We collected 681 small mammals during the field survey. Of these, we analyzed 214 for this study on the basis of RNA availability at the time of the study (Technical Appendix Table 2). These samples were obtained from rodents captured at 13 locations in 3 districts within the southern and eastern provinces of Sierra Leone (Figure 1; Technical Appendix Table 1). The rodents belonged to 6 genera; 199 were identified as M. natalensis, which is consistent with published data on the ubiquitous presence of this species in domestic environments in West Africa (11,12). The other rodents (identified to genus level only) were Rattus sp. (9 [4.2%] rodents); Cricetomys sp. (3 [1.4%] rodents); and Mus sp., Praomys sp., and Hylomyscus sp. (1 [0.5%] rodent each).

We screened all samples for LASV by RT-PCR using pan–Old World arenavirus (OWA) primers (17), which amplify a 670-nt section of NP gene. Because of poor results of NP amplification using OWA primers (inefficient amplification, multiple bands), we modified the screening protocol to include a second nested PCR using primers internal to the OWA amplification product, designed to amplify 650-nt segment of NP sequence and be more specific for lineage IV NP gene sequences. We obtained sequencing-quality NP amplicons from 39 samples using this protocol. For all of these NP-positive samples, we attempted to obtain RT-PCR amplicons for fragments of GPC and L genes using previously published or modified primers (17,24,28). The GPC and L amplifications were successful for most NP-positive samples and failed only in 5 and 6 samples, respectively (Table 2).

In addition, we screened a randomly selected subset of 51 samples (representing 8 collection sites and 4 rodent genera; Technical Appendix Table 2) for nucleic acids of 84 different pathogens using RPM-TEI v. 1.0. Of the 51 samples tested using RPM-TEI, 9 were positive for LASV, and no other pathogens were detected in the analyzed samples. Although the percentage of positive samples was similar to that from RT-PCR results, the RPM-TEI failed to detect LASV in 3 samples that were positive for the NP gene by 2-step RT-PCR (LM34, LM58, and LM68). All these samples originated from the same village (Bumpeh), and no sample from this site was RPM-TEI positive, which suggests that detection failures might have been caused by inefficient amplification of the sequence variant of LASV circulating in the Bumpeh area with the RPM-TEI primers. On the other hand, RPM-TEI detected viral RNA in 2 RT-PCR–negative samples (LM591 and LM649), bringing to 41 the number of samples positive for LASV nucleic acids (Table 2).

In summary, 41 LASV-positive samples were obtained from animals captured in 8 locations: Barlie (13 samples), Largo (8 samples), Bumpeh (7 samples), Ngiehun (4 samples), Koi and Yawei (3 samples each), Taiama (2 samples), and Saama (1 sample) (Figure 1). No LASV RNA was detected in samples collected from Gouma, Joru, Kenema, Panguma, or Segbwema. Lack of LASV detection in Kenema, Panguma, and Segbwema, areas well known to have regular LASV transmission, might be due to the small number of traps used. The town of Joru was extensively trapped, and no LASV was found. This finding is not surprising because Joru is south of the area where LASV is usually found.

All positive samples came from multimammate rats, which is considered the sole vector species for LASV (13). The results of LASV detection using several different RT-PCR strategies and a broad-range resequencing microarray (RPM-TEI v. 1.0) showed that none of the techniques applied alone detected viral RNA in all positive samples. This result underscores the difficulty of developing a truly universal diagnostic assay for this highly variable virus, even in the case of closely related strains belonging to lineage IV.

The analysis of the new sequences of LASV strains circulating in rodents in Sierra Leone indicated that the viral genome diversity is higher than previously estimated (17). For all available Sierra Leone sequences (including this study) the mean difference calculated for partial NP, GPC, and L sequences was 7.01% nt, 8.92% nt, and 9.83% nt, respectively, and 2.82% aa, 4.06% aa, and 0.71% aa, respectively (Table 3). These differences are higher than the reported 4.6% nt and 1.7% aa differences based on partial NP sequences in a study with fewer isolates (17). The L gene fragment seemed to vary the most at the nucleotide level, followed by GPC and NP, which is consistent with previous observations (33). However, at the amino acid level, the GPC gene varied most, followed by the NP and L genes. The high conservation of the protein sequence of L gene fragment analyzed in this study (0.71% mean difference) seemed have resulted from selection of a highly conserved part of L gene (located in RNA polymerase domain) when the diagnostic assay was designed (28).

The analysis of phylogenetic trees constructed by using all available partial sequences of NP, GPC, and L genes from Sierra Leone confirmed previous findings that the strains circulating in this country belong to lineage IV and are closely related to each other (17,19). The topology of the largest NP-based tree (Figure 2) strongly supports the hypothesis that the isolates from Sierra Leone belong to at least 3 distinct major clades (posterior probability 1.00 in all cases): the first clade (A), including a large cluster of strains originating from a group of villages to the north and east of Kenema in the Eastern Province (Bumpeh, Gondama, Koi, Konia, Largo, Ngiehun, Panguma, Segbwema, Taiama, Tongo, and Yawei; Figure 1); the second clade (B), including several strains isolated from rodents captured in Barlie (located a few kilometers southeast of Bo) and 1 isolate from Saama (located northeast of Kenema); and the third clade (C) represented by just 2 older human isolates from Mano and Mobai.

