistics Database, 2010 [http: //comtrade.un.org/db/]). No official records are available to substantiate the importation of cattle from the African mainland to Madagascar; however, the legal and illegal importation of livestock and goods is a common practice throughout the region and has already been implicated as the possible cause of a case of RVFV in Comoros in 2007 (27). Once infected livestock are introduced into an area, the disease spreads rapidly, particularly in countries, such as Madagascar, where the large-scale movement of cattle is common.
Detailed examination of the phylogenetic trees indicates considerable movement of virus-infected livestock or mosquitoes on the African mainland. It is obvious that during 2006 to 2008, the Kenya-I lineage ( Fig. 1; see also Fig. S1 in the supplemental material) was distributed across the large distances separating Kenya, Tanzania, Madagascar, and South Africa. In addition, some of the Central African viruses are dispersed in different RVFV lineages, as are Zimbabwe lineages and South African viruses. Clearly, distant movements of RVFVs are not infrequent events.
In conclusion, the generation of whole genomic sequences and advances in Bayesian coalescent methods enabled us to conclude that the reported RVFV events in Madagascar can be attributed to importation or expansion from mainland Africa. These findings underscore the importance of maintaining accurate and complete import/export records and remaining vigilant for the signs and symptoms of RVFV infection, including unexpected deaths and/or abortions, in livestock and humans.