Has Rift Valley fever virus evolved with increasing severity in human populations in East Africa?

Abstract

Rift Valley fever (RVF) outbreaks have occurred across eastern Africa from 1912 to 2010 approximately every 4-15 years, most of which have not been accompanied by significant epidemics in human populations. However, human epidemics during RVF outbreaks in eastern Africa have involved 478 deaths in 1998, 1107 reported cases with 350 deaths from 2006 to 2007 and 1174 cases with 241 deaths in 2008. We review the history of RVF outbreaks in eastern Africa to identify the epidemiological factors that could have influenced its increasing severity in humans. Diverse ecological factors influence outbreak frequency, whereas virus evolution has a greater impact on its virulence in hosts. Several factors could have influenced the lack of information on RVF in humans during earlier outbreaks, but the explosive nature of human RVF epidemics in recent years mirrors the evolutionary trend of the virus. Comparisons between isolates from different outbreaks have revealed an accumulation of genetic mutations and genomic reassortments that have diversified RVF virus genomes over several decades. The threat to humans posed by the diversified RVF virus strains increases the potential public health and socioeconomic impacts of future outbreaks. Understanding the shifting RVF epidemiology as determined by its evolution is key to developing new strategies for outbreak mitigation and prevention of future human RVF casualties.

Figure 1

Map of eastern African countries indicating frequencies of major RVF outbreaks over the past century (1912–2010). The numbers of outbreaks in specific countries are indicated in the gray spheres.

Figure 2

Maximum likelihood phylogenetic tree of complete RVFV genome sequences (combined L (GenBank accessions: DQ375400–1, DQ375406, DQ375410, DQ375427, DQ375429, EU574004, EU574017, EU574020, EU574029, HM586953–60, JF311371, JF311375–6 and JQ820488–91), M (GenBank accessions: DQ380190–1, DQ380196–7, DQ380200, DQ380205, JQ820488–91, EU574031, EU574044, EU574047, EU574055, HM586964–71, JF311368–9, JF311371, JF311375–6 and JQ820483–6) and S segments (GenBank accessions: DQ380145, DQ380149, DQ380156, DQ380169–70, DQ380176, EU574057, EU574086, EU574072, EU574075, HM586975–82, JF311386–7, JF311389, JF311393–4, JQ820472, JQ820474, JQ820476 and JQ820477)) analyzed using PhyML v. 3.0. The phylogenies employed the General Time-Reversible nucleotide substitution (rate categories=4) model, in which the base frequencies and the relative substitution rates between them were estimated by maximizing the likelihood of the phylogeny. For estimating the tree topology, both nearest-neighbor interchange and sub-tree pruning and regrafting improvements were used. Country (bold), isolate identification (italics), year and host are indicated for all 27 sequences analyzed. Bootstrap values at the major nodes are expressed as percentage agreement among 1000 replicates. The branch length scale represents substitutions per site. A red sequence indicates a reassortant isolate.

Figure 3

Maximum likelihood phylogenetic tree of complete RVFV M segment sequences. Country (bold), isolate identification (italics), GenBank accession (in brackets), year and host are indicated for all 27 sequences analyzed. Bootstrap values at the major nodes are expressed as percentage agreement among 1000 replicates. The branch length scale represents substitutions per site. A red sequence indicates a reassortant isolate.