Phylogenetic trees based on GPC and L sequences (Figures 3, 4) had similar topology and supported existence of clades A (with posterior probabilities 0.74 and 1.00, respectively) and B (with posterior probabilities 1.00 for both trees). However, the clade C was not present because the sequences for GPC and L gene fragments were not available for the strains forming this cluster in the NP-based tree. In addition to clades A and B, GPC- and L-based trees suggested existence of 2 additional and distinct clades. Clade D was represented by 2 sequences from human isolates SL25 and SL26, which formed a separate cluster (posterior probability 1.00 for both trees), and clade E represented by sequences obtained from a single strain isolated in 2006 (SL06–2057). These clades are defined by a very small number of sequences, and the GPC- and L-based trees disagree on the order of their separation from other clades. In addition, no data have been published on geographic origin of clade D and E samples. More data are needed (including corresponding NP sequences) to establish the existence and position of clades D and E with more certainty.

All of the trees indicate a high degree of geographic clustering of the strains. This kind of clustering has been reported previously over large geographic distances and is believed to have resulted from limited dispersal and migration of the host species (17,19). Results of this study show that this phenomenon also can be observed over relatively short distances. Isolates originating from multimammate rat specimens obtained in a particular location tended to cluster, and conversely sequences present in specific branches of the trees in many cases originated from a single location or few locations not far from each other. This kind of clustering could be observed especially well in samples from Barlie, Largo, Bumpeh, Konia, and Yawei (Figures 2–4).

In addition to the general pattern of geographic clustering, in several cases single isolates clustered with strains from different locations. For example, the sequence from a single sample from Saama (LM513) was closely related to that of strains from Barlie (based on NP sequence analysis). In another example, 1 GPC sequence originating from Liberia (523) clustered with Sierra Lone clade A sequences. In some cases (e.g., Saama sample LM513), such unusual clustering patterns may be explained by cross contamination or mislabeling of the samples. They also might result from relative proximity of all sampling sites and inadvertent anthropogenic transfer of rodents. Massive population movements that occurred in Sierra Leone during the 1991–2002 civil war could contribute to the process of mixing multimammate rat subpopulations carrying different LASV strains (34).

The geographic location of human cases at such a fine spatial scale can be problematic because humans can move large distances after exposure before disease is detected. For the human isolates, the clustering inconsistent with geographic location might have resulted from recording of the hospital location or patient’s current location as strain’s origin instead of the actual location of rodent–human transmission. For example, the NP-based phylogenetic tree indicates that human isolates from Segbwema and Gondama (obtained in 1996 and 1977, respectively) most likely originated from the Yawei village area because they cluster closely. A few other human isolates (SL15, SL20, and SL21) for which no location information is available also clustered with Yawei isolates on the basis of GPC and L sequences, suggesting their origin in the same area. These sequences were obtained in 2002 from United Nations peacekeepers stationed in this part of Sierra Leone (28,35).

Recent epidemiologic data show that LF was detected in 10 of 13 districts in Sierra Leone, which suggests that the infection is much more common that previously recognized (36). Phylogenetic analysis of the sequences revealed that strains circulating in districts to the west of the traditional hyperendemic area from which most sequence information is available differ significantly (clade B), which suggests that these could be distinct LASV strains that circulated in local multimammate rat populations for a long time since diverging from a common ancestor and are unlikely to have resulted from recent expansion of this rodent to new areas, as was recently suggested to explain emergence of cases from districts in which LF was not previously reported (36). Furthermore, the presence of LASV in Barlie with such high prevalence was surprising because this area historically has had few reports of LASV until 2 human LF cases reported in 2009 (L.M. Moses, unpub. data). The lack of reported LF cases from this area leads to speculation that clade B may be a less pathogenic form of LASV, and transmission to humans might have occurred previously but went unrecognized because of milder, nonhemorrhagic symptoms. In fact, the idea of broader area of LASV endemicity in Sierra Leone is consistent with results of serosurveys conducted during the 1980s by McCormick, who found seroprevalence levels ranging from 8% in southern coastal areas to 15% in villages in Northern Province (6).

Molecular characterization of isolates from a wider geographic area of the country is needed to fully understand the diversity of the LASV strains in Sierra Leone and its impact on disease distribution and risk. Such information would be useful for developing efficient viral detection technologies, for example, enabling design of PCR primers and antibodies specific for a broad range of LASV types. These diagnostic tests are extremely relevant to disease surveillance and monitoring and evaluation of interventions to prevent primary LASV infection in humans. More extensive information about sequence diversity affecting the antigenicity of the virus or the function of its RNA-dependent RNA polymerase may help in the development of vaccines and antiviral drugs. It will also lead to deeper understanding of the biology and pathogenesis of LASV.

Rodent trapping procedures; coordinates of sampling locations; and list of all analyzed samples